{"@context":{"@language":"en","Affiliation":"http:\/\/vivoweb.org\/ontology\/core#departmentOrSchool","AggregatedSourceRepository":"http:\/\/www.europeana.eu\/schemas\/edm\/dataProvider","Campus":"https:\/\/open.library.ubc.ca\/terms#degreeCampus","Creator":"http:\/\/purl.org\/dc\/terms\/creator","DateAvailable":"http:\/\/purl.org\/dc\/terms\/issued","DateIssued":"http:\/\/purl.org\/dc\/terms\/issued","Degree":"http:\/\/vivoweb.org\/ontology\/core#relatedDegree","DegreeGrantor":"https:\/\/open.library.ubc.ca\/terms#degreeGrantor","Description":"http:\/\/purl.org\/dc\/terms\/description","DigitalResourceOriginalRecord":"http:\/\/www.europeana.eu\/schemas\/edm\/aggregatedCHO","FullText":"http:\/\/www.w3.org\/2009\/08\/skos-reference\/skos.html#note","Genre":"http:\/\/www.europeana.eu\/schemas\/edm\/hasType","GraduationDate":"http:\/\/vivoweb.org\/ontology\/core#dateIssued","IsShownAt":"http:\/\/www.europeana.eu\/schemas\/edm\/isShownAt","Language":"http:\/\/purl.org\/dc\/terms\/language","Program":"https:\/\/open.library.ubc.ca\/terms#degreeDiscipline","Provider":"http:\/\/www.europeana.eu\/schemas\/edm\/provider","Publisher":"http:\/\/purl.org\/dc\/terms\/publisher","Rights":"http:\/\/purl.org\/dc\/terms\/rights","RightsURI":"https:\/\/open.library.ubc.ca\/terms#rightsURI","ScholarlyLevel":"https:\/\/open.library.ubc.ca\/terms#scholarLevel","Supervisor":"http:\/\/purl.org\/dc\/terms\/contributor","Title":"http:\/\/purl.org\/dc\/terms\/title","Type":"http:\/\/purl.org\/dc\/terms\/type","URI":"https:\/\/open.library.ubc.ca\/terms#identifierURI","SortDate":"http:\/\/purl.org\/dc\/terms\/date"},"Affiliation":[{"@value":"Business, Sauder School of","@language":"en"},{"@value":"Finance, Division of","@language":"en"}],"AggregatedSourceRepository":[{"@value":"DSpace","@language":"en"}],"Campus":[{"@value":"UBCV","@language":"en"}],"Creator":[{"@value":"Guan, Xuejing","@language":"en"}],"DateAvailable":[{"@value":"2022-08-23T17:11:49Z","@language":"en"}],"DateIssued":[{"@value":"2022","@language":"en"}],"Degree":[{"@value":"Doctor of Philosophy - PhD","@language":"en"}],"DegreeGrantor":[{"@value":"University of British Columbia","@language":"en"}],"Description":[{"@value":"This thesis consists of three essays studying the impacts of economic uncertainty on the financial markets. The first essay examines the impact of economic uncertainty on firms\u2019 decisions to go private. Using an instrumental variable approach, I show that firms are more likely to go private following economic uncertainty shocks. The effect is stronger for firms prone to severe agency conflicts. After going private, the cost of debt decreases. These results are consistent with uncertainty exacerbating agency frictions faced by public companies. Firms go private to alter their capital structures to be less prone to agency frictions: ones with a small number of dominant stakeholders with aligned interests. The agency frictions are mitigated through going private, resulting in a decrease in the cost of debt.\r\n        The second essay examines how the money creation function of banks affects the relative cost of firm financing in the bank loan vs. bond market \u2013 the loan-bond spread. Using a sample of loans and bonds issued by the same firm, the essay finds a lower loan-bond spread for firms impacted by positive information cost shocks. We call this decline in the relative cost of bank credit induced by firm information cost shock the opacity discount and show that it is consistent with the \u201cmoney creation\u201d hypothesis in the financial intermediation theory, which suggests that banks need to keep information about their assets secret to produce private money.\r\n        The third essay studies how firms use earnout, a contingent payment contract in M&A, to manage valuation risks under uncertainty. I find that the usage of earnouts positively correlates with target uncertainty. The likelihood of deal completion increases significantly with earnouts. Despite the benefits of bridging the valuation gap, an earnout can introduce incentive misalignment problems in the post-transaction period. After the transaction, the acquirer\u2019s objective is to maximize firm value, while the target\u2019s objective is to maximize earnout payments. Such incentive misalignments can destroy firm value. The essay documents a negative impact on acquirer wealth gains when earnouts are not used to manage valuation risks.","@language":"en"}],"DigitalResourceOriginalRecord":[{"@value":"https:\/\/circle.library.ubc.ca\/rest\/handle\/2429\/82456?expand=metadata","@language":"en"}],"FullText":[{"@value":"Essays in Empirical Corporate Finance:The Impact of Economic Uncertainty on the Financial MarketsbyXuejing GuanB.COM., University of Toronto, 2013M.A., University of Toronto, 2014A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THEDEGREE OFDOCTOR OF PHILOSOPHYinTHE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES(Business Administration - Finance)THE UNIVERSITY OF BRITISH COLUMBIA(Vancouver)August 2022\u00a9 Xuejing Guan, 2022The following individuals certify that they have read, and recommend to the Faculty of Graduateand Postdoctoral Studies for acceptance, the dissertation entitled:Essays in Empirical Corporate Finance: The Impact of Economic Uncertainty on the FinancialMarketssubmitted by Xuejing Guan in partial fulfillment of the requirements forthe degree of Doctor of Philosophyin Business Administration - FinanceExamining Committee:Professor Jan Bena, Finance Division, Sauder School of Business, UBCSupervisorProfessor Murray Carlson, Finance Division, Sauder School of Business, UBCCo-supervisorProfessor Viktoriya Hnatkovska, Vancouver School of Economics, UBCUniversity ExaminerProfessor Jenny Zhang, Sauder School of Business, UBCUniversity ExaminerProfessor Alfred Lehar, Haskayne School of Business, University of CalgaryExternal ExaminerAdditional Supervisory Committee Members:Professor Michael Devereux, Vancouver School of Economics, UBCSupervisory Committee MemberProfessor Jack Favilukis, Finance Division, Sauder School of Business, UBCSupervisory Committee MemberProfessor Will Gornall, Finance Division, Sauder School of Business, UBCSupervisory Committee MemberiiAbstractThis thesis consists of three essays studying the impacts of economic uncertainty on the finan-cial markets. The first essay examines the impact of economic uncertainty on firms\u2019 decisionsto go private. Using an instrumental variable approach, I show that firms are more likely to goprivate following economic uncertainty shocks. The effect is stronger for firms prone to severeagency conflicts. After going private, the cost of debt decreases. These results are consistentwith uncertainty exacerbating agency frictions faced by public companies. Firms go private toalter their capital structures to be less prone to agency frictions: ones with a small number ofdominant stakeholders with aligned interests. The agency frictions are mitigated through goingprivate, resulting in a decrease in the cost of debt.The second essay examines how the money creation function of banks affects the relativecost of firm financing in the bank loan vs. bond market \u2013 the loan-bond spread. Using a sampleof loans and bonds issued by the same firm, the essay finds a lower loan-bond spread for firmsimpacted by positive information cost shocks. We call this decline in the relative cost of bankcredit induced by firm information cost shock the opacity discount and show that it is consis-tent with the \u201cmoney creation\u201d hypothesis in the financial intermediation theory, which sug-gests that banks need to keep information about their assets secret to produce private money.The third essay studies how firms use earnout, a contingent payment contract in M&A, tomanage valuation risks under uncertainty. I find that the usage of earnouts positively correlateswith target uncertainty. The likelihood of deal completion increases significantly with earnouts.Despite the benefits of bridging the valuation gap, an earnout can introduce incentive misalign-ment problems in the post-transaction period. After the transaction, the acquirer\u2019s objective isto maximize firm value, while the target\u2019s objective is to maximize earnout payments. Suchincentive misalignments can destroy firm value. The essay documents a negative impact onacquirer wealth gains when earnouts are not used to manage valuation risks.iiiLay SummaryThis thesis contains three essays in empirical corporate finance, with a focus on the impacts ofeconomic uncertainty on the financial markets. The first essay investigates how economic un-certainty affects companies in the equity market. It documents that companies use private eq-uity to opt out of public markets to enhance corporate governance and lower their cost of cap-ital. The second essay studies the impact of uncertainty on firms\u2019 relative cost of debt throughthe money creation function of banks. The essay shows that firms experiencing positive uncer-tainty shocks receive relatively lower cost of debt from banks than from the bond market. Thethird essay focuses on the impact of economic uncertainty in the M&A market. The essay findsthat firms adopt contingent payment contracts to manage valuation risks during high uncertainperiods. The contingent payment contracts help facilitate deal completion while introducingnew moral hazard problems in the post-transaction periods.ivPrefaceChapters 2 and 4 of this thesis are solely my own work. Chapter 3 is a co-authored project withProfessor Jan Bena and Professor Isha Agarwal. We contributed equally to this project.vTable of ContentsAbstract. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiLay Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xList of Figures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiAcknowledgements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiiDedication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 Economic Uncertainty and Going Private Transactions: The Corporate GovernanceChannel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42.2 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122.2.1 Going Private Sample. . . . . . . . . . . . . . . . . . . . . . . . . . . 122.2.2 Descriptive Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . 152.3 Empirical Methodology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172.3.1 Cox Proportional Hazards Model . . . . . . . . . . . . . . . . . . . . . 172.3.2 Measuring Uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . . 182.3.3 Control Function Approach . . . . . . . . . . . . . . . . . . . . . . . 222.3.4 Matching Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23vi2.3.5 Difference-in-differences Analysis of the Impacts of Going Private on LoanRate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252.4.1 Main Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252.4.2 Estimation Results Based on Matched Control Samples . . . . . . . . . 262.4.3 Agency Problems and Going Private Transactions . . . . . . . . . . . . 272.4.4 Impacts of Going Private on Loan Rate . . . . . . . . . . . . . . . . . . 302.4.5 Subsample Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 302.5 Alternative Explanations and Robustness Tests. . . . . . . . . . . . . . . . . . 312.5.1 Alternative Explanations . . . . . . . . . . . . . . . . . . . . . . . . . 322.5.2 Robustness Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353 Relative Pricing of Private and Public Debt: The Role of Money Creation Channel . . 603.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 603.2 Conceptual Framework and Hypotheses . . . . . . . . . . . . . . . . . . . . . 693.3 Sample, Data, and Empirical Methodology . . . . . . . . . . . . . . . . . . . . 713.3.1 Data and Sample Construction . . . . . . . . . . . . . . . . . . . . . . 713.3.2 Summary Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 733.3.3 Information Cost Shock and the Loan-bond Spread . . . . . . . . . . . 753.4 Main Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 793.4.1 Information Cost and the Loan-bond Spread . . . . . . . . . . . . . . 793.4.2 Evidence on the Money Creation Channel . . . . . . . . . . . . . . . . 823.5 Alternative Explanations and Robustness Tests. . . . . . . . . . . . . . . . . . 873.5.1 Alternative Explanations . . . . . . . . . . . . . . . . . . . . . . . . . 87vii3.5.2 Robustness Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 913.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 924 Earnout: Managing Valuation Risks in Mergers and Acquisitions under Uncertainty 1074.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1074.2 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1134.2.1 Sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1134.2.2 Measuring Uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . . 1154.2.3 Descriptive Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . 1164.3 Empirical Methodology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1184.3.1 Earnout and M&A Deal Completion . . . . . . . . . . . . . . . . . . . 1184.3.2 Earnout and Acquirer Wealth Gains . . . . . . . . . . . . . . . . . . . 1194.3.3 Earnout and Target Industry Uncertainty . . . . . . . . . . . . . . . . 1204.3.4 Matching Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1214.3.5 Earnout Misuse. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1224.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1234.4.1 Earnout and M&A Deal Completion: Logistic Estimation Results . . . . 1234.4.2 Earnout and Acquirer Wealth Gains: OLS Estimation Results . . . . . . 1244.4.3 Estimation Results on Earnout and Target Industry Uncertainty . . . . . 1254.4.4 Results on Matching Analysis. . . . . . . . . . . . . . . . . . . . . . . 1264.4.5 Earnout Misuse and Acquirer Wealth Gains: OLS Estimation Results . . 1274.4.6 Robustness Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1284.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1305 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157viiiAppendices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166Appendix A. Appendix to Chapter 2 . . . . . . . . . . . . . . . . . . . . . . . . . . 166A.1 Variable Definition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166A.2 Additional Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168Appendix B. Appendix to Chapter 3 . . . . . . . . . . . . . . . . . . . . . . . . . . 175B.1 Variable Definition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175B.2 Instrument Variables Construction . . . . . . . . . . . . . . . . . . . . . 177B.3 Additional Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178Appendix C. Appendix to Chapter 4 . . . . . . . . . . . . . . . . . . . . . . . . . . 190C.1 Variable Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190C.2 Measuring Macroeconomic Uncertainty . . . . . . . . . . . . . . . . . . 192C.3 Additional Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193ixList of Tables2.1 Descriptive Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 442.2 Uncertainty Shocks and Going Private Transactions . . . . . . . . . . . . . . . . . . 462.3 Uncertainty Shocks and Going Private Transactions: Matching Analysis on IPOand Pre-delisting Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 472.4 Shareholder Conflicts and Going Private Transactions . . . . . . . . . . . . . . . . 502.5 Shareholder-creditor Conflicts and Going Private Transactions . . . . . . . . . . . 522.6 Bank Loan Rates of the Going-Private Firms . . . . . . . . . . . . . . . . . . . . . . 552.7 Uncertainty Shocks and Going Private Transactions: Subsample Analysis . . . . . 562.8 Alternative Explanations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 583.1 Summary Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 953.2 Firm Information Cost Shock and the Loan-Bond Spread . . . . . . . . . . . . . . 963.3 Firm Information Cost and the Loan-Bond Spread: Alternative Measures . . . . . 993.4 Money Creation Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1003.5 Relationship Lending and the Loan-Bond Spread . . . . . . . . . . . . . . . . . . . 1033.6 Information Cost and Loan Share . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1053.7 Information Cost and Bank Participation in the Loan Syndicate . . . . . . . . . . . 1064.1 Descriptive Statistics: Full Sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1394.2 Earnout and M&A Deal Completion . . . . . . . . . . . . . . . . . . . . . . . . . . . 1414.3 Earnout and Acquirer Announcement Returns of M&A Transactions . . . . . . . . 1434.4 Earnout and Target Industry Uncertainty Shock . . . . . . . . . . . . . . . . . . . . 1454.5 Matching Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1494.6 Earnout Misuse and Acquirer Announcement Returns of M&A Transactions . . . 154xList of Figures2.1 Going Private Transactions by Industry: 1994-2017 . . . . . . . . . . . . . . . . . . 372.2 Capital Structure before vs. after Going Private . . . . . . . . . . . . . . . . . . . . 382.3 Differences in Loan Rates Between Going Private Firms and Control Firms . . . . 402.4 Cumulative Abnormal Returns of the Going Private Companies . . . . . . . . . . . 412.5 Uncertainty Shocks by Industry: 1994-2017 . . . . . . . . . . . . . . . . . . . . . . . 423.1 The Loan-Bond Spread: 1995-2020 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 933.2 The Loan-Bond Spread across Firm Information Cost Shock Quartiles . . . . . . . 944.1 Fraction of M&A Transactions with Earnout: 1991-2019 . . . . . . . . . . . . . . . 1314.2 Fraction of M&A Transactions with Earnout: Within Industry . . . . . . . . . . . . 1344.3 Fraction of Earnout Transactions by Industry . . . . . . . . . . . . . . . . . . . . . 137xiAcknowledgementsI would like to express my deepest gratitude to my advisors. This work would not have beenpossible without their tremendous guidance and support. I am forever indebted to my mainsupervisor, Jan Bena, for his help throughout my doctoral study, and especially for his confi-dence in me. I learned immensely about research by interacting and working with him. Moreimportantly, his positive attitude towards work and life encourages me to be a better version ofmyself. I would also like to sincerely thank my co-supervisor, Murray Carlson, who guided methrough the theoretical foundations of my work and taught me to think about the economicsbehind the empirical findings. His intriguing questions encouraged me to always stay curious.I thank Will Gornall for providing invaluable advice through every stage of my Ph.D., from thefirst-year summer paper to the job market experience. I benefit significantly from his com-ments and suggestions. I thank Jack Favilukis for sharing his insights on the topic of economicuncertainty, and for providing helpful feedback on my presentations. I thank Michael Devereuxfor the advice on my research with his expertise in international economics and finance.I also thank the other faculty members in the Finance division. Each of them helped me indifferent ways throughout my years at Sauder. In particular, I thank Kai Li and Hernan Ortiz-Molina for their support, especially in the early years of my Ph.D. studies. I also thank IshaAgarwal and Bo Bian for their help during the job market process. I thank Caren Lombard forbeing my mentor in teaching. I learned a great deal from her, both about teaching and life. Andthanks to Sally Bei and Elaine Cho for their extraordinary responsiveness and help.My time in Vancouver was a memorable experience. I am fortunate to have amazing friendsand colleagues to walk along this journey. I would like to thank my friends and colleagues inthe Ph.D. program, with particular thanks to the job market squad: Ellen, Iris, and Valentina.I enjoyed our discussions a lot. Thanks to Ella and Kevin for sharing the laughter and tearstogether. Thanks to Weiyi for always being there, encouraging me, and listening to my thoughts.xiiDedicationI would like to dedicate this thesis to my family. I thank my parents, Xue and Mingyang, for theirunconditional love and support. Thanks to my husband, Changbo, for always standing by me,believing in me, cheering me up, and going through the ups and downs together with me.In memory of my grandmother, Guoqing Shi, who has always saved the best for me. I hopeyou are proud of me.xiiiChapter 1IntroductionEconomic uncertainty plays a vital role in economic outcomes, especially during economicdownturns. This thesis is a collection of three essays studying the impact of economic uncer-tainty on the financial markets. In particular, it empirically investigates how economic uncer-tainty affects companies in the equity market, the debt market, and in mergers and acquisitions.In the first essay, I study how economic uncertainty affects companies\u2019 choice of public vs. pri-vate equity. I show that uncertainty exacerbates the agency frictions faced by public compa-nies. As a response, firms go private to mitigate these agency frictions. They use private equityto opt out of public markets to enhance corporate governance and lower their cost of capi-tal. The second essay focuses on the impact of economic uncertainty on firms\u2019 relative cost ofdebt. We show that firms experiencing positive uncertainty shocks receive relatively lower costof debt from banks than from the bond market. The results are consistent with the financial in-termediation theory: banks offer opacity discounts to firms with high information productioncosts because lending to such companies reduces banks\u2019 cost of private money creation. Thethird essay focuses on the impact of economic uncertainty in the M&A market. I find that firmsadopt contingent payment contracts to manage valuation risks during high uncertain periods.The contingent payment contracts help facilitate deal completion while introducing new moralhazard problems in the post-transaction periods.In the first essay, \u201cEconomic Uncertainty and Going Private Transactions: The CorporateGovernance Channel\u201d, I investigate how firms change their capital structures to ones that areless prone to agency frictions to alleviate the negative impacts of uncertainty. I show that firmsare more likely to go private following economic uncertainty shocks. This effect is stronger for1firms prone to severe agency conflicts: firms with dual-class structure, less institutional own-ership, lower asset redeployability, lower loan-to-bond ratio, and for firms in financial distress.After going private, the cost of debt decreases. The results are consistent with the corporate gov-ernance hypothesis, where uncertainty exacerbates the agency frictions faced by public com-panies and increases the agency cost of capital. To alleviate the negative impacts of uncertainty,firms go private to alter their capital structures from dispersed to ones with a very small num-ber of dominant stakeholders with aligned interests. The agency frictions are mitigated throughgoing private, resulting a decrease in the cost of debt.The second essay, \u201cRelative Pricing of Private and Public Debt: The Role of Money CreationChannel\u201d, examines how the money creation function of banks affects the relative cost of firmfinancing in the bank loan v.s. public bond market. Using economic uncertainty and othermeasures as proxies for the cost of information production, we show that firms impacted bypositive information cost shocks have lower cost of bank loans relative to the cost of corporatebonds. We call this decline in the relative cost of bank credit induced by firm information costshock the opacity discount. We argue that it is consistent with the \u201cmoney creation\u201d hypoth-esis in the theory of financial intermediation: To produce private money, banks need to keepinformation about their assets secret. Therefore, they offer discounts when lending to opaquefirms.In the third essay, \u201cEarnouts: Managing Valuation Risks in Mergers and Acquisitions Un-der Uncertainty\u201d, I study how firms respond to increased valuation risks following uncertaintyshocks in mergers and acquisitions. I find that firms are more likely to use earnouts, a con-tingent payment contract, when target uncertainty is high. The usage of earnouts increasesdeal completion rates significantly. Despite the benefits of bridging the valuation gap betweenbuyers and sellers, acquirers announcement returns are insignificantly different from those ofthe deals without earnout. This suggests that there can be costs associated with the earnoutcontracts. The contingent payment mechanism can introduce agency conflicts in the post-acquisition period. After the transaction, acquires\u2019 objective is to maximize firm value, while2targets\u2019 goal is to maximize earnout payments. Such incentive misalignment can destroy firmvalue. I find that acquirers experience negative cumulative abnormal returns when earnoutsare not used to manage the valuation risks.Since the three essays comprising this thesis are in separate topics, chapters are designed tobe self-contained. Each chapter discusses the relevant literature and contains its own introduc-tion and conclusion. A general conclusion of the three chapters is provided at the end of thisthesis.3Chapter 2Economic Uncertainty and Going PrivateTransactions: The Corporate GovernanceChannel2.1 IntroductionEconomic uncertainty plays a vital role in economic outcomes, especially during downturns.Uncertainty shocks reduce economic growth, hamper stock market performance, and makefirms reduce investment and employment leading to lower sales growth and profitability.1 Thenegative impact of economic uncertainty is amplified by the real and financial frictions faced byfirms: Alfaro et al. (2021) show that, in the presence of these frictions, uncertainty shocks leadto larger recessions with slower recovery. While prior work documents the negative impact ofuncertainty shocks on firms, our understanding of how firms respond to such shocks in orderto lessen their impacts is minimal.2In this chapter, I investigate how firms change their capital structures to ones that are lessprone to agency frictions to alleviate the negative impacts of uncertainty shocks. Specifically,I study whether going private transactions\u2014events in which firms\u2019 capital structures are al-tered from dispersed to ones with a very small number of dominant stakeholders with alignedinterests\u2014is a possible mechanism by which firms respond to uncertainty shocks. The level1Bloom (2009); Mian and Sufi (2010); Pastor and Veronesi (2012); Kahle and Stulz (2013); Alfaro et al. (2021).2Im et al. (2017) and Alfaro et al. (2021) find that firms adopt more conservative corporate policies such asmore cash holdings and less dividend payouts.4of economic uncertainty has risen significantly in the wake of the COVID-19 pandemic.3 Inthis period, we saw a resurgence of the going private transactions. These transactions receiverecord-high premiums in the years of 2020-21. The media describe the relationship betweenuncertainty and going private as follows: \u201cGoing-private transactions are cyclical in nature andtend to increase in number during economic downturns, where a variety of factors can causethe share price of a listed company to trade at a discount to its net asset value per share. 2020 isa case in point, as global stock markets saw increased volatility due to the Covid-19 pandemicand macroeconomic uncertainty.\u201d4Agency frictions constitute a theoretically important cost for public companies. The sep-aration of ownership and control creates conflicts between managers and shareholders, andbetween creditors and shareholders (Jensen and Meckling 1976). Conflicts of interest also existbetween controlling and minority shareholders. These agency problems can generate financialfrictions and increase the cost of external capital. Existing literature documents that investorsand lenders require higher rate of returns to compensate for the agency costs (La Porta et al.2002; Aslan and Kumar 2012).Uncertainty can exacerbate firms\u2019 agency problems through a variety of channels.5 First, itcan aggravate information asymmetry, increasing the costs of signaling and monitoring. Previ-ous studies show that firms increase voluntary disclosure to reduce information asymmetry inresponse to uncertainty shocks (Balakrishnan et al. 2014; Guay et al. 2016).Second, moral hazard problems between shareholders and creditors and between managersand shareholders are more severe with high uncertainty. Cash flows become more volatile, cre-ating risk-shifting incentives for shareholders to exploit creditors. Managers may also expro-priate more from shareholders when outcomes are uncertain. In addition, firms tend to havemore cash holdings following uncertainty (Im et al. 2017), which can be easily turned into pri-3https:\/\/voxeu.org\/article\/economic-uncertainty-wake-covid-19-pandemic4Finanical Times, Oct 2021.5While most agency problems are exacerbated with uncertainty, the underinvestment problem is mitigatedfollowing uncertainty shocks. Uncertainty increases the outcome dispersion of investment opportunities, whichincreases the potential returns to shareholders and reduces their incentives to forego valuable projects.5vate benefits by management. The equity-based incentive mechanism may also become lesseffective since firm performance is highly volatile despite management efforts.Third, uncertainty magnifies the coordination frictions among managers, shareholders, andcreditors. Garlappi et al. (2017, 2021) find that heterogeneous priors can lead to inefficiencieswhen decisions are made collectively by a group of agents. The coordination frictions are moresevere following uncertainty shocks because agents\u2019 beliefs about future outcomes may becomemore dispersed. Moreover, these coordination frictions make firms less responsive to uncer-tainty shocks. While uncertainty triggers the need for companies to restructure their assets,disagreements among agencies make the negotiation process difficult. The frictions need tobe resolved before firms can implement the changes. Garlappi et al. (2017) show that the inef-ficiencies due to coordination frictions may be resolved when agents can trade among them-selves or collectively trade with outside investors.Due to these agency frictions, firms experience higher costs of capital during periods of highuncertainty (P\u00e1stor and Veronesi 2013; Gilchrist et al. 2014; Ashraf and Shen 2019; Kaviani et al.2020). Alfaro et al. (2021) show that the financial frictions amplify, prolong, and propagate thenegative impact of uncertainty shocks. They argue that even small financial adjustment costscould generate significant impacts. The elevated agency costs following uncertainty shocks cre-ate an incentive for firms to address the agency problems. In this chapter, I postulate that onepossible way to mitigate the agency frictions is to restructure the capital via going private. Ingoing private transactions, firms alter their capital structure from dispersed to ones with a verysmall number of dominant stakeholders with aligned interests. Based on these arguments, Ihypothesize that firms are more likely to go private following uncertainty shocks. The effectsare expected to be stronger for firms that are prone to severe agency problems.To study the impacts of economic uncertainty on going private, I collect a sample of firmsthat went private from 1994 to 2017 and compare them with those that remain public. FollowingDeAngelo et al. (1984), Leuz et al. (2008) and Bharath and Dittmar (2010), I identify the goingprivate sample as those that filed Schedule 13E-3 and delisted from the stock exchange within6two years. A publicly-traded company must file Schedule 13E-3 if the company or an affiliatevoluntarily engages in a transaction resulting in the delisting of the company\u2019s shares. Figure2.1 illustrates the number of going private transactions across industries from 1994 to 2017.The period 1994-2006 saw a boom in the going private transactions due to the development ofthe private equity market and the increase in compliance costs for public companies after theSarbanes-Oxley (SOX) Act. The number of going private transactions decreased dramaticallyafter the financial crisis because of the contractionary debt market. Going private transactionsexperienced a resurgence in recent years attributed to the high level of uncertainty.By reading through the Schedule 13E-3 filings of the going private firms, I find significantchanges in firms\u2019 capital structures through the going private process. Figure 2.2 illustrates thechanges. Panel A and B compare the capital structures of a representative company, Ameri-can Greetings Corp., before and after it went private. Before going private, the company haddual class shares with a large number of institutional and dispersed shareholders. After goingprivate, the company was owned entirely by management and a private equity investor. Thedebt structure also became less complex after going private. Existing loans were paid off withnew loan facilities arranged by one syndicate with previous lending relationships with the PEinvestor. Panel C of Figure 2.2 illustrates the capital structure of the average company after go-ing private, demonstrating a similar pattern as in Panel B. Firms alter their capital structuresthrough the going private process. The capital structure before going private is prone to severeagency frictions. After going private, agency problems are mitigated since the management, thePE investor, and the creditors share aligned interests.I measure firm uncertainty using changes in realized stock return volatility. Using a Coxproportional hazards model, I find that firms are more likely to go private with high uncer-tainty. An one standard deviation increase in the change of annualized stock return volatilityleads to a 14% increase in the hazard rate of going private. One concern with using changesin stock return volatility as a proxy for uncertainty is that firm characteristics can simultane-ously affect stock return volatility and the going private decisions. For example, stock liquidity7affects stock return volatility, and firms may choose to delist due to the lack of liquidity. Inaddition, the decision to go private may affect stock return volatility reversely. To address theendogeneity concern, I employ an instrumental strategy following Alfaro et al. (2021). I con-struct the instruments exploiting firms\u2019 differential exposure to aggregate uncertainty shocks inenergy, currency, policy, and U.S. Treasury notes. The instruments, by construction, capture thechanges in firm-level stock return volatility which are induced by exogenous uncertainty shocksto macro variables. Using a control function approach with the instrumental variables, I findthat firms experiencing high uncertainty which is induced by aggregate economic uncertaintyshocks are more likely to go private. The results are robust when I control for macroeconomicconditions such as GDP growth, investor sentiment, indicators for NBER recessions, VIX, or theterm premium. The results also hold when I conduct a propensity score matching based on theinitial conditions at IPO and firm characteristics three years before going private.I exploit heterogeneity in firm characteristics to investigate the economic mechanism driv-ing the results. Consistent with the corporate governance hypothesis, I find the positive effectsof uncertainty on going private to be stronger for firms subject to severe manager-shareholderconflicts. Masulis et al. (2009) show that the dual-class structure aggravates the agency prob-lems between managers and shareholders. For such firms, incentives to resolve agency con-flicts following uncertainty shocks are expected to be higher. I find that the impacts are moreprominent for firms with dual class shares. The going private filings indicate that the dual-classstructure is eliminated after going private in most cases. For firms which still have dual classshares after delisting, management and PE investors own the same proportions for both shareclasses. The impacts of economic uncertainty shocks on going private are also stronger for firmswith less ownership by institutional blockholders. Literature on corporate governance (Agrawaland Mandelker 1990; Mehran 1995; Core et al. 1999) shows that blockholders provide effectivemonitoring for public firms. Therefore, the agency problems should be less of a concern forfirms with more institutional blockholders.The positive effects of uncertainty on going private are also more pronounced in companies8with more creditor-shareholder conflicts. Agency problems between creditors and sharehold-ers may be more severe for firms in financial distress. Using Altman Z-Score as a measure forfinancial distress, I find the effects to be stronger for firms in financial distress. The effects alsoconcentrate in firms with lower asset redeployability, that is, when the collateral value is lowerfor firms whose assets are more difficult to sell in the secondary market. For such firms, theconflicts are more severe because creditors experience lower recovery rates in bankruptcy. Theeffects of uncertainty on going private are also stronger for firms with a higher ratio of corporatebonds to bank loans. One advantage of bank loans to corporate bonds is the flexibility of rene-gotiation (Chemmanur and Fulghieri 1994). Firms\u2019 incentives to resolve the agency conflicts ofdebt are higher if they experience difficulties renegotiating with current debtholders.I conduct a subsample analysis to further investigate the corporate governance hypothesisby classifying the going private transactions into management buyouts, private equity buyouts,and buyouts with no management or PE participation. Consistent with the corporate gover-nance hypothesis, I find the effects stronger when management and\/or private equity investorsparticipate in the going private transactions. Management has a better understanding of theagency frictions faced by the companies, and stronger incentives to resolve the agency frictionsfollowing uncertainty shocks. In terms of buyouts by private equity investors, the effects of un-certainty on going private are more substantial because incentive alignment is one of the mostimportant value drivers for PE deals.If the conflicts between shareholders and creditors are mitigated through going private, thecost of debt is expected to decrease. I conduct a difference-in-differences analysis to examinethe impact of going private on the cost of debt. Specifically, I compare the difference in bankloan spreads of the going private firm in the pre- and post-delisting period with that of a groupof matched firms that remain public. I find that the costs of debt are significantly lower for goingprivate firms after they delist. Figure 2.3 shows that going private firms pay significantly higherloan spreads relative to the control group before going private, but the loan spreads becomecomparable after delisting. By realigning the control rights and cash flow rights through going9private, the agency problems are mitigated. As a result, the cost of debt decreases.Collectively, the evidence suggests that firms go private to resolve the heightened agencyfrictions following uncertainty shocks. Companies are more likely to go private in the pres-ence of high uncertainty, and the positive effects are more prominent for firms prone to se-vere agency problems, both between managers and shareholders and between creditors andshareholders. Uncertainty can exacerbate financial frictions and increase the cost of externalcapital. Going private helps alleviate the problems by aligning incentives of the management,new shareholders, and new creditors. As a result, firm receives lower cost of capital after goingprivate.I investigate several alternative explanations for the results. One possible explanation isthe market timing hypothesis, where managers and private equity investors take firms privatewhen they are undervalued. Undervaluation may be more likely following uncertainty shocks,since it becomes more difficult for investors to evaluate firm fundamentals. Using firm Tobin\u2019s Qrelative to the industry average as a proxy for undervaluation, I show that impacts of uncertaintyon going private are indifferent between undervalued and fairly valued firms.Another alternative explanation is the market distraction hypothesis. Changes in stock pricesin the public market distract controlling shareholders and employees. Managers may decide totake the firm private to enjoy a quiet life. Following Easton and Zmijewski (1989), I constructthe earnings response coefficient (ERC) to measure sensitivity of stock returns to earning an-nouncements. Based on the market distraction hypothesis, managers of the companies whosestock returns are more sensitive to earning news should be more likely to take the firm privatein uncertain times. However, I find the effects do not vary with ERC.The positive impacts of economic uncertainty on going private may also be driven by theheightened cost of information production in uncertain times. Subrahmanyam and Titman(1999) discuss the costs of duplication of information production for public companies. It ismore costly for investors to produce information during periods of high uncertainty. Using an-alyst coverage as a proxy for the cost of information production, I show that the results do not10vary across firms with different information production costs. In summary, the results suggestthat the positive impacts of economic uncertainty on going private are not driven by underval-uation, market distraction, or higher information production costs following uncertainty.The chapter is related to two veins of literature. First, it relates to the growing literature oneconomic uncertainty. A large number of studies document the negative impacts of uncertaintyon corporations. They show that uncertainty negatively impacts firm performance and growth(Gulen and Ion 2015; Alfaro et al. 2021). Firms reduce investment and employment (Bernanke1983; Leahy and Whited 1996; Guiso and Parigi 1999; Bloom 2009; Fern\u00e1ndez-Villaverde et al.2011; Bachmann and Bayer 2013; Stein and Stone 2013; Fern\u00e1ndez-Villaverde et al. 2015; Alfaroet al. 2021), and adopt conservative corporate policies (Chen et al. 2014; Chen 2016; Im et al.2017; Alfaro et al. 2021) following uncertainty shocks. As for financial consequences, studiesshow that higher uncertainty leads to higher cost of bank loans (Ashraf and Shen 2019), corpo-rate bond spreads (Kaviani et al. 2020) and the cost of equity (P\u00e1stor and Veronesi 2013). Alfaroet al. (2021) show that financial frictions amplify the impacts of uncertainty in the real econ-omy. This chapter contributes to this literature by reporting novel evidence of how economicuncertainty can directly exacerbate financial frictions. To the best of my knowledge, this is thefirst chapter providing empirical evidence on the impacts of economic uncertainty on corpo-rate governance. In addition, this chapter documents new findings that firms alter their capitalstructures via going private to moderate the high agency frictions following uncertainty.Second, the chapter relates to the large body of literature investigating the choice betweenpublic and private ownership structure (Shah and Thakor, 1988; Zingales, 1995, Chemmanurand Fulghieri, 1999; Boot et al., 2006). Studies show that firms choose the public status whenthe benefits outweigh the costs. The benefits as a public company include liquidity, easy accessto capital (Brav, 2009), risk sharing (Chemmanur and Fulghieri, 1999), etc. The costs of listinginclude the agency costs due to dispersed ownership and the separation of ownership and con-trol (Jensen and Meckling, 1976), loss of control in decision making (Boot et al., 2006), and thecompliance and disclosure costs (Engel et al., 2007). Within this literature, the chapter is closely11related to the studies on going private decisions. Jensen (1986) argue that delisting can be usedto reduce agency problems between managers and shareholders. Maupin et al. (1984), Lehnand Poulsen (1989) and Opler and Titman (1993) supports this argument by showing that firmswith more free cash flows are more likely to go private. Bolton and Von Thadden (1998) andBharath and Dittmar (2010) show that firms use private equity to opt out public markets for in-formation and liquidity considerations. Mehran and Peristiani (2009) finds that firms go privatewhen they lack financial visibility and fail to attract investor attention. Engel et al. (2007) arguethat firms go private to avoid compliance costs. Firm characteristics that affect the going privatedecision include size, market to book ratio, growth prospects, performance, and leverage. (Kimand Lyn, 1991; Kieschnick, 1998; Caprio et al., 2011; Martinez and Serve, 2011; Thomsen andVinten, 2014;). The chapter contributes to the literature by identifying economic uncertainty asa missing factor that can help explain going private transactions. In addition, the chapter findsthat going private can not only resolve agency problems between managers and shareholders,but also the conflicts between shareholders and creditors.The chapter is organized as follows. Section 2.2 presents the sample and data. Section 2.3describes the empirical methodology. Section 2.4 summarizes the main results. Section 2.5discusses the alternative explanations and the robustness tests. Section 2.6 concludes.2.2 DataThis section describes the data used to study the impacts of economic uncertainty on goingprivate transactions. I first describe the sample construction process. I then discuss summarystatistics of the going private sample and the deal structure of the going private transactions.2.2.1 Going Private SampleI follow SEC\u2019s legal definition of going private to construct the going private sample. Accordingto Rule 13E-3 of the Securities Exchange Act of 1934, a public company must file Schedule 13E-312if the company or an affiliate is engaged in the transactions which will cause a class of equitysecurities to become eligible for deregistration or delisting. I follow the SEC rule because thereis no ambiguity with this definition. In practice, going private transactions can be quite het-erogeneous. A broad range of transactions can fall into this definition, including managementbuyouts (MBO), non-leveraged or leveraged buyouts (LBO) by private equity firms, or strategicbuyouts by private operating companies. Schedule 13E-3 filings have also been used to identifygoing private transactions by other studies (DeAngelo et al., 1984; Engel et al., 2007; Leuz et al.,2008; Mehran and Peristiani, 2009; Bharath and Dittmar, 2010)To construct the going private sample, I retrieve all Schedule 13E-3 filings from 1994 to 2017.To ensure the transactions are completed, I cross-check with SEC Form 15 and Form 25 filings,which are filed when the securities are officially delisted. In addition, I check CRSP to ensurecompanies are no longer publicly traded within two years after they filed Schedule 13E-3. I alsoscreen the sample firms to ensure they are not traded on the pink sheets or over-the-counter.By doing so, I exclude the firms that \"go dark\", which refers to the action to deregister from SEC,but continue to trade on the pink sheets or over-the-counter. According to Leuz et al. (2008),going dark and going private are very different corporate events with different economic conse-quences. Firms usually go dark due to poor prospects and high compliance costs. Controllinginsiders may also deregister the firm to extract private benefits and escape from public scrutiny.Therefore, going dark is usually associated with negative cumulative abnormal returns. Goingprivate, on the other hand, is mostly associated with positive cumulative abnormal returns. Inthis chapter, I exclude the going dark firms and focus purely on the going private transactions.1,453 firms filed Schedule 13E-3 from 1994 to 2017. Among these deals, 188 deals were with-drawn (voluntarily or rejected by shareholders). 1,265 firms delisted within two years after theinitial filing. Companies from financial and utilities industries are excluded from the sample,decreasing the sample size to 935 companies. To calculate firm-level uncertainty shock, firmsneed to have two consecutive years of stock return data available before delisting. The samplesize drops to 525 firms due to data availability. The control sample in the main analysis is the13firms that remain public until the end of 2017. I also conduct a matching analysis to account forthe selection bias of the going private sample. Details of the matching process are discussed inSection 2.3.4. The final sample consists of 525 going private firms and 2,659 control firms, with48,060 firm-year observations.Table A.2.1 Panel A illustrates the industries in which the going private firms operate, basedon Fama-French twelve industry classifications.6 Industries that experience most going pri-vate transactions are business equipment, which includes computers, software and electronicequipment, and shops including wholesale, retail, and some services. Table A.2.1 Panel B de-scribes the sample composition by year. The year of going private is identified by the year firmsfile for going private, rather than the year they delist. Most firms delist at the same year orwithin one year after they file Schedule 13E-3. Figure 2.1 illustrates the time series trend of thegoing private transactions across industries7 from 1994 to 2017. The period 1994-2006 saw aboom in the going private transactions due to the development of the private equity market.The Sarbanes-Oxley Act of 2002, which increases the compliance costs of public companies,also contributes to this trend. The number of going private transactions decreased after the fi-nancial crisis in 2007-2009, since the debt markets have become more cautious in participatingin leveraged buyout deals. The number of going private transactions has increased again in re-cent years due to the heightened level of uncertainty. The going private trend also varies acrossindustries. While the manufacturing and hi-tech companies experienced steady growth in thegoing private transactions from 1994 to 2006, the number of going private transactions fluctu-ated in the consumer, healthcare, and other industries over the years. Figure 2.4 demonstratesthe cumulative abnormal return (CAR) of the going private companies at the announcementdate. On average, the going private companies receive a 25% cumulative abnormal return overthe announcement period.Deal-specific information on the going private transactions is retrieved from Schedule 13E-3 filings. Accounting variables are from Compustat and variables in the stock market are from6http:\/\/mba.tuck.dartmouth.edu\/pages\/faculty\/ken.french\/Data_Library\/det_12_ind_port.html7The industries are classified based on Fama-French five industry classifications.14CRSP. IPO dates are from SDC New Issues database. The going private deal announcementdates and deal classifications are from SDC M&A database. Data on currency exchange rates,crude oil prices and Treasury returns are from Bloomberg. Measurement of economic politicaluncertainty is from Baker et al. (2016). Asset redepoyability measures are from Kim and Kung(2017). Information on institutional ownership is from SEC 13F holdings. Information on thefirm\u2019s debt structure is from Capital IQ. All variables are winsorized at the 1% and 99% levels.Detailed definitions of the variables are discussed in Appendix A.1.2.2.2 Descriptive StatisticsTable 2.1 Panel A compares firm characteristics of the going private sample and a control sam-ple of surviving firms over their entire public life cycle. The control sample constitutes firmsthat remain public until the end of 2017. There are 525 going private firms and 2,659 surviv-ing firms. The going private sample experiences lower but more volatile stock returns. Firmsthat go private are significantly smaller in size and have lower Tobin\u2019s Q. However, performancemeasured by return on assets is similar between the two groups. The going private sample hashigher leverage, possesses fewer intangible assets, and demonstrates a higher tax to assets ratioon average.To better understand the going private transactions, I read the Schedule 13E-3 filings of thegoing private transactions in detail. The going private company is required to discuss the pur-poses of the transaction, any alternatives that the company considered, and whether the trans-action is fair to unaffiliated shareholders in the Schedule 13E-3 filings. Most companies alsodisclose the source of deal financing, the ownership structure before and after the transactionin the Schedule 13E-3 filings.Since the study focuses on the capital restructuring process, I focus on the Schedule 13E-3filings of a subset of going private firms with outstanding bank loans before the going privatetransaction. Within the 252 firms with non-missing control variables in the main analysis, 120firms had bank loans outstanding before they went private. I go through their Schedule 13E-153 (13E-3, DEF13E-3, PRE13E-3) filings and obtain detailed descriptions of 84 transactions. Ialso go through the Schedule TO filings, which are filed if the going private transactions arecompleted through tender offers. Panel B of Table 2.1 reports summary statistics of these trans-actions. The deal value is $544.6 million on average. Bidders pay an average takeover premiumof 34.5% to the pre-deal share price. Deals are usually financed by a combination of bank loans,cash on the company\u2019s balance sheet, and equity contributions by a PE firm and the manage-ment. On average, 61% of the deal is financed by debt. 84% of that debt comes from bank loans,usually a term loan facility and a revolving credit facility. The bank loans are usually arrangedby a syndicate of banks with lending relationships with the PE investors. Sometimes, the goingprivate company also issues corporate bonds to finance the deal, accounting for the remaining16% of the debt. The remaining 39% of the deal is financed by cash on the company\u2019s balancesheet (16%), and equity contributed by a PE firm (63%) and the management (21%). After delist-ing, the PE firm owns 64% of the company on average. Management owns 35% of the company,with the remaining 1% owned by other existing shareholders before going private.Figure 2.2 illustrates the changes in the companies\u2019 capital structures through the going pri-vate process. Panel A and B compare the capital structures of a representative company, Amer-ican Greetings Corp., before and after it went private. The company was held by the manage-ment, several institutional investors, and many dispersed shareholders with dual-class sharesbefore going private. After going private, the company was entirely held by the managementand a private equity investor. The existing debts, which included four different loan facilities ar-ranged by two syndicates, were paid off with newly issued debts. The new debts were arrangedby a loan syndicate that had previous lending relationships with the private equity investor.Panel C of Figure 2.2 illustrates the capital structure of the average company after going private,which demonstrates a similar pattern as in Panel B.162.3 Empirical MethodologyIn this section, I describe the empirical methodology used to study the impacts of economic un-certainty on going private. First, I describe the Cox proportional hazards model. Second, I dis-cuss how I measure economic uncertainty and the identification strategy. Then I describe thecontrol function approach to instrument economic uncertainty in the Cox proportional haz-ards model. I also discuss the matching analysis to address the sample selection bias, and thedifference-in-differences analysis to study the impact of going private on the loan rate in thissection.2.3.1 Cox Proportional Hazards ModelFollowing Mehran and Peristiani (2009) and Bharath and Dittmar (2010), I use a hazards modelto study firms\u2019 decisions to go private. Hazards models are widely applied in survival analysis.Shumway (2001) shows that they are more appropriate to analyze survival data compared tostatic models. A hazards model is suitable to analyze going private decisions in the followingtwo ways.First, hazard models trace down firms\u2019 decisions over their entire life cycles. In hazard mod-els, each firm is treated as one observation during its entire life span. The time-varying firmcharacteristics allow me to study both cross-sectional and time-series effects of uncertainty ongoing private. Second, hazards models can handle censored data, which is a crucial feature ofthe going private sample. The sample period ends in 2017. Firms are still at risk of going privateafter the sample period ends. Instead of treating these firms as surviving as done by static mod-els, hazards models treat all firms as being dropped out of the sample at the end of the sampleperiod.I use a Cox proportional hazards model because it does not impose any restriction on thebaseline hazard rate. The model to estimate ish(t ,X t\u22121)= h(t ,0)exp(\u03b2\u2032X t\u22121+\u03be) (2.1)17where h(t ) is the hazard rate for a firm with covariates X t\u22121 to go private at time t . h(t ,0) isthe baseline hazard rate. The coefficient vector to be estimated is \u03b2. Cox proportional hazardsmodel allows me to estimate \u03b2 without estimating the baseline hazard rate h(t ,0). A positive \u03b2means that the hazard rate of going private is higher when x is higher. The hazard ratio exp(\u03b2)indicates the increase in the hazard rate when there is one unit change in the independentvariable.2.3.2 Measuring UncertaintyFollowing the uncertainty literature, I measure firm-level uncertainty using realized stock re-turn volatility \u03c3i ,t , which is the standard deviation of daily dividend cumulative stock returnswithin a fiscal year. Uncertainty shock is defined as the change in annualized stock returnvolatility \u2206\u03c3i ,t = (\u03c3i ,t \u2212\u03c3i ,t\u22121)\/( 12\u03c3i ,t + 12\u03c3i ,t\u22121) for firm i at a given year t.Stock return volatility is an endogenous variable that can be related to various aspects ofa firm. First of all, it may correlate with other omitted variables which drive firms\u2019 going pri-vate decisions. For example, stocks of firms with less analysts coverage can be very volatile.Meanwhile, firms with less analysts coverage may decide to go private due to their lack of fi-nancial visibility in the public market. Second, if investors anticipate the firm to go privatesoon, its stock prices can move dramatically within a short period. This generates an issue ofreverse causality. Indeed, previous literature finds contradictory effects of stock return volatilityon going private, indicating that stock return volatility contains various aspects of information,which affects the going private decisions in different directions. Therefore, to study the impactsof economic uncertainty on going private, it is crucial to identify the component in changes ofstock return volatility due to exogenous uncertainty shocks.IdentificationI follow the identification strategy in Alfaro et al. (2021) to construct instruments for uncertaintyshocks. To be more specific, I employ firms\u2019 differential exposure to aggregate macroeconomics18uncertainty shocks to capture shocks to firm-level uncertainty. The identification strategy issimilar to Bartik (1991), which utilizes local industry share and overall industry growth of thecountry to measure local development. In this chapter, I use uncertainty shocks to oil prices,economic political uncertainty, US 10-year treasury notes and seven major currency exchangerates defined by the Federal Board8. In the following of the chapter, I refer to these ten macroe-conomic factors as commodities. The intuition is as follows. Suppose there are two firms, oneoperates in an industry which is highly government-dependent, such as health care or defense,while the other is a local retailer. When political uncertainty rises, the first company will be af-fected significantly while the latter will be barely affected. Similarly, firms operating in energyindustries will experience high uncertainty following an uncertainty shock to oil prices.Construction of the instruments follows two steps. First, I estimate firms\u2019 exposure to ag-gregate macroeconomic conditions. Second, I calculate firm-level uncertainty shocks as theproduct of firm exposure and aggregate uncertainty shocks.Exposure to Aggregate Uncertainty ShocksFirm exposure to currencies, energy, treasury and policy is obtained by regressing risk adjustedstock returns on the changes in prices of the 10 commodities:ri ,t =\u03b1 j ,t +\u2211c\u03b2cj \u00b7 r ct +\u03f5i ,t (2.2)ri ,t is the daily risk-adjusted stock return of firm i, which is the residual, \u03b5i ,t , of Equation(2.3). r ct is the daily change in prices of commodity c. Firm exposure to commodity c is thecoefficient \u03b2cj , which measures the sensitivity of stock price to commodity c. \u03b2cj is estimated atSIC 3-digit level, on a rolling basis with daily stock returns in the past ten years9.The daily risk-adjusted return of firm i is the Carhart (1997) four factor risk adjusted return,8The seven \"major\" currencies defined by the Federal Board includes the euro, Canadian dollar, Japanese yen,British pound, Swiss franc, Swedish krona and the Australian dollar9\u03b2cj is estimated at SIC 3-digit level to reduce idiosyncratic noise in firm-level stock returns, which increasesestimation precision.19which is the residual of the following equation:r excessi ,t =\u03b1i +\u03b2i ,MKT \u00b7MKTt +\u03b2i ,HML \u00b7HMLt +\u03b2i ,SMB \u00b7SMBt +\u03b2i ,UMD \u00b7UMD t +\u03b5i ,t , (2.3)where r excessi ,t is firm i\u2019s daily stock return in excess of risk free rates. MKT is the value weightedmarket index in excess of risk free rate. HML is the book to market factor. SMB is the size factor,and UMD is the momentum factor. Risk adjusted returns are used to estimate sensitivities sothat \u03b2cj captures firm exposure to commodities rather than systematic risks. I also estimate thesensitivities using raw returns and returns adjusted by other risk models. The results are similarand discussed in section 2.5.2.Construction of Instrument VariablesFirm-level uncertainty shocks are constructed using the sensitivities of stock returns to factorprices and aggregate uncertainty shocks:IV cj ,t = |\u03b2c,weightedj ,t\u22122 | \u00b7\u2206\u03c3ct , (2.4)where \u03b2c,weightedj ,t\u22122 is a weighted value of sensitivity estimated in the first step (discussedbelow). \u03c3ct is the standard deviation of daily changes in the price of commodity c within ayear. \u2206\u03c3ct is the change of \u03c3ct , which is calculated in a similar way as \u2206\u03c3i ,t . I adjust each \u03b2cjby its significance level to obtain the significance weighted sensitivities. To be more specific,\u03b2c,weightedj = \u03c9cj \u00b7\u03b2cj , where \u03c9cj =|t cj |\u2211c |t cj | and |t cj | is the absolute value of t-statistics estimated in(2.2) for commodity c. \u03c9cj is calculated as the ratio of |t cj | to the sum of absolute t-statistics forall commodities. The significance weighted sensitivities capture both economic and statisticalsignificance of firms\u2019 exposure to the commodities.To ensure the instruments capture the effects of uncertainty shocks other than economicconditions, I also include the first moment variables as control variables in the regressions. Thefirst moment variables are calculated as \u03b2c,weightedj \u00b7 r ct , where r ct is the annual growth of the 1020commodity prices.Figure 2.5 demonstrates how oil, interest rate, exchange rate, and economic policy uncer-tainty vary across industries from 1994 to 201710. Panel A, B, C, and D show the value-weightedaverage of the instruments for each industry constructed based on oil, interest rate, exchangerate, and economic policy uncertainty shocks respectively. Figure 2.5 Panel A shows that man-ufacturing and energy companies experience the highest level of oil uncertainty shocks amongthe five sectors, while companies operating in hi-tech and healthcare industries experience avery low level of oil uncertainty shocks. Panel B and C illustrate that interest rate and exchangerate uncertainty are highly correlated across sectors. Companies experience very high interestrate and exchange rate uncertainty shocks during the financial crisis. Panel D shows that the hi-tech industry demonstrates the highest economic policy uncertainty level, while the healthcareindustry experiences the lowest level of economic policy uncertainty.The instruments satisfy the exogeneity condition for the following two reasons. First, aggre-gate uncertainty shocks are very unlikely to be driven by firm characteristics. Second, sensitiv-ities are estimated two years ahead of time to capture the pre-shock exposure and avoid anylooking forward bias. Table A.2.2 illustrates the first stage results of IV regressions. The depen-dent variable is the changes in stock return volatility and the independent variables are the teninstruments. Column (1) shows the results without any control variable. Column (2) reportsthe results with firm characteristics and the ten first moment variables as controls. Columns(3) and (4) report the results with industry fixed effects, and with industry and year fixed effectsrespectively. All the coefficients are positive and statistically significant except for instrumentsof the British pound. Results of the test statistics indicate that the instruments pass both theunderidentification tests and the overidentification tests.10The industries are classified based on Fama-French five industry classifications.212.3.3 Control Function ApproachThe standard two-stage least squares (2SLS) estimations cannot be applied in non-linear mod-els like the Cox proportional hazards model. To instrument for uncertainty shocks in the Coxproportional hazards model, I use a control function approach. The control function approachfollows a two-step estimation procedure. First, I regress firm-level uncertainty shocks on the 10instruments and obtain the residuals.\u2206\u03c3i ,t =\u03b10+\u03b11Xi ,t +\u03b12Z1, j ,t +\u03b13Z2, j ,t +\u03b6i ,t (2.5)Xi ,t are the control variables of firm characteristics. Z1, j ,t are the first moment effects at SIC 3digit level discussed in the identification section. Z2, j ,t are the 10 instruments constructed atSIC 3-digit level.The residual has two components:\u03b6i ,t = \u03b4\u03bei ,t +\u03b7\u2032i ,t (2.6)The first component \u03b4\u03bei ,t contains the endogenous part in \u2206\u03c3i ,t , while the second component\u03b7\u2032i ,t is orthogonal to it. Rewrite Equation (2.6), we get:\u03bei ,t =\u03bb\u03b6i +\u03be\u2032i (2.7)where\u03bb= 1\/\u03b4 and \u03be\u2032i =\u2212\u03b7\u2032i ,t\/\u03b4. By running the Cox hazard proportional model with the residual\u03b6i ,t obtained from Equation (2.5) as an explanatory variable:h(t ,\u2206\u03c3i ,Xi ,Z1, j )= h(t ,0)exp(\u03b21\u2206\u03c3i +\u03b22Xi +\u03b23Z1, j +\u03bb\u03b6i +\u03be\u2032i ) (2.8)The new error term \u03be\u2032i is orthogonal to the change in stock return volatility. Therefore, \u03b21 is anunbiased estimator of the effects of uncertainty shocks on going private.222.3.4 Matching AnalysisComparison between the going private companies and the surviving companies in Table 2.1Panel A indicates that firm characteristics are significantly different between the two groups. Toaddress the concern that the going private companies are fundamentally different from the sur-viving firms, I investigate the effects of uncertainty on going private through a matching analy-sis. The matched control samples are constructed based on firm characteristics one year afterIPO and three years before going private. Bharath and Dittmar (2010) finds that firm character-istics at the time of IPO are important determinants for the decision to go private. Therefore,I construct alternative control samples based only on IPO characteristics as a robustness test.The variables to match include industry, firm size, Tobin\u2019s Q, and annual stock returns. Amongall the companies that remain public until the end of 2017, I first restrict the matched group tothose that went public in the same year as the going private firm. For each going private firm,I then construct different control samples by selecting the firms operate in the same 2 digit SICindustry, whose log sales, Tobin\u2019Q or annual stock return is within +\/- 10% of the delisted firm. Ialso conduct a propensity score matching based on these characteristics. For each going privatecompany, I select up to five companies that remain public at the end of 2017, and operate in thesame Fama-French 12 industry and went public in the same year as the going private company.105 going private companies are matched with 410 control companies. Panel A and B of Table2.3 present the at-IPO and pre-delisting comparisons of the delisted firms and the control sam-ple based on the propensity score matching. The summary statistics in Panel A and Panel B ofTable 2.3 indicate that the differences in firm characteristics between the going private sampleand the matched control group are insignificant, both at the time of IPO and in the year beforegoing private.232.3.5 Difference-in-differences Analysis of the Impacts of Going Private onLoan RateIf agency problems between creditors and shareholders are mitigated through going private,the agency cost of debt should decrease after going private. To further investigate the economicmechanism driving the results, I conduct a difference-in-differences analysis to investigate theimpact of going private on loan rates. The sample construction process of the difference-in-differences analysis is as follows.I select the subsample of going private companies which have loan facilities both before andafter the going private transaction. To minimize changes in firm fundamentals between the twoloans, I restrict the loan facilities to those within two years of the going private date. If more thanone loan facility satisfies the criteria, I select the one closest to the delisting date. Loans used tofinance the going-private transaction are excluded from the sample. The loan pair allows me tocompare the cost of two bank loans with little change in the firm\u2019s fundamentals except for thepublic status.For each pre-delisting and post-delisting loan pair issued by the going private firm, loanpairs issued by public firms operating in the same 2-digit SIC industry are matched. The loanpairs must start within one year of the loan pair of the going private firm so that the loan ratesare not affected by market conditions. Among all the matched firms with available loan pairs, Iconduct a propensity score matching based on firm size, stock return, and stock return volatil-ity. Due to the restrictive criteria, the number of matched control companies is much smallercompared to the matching analysis in Section 2.3.4. To have a balanced sample, I select upto two control companies for each going private firm in this difference-in-differences analy-sis. Panel A of Table 2.6 reports summary statistics of the going private firms and the controlfirms with matched loan pairs. The going private sample demonstrates lower stock returns andhigher stock return volatility before going private. However, the differences are statistically in-significant.I estimate the effect of going private on the loan rate using the following difference-in-24differences regression:LoanRatei ,t =\u03b21+\u03b22GoingPr i vatei\u00d7Postt+\u03b23GoingPr i vatei+\u03b24Postt+\u03b25Loani ,t+\u03b8p+\u03c8t+\u03f5i ,p,t(2.9)whereGoing Pr i vatei is an indicator variable that equals one for going private firms, and zerootherwise. Postt is a dummy variable that equals one if the loan facility starts after delisting, oris matched to a post-delisting loan. I include year fixed effects, \u03c8t , to control for any macroe-conomic factors affecting the loan spread. I also include fixed effects for each matched pair, \u03b8p ,to control for unobserved matched pair characteristics that might affect the loan spread. Loancharacteristics are also included in the regression to control for heterogeneity in loan facilities.2.4 ResultsI discuss the results in this section. First, I discuss the main results of impacts of uncertaintyshocks on going private transactions. After that, I provide evidence of the corporate governancemechanism exploiting heterogeneity in the level of agency problems firms face. Lastly, I discusshow going private affects firms\u2019 cost of bank loans.2.4.1 Main ResultsTable 2.2 reports the results of Cox proportional hazards model for time to go private. The majorindependent variable is economic uncertainty shock, which is measured as the year-on-yearchange in stock return volatility. The dependent variable is the hazard rate of going private.Columns (1) and (2) show results of the baseline Cox proportional hazards estimations. Col-umn (1) shows the univariate results, and column (2) reports the results with control variables.Both coefficients of the change in stock return volatility and volatility are positive and statisti-cally significant, indicating that firms are more likely to go private following uncertainty shocks.Regarding economic magnitudes, results in column (2) indicate that a one standard deviation25increase in uncertainty shocks increases the hazard rate of going private by 14%.Columns (3) to (6) report the IV results of the Cox proportional hazards models for timeto go private, using a control function approach. Column (3) reports the univariate results.Column (4) includes additional firm characteristics and first moment macroeconomic variablesas control variables. Column (5) includes Fama-French 12 industry fixed effects, and column(6) includes industry and year fixed effects. The impacts of uncertainty shocks on going privateare positive and statistically significant at 1% level for all specifications. Results in column (6)indicate that a one standard deviation increase in the\u2206Volatility induced by macro uncertaintyshocks increases the hazard rate of going private by 22.8%.Firms with lower stock returns are more likely to go private. The coefficient on log sales isnegative, meaning that smaller firms are more likely to go private. It is easier for larger firms toamortize fixed costs, and smaller firms go private to avoid compliance costs. The coefficient onTobin\u2019s Q is negative, showing that firms with fewer growth opportunities are more likely to goprivate. It also suggests that undervalued companies are more likely to become going privatetargets. Asset intangibility positively affects the likelihood of going private. Firms with moreintangible assets may be more likely to be misvalued. Alternatively, there may be more dis-agreement between public investors and firm insiders in these companies, creating incentivesto go private. Consistent with Mehran and Peristiani (2009), firms that went private demon-strate higher return on assets before delisting.2.4.2 Estimation Results Based on Matched Control SamplesTable 2.3 Panel C reports the matching results of the Cox proportional hazards models with thecontrol function. The control sample constitutes public firms that went public in the same yearas the going private firms, and matched on various firm characteristics at the time of IPO andthree years before delisting. The control firms in column (1) operate in the same SIC 2-digitindustries as the going private firms. Control firms in columns (2) and (3) are matched on logsales and Tobin\u2019s Q respectively. Firms with characteristics +\/- 10% of the going private firms26are included in the matched sample. The control sample in column (4) is constructed basedon a one-to-five propensity score matching on SIC 2-digit industry, log sales, Tobin\u2019s Q, andannual stock returns. Control variables and first moment macro variables are included in allspecifications. Year and Fama-French 12 industry fixed effects are included to account for thetime varying effects and heterogeneity across industries. The effects of uncertainty shocks onthe hazard rates of going private are positive and significant, consistent with Table 2.2. Resultsof the propensity score matching in column (4) indicate that one standard deviation increase inuncertainty induced by macro uncertainty shocks increases the likelihood of going private by13%.2.4.3 Agency Problems and Going Private TransactionsBased on the agency hypothesis, incentives for firms to go private should be higher when thereare more agency problems associated with the firms\u2019 capital structures. Therefore, the like-lihood of going private should be higher for firms with more agency problems, both amongshareholders, and between shareholders and creditors. To investigate this hypothesis, I ex-ploit heterogeneity in firms\u2019 agency problems associated with their capital structures, and testwhether firms with more agency problems are more likely to go private under uncertainty shocks.Specifically, I re-estimate Equation (2.1) by including the interaction terms of economic uncer-tainty shocks and various proxies for agency frictions.Shareholder Conflicts and Going Private TransactionsI first study firms\u2019 ownership structures and investigate whether firms are more likely to go pri-vate with uncertainty shocks when there are more agency conflicts between managers\/controllingshareholders and minority shareholders.Masulis et al. (2009) show that the dual-class structure aggravates the agency problems be-tween managers\/controlling shareholders and minority shareholders. The divergence betweencontrol rights and cash flow rights allows managers and controlling shareholders to extract pri-27vate benefits without bearing the financial consequences. Because of these agency frictions,firms with dual-class structure bear higher cost of capital (Masulis et al. 2009), and experiencelower firm value and stock returns (Claessens et al. 2002; Lemmon and Lins 2003). Therefore,the incentives to go private under uncertainty shocks to resolve the agency conflicts are ex-pected to be higher for the firms with dual-class structure. I also investigate whether the im-pacts of uncertainty on going private vary for firms with different levels of institutional own-ership. Literature on corporate governance (Shleifer and Vishny 1986; Agrawal and Mandelker1990; Shleifer and Vishny 1997) suggests that institutional blockholders provide effective moni-toring for public firms. Therefore, the agency problems should be bigger for firms with fewer in-stitutional blockholders, increasing the incentives for firms to go private following uncertaintyshocks.Table 2.4 reports the results on shareholder conflicts and going private transactions. PanelA shows results on dual-class structure and going private transactions. Dual class is an indica-tor variable that equals to one if the company has dual class shares that year. Consistent withthe results in Table 2.2, I find that the coefficients of uncertainty shocks on going private arepositive and statistically significant across all specifications. The effects are stronger for firmswith dual class shares. The Schedule 13E-3 filings indicate that the dual-class structure is usu-ally eliminated after delisting. For the firms with dual class shares after delisting, managementand the PE firm usually own the same proportions for both classes. Panel B presents results oninstitutional ownership and going private transactions. Institutional investor is the percentageownership by institutional blockholders. Consistent with Table 2.2, firms are more likely to goprivate under uncertainty shocks. The impacts are stronger for firms with less ownership byinstitutional blockholders. Results in Table 2.4 suggest that firms with potential agency prob-lems between managers\/controlling shareholders and minority shareholders are more likely togo private under uncertainty shocks, consistent with the agency hypothesis.28Shareholder-creditor Conflicts and Going Private TransactionsI also investigate whether the effects of uncertainty on going private concentrate on firms withmore agency costs of debt. The agency problems between shareholders and creditors are par-ticularly costly when firms are in financial distress. Therefore, incentives to resolve the agencyconflicts of debt are higher for the firms in financial distress. I also exploit heterogeneity infirms\u2019 asset redeployability to investigate the economic mechanism. Theories suggest that col-lateral alleviates financial frictions of debt. Creditors bear significantly fewer risks if the as-sets are easy to resell in the secondary markets. Benmelech and Bergman (2009) shows thatthe ability to pledge redeployable collateral lowers the cost of external financing and increasesdebt capacity. Based on this argument, firms with less redeployable assets should demonstratehigher agency costs of debt and therefore have larger incentives to go private under uncertaintyshocks. I also investigate how a firm\u2019s debt structure affects the impact of uncertainty on goingprivate. Specifically, I investigate whether the positive impacts of uncertainty on going privatevary with the ratio of bank loans to corporate bonds. Chemmanur and Fulghieri (1994) showthat bank loans are more flexible for renegotiation in the event of financial distress. Firms\u2019 in-centives to resolve the agency conflicts of debt are higher if they experience difficulties in therenegotiation process.Table 2.5 reports the results on shareholder-creditor conflicts and going private transac-tions. Consistent with the main results, the positive impacts of uncertainty on going private arepositive and statistically significant in all the results. Panel A reports results on asset redeploy-ability and going private transactions. Asset redeployability is the standardized value-weightedasset redeployability index from Kim and Kung (2017) times minus one. The results suggestthat firms with less redeployable assets are more likely to go private under uncertainty shocks.Panel B reports the results of financial distress and going private transactions. Financial distressis an indicator variable if the Altman Z-Score is lower than 1.8. Results in Panel B indicate thatthe positive impacts of uncertainty on going private are stronger for firms in financial distress.Panel C of Table 2.5 reports the debt structure and going private transactions. Loan to bond29ratio is the ratio of outstanding bank loans to corporate bonds of the firms. Results in Table 2.5Panel C indicate that firms are more likely to go private under uncertainty when they have morecorporate bonds than bank loans. Overall, the results in Table 2.5 suggest that firms are morelikely to go private when they face more shareholder-creditor conflicts.2.4.4 Impacts of Going Private on Loan RateFigure 2.3 illustrates the comparisons of loan rates between going private firms and controlfirms before and after delisting. Before delisting, going private firms pay significantly highercosts for bank credit compared to the control group. After delisting, the difference becomesinsignificant. Panel B of Table 2.6 summarizes the differences in loan rates. The difference isclose to zero when we compare the loan rate residuals, which are residuals from the regressionof loan rates on year and matched pair fixed effects, in the post-delisting period.Panel C of Table 2.6 reports regression results of the difference-in-differences analysis fromEquation (2.9). Year fixed effects are included in all specifications. Columns (2) to (4) includematched pair fixed effects. On average, going private firms pay more for bank credit comparedto matched control firms. However, their relative cost of bank credit decreases after delist-ing, because they had worse ex-ante credit quality, which was improved through going private.The difference-in-differences coefficient in column (4) suggests that, compared to the matchedsample, going private firms pay 230 bps less for bank credit after they delist\u2014a significant de-crease in economic terms. This result provides further support to the agency hypothesis thatgoing private resolves the agency conflicts of debt. As a result, the cost of bank loans decreases.2.4.5 Subsample AnalysisTo further investigate the economic mechanism, I classify the going private deals into manage-ment buyouts and the buyouts by private equity investors and investigate whether the effectsvary when management or private equity investors participate. Based on the corporate gover-nance hypothesis, companies go private to resolve the heightened agency frictions following30uncertainty shocks. The effects are expected to be stronger when management is involvement,because management has a better understanding of the agency frictions faced by the company.If the agency frictions are exacerbated following uncertainty shocks, management should bemore likely to take the company private to resolve the issues. The effects are also expected tobe stronger for private equity buyouts since incentive alignment is one of the most importantvalue drivers for the deals. The buyout classifications are from SDC M&A database.Table 2.7 reports the results of the subsample analysis. The going private sample in Panel Aand B involve management buyouts and deals without management participation respectively.The going private sample in Panel C and D constitute the buyouts with and without privateequity investors respectively. The control sample includes the companies that remain publicuntil the end of 2017. Results are estimated based on the Cox proportional hazards model witha control function approach. Results in Panel A and B indicate that the effects of uncertaintyon going private are slightly stronger when there is management participation. Panel C and Dsuggest that the effects are more prominent when the buyouts involve private equity investors.A comparison of the results in column (6) shows that the positive impacts of uncertainty ongoing private are 60% higher when there is PE participation. Overall, the results are consistentwith the corporate governance hypothesis.2.5 Alternative Explanations and Robustness TestsThis section investigates three alternative explanations, which may drive the positive impactsof uncertainty shocks on going private: undervaluation, market distraction, and the cost ofinformation production. It also presents several robustness tests.312.5.1 Alternative ExplanationsUndervaluationWhen firms experience uncertainty shocks, it may become more difficult for investors to evalu-ate the fundamentals of the firms. Firms are more likely to be misvalued. Previous studies showthat managers and private equity investors are more likely to take firms private when they be-lieve the firms are undervalued. If undervaluation is the primary channel that drives the results,the impacts should be stronger for firms that are undervalued. I use relative Tobin\u2019s Q, which isfirm Tobin\u2019s Q divided by industry Tobin\u2019s Q at SIC 3-digit level, as a proxy for undervaluation.Column (1) in Table 2.8 reports the results with relative Tobin\u2019s Q as an additional controlvariable in the regression. The coefficient on uncertainty shock is similar compared to the mainresults in Table 2.2. The negative coefficient on relative Tobin\u2019s Q indicates that firms are morelikely to go private when they are undervalued. Column (2) adds an interaction term of relativeTobin\u2019Q and uncertainty shock into the regression. The coefficient on the interaction term isinsignificant, indicating that the effects are indifferent between undervalued and fairly pricedfirms. The results suggest that undervaluation is a major reason for firms to go private. However,undervaluation does not drive the impacts of uncertainty on going private.Market DistractionAnother possible explanation of the results is market distraction. Uncertainty increases thevolatility of stock prices, which can be a distraction to controlling shareholders and employees.Following Easton and Zmijewski (1989) in the accounting literature, I construct the earnings re-sponse coefficient (ERC) to measure the sensitivity of stock returns to earning announcements.ERC is estimated as the coefficient of regressing size-adjusted abnormal returns around theannouncement date on unexpected earnings at SIC 3-digit level. ERC measures market respon-siveness to earning news. The underlying reasoning is as follows. Managers of the companieswhose stock returns are more sensitive to earning news are more likely to take the firm private32to enjoy a quiet life. When stock return volatility increases due to uncertainty, the need to takethe firm private becomes higher.Column (3) in Table 2.8 shows the results with log ERC as an additional control variable. Thecoefficient of log ERC on the hazard rate of going private is insignificantly different from zero.Column (4) adds the interaction term to the regressions. The results indicate that the effects ofuncertainty on going private are indifferent for firms with high and low ERC.Cost of Information ProductionAnother possible explanation is the elevated cost of information production under uncertainty.Subrahmanyam and Titman (1999) highlights the cost of duplication of information productionby dispersed investors of public firms. Their paper suggests that more firms would go privateif the cost of information production increases. With economic uncertainty shocks, investors\u2019costs of information production are higher. Therefore, the positive effect of economic uncer-tainty on going private may be attributed to the increased cost of information production un-der uncertainty. When a large number of analysts follow the company, the cost of duplicationof information production is mitigated because the analysts produce more publicly availableinformation. If the effects of uncertainty on going private are driven by the cost of informationproduction, the effects should concentrate in the firms followed by fewer analysts.Column (5) and (6) in Table 2.8 shows the results investigating the cost of information pro-duction hypothesis. Column (5) includes analyst coverage of the firm as an additional controlvariable, and column (6) includes the interaction term in the regression. Results indicate thatanalyst coverage negatively affects the hazard rate of going private. Analyst coverage representsfinancial visibility of the company. Mehran and Peristiani (2009) finds that firms with a lack offinancial visibility choose to go private since they have fewer benefits of being public. The in-teraction term of analyst coverage with uncertainty shock is insignificantly different from zero,suggesting that the cost of information production is not driving the results. In conclusion, re-sults in Table 2.8 indicate that the alternative explanations do not drive the positive effects of33uncertainty shocks on going private.2.5.2 Robustness TestsThis section investigates robustness of the findings. The effects are re-estimated controlling formacroeconomic conditions. I also re-examine Equation (2.1) using alternative factor modelsfor risk-adjusted returns.Effects of Macroeconomic ConditionsStudies show that uncertainty is counter-cyclical. Therefore, the results may be driven by busi-ness cycles rather than uncertainty shocks. To ensure the results are not driven by businesscycles, I include the 10 first moment variables on changes in commodity prices as controls. Tofurther address the concern, I add macroeconomic variables in the regressions. Table A.2.3 re-ports the impacts of macroeconomic factors on the hazard rate of going private. The hazardrate of going private is higher when investor sentiment is high. The hazard rate of going privateis lower when yield curve is steeper. Supply of debt in the credit is an important determinant forgoing private since many going private transactions are completed through leveraged buyouts.Consistent with the main results, VIX positively affect the hazard rate of going private. Resultsin column 5 show that recessions do not play a role in the probability of going private. The pos-itive effect of GDP growth on going private is somehow surprising. The result may be an artifactsince changes in prices of the 10 commodities, which are highly correlated with GDP growth,are already included as controls in the regressions. Results in OA3 suggest that the positive ef-fects of uncertainty shocks on going private are not driven by business cycles.Different Factor Models for Risk Adjusted ReturnThe risk factors may be correlated with macroeconomic uncertainty. To ensure the effects arenot driven by different risk factors, I re-construct the instruments using risk-adjusted returnsestimated based on different factor models. Table A.2.4 demonstrates the results using different34risk-adjusted returns to estimate firm exposure to aggregate uncertainty shocks. Panel A showsthe first stage results. Similar to Table A.2.2, the 10 instruments positively predict firm-leveluncertainty shocks. All of the specifications pass the Kleibergen-Paap underidentification testand Hansen-Sargan J overidentification test. Panel B shows the main results of Cox proportionalhazards models with risk adjusted returns by different factor models. Column (1) shows theresults with raw returns. Column (2)-(4) report results with CAPM, Fama-French 3-factor modeland Fama-French 5-factor model respectively. The coefficients of uncertainty on going privateare significantly positive across all specifications. The economic magnitudes are similar to themain results.2.6 ConclusionIn this chapter, I investigate the effect of economic uncertainty on going private. I find that firmsare more likely to go private following uncertainty shocks. The positive correlation betweenuncertainty and going private is robust to controlling for firm and macroeconomics character-istics such as GDP growth, investor sentiment, indicators for NBER recession, VIX or the termpremium. Moreover, the results are not sensitive to sample composition, or to controls for en-dogeneity problems using a control function analysis with instrumental variables.In additional analyses, I find the positive effects of uncertainty on going private concentrateon firms with more agency conflicts. Specifically, the effects are more substantial for firms withdual-class structure and with less institutional ownership. Also, the effects are more prominentfor firms with more credit-shareholder conflicts: firms with lower asset redeployability, firmsin financial distress, and firms with low loan-to-bond ratio. Results of the subsample analysisindicate that effects are stronger when management and\/or private equity investors participatein the going private transactions. A difference-in-differences analysis indicates that the cost ofdebt decreases after going private. The results are consistent with the corporate governancehypothesis. Uncertainty exacerbates the agency frictions faced by public companies. It gen-35erates more information asymmetry, and amplifies moral hazard problems and coordinationfrictions among managers, shareholders and creditors. As a response, firms alter their capitalstructures via going private to address the financial frictions and lessen the negative impacts ofuncertainty. After agency frictions are mitigated through going private, firms obtain lower costsof debt.The chapter documents uncertainty as a missing factor which can explain going privatetransactions. More importantly, the chapter provides novel evidence on the impacts of un-certainty on corporate governance. The chapter proposes one possible response by firms toaddress the impacts of uncertainty shocks. The impacts of uncertainty on firms are well docu-mented in the literature, while firms\u2019 responses to uncertainty shocks are less studied. Im et al.(2017) and Alfaro et al. (2019) find that firms adopt more conservative corporate policies suchas more cash holdings and fewer dividend payouts. This chapter, on the other hand, docu-ments a different kind of response: capital restructuring through going private. Studying firms\u2019responses to the uncertainty shocks helps us better understand economic uncertainty and howto recover from the negative impacts of uncertainty.36Figure 2.1 Going Private Transactions by Industry: 1994-2017The figure plots the percentage of going private companies across industries in the sample overthe period 1994-2017. The industries are classified based on Fama-French five industry classifi-cations. The shaded vertical bars represent NBER recessions.37Figure 2.2 Capital Structure before vs. after Going PrivateThe figure compares capital structures of the company before and after going private. Panel Aand Panel B illustrate the capital structures of American Greetings Corp. before and after itwent private in 2013. Panel C illustrates the average post-delisting capital structure of the goingprivate firms.Panel A. Capital Structure of American Greetings Corp. on Dec 31, 201238Panel B. Capital Structure of American Greetings Corp. after Going PrivatePanel C. Capital Structure of the Average Company after Going Private39Figure 2.3 Differences in Loan Rates Between Going Private Firms and Control FirmsThe figure illustrates the differences in loan rates between going private firms and matchedcontrol firms that remain public, in the pre-delisting period and the post-delisting period. Thecontrol firms are selected based on a propensity score matching on firm size, stock return, andstock return volatility of the year before a firm goes private.40Figure 2.4 Cumulative Abnormal Returns of the Going Private CompaniesThe figure plots the cumulative abnormal returns of the going private companies within the [-30d,+30d] period relative to the going private announcement date. The sample constitutes asubsample of the going private transactions which can be identified in the SDC M&A Database.41Figure 2.5 Uncertainty Shocks by Industry: 1994-2017The figure plots the oil, interest rate, exchange rate, and economic policy uncertainty acrossindustries from 1994 to 2017. Panel A, B, C, and D show the industry value-weighted averageof the instruments constructed based on oil, interest rate, exchange rate, and economic policyuncertainty shocks respectively. The industries are classified based on Fama-French five industryclassifications. The shaded vertical bars represent NBER recessions.Panel A. Oil Uncertainty ShocksPanel B. Interest Rate Uncertainty Shocks42Panel C. Exchange Rate Uncertainty ShocksPanel D. Economic Policy Uncertainty Shocks43Table 2.1 Descriptive StatisticsPanel A. Comparison of Firm Characteristics between Going Private Firms andthe Firms Remaining PublicThis table compares going private firms with a control sample of surviving firms over the periodof 1994-2017. The going private sample is the firms that filed for a Schedule 13E-3 (the goingprivate statement) and delisted within two years after the filing. The control sample constitutesthe firms that remain public at the end of 2017. Companies from financial and utility industriesare excluded from the sample. The summary statistics summarize firm characteristics over theentire public life cycle. *, **, and *** indicate statistical significance at the 10%, 5%, and 1%levels, respectively. Variables are defined in Appendix A.1.Going Private Firms Control FirmsMean SD Mean SD DifferenceStock Return Variables\u00a2Volatility 0.001 0.284 -0.005 0.267 0.006Volatility 0.616 0.435 0.476 0.353 0.140***Stock Return 0.120 0.634 0.178 0.614 -0.058***Control VariablesLog Sales 4.777 1.792 5.946 2.341 -1.168***Tobin\u2019s Q 1.537 1.848 2.101 2.494 -0.563***Leverage 0.201 0.199 0.174 0.181 0.028***Intangible Assets 0.114 0.158 0.140 0.177 -0.026***Return on Assets -0.014 0.197 -0.013 0.238 -0.001Tax Ratio 0.024 0.042 0.021 0.031 0.003***Other Firm CharacteristicsDual Class 0.026 0.160 0.033 0.179 -0.007**Institutional Ownership 0.113 0.140 0.173 0.144 -0.059***Asset Redeployability 0.421 0.102 0.402 0.101 0.018***Financial Distress 0.408 0.492 0.263 0.440 0.145***Loan to Bond Ratio 0.581 0.415 0.461 0.425 0.121***Log Relative Tobin\u2019s Q -0.109 0.871 0.088 0.795 -0.197***Log ERC -0.328 1.688 -0.264 1.865 -0.065*Analyst Coverage 5.109 5.166 8.711 7.450 -3.602***No. of Firms 525 2,659 3,184Firm-year Observations 4,915 43,145 48,06044Panel B. Summary Statistics of Going Private TransactionsThe table reports summary statistics of the going private transactions. The sample in-cludes a subsample of going private transactions for firms with debt outstanding before delisting,and with available information in the going private filings (13E-3, DEF13E-3, PRE13E-3 andSchedule TO). Variables are defined in Appendix A.1.Mean SD P10 P50 P90 Obs.Deal CharacteristicsDeal Value ($MM) 544.6 767.8 77.7 188 1500 84Premium (%) 34.5 14.9 19.2 32.1 68 84Post-delisting Equity StructureManagement Ownership 0.35 0.33 0.08 0.20 1 84Private Equity Ownership 0.64 0.33 0 0.78 0.90 84Other Existing Shareholder Ownership 0.01 0.04 0 0 0.06 84Source of Deal FinancingLeverage 0.61 0.20 0.31 0.66 0.81 84Bank Loan\/Total Debt 0.84 0.25 0.43 1 1 84Corporate Bond\/Total Debt 0.16 0.25 0 0 0.57 84Private Equity\/Total Equity 0.63 0.37 0 0.75 1 84Equity by Management\/Total Equity 0.21 0.31 0 0.06 0.95 8445Table 2.2 Uncertainty Shocks and Going Private TransactionsThis table reports results of the Cox proportional hazards models for time to go private,estimated using Equation (2.1). The sample includes going-private firms over the period of1994-2017 and a group of control firms that remain public. The dependent variable is the hazardrate of going private. In the Cox proportional hazards models, the firm-year observations aretreated as recurring censored events until the firm goes private or the end of 2017. Columns (1)and (2) report estimates from the Cox proportional hazards models, assuming that \u00a2Volatilityis exogenous. Columns (3)-(6) present control function estimates of the Cox proportional hazardmodels treating \u00a2Volatility as endogenous. Standard errors (in parentheses) are clustered atSIC 3-digit level. Columns (3)-(6) report bootstrapped standard errors with 300 replications. *,**, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. Variablesare defined in Appendix A.1.Cox Proportional Cox Proportional Hazards ModelHazards Model with Control Function(1) (2) (3) (4) (5) (6)\u00a2Volatilityi,t-1 0.41*** 0.40** 1.79*** 1.80*** 1.55*** 1.34***(0.15) (0.20) (0.17) (0.24) (0.23) (0.28)Volatilityi,t-2 0.93*** 0.75*** 1.33*** 1.28*** 1.11*** 0.82***(0.09) (0.14) (0.11) (0.19) (0.20) (0.23)Stock Returni,t-1 -0.54*** -0.33*** -0.44*** -0.37*** -0.41*** -0.44***(0.07) (0.09) (0.08) (0.09) (0.10) (0.11)Log Salesi,t-1 -0.15*** -0.10** -0.15*** -0.17***(0.04) (0.04) (0.04) (0.04)Tobin\u2019s Qi,t-1 -0.23*** -0.17* -0.12* -0.15**(0.08) (0.09) (0.07) (0.07)Leveragei,t-1 0.61** 0.34 0.16 0.26(0.30) (0.68) (0.42) (0.40)Intangible Assetsi,t-1 0.87*** 0.44 0.31 0.44(0.30) (0.62) (0.56) (0.58)Return on Assetsi,t-1 0.71** 0.91** 0.54** 0.51**(0.32) (0.39) (0.26) (0.24)Taxi,t-1 0.11 3.35 3.07 2.91(1.77) (2.36) (2.27) (2.42)1st Moment 10 IVi,t\u00b01 No No No Yes Yes YesIndustry FE No No No No Yes YesYear FE No No No No No YesFirm-year Observations 48,060 36,452 33,711 26,034 26,034 26,034No. of Firms 3,184 2,893 2,996 2,620 2,620 2,620No. of Going Private Firms 525 378 356 252 252 252Wald \u00ac2 171.0*** 132.7*** 133.4*** 166.6*** 364.6*** 2207.8***46Table 2.3 Uncertainty Shocks and Going Private Transactions: Matching Analysis onIPO and Pre-delisting CharacteristicsPanel A. At IPO ComparisonThis table compares firm characteristics between the going private firms and the control firmstwo years after IPO. The going private sample is the firms that filed for a Schedule 13E-3 (thegoing private statement) and delisted within two years after the filing. The control sample isconstructed with propensity score matching on firm characteristics (Fama-French 12 industry,log sales, Tobin\u2019s Q, and stock return) one year after IPO and three years before going private. *,**, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. Variablesare defined in Appendix A.1.Going Private Firms Matched Control FirmsMean SD Mean SD DifferenceStock Return Variables\u00a2Volatility 0.023 0.299 0.006 0.268 0.017Volatility 0.699 0.487 0.605 0.403 0.094Stock Return 0.121 0.723 0.185 0.747 -0.064Control VariablesLog Sales 4.305 1.908 4.417 1.779 -0.112Tobin\u2019s Q 2.557 2.981 2.715 3.420 -0.159Leverage 0.182 0.201 0.176 0.189 0.006Intangible Assets 0.095 0.138 0.096 0.161 -0.001Return on Assets -0.058 0.288 0.000 0.190 -0.058Tax Ratio 0.015 0.028 0.022 0.033 -0.006Other Firm CharacteristicsDual Class 0.067 0.251 0.078 0.269 -0.011Institutional Ownership 0.097 0.126 0.100 0.124 -0.003Asset Redeployability 0.422 0.107 0.424 0.112 -0.002Financial Distress 0.455 0.501 0.363 0.482 0.092Loan to Bond Ratio 0.614 0.358 0.644 0.401 -0.030Log Relative Tobin\u2019s Q 0.225 0.959 0.279 0.923 -0.054Log ERC -0.958 1.769 -0.277 1.765 -0.681**Analyst Coverage 4.087 3.221 4.208 3.203 -0.121No. of Firms 105 410 51547Pane B. Pre-delisting ComparisonThis table compares firm characteristics between the going private firms and the controlfirms one year before delisting. The going private sample is the firms that filed for a Schedule13E-3 (the going private statement) and delisted within two years after the filing. The controlsample is constructed with propensity score matching on firm characteristics (Fama-French 12industry, log sales, Tobin\u2019s Q, and stock return) one year after IPO and three years before goingprivate. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.Variables are defined in Appendix A.1.Going Private Firms Matched Control FirmsMean SD Mean SD DifferenceStock Return Variables\u00a2Volatility 0.002 0.308 0.001 0.268 0.002Volatility 0.588 0.428 0.554 0.393 0.033Stock Return -0.062 0.447 0.196 0.680 -0.258***Control VariablesLog Sales 5.315 1.735 5.526 1.757 -0.210Tobin\u2019s Q 1.380 1.603 1.778 2.024 -0.398*Leverage 0.197 0.218 0.195 0.193 0.002Intangible Assets 0.142 0.178 0.134 0.184 0.008Return on Assets -0.041 0.273 0.010 0.172 -0.050Tax Ratio 0.020 0.042 0.022 0.033 -0.002Other Firm CharacteristicsDual Class 0.069 0.255 0.082 0.275 -0.013Institutional Ownership 0.139 0.142 0.174 0.141 -0.035*Asset Redeployability 0.415 0.105 0.415 0.112 -0.000Financial Distress 0.347 0.478 0.246 0.431 0.100Loan to Bond Ratio 0.598 0.396 0.596 0.409 0.003Log Relative Tobin\u2019s Q -0.227 0.795 -0.073 0.865 -0.154Log ERC -0.098 1.701 -0.232 1.686 0.134Analyst Coverage 4.875 5.395 6.174 5.601 -1.298No. of Firms 105 410 51548Panel C. Cox Proportional Hazards Models for Time to Go PrivateThis table reports results of the Cox proportional hazards models for time to go private,estimated using Equation (2.1). The sample includes going-private firms over the period of1994-2017 and control firms that matched on firm characteristics both one year after IPO andthree years before delisting. The dependent variable is the hazard rate of going private. In theCox proportional hazards models, the firm-year observations are treated as recurring censoredevents until the firm goes private or the end of the sample period. The control samples incolumns (1)-(3) are matched on SIC 2-digit industry, log sales, and Tobin\u2019s Q respectively. Thecontrol sample in column (4) is constructed with propensity score matching on Fama-French 12industry, log sales, Tobin\u2019s Q, and stock return. Standard errors (in parentheses) are clusteredat SIC 3-digit level and bootstrapped with 300 replications. *, **, and *** indicate statisticalsignificance at the 10%, 5%, and 1% levels, respectively. Variables are defined in Appendix A.1.Cox Proportional Hazards Model with Control FunctionSIC2 Log Sales Tobin\u2019s Q P-score(1) (2) (3) (4)\u00a2Volatilityi,t-1 2.28*** 1.90*** 2.92*** 2.35***(0.43) (0.54) (0.93) (0.39)Volatilityi,t-2 1.39*** 1.07** 1.42* 1.06***(0.38) (0.46) (0.73) (0.41)Stock Returni,t-1 -0.42*** -0.62*** -0.24 -0.66***(0.16) (0.21) (0.40) (0.21)Log Salesi,t-1 -0.12** -0.11 -0.02 -0.12(0.06) (0.09) (0.15) (0.08)Tobin\u2019s Qi,t-1 -0.16 -0.27 -0.06 -0.14(0.11) (0.17) (0.23) (0.15)Taxi,t-1 2.96 1.19 6.50 3.39(3.50) (3.45) (6.68) (3.73)Leveragei,t-1 0.33 0.66 0.90 0.07(0.52) (0.73) (1.22) (0.59)Return on Assetsi,t-1 0.51 0.30 -0.37 0.15(0.40) (0.62) (0.98) (0.57)Intangible Assetsi,t-1 -0.21 -0.13 -0.72 0.43(0.72) (0.76) (1.60) (0.67)Control Variables Yes Yes Yes Yes1st Moment 10 IVi,t\u00b01 Yes Yes Yes YesIndystry FE Yes Yes Yes YesYear FE Yes Yes Yes YesFirm-year Observations 13,774 8,353 2,163 6,055No. of Firms 1,132 716 202 515No. of Going Private Firms 140 105 50 105Wald \u00ac2 2542.7*** 1488.6*** 96743.8*** 99444.8***49Table 2.4 Shareholder Conflicts and Going Private TransactionsThe table presents evidence of the economic mechanism, focusing on shareholder conflicts of thefirms. Results are estimated using Cox proportional hazards models with control functions. Thesample includes going-private firms over the period of 1994-2017 and a group of control firmsthat remain public. The dependent variable is the hazard rate of going private. Dual class is anindicator variable which equals to one if a firm has dual class shares in the year before goingprivate. Inst. Ownership is the percentage ownership by institutional blockholders. In the Coxproportional hazards models, the firm-year observations are treated as recurring censored eventsuntil the firm goes private or the end of the sample period. Standard errors (in parentheses)are clustered at SIC 3-digit level and bootstrapped with 300 replications. *, **, and *** indicatestatistical significance at the 10%, 5%, and 1% levels, respectively. Variables are defined inAppendix A.1.Panel A. Dual Class Shares Status and Going Private TransactionsCox Proportional Hazards Modelwith Control Function(1) (2) (3) (4)\u00a2Volatilityi,t-1 1.74*** 1.60*** 1.31*** 1.04***(0.18) (0.25) (0.24) (0.29)\u00a2Volatilityi,t-1 \u00a3 Dual Classi,t-1 2.46*** 5.50*** 6.26*** 6.90***(0.56) (0.96) (0.97) (0.96)Dual Classi,t-1 0.58** 0.34 0.14 0.09(0.28) (0.41) (0.40) (0.40)Volatilityi,t-2 1.35*** 1.26*** 1.08*** 0.77***(0.09) (0.19) (0.20) (0.23)Stock Returni,t-1 -0.43*** -0.36*** -0.41*** -0.45***(0.07) (0.09) (0.10) (0.16)Control Variables No Yes Yes Yes1st Moment 10 IVi,t\u00b01 No Yes Yes YesIndustry FE No No Yes YesYear FE No No No YesFirm-year Observations 33,711 26,034 26,034 26,034No. of Firms 2,996 2,620 2,620 2,620No. of Going Private Firms 356 252 252 252Wald \u00ac2 143.9*** 176.7*** 377.9*** 2325.5***50Panel B. Institutional Blockholders and Going Private TransactionsCox Proportional Hazards Modelwith Control Function(1) (2) (3) (4)\u00a2Volatilityi,t-1 2.57*** 1.97*** 1.82*** 1.40***(0.26) (0.35) (0.33) (0.36)\u00a2Volatilityi,t-1 \u00a3 Inst. Ownershipi,t-1 -2.75** -4.64** -4.79*** -5.03***(1.27) (1.86) (1.83) (1.78)Inst. Ownershipi,t-1 -1.88*** -2.61*** -2.37*** -1.97***(0.44) (0.65) (0.66) (0.64)Volatilityi,t-2 1.57*** 1.09*** 1.00*** 0.60***(0.12) (0.21) (0.20) (0.22)Stock Returni,t-1 -0.52*** -0.54*** -0.56*** -0.57***(0.09) (0.12) (0.12) (0.13)Control Variables No Yes Yes Yes1st Moment 10 IVi,t\u00b01 No Yes Yes YesIndustry FE No No Yes YesYear FE No No No YesFirm-year Observations 22,982 16,581 16,581 16,581No. of Firms 2,382 2,000 2,000 2,000No. of Going Private Firms 301 197 197 197Wald \u00ac2 169.6*** 215.2*** 385.7*** 4905.5***51Table 2.5. Shareholder-creditor Conflicts and Going Private TransactionsThe table presents evidence of the economic mechanism, focusing on shareholder-creditorconflicts of the firms. Results are estimated using Cox proportional hazards models with thecontrol function approach. The sample includes going-private firms over the period of 1994-2017and a group of control firms that remain public. The dependent variable is the hazard rate ofgoing private. Asset redeployability is minus one times the asset redeployability index from Kimand Kung (2017). Financial distress is an indicator variable if the Altman Z-score is lower than1.8. Loan to bond ratio is the ratio of outstanding bank loans to corporate bonds. In the Coxproportional hazards models, the firm-year observations are treated as recurring censored eventsuntil the firm goes private or the end of the sample period. Standard errors (in parentheses)are clustered at SIC 3-digit level and bootstrapped with 300 replications. *, **, and *** indicatestatistical significance at the 10%, 5%, and 1% levels, respectively. Variables are defined inAppendix A.1.Panel A. Asset Redeployability and Going Private TransactionsCox Proportional Hazards Modelwith Control Function(1) (2) (3) (4)\u00a2Volatilityi,t-1 1.43*** 1.37*** 1.12** 0.77(0.43) (0.49) (0.50) (0.83)\u00a2Volatilityi,t-1 \u00a3 Asset Redeployabilityi,t-1 0.97 1.69** 1.50** 1.45**(0.66) (0.76) (0.66) (0.74)Asset Redeployabilityi,t-1 -0.16* -0.17 -0.06 -0.05(0.09) (0.10) (0.10) (0.10)Volatilityi,t-2 1.20*** 1.23*** 1.07*** 0.76*(0.23) (0.29) (0.29) (0.42)Stock Returni,t-1 -0.49*** -0.45*** -0.48*** -0.51***(0.09) (0.13) (0.13) (0.14)Control Variables No Yes Yes Yes1st Moment 10 IVi,t\u00b01 No Yes Yes YesIndustry FE No No Yes YesYear FE No No No YesFirm-year Observations 30,892 21,705 21,705 21,705No. of Firms 2,640 2,264 2,264 2,264No. of Going Private Firms 333 214 214 214Wald \u00ac2 127.9*** 146.8*** 315.6*** 3507.7***52Panel B. Financial Distress and Going Private TransactionsCox Proportional Hazards Modelwith Control Function(1) (2) (3) (4)\u00a2Volatilityi,t-1 1.41*** 1.34*** 1.06*** 0.78**(0.27) (0.33) (0.33) (0.36)\u00a2Volatilityi,t-1 \u00a3 Financial Distressi,t-1 0.58* 0.89** 0.89** 0.79(0.34) (0.44) (0.44) (0.49)Financial Distressi,t-1 0.07 0.10 0.12 0.31(0.15) (0.20) (0.20) (0.20)Volatilityi,t-2 1.28*** 1.25*** 1.07*** 0.72***(0.10) (0.19) (0.21) (0.23)Stock Returni,t-1 -0.42*** -0.35*** -0.40*** -0.43***(0.07) (0.09) (0.10) (0.11)Control Variables No Yes Yes Yes1st Moment 10 IVi,t\u00b01 No Yes Yes YesIndustry FE No No Yes YesYear FE No No No YesFirm-year Observations 32,867 25,522 25,522 25,522No. of Firms 2,990 2,615 2,615 2,615No. of Going Private Firms 355 252 252 252Wald \u00ac2 133.2*** 174.5*** 411.8*** 2870.3***53Panel C. Bank Loans, Corporate Bonds and Going Private TransactionsCox Proportional Hazards Modelwith Control Function(1) (2) (3) (4)\u00a2Volatilityi,t-1 3.49** 3.02** 3.05** 4.19**(1.47) (1.49) (1.51) (1.83)\u00a2Volatilityi,t-1 \u00a3 Loan to Bond Ratioi,t-1 -4.19* -4.28* -4.28* -5.12**(2.48) (2.45) (2.44) (2.57)Loan to Bond Ratioi,t-1 0.23 0.20 0.10 0.27(0.21) (0.28) (0.26) (0.25)Volatilityi,t-2 1.00*** 0.36 0.35 0.59(0.33) (0.49) (0.52) (0.71)Stock Returni,t-1 -0.37*** -0.36* -0.41* -0.44*(0.14) (0.22) (0.23) (0.23)Control Variables No Yes Yes Yes1st Moment 10 IVi,t\u00b01 No Yes Yes YesIndustry FE No No Yes YesYear FE No No No YesFirm-year Observations 14,404 11,750 11,750 11,750No. of Firms 1,984 1,746 1,746 1,746No. of Going Private Firms 134 96 96 96Wald \u00ac2 48.6*** 103.4*** 664.1*** 2687.6***54Table 2.6 Bank Loan Rates of the Going-Private FirmsThe table compares loan rates of the going-private firms in the pre- and post-delisting pe-riods. Panel A reports summary statistics of the going private firms and a matched sampleof firms that remain public. Panel B compares the loan rates between the going private firmsand control firms in the pre-delisting and post-delisting periods. Panel C reports results ofthe difference-in-differences analyses studying the impacts of going private on loan rate. Thedependent variable is the loan rate. GP is a dummy variable which equals 1 if the firm goesprivate. Post is a dummy variable that equals one if the loan starts after the firm delists (or amatched loan for the control firm). All columns include year fixed effects. Columns (2)-(4) includematched pair fixed effects. The standard errors are in parentheses. The standard errors in PanelC are clustered at firm level. *, **, and *** denote significance at the 10%, 5%, and 1% levels,respectively. Variables are defined in Appendix A.1.Panel A. Summary StatisticsGoing Private Firms (GP) Control Firms DifferenceStock Return Volatility 0.64 0.58 0.06(0.13) (0.08) (0.14)Stock Return -0.26 -0.14 -0.12(0.13) (0.10) (0.17)Total Assets ($B) 3.30 2.36 0.94(1.96) (0.55) (1.60)Panel B. Loan Rate ComparisonsPre-delisting Post-delistingGP Control Difference GP Control DifferenceLoan Rate 7.34 5.89 1.46* 6.66 6.50 0.16(0.98) (0.35) (0.85) (0.66) (0.54) (0.89)Loan Rate Residual 1.44 -0.31 1.75** -0.01 -0.16 0.15with Year FE (0.79) (0.38) (0.76) (0.37) (0.31) (0.51)Loan Rate Residual 1.40 -0.12 1.52** -0.23 -0.23 0.00with Year & Matched Pair FE (0.64) (0.32) (0.64) (0.19) (0.31) (0.46)Panel C. Impact of Going Private on the Loan Rate(1) (2) (3) (4)GP = 1 \u00a3 Post = 1 -2.13* -2.46** -2.72** -2.30*(1.10) (1.01) (1.13) (1.19)GP = 1 2.19*** 1.72** 1.52** 1.28*(0.71) (0.66) (0.65) (0.68)Post = 1 0.39 -0.98 -1.10 -0.65(0.71) (0.95) (1.05) (1.13)Term Loan 0.88 0.93(0.68) (0.68)Secured Loan 1.09* 1.00(0.59) (0.59)Loan Maturity 0.04 0.03(0.19) (0.19)Log Loan Amount -0.33(0.30)Year FE Yes Yes Yes YesMatched Pair FE No Yes Yes YesAdjusted R2 0.39 0.53 0.56 0.56Observations 70 70 68 6855Table 2.7 Uncertainty Shocks and Going Private Transactions: Subsample AnalysisThe table reports results of the subsample analysis, estimated using Cox proportional hazardsmodels with control functions. The going private sample in Panel A and B include managementand non-management buyouts. The going private sample in Panel C and D involve privateequity and non-private equity buyouts. The control sample constitutes companies that remainpublic at the end of 2017. The dependent variable is the hazard rate of going private. In the Coxproportional hazards models, the firm-year observations are treated as recurring censored eventsuntil the firm goes private or the end of the sample period. Standard errors (in parentheses)are clustered at SIC 3-digit level and bootstrapped with 300 replications. *, **, and *** indicatestatistical significance at the 10%, 5%, and 1% levels, respectively. Variables are defined inAppendix A.1.Panel A. Management BuyoutsCox Proportional Hazards Modelwith Control Function(1) (2) (3) (4)\u00a2Volatilityi,t-1 2.60*** 2.85*** 2.85*** 3.41***(0.47) (0.79) (0.82) (1.02)Volatilityi,t-2 1.90*** 1.75*** 1.79*** 1.53**(0.21) (0.50) (0.53) (0.73)Stock Returni,t-1 -0.30** -0.38 -0.45** -0.44(0.14) (0.24) (0.23) (0.33)Control Variables No Yes Yes Yes1st Moment 10 IVi,t\u00b01 No Yes Yes YesIndustry FE No No Yes YesYear FE No No No YesFirm-year Observations 30,863 24,048 24,048 24,048No. of Firms 2,561 2,265 2,265 2,265No. of Going Private Firms 70 40 40 40Wald \u00ac2 39.6*** 124.4*** 214.9*** 13776.6***Panel B. Non-management BuyoutsCox Proportional Hazards Modelwith Control Function(1) (2) (3) (4)\u00a2Volatilityi,t-1 2.05*** 2.07*** 1.78*** 3.49***(0.36) (0.48) (0.49) (0.45)Volatilityi,t-2 1.48*** 1.30*** 1.08*** 1.59***(0.19) (0.33) (0.35) (0.37)Stock Returni,t-1 -0.32** -0.16 -0.20 -0.29(0.15) (0.19) (0.19) (0.21)Control Variables No Yes Yes Yes1st Moment 10 IVi,t\u00b01 No Yes Yes YesIndustry FE No No Yes YesYear FE No No No YesFirm-year Observations 31,509 24,575 24,575 24,575No. of Firms 2,639 2,345 2,345 2,345No. of Going Private Firms 128 101 101 101Wald \u00ac2 46.3*** 134.2*** 173.2*** 11036.7***56Panel C. Private Equity BuyoutsCox Proportional Hazards Modelwith Control Function(1) (2) (3) (4)\u00a2Volatilityi,t-1 2.03*** 2.67*** 2.48*** 4.56***(0.45) (0.68) (0.69) (0.81)Volatilityi,t-2 1.85*** 1.82*** 1.74*** 2.27***(0.20) (0.44) (0.44) (0.53)Stock Returni,t-1 -0.24 -0.34 -0.43 -0.58*(0.17) (0.27) (0.29) (0.31)Control Variables No Yes Yes Yes1st Moment 10 IVi,t\u00b01 No Yes Yes YesIndustry FE No No Yes YesYear FE No No No YesFirm-year Observations 30,866 24,074 24,074 24,074No. of Firms 2,560 2,267 2,267 2,267No. of Going Private Firms 74 46 46 46Wald \u00ac2 30.2*** 184.3*** 4294.1*** 97049.6***Panel D. Non-private Equity BuyoutsCox Proportional Hazards Modelwith Control Function(1) (2) (3) (4)\u00a2Volatilityi,t-1 2.34*** 2.08*** 1.80*** 2.69***(0.33) (0.45) (0.45) (0.46)Volatilityi,t-2 1.51*** 1.27*** 1.05** 1.24**(0.16) (0.34) (0.37) (0.42)Stock Returni,t-1 -0.37*** -0.18 -0.23 -0.28(0.13) (0.18) (0.19) (0.21)Control Variables No Yes Yes Yes1st Moment 10 IVi,t\u00b01 No Yes Yes YesIndustry FE No No Yes YesYear FE No No No YesFirm-year Observations 31,506 24,549 24,549 24,549No. of Firms 2,640 2,343 2,343 2,343No. of Going Private Firms 124 95 95 95Wald \u00ac2 58.6*** 117.7*** 167.0*** 7737.4***57Table 2.8 Alternative ExplanationsThe table reports results investigating the alternative hypotheses. Columns (1) and (2) examine the undervaluation hypothesis. Columns (3)and (4) examine the market distraction hypothesis. Columns (5) and (6) examine the information production hypothesis. Results are estimated usingCox proportional hazards models with control functions. The sample includes going private firms over the period of 1994-2017 and a group of controlfirms that remain public. The dependent variable is the hazard rate of going private. Relative Tobin\u2019s Q is the log of firm Tobin\u2019s Q relative to theindustry average. Log ERC is the log of earnings response coefficient. Analyst coverage is the number of analysts following the company. In the Coxproportional hazards models, the firm-year observations are treated as recurring censored events until the firm goes private or the end of the sampleperiod. Standard errors (in parentheses) are clustered at SIC 3-digit level and bootstrapped with 300 replications. *, **, and *** indicate statisticalsignificance at the 10%, 5%, and 1% levels, respectively. Variables are defined in Appendix A.1.Cox Proportional Hazards Model with Control FunctionUndervaluation Market Distraction Cost of Info. Production(1) (2) (3) (4) (5) (6)\u00a2Volatilityi,t-1 1.22*** 1.29*** 1.41*** 1.35*** 1.27*** 1.15**(0.28) (0.30) (0.28) (0.28) (0.36) (0.52)Relative Tobin\u2019s Qi,t-1 -0.36** -0.36**(0.15) (0.15)\u00a2Volatilityi,t-1 \u00a3 Relative Tobin\u2019s Qi,t-1 0.15(0.25)Log ERCi,t-1 0.05 0.06(0.07) (0.07)\u00a2Volatilityi,t-1 \u00a3 Log ERCi,t-1 -0.12(0.16)Analyst Coveragei,t-1 -0.09*** -0.09***(0.03) (0.03)\u00a2Volatilityi,t-1 \u00a3 Analyst Coveragei,t-1 0.03(0.11)Volatilityi,t-2 0.74*** 0.74*** 0.86*** 0.86*** 0.65** 0.65**(0.24) (0.24) (0.25) (0.28) (0.32) (0.33)Stock Returni,t-1 -0.43*** -0.43*** -0.44*** -0.44*** -0.63*** -0.63***(0.10) (0.10) (0.12) (0.12) (0.14) (0.15)58Table 2.8 Continued(1) (2) (3) (4) (5) (6)Control variables Yes Yes Yes Yes Yes Yes1st Moment 10 IVi,t\u00b01 Yes Yes Yes Yes Yes YesYear, Industry FE Yes Yes Yes Yes Yes YesFirm-year Observations 26,034 26,034 22,200 22,200 16,635 16,635No. of Firms 2,620 2,620 2,580 2,580 1,950 1,950No. of Going Private Firms 252 252 223 223 144 144Wald \u00ac2 2128.9*** 2200.8*** 2203.2*** 2209.4*** 2439.2*** 2504.8***59Chapter 3Relative Pricing of Private and Public Debt:The Role of Money Creation Channel3.1 IntroductionBank loans and public bonds are the two most important sources of debt for non-financialfirms. While there is a large literature studying the optimal debt structure of firms and het-erogeneity in firms\u2019 reliance on bank loans and bonds (Diamond, 1991; Rajan, 1992; Bolton andFreixas, 2000; Rauh and Sufi, 2010; Becker and Ivashina, 2014), less is known about determi-nants of the relative cost of raising funds in the private versus public debt market. The chapterfills this gap by offering evidence on the role of the bank money creation channel in explainingthe relative pricing of bank loans and public bonds in the primary market.Our work is motivated by the theoretical literature that explains how banks create demanddeposits \u2013 money-like securities that are redeemable at par and liquid. For a security to bemoney-like, its value needs to be insensitive to information. Dang et al. (2015) show that whendebt, a relatively information insensitive financial claim, is used as a collateral for another debtcontract, the latter debt contract is the least sensitive to information. By lending to firms, banksuse this \u2018debt-on-debt\u2019 structure to create demand deposits that are money-like. Dang et al.(2017) show that banks\u2019 money-creation function can be further enhanced if they issue loansto borrowers that are unlikely to fail and opaque, and when banks keep the information theyproduce about the borrowers secret. If banks\u2019 assets are safe but hard to evaluate by outsideinvestors, the return to outsiders from exerting effort to learn about such opaque assets is low.60Little information is produced outside of banks, which ensures that banks\u2019 demand depositsare insensitive to information and thus money-like \u2013 by lending to safe and opaque projects,banks lower their costs of private money production.Guided by this theory, we argue that the production of private money creates an incentive tolend to firms that are safe and opaque, which is unique to banks. Since banks benefit from theopacity of their borrowers, it should be less expensive for banks, relative to non-bank lenders, tofinance such borrowers. From a firm\u2019s perspective, this reasoning implies that the cost of bankcredit relative to the cost of raising funds in the public debt market should depend on the firm\u2019sopacity \u2013 firms whose assets are harder to evaluate by outside investors will have a relativelylower cost of capital when they borrow from banks.To quantify the effect of this money creation channel on a firm\u2019s cost of borrowing, we ex-amine how the relative pricing of loans and bonds changes in response to shocks to the firm\u2019sopacity, or in other words, shocks to investors\u2019 information acquisition cost about the firm\u2019s as-sets. We construct a granular dataset on loan-bond pairs issued by the same firm in the primarymarket, with investment-grade credit rating, with the same maturity and seniority, and at thesame time. We define our main variable of interest \u2013 the loan-bond spread \u2013 as the within firmdifference in the price of bank credit relative to public debt for each loan-bond pair in the sam-ple. We consider several alternative proxies to construct the main independent variable \u2013 thechange in the cost of acquiring information about a firm\u2019s assets.1 This shock should capturehow hard it is for outside investors to learn about the fundamental value of the firm\u2019s assets.In our baseline analysis, following the growing literature on the impact of economic un-certainty on the corporate sector, we measure information acquisition cost using changes involatility of the firm\u2019s stock returns, as well as changes in volatility of the firm\u2019s stock returns thatare induced by aggregate uncertainty shocks. An increase in equity volatility and an increase inequity volatility due to aggregate uncertainty shocks plausibly makes it harder for outside in-vestors to learn about the firm and thus can be considered an increase in a firm\u2019s opacity in1We refer to firm opacity and the cost of acquiring information by outside investors synonymously throughoutthe chapter.61accordance with the notion of opacity in Dang et al. (2017). Our main hypothesis is that a pos-itive information cost shock leads to a larger increase in the firm\u2019s cost of public debt relativeto that of bank credit, that is, the loan-bond spread decreases. This is because a positive infor-mation cost makes it harder for outside investors to learn about the firm\u2019s assets, which reducesthe return on producing information about those assets, thereby making such assets relativelyeasier to be funded by the banking sector.Using a sample of matched loan-bond pairs, we find that a positive firm-level informationcost shock reduces the loan-bond spread, suggesting that higher cost of acquiring informationmakes bank credit relatively cheaper compared to public debt. In terms of economic magni-tudes, one standard deviation increase in the level of a firm\u2019s opacity is associated with a re-duction in the loan-bond spread of 22 bps, which is economically significant given an averagespread of 123 bps. We refer to this reduction in the loan-bond spread in response to the firm-level information cost shock as the \u2018opacity discount\u2019.Since we focus on the within-firm response of the loan-bond spread to the information costshock, our main finding is not influenced by any issuer-level time-varying characteristics, bothobservable and unobservable, such as firm credit risk, growth opportunities, or governance,that could affect the pricing of either of the two debt contracts. Furthermore, our result surviveswhen we include different sets of fixed effects, for example, when we control for unobservablebank\/underwriter time-invariant characteristics, or when we include different sets of controlvariables, mainly to capture detailed characteristics of loan and bond contracts used in anyspecific deal.One concern with using changes in the volatility of stock returns as a proxy for informationcost is that unobserved firm-level factors can simultaneously affect the volatility of stock returnsand the loan-bond spread. For example, a departure of a firm\u2019s CEO can lead to a change in thefirm\u2019s stock return volatility and, at the same time, to a change in the loan rate due to the lossof a bank relationship. In this case, the negative relationship between an increase in volatilityand the loan-bond spread cannot be attributed to the information acquisition cost channel. To62establish the causal effect of a firm-level information cost shock on the relative cost of bankcredit, we employ an instrumental variable estimation approach following Alfaro et al. (2021).Specifically, we instrument changes in firm-level volatility using firms\u2019 differential exposuresto changes in aggregate volatility of the macro variables such as energy, currency, policy, andU.S. Treasury notes. The instruments, by construction, capture only those changes in firm-level volatility that are induced by changes in aggregate volatility of the macro variables. Hence,this approach allows us to rule out alternative factors, such as changes in firm fundamentals,which can simultaneously lead to an increase in firm-level volatility and the loan-bond spread.The results from the instrumental variable estimation are consistent with the baseline analysis.We find that firms experiencing positive information cost shock induced by aggregate volatilityshocks receive opacity discounts on bank debt.To provide further evidence that our information cost measure captures changes in equityvolatility induced by exogenous factors, we look at the impact of the 9\/11 uncertainty shockevent on firms\u2019 cost of borrowing from banks relative to that from the public bond market. The9\/11 shock was exogenous to firm fundamentals but it had a significant impact on firm-levelvolatility. We show that firms that experienced a larger increase in volatility after the 9\/11 shockhad lower loan-bond spreads in the post-shock period. This finding again lends support tothe hypothesis that an increase in the information acquisition cost made it harder for outsideinvestors to learn about the firms\u2019 fundamentals, which in turn made it relatively easier forbanks to fund projects of these firms.We consider two alternative measures of firm information acquisition cost. First, we followAnderson et al. (2009) to construct a firm-level opacity index, which ranks the relative opacity offirms in our sample based on four proxies for opacity: bid-ask spread, trading volume, analystcoverage, and analyst forecast errors. The firm-level opacity index is the sum of the rankingsbased on these four variables, normalized by 20. A higher opacity index indicates larger infor-mation asymmetry which should make it more difficult for outside investors to evaluate thefirm\u2019s assets. Second, we follow Morgan (2002) and measure information acquisition cost using63disagreement in ratings assigned to a firm by Standard & Poor\u2019s and Moody\u2019s. If a firm\u2019s assetsare harder to evaluate, there will be more disagreement among rating agencies about the truevalue of the firm\u2019s assets. Following this argument, we define a proxy for information acquisitioncost as the absolute difference in ratings assigned to a firm by Standard & Poor\u2019s and Moody\u2019s.Using both these alternative measures of firm-level information acquisition cost we obtain re-sults that are consistent with our hypothesis: as the information cost increases, the loan-bondspread shrinks.Next, we provide direct evidence on the economic mechanism driving our results. Accord-ing to the financial intermediation theory, the need for banks to maintain opacity should belarger when private money creation is not backed by the government. Specifically, deposit in-surance provided by the government on demandable debt produced by banks makes such in-sured deposits insensitive to information, reducing banks\u2019 need to maintain opacity throughtheir lending decisions. Supporting this hypothesis, Chen et al. (2020) show that uninsured de-posits are more responsive to negative information about banks\u2019 assets. We hypothesize thata higher opacity discount should be offered by banks that create relatively more liquidity inthe form of uninsured deposits and that experience outflows due to the uncertainty about thevalue of banks\u2019 assets. To preserve the value of uninsured deposits and to prevent further de-posit outflows, these banks should invest in firms whose assets are hard to evaluate by outsideinvestors.To test this hypothesis, we exploit variation in the ratio of uninsured deposits and uninsureddeposit outflows across banks and test whether banks with a higher ratio of uninsured depositsoffer a larger opacity discount when they see a larger outflow of uninsured deposits. We find thatbanks relying more on uninsured deposits offer significantly larger opacity discounts to firmsafter they experience large uninsured deposits outflows. This result supports the presence ofthe money creation mechanism.To provide further evidence that these deposit outflows are induced by investors\u2019 confidencein the asset quality of the bank, we use the money market dollar funding shock of April 2011.64European banks in the U.S. raise most of their dollar funding from uninsured sources, such asthe commercial paper market while the dollar funding of U.S. banks is mostly sourced from in-sured retail deposits (Ivashina et al., 2015). In April 2011, money market funds started becomingconcerned about European banks\u2019 exposure to Greek sovereign debt and they reduced their ex-posure to the Eurozone banks active in the U.S., which led to uninsured deposit outflows fromthese banks. We hypothesize that following the money market funding shock, the Europeanbanks should offer larger opacity discounts to firms whose assets are harder to evaluate sincethose banks have a greater need to keep information about their assets secret to prevent furtherwithdrawals. Our results lend support to this hypothesis. Using difference-in-differences anal-ysis, we find that a relatively larger opacity discount was offered by European banks to firmswhose assets became harder to evaluate in the post-shock period.We also exploit the role of implicit government guarantees after crisis to further isolate therole of the money creation mechanism. Following the 2008-09 financial crisis, the Financial Sta-bility Board started to publish a list of global systemically important banks (G-SIBs). Such banksreceive implicit \u2018too-big-to-fail\u2019 guarantees. These guarantees reduce banks\u2019 need to managethe opacity of their assets, and G-SIBs should thus offer lower opacity discounts to their bor-rowers. To test this hypothesis, we conduct a difference-in-differences analysis to investigatewhether banks offer lower opacity discounts after they are classified as G-SIBs. We show thatnon-G-SIBs offer a significantly larger opacity discount to firms following a positive informa-tion cost shock after 2009 compared to G-SIBs. This result suggests that while banks no longerneed to manage opacity after they are classified as G-SIBs, the need for using lending decisionsto manage opacity remains present for non-G-SIBs. Both these results lend further support tothe presence of the money creation mechanism.Can relationship lending explain our results? Existing studies have shown that firms withlonger bank relationships pay lower interest rates (Berger and Udell, 1995) and, during periodsof crises, firms are able to receive cheaper bank credit from relationship banks compared tobanks with which firms have only transaction lending relationships (Bolton et al., 2016). It is65thus possible, that the lower loan-bond spread in response to firm information cost shocks isdriven by relationship lending and not by the money creation channel. We test this hypothesisby evaluating how the loan-bond spread responds to the interaction between firm informationcost and length of the firm\u2019s relationship with the bank. We find that the opacity discount of-fered by banks does not depend on the length of the firm\u2019s relationship with the bank. Thisfinding suggests that relationship lending is not driving our results.Is it possible that firms\u2019 demand for bank credit relative to public bonds explains our re-sults? If firms that experience larger volatility in their stock returns lower their demand for bankcredit and increase their demand for public debt, this relative fall in demand for bank creditmay translate into a lower cost of bank credit and a higher cost of public debt and may therebygenerate a negative relationship between firm opacity and the loan-bond spread. If firms ac-tively switch to public debt when they are hit by an opacity shock, we should see a fall in theshare of bank loans to the total credit by firms that experience positive opacity shocks. Our re-sults show that this is not the case. We do not find any significant effect of the firm informationcost shock on the share of loans in total amount borrowed by the firm.Another alternative explanation of our results may be that banks could participate less inloan syndicates when borrowers are opaque, which may lower their exposure and thereby lowertheir cost of lending to such borrowers. To investigate this alternative hypothesis, we examinethe effects of firm information cost shock on bank participation in loan syndicates and find nosignificant relationship between the two.In summary, our evidence suggests that the reduction in the loan-bond spread for firmswhose assets become harder to evaluate by outside investors reflects the money creation chan-nel. A key takeaway from our findings is that the need for banks to maintain opacity to performtheir core function of money creation does have meaningful benefits for non-financial firmsthat are relatively more opaque but safe. This result has important policy-relevant implica-tions. Mainly, regulations that promote greater public disclosure of banks\u2019 assets may adverselyaffect cost of capital for safe but opaque firms, especially for those that are not able to access66the public bond market.The chapter contributes to the literature on the relative pricing of private and public debt.Schwert (2020) finds that loans are relatively overpriced compared to a bond-implied creditspread. Relative to Schwert (2020), our contribution to the literature is twofold. First, we com-pare the at issuance price of bonds with the cost of banks loans while Schwert (2020) comparesthe pricing of loans with a model implied spread based on secondary market quotes of tradedbonds. Hence, our analysis provides a direct empirical estimate of the difference in firms\u2019 costof raising funds in the form of bank loans vs bonds. Second, we identify an economic channelsuggested by theory \u2013 the money creation channel \u2013 that we show partially explains the dif-ference in the pricing of bank loans and bonds. To the best of our knowledge, ours is the firstchapter to provide empirical evidence on the link between the money creation channel and therelative pricing of private and public debt.The chapter builds on the recent literature (Dang et al., 2017) that endogenizes bank opacityand studies how banks optimally manage their opacity to support their core function of moneycreation. This recent literature belongs to the larger literature on the costs and benefits of bankopacity. Cost of bank opacity is studied in asset-based theories of financial intermediation thathighlight the disciplining role of bank transparency (Diamond, 1984; Calomiris and Kahn, 1991;Diamond and Rajan, 2001). Empirical studies documenting the negative effects of bank opac-ity focus on the impact on financial stability (Jones et al., 2012; Flannery et al., 2013; Acharyaand Ryan, 2016) and bank lending (Zheng, 2020). On the other hand, benefits of bank opacityare highlighted in liability based theories of financial intermediation (Gorton and Pennacchi,1990; Dang et al., 2017). More broadly, the chapter contributes to the literature on costs andbenefits of public disclosure of information. Starting from Hirshleifer (1971), a large theoreticalliterature has argued that more information is not always better and, in some instances, opacitycan be socially optimal.2 In the context of banking, Dang et al. (2017) argue that financing ofopaque projects helps the money creation function of banks. The chapter adds to this literature2See, for example, Kaplan (2006); Monnet and Quintin (2017); Andolfatto (2010); Pagano and Volpin (2012).67by being the first to provide empirical support to the endogenous bank opacity for money cre-ation hypothesis of Dang et al. (2017), and by showing the potential benefits of this endogenousbank opacity on the cost of credit for non-financial firms.The costs of public disclosure of information have also been extensively studied in the con-text of supervisory stress-tests for banks. Most studies arguing against public disclosure of su-pervisory test results focus on the proprietary costs for banks whose information is being dis-closed (Dye, 1986; Darrough and Stoughton, 1990; Gigler, 1994) or the negative effects on risksharing (Hirshleifer, 1971; Goldstein and Sapra, 2014), or inefficient ex post reaction to disclo-sure (Morris and Shin, 2002). We highlight a new channel \u2013 cost of borrowing for opaque andsafe firms \u2013 through which public disclosure of stress test results can negatively affect the realeconomy.Our work also adds to the literature studying the effects of disclosure by non-financial firms.Diamond and Verrecchia (1991) show that while more information revelation can reduce thecost of capital for firms, it can negatively affect liquidity in the secondary market. Some stud-ies have shown that greater disclosure can have a negative effect on non-financial firms by re-vealing trade secrets to competitors (Bernard, 2016; Li et al., 2018). Jayaraman and Wu (2019)show that mandated disclosure can have a negative effect on managerial learning. Agarwalet al. (2018) show that greater transparency of a firm\u2019s assets through mutual funds\u2019 portfoliodisclosures leads to myopic corporate investment behaviour and leads to a negative effect oncorporate innovation. The chapter shows that another important channel through which infor-mation revelation can affect firms is the bank credit channel.Finally, we contribute to the literature examining the effects of bank frictions on corporatelending. Adverse capital and liquidity shocks to banks are transmitted to borrowers throughreductions in credit supply and stricter loan contracts (Peek and Rosengren, 1997, 2000; Khwajaand Mian, 2008; Paravisini, 2008; Chava and Purnanandam, 2011; Murfin, 2012; Schandlbauer,2017). Other studies emphasize the role of bank capital for bank lending behavior (Thakor,1996; Gambacorta and Mistrulli, 2004; Behn et al., 2016; Fraisse et al., 2020). Gornall and Stre-68bulaev (2018) model the joint capital structure decisions of banks and their borrowers and ar-gue that bank leverage and firm leverage are both strategic substitutes and complements. Ourresearch points out that banks\u2019 role as liquidity providers may affect loan pricing even in theabsence of frictions.The chapter is organized as follows. Section 3.2 discusses the conceptual framework and themain hypotheses. Section 3.3 describes the sample, data, and our empirical strategies. Section3.4 summarizes the main results. Section 3.5 shows results for a variety of robustness tests, andSection 3.6 concludes.3.2 Conceptual Framework and HypothesesThe conceptual framework underlying our main hypothesis is derived from the financial in-termediation theory which argues that banks need to be opaque with respect to third partiesin order to perform one of their core functions \u2013 creation of demand deposits, that is, privatemoney in the form of securities redeemable at par. This money creation is facilitated by opacityof banks\u2019 assets. The argument for why money is created by banks and why banks\u2019 endogenousopacity facilitates money creation proceeds in two steps.First, Gorton and Pennacchi (1990); Holmstr\u00f6m (2009); Dang et al. (2015); Holmstr\u00f6m (2015)note that, for a security to be money-like it should have the desirable features of liquidity andsafety, that is, agents should be able to use money for economic transactions without worryingabout the fact that its value will change over time due to trading by privately informed agents.3These studies argue that one way to achieve such value-invariance is to make money insensitiveto information, either public or privately produced. This implies that optimal design of thesesecurities should be such that agents have the least incentives to produce private informationabout the payoffs of these securities.Dang et al. (2015) show that when debt is used as a collateral for another debt contract, the3As an example, tradable shares of a non-financial firm are not money-like because the value of tradable shareschanges over time as a result of trading by privately informed agents.69\u2018debt-on-debt\u2019 structure is least sensitive to information. This is because it preserves symmetricignorance optimally from the security design perspective \u2013 the debt-on-debt contract structureminimizes the incentive of third parties to produce private information about the payoffs. Theidea is that if the collateral value protecting the debt contract is sufficiently high relative to theface value of the debt contract, producing costly information about the exact value of the col-lateral is not worthwhile. If a particular debt contract is almost always information-insensitive,then using that debt contract as collateral for another debt contract makes the second debtcontract even more information-insensitive. For this reason, bank\u2019s demand deposits (whichare essentially debt claims) issued against loans on the bank\u2019s asset side (which are also debtclaims) are the least information-sensitive, which makes them liquid and safe.The second conceptual point, made in Dang et al. (2017), is that, in order for the abovemechanism to work, a bank needs to be able to keep the information about its assets secret.This implies that a bank needs to select its assets so that expert outsiders do not have an incen-tive to produce private information about the value of the bank\u2019s assets. This can be accom-plished if the bank makes loans that are costly for outsiders to learn about, such as loans tosmall businesses or firms that are opaque. Opacity makes it costly for an expert investor to findout information about the details of the bank\u2019s balance sheet, eliminating the expert\u2019s informa-tional advantage and thereby facilitating the value-invariance of money. Outside investors willalso have a low incentive to produce private information if the projects that the bank funds areunlikely to fail. As a result, to support their function of private money creation, banks have,relative to other financial intermediaries, a unique incentive to issue loans to projects that areunlikely to fail and opaque because such assets lower the cost of producing private money themost.From the firms\u2019 perspective, banks\u2019 intrinsic need to maintain opacity implies that the costof bank credit relative to that of public debt, as an alternative form of debt financing, shoulddepend on firms\u2019 opacity. Specifically, all else equal, firms whose assets are harder to evaluateby outside investors will have a lower cost of capital when they borrow from banks. This effect70does not rely on any comparative advantage that banks may have in evaluating and overseeingprojects. Based on these theories, we develop our main hypothesis that an increase in the cost ofacquiring information about a firm\u2019s assets leads to a larger increase in the firm\u2019s cost of publicdebt relative to bank credit. In other words, banks offer \u2018opacity discounts\u2019 to firms experiencingpositive information cost shocks, and larger opacity discounts should be offered by banks witha greater need to finance opaque projects.3.3 Sample, Data, and Empirical Methodology3.3.1 Data and Sample ConstructionOur main analyses rely on a unique sample design which facilitates within firm comparisonof the pricing of new loans and bond issuances. We construct loan-bond pairs issued by non-financial public firms in the U.S. using data on new loan facilities from the Dealscan databaseand data on bond issuances in the primary market from the FISD database over 1995-2019.4Within firm pairing of loan contracts with bond issuances is based on date of issuance andmaturity of the contracts. For each firm, we pair new loan facilities and bond issuances withthe same maturity and those that are issued within a window of 60 days. If one loan facility ismatched to multiple bond issuances, we retain the one with the closest issuance date to theloan facility start date. We restrict the sample to senior loans and bonds denominated in USD.For each loan-bond contract pair, we create our main dependent variable as the differencebetween the loan spread and the bond spread \u2013 the loan-bond spread. Our dependent variablemeasures a firm\u2019s cost of bank credit relative to that of public debt. By constructing loan-bondpairs issued by the same firm at the same time, we control for any issuer-level time-varyingcharacteristics, both observed and unobservable, like firm credit risk, growth opportunities,or governance, that affect the pricing of either of the two debt contracts. We also control for4Our sample starts in starts in 1995 to exclude the period of high volatility in the bond market before 1994(Lemmon and Roberts, 2010).71maturity, the key contract level characteristic affecting pricing, by selecting the loan-bond pairwith the same maturity category, i.e., short-term, mid-term or long-term loans and bonds.Another important factor that can affect the loan-bond spread is the seniority of bank debtin bankruptcy. A higher recovery rate for bank loans compared to bonds in the state of defaultimplies a relatively larger cost of public debt. This difference in recovery rate is especially rele-vant for firms that are likely to default but should matter less for safer firms. This is supportedby Schwert (2020) who shows that the loan and bond spreads are statistically indistinguish-able from each other when firms are far from default. To alleviate concerns that seniority ofbank loans in bankruptcy could drive our results, we restrict our sample to investment graderated firms since these firms are the least likely to default, which should make the seniority ofdifferent debt contracts relatively less important as a pricing factor. Restricting the sample toinvestment grade firms is also consistent with the financial intermediation theory of endoge-nous bank opacity and money creation as the theory suggests that the banks\u2019 function of moneycreation is supported by investment in projects that are unlikely to fail and opaque.After imposing these restrictions, our final sample consists of a quarterly panel of 1,597 loan-bond pairs issued by 414 firms in 1995-2019. On average, each firm has 3.86 loan-bond pairsthroughout the sample period. We describe our sample construction process in detail in TableB.3.1.We use several measures to construct the main independent variable \u2013 firm informationcost shock. First, following the literature that examines consequences of economic uncertainty,we construct a quarterly measure of firm information cost shock as the year-on-year changein the annualized stock return volatility of the firm for each quarter. Second, to address con-cerns that changes in firm fundamentals can simultaneously affect firm-level volatility and theloan-bond spread, we employ an instrumental variable approach following Alfaro et al. (2021).Specifically, we instrument firm opacity shocks using aggregate volatility shocks of macro vari-ables such as currency, energy, policy, and U.S. Treasury. We use data from Bloomberg, the St.Louis Fed, and the Economic Policy Uncertainty Index from Baker et al. (2016) to construct the72instruments. Third, we use the 9\/11 shock and define the information cost shock as the changein firm-level volatility in a small window around the 9\/11 event. Finally, we use two alterna-tive measures to measure firm-level information cost and construct an opacity index measurefollowing Anderson et al. (2009) and Morgan (2002). Section 3.3.3 discusses these measures indetail.Data on firm characteristics and stock returns are from Compustat and CRSP. Data on loan-and bond-level characteristics are from Dealscan and FISD. Information on bank lenders isfrom the FR Y-9C reports filed by bank holding companies. See Appendix B.1 for variable defi-nitions.3.3.2 Summary StatisticsTable 3.1 reports sample summary statistics. All the variables are winsorized at 1% and 99%level. Our sample consists of a quarterly panel of 1,597 loan-bond pairs issued by 414 firmsin 1995-2019. On average, each firm has 3.86 loan-bond pairs throughout the sample period.Since we focus on firms that access both the loan and bond market, and issue investment-gradepublic debt, they are large and have strong fundamentals. The average size of firms in our sam-ple is 1.4 billion USD (in terms of asets). On average, firms in our sample exhibit positive returnon assets in the quarter before the loan-bond pair origination, and the average firm exhibits astock return of 16% in the year before the loan-bond pair origination.Since we restrict the sample to firms with investment-grade bonds, the firms are far fromdefault with the implied probability of default (Bharath and Shumway, 2008) close to zero. Theaverage bond in our sample is rated BBB+, with the vast majority of bonds falling between A toBBB- rating. On average, the cost of bank loans is 124 bps lower than the costs of borrowing inthe bond market. The average loan facility amount is 1.34 billion and the bond face value is 639million. On average, the loan facility accounts for 61% of the total borrowing amount within theloan-bond pair.Most loans in our sample are syndicated loans; 16.5% of them are term loans and 15.4% are73secured loans. The fraction of secured loans is lower compared to that in Schwert (2020), con-sistent with the fact that loans to investment-grade firms are generally unsecured. Only 3% ofthe bonds in our sample are secured bonds. In our empirical analyses, we control for whetherloans and bonds are secured to make sure our results are not driven by differences in colla-terizability of the debt contract. The fraction of redeemable bonds is 89%. 27% of the bondshave embedded investor options. 64% have bondholder protective covenants, and 66% havenegative-pledge covenants. We include an indicator variable for each of these bond character-istics in our regressions to control for any difference in bond pricing due to variation in bondfeatures.In Figure 3.1, we illustrate the time series patterns of the loan-bond spread. The spreadis close to zero in normal times and it becomes significantly negative in recessions. Duringeconomic recoveries, the loan-bond spread gradually approaches zero as the loan rate approx-imates bond rate. This pattern suggests that changes in loan-bond spread over time are mainlydriven by fluctuations in the loan rate, rather than the bond rate. Overall, the time-series pat-tern of the loan-bond spread is consistent with the money creation channel according to whichthe spread should be more negative in periods of stress when opacity is needed the most forbanks to maintain the value-invariance of demand deposits.Figure 3.2 demonstrates the cross-sectional properties of the loan-bond spread. It showshow the loan-bond spread varies across the firm cost shock distribution. We divide the loan-bond pairs in our sample into quartiles based on the firm information cost shock. The firminformation cost shock is measured using firm-level changes in return volatility. To isolate theeffects of firm information cost shocks from other factors, we plot loan rate residuals and bondyield residuals from a regression of loan rates and bond yields on loan\/bond maturity, stockvolatility, and year fixed effects. We discuss two key observations from Figure 3.2.First, Figure 3.2 shows that firms in the bottom quartile of information cost shock pay morefor bank credit than for public debt. This finding is consistent with the bank money creationhypothesis. The cost of private money production is higher when banks\u2019 assets are less opaque.74Therefore, banks can finance such projects only when they receive higher compensation fromfunding these projects than that received in the bond market.Second, Figure 3.2 shows that as the firm-level information cost increases, it becomes moreexpensive for banks as well as the public bond market to finance such firms \u2013 both the bond rateand loan rate increases with the magnitude of the information cost shock. However, banks alsobenefit from the opacity of their borrowers as it helps their money creation function, therefore,even though the loan rate increases with the level of firm-level uncertainty, the rate of increaseis much lower compared to the rate of increase in the bond rate. As a result, the loan-bondspread decreases with higher firm-level information acquisition cost. On average, firms in thehighest information cost shock quartile exhibit a 29 bps lower loan-bond spread compared tothose in the bottom quartile.3.3.3 Information Cost Shock and the Loan-bond SpreadIn this section, we consider several different alternatives to measure shocks to information ac-quisition cost of firms\u2019 assets and study the subsequent effect on the loan-bond spread.Firm-level Uncertainty Shock: OLS EstimationFollowing the growing literature on the impact of economic uncertainty on the corporate sector,we use uncertainty shocks to capture changes in the cost of information acquisition for outsideinvestors. We argue that it becomes more costly for investors to learn about the fundamentalvalue of firms whose stock experiences a larger increase in volatility. In particular, to study theimpact of an information cost shock on the loan-bond spread, we estimate:(loan\u2212bond)i ,t =\u03b21+\u03b22\u00b7\u2206\u03c3i ,t\u22121+\u03b23\u00b7\u03c3i ,t\u22125+\u03b24\u00b7ri ,t\u22121+\u03b25\u00b7Bondi ,t+\u03b26\u00b7Loani ,t+\u03c6 j+\u03c8t+\u03f5i j ,t .(3.1)The dependent variable is the within-firm difference between loan and bond spread for firmi at quarter t when the loan facility starts. \u2206\u03c3i ,t\u22121 is our measure of firm information cost shock,75which is a year-on-year change in the annualized equity volatility lagged by one quarter fromthe facility start date. Based on the bank opacity hypothesis, an increase in equity volatilityshould make it harder for outside investors to acquire precise information about the firm and,therefore, should lower the relative cost of bank credit to public debt. This implies that theestimate of \u03b22 should be negative and statistically significant.We include lagged stock return ri ,t\u22121 and the annual stock return volatility measured beforethe information cost shock, \u03c3i ,t\u22125, to control for firm credit risk. Bondi ,t is a vector of bond-level attributes such as bond rating, indicators for whether a bond is secured, redeemable,with embedded investor options, as well as whether the bond has bondholder-protective ornegative-pledge covenants. Loani ,t is a vector of attributes for loans, including indicators forterm and secured loans, and the loan amount. The loan-bond pairs are matched on maturitycategory, that is, short-term, mid-term, and long-term. In addition, we control for the exact dif-ference in maturity between the loan facility and the bond issuance. We include year-by-quarterfixed effects, \u03c8t , to control for any macroeconomic factors affecting the differential pricing ofloans versus bonds, such as the state of a business cycle or monetary policy shocks. Last, weinclude fixed effects for lead banks or bank holding companies, \u03c6 j , to control for unobservedlender\/underwriter time-invariant characteristics that might affect the loan-bond spread.Firm-level Uncertainty Shock: IV EstimationThe estimated results from Equation (3.1) could be biased if unobserved factors simultaneouslyaffect the change in the firm\u2019s equity volatility and the loan-bond spread. For example, thedeparture of a firm\u2019s CEO may result in an increase in stock return volatility and an increase inthe loan rate due to the loss of bank relationship at the same time. To isolate the causal effectof information cost shocks driven by factors exogenous to firm fundamentals, we follow Alfaroet al. (2021) and employ an instrumental variable estimation that allows us to capture changesin firm information cost caused by factors orthogonal to unobservable firm characteristics.76We construct instruments for firm uncertainty shocks by exploiting firms\u2019 differential ex-posures to aggregate volatility shocks of multiple variables such as crude oil, currencies, the10-year U.S. Treasury note, or aggregate economic policy uncertainty. By construction, the in-strumented firm-level uncertainty shocks capture the changes in firm stock return volatilitythat are induced by aggregate uncertainty shocks. For instance, when an aggregate variable,such as the 10-year Treasury note, experiences a volatility shock, firms with different levels ofexposure to the 10-year Treasury note will experience different degrees of uncertainty shocks.We first estimate a firm\u2019s exposure to each aggregate variable c as the sensitivity of the firm\u2019sstock returns to the price changes of the aggregate variable. Then, we construct the instrumentvariable for each aggregate variable c as the product of the estimated sensitivity and the year-on-year change in the standard deviation of daily price changes for c.5Aggregate volatility shocks are unlikely to be driven by firm characteristics. For this reason,the instruments, by construction, do not correlate with any unobservable firm characteristics.The instruments together are strong predictors of changes in firm-level stock return volatilities.We report the 1st stage results of the 2SLS estimator in Appendix B.3.2. The 2nd stage of the 2SLSestimation studying the impact of information cost shock on the loan-bond spread is:(loan\u2212bond)i t =\u03b21+\u03b22 \u00b7 \u2206\u0302\u03c3i ,t\u22121+\u03b23 \u00b7\u03c3i ,t\u22125+\u03b24 \u00b7 ri ,t\u22121+\u03b25 \u00b7Bondi ,t +\u03b26 \u00b7Loani ,t+\u03b27Agg ck,t +\u03c6 j +\u03c8t +\u03f5i , j ,t ,(3.2)where \u2206\u03c3i ,t\u22121 is the instrumented firm-level uncertainty measure using the set of instruments.We control for the direct impact of aggregate price changes by including the aggregate first mo-ment effects, Agg ck,t . For each aggregate variable c and industry k, Aggck,t is calculated as theproduct of the estimated sensitivity of industry k to the aggregate variable c and the annualprice changes of c. Controlling for Agg ck,t allows us to isolate the effects of aggregate volatilityshocks from the changes in levels of aggregate quantities on firm-level changes in volatility. Inthis way, the instruments capture only those changes in firm-level volatility that are driven by5See Appendix B.2 for a full description of the construction of instruments.77changes in aggregate volatility shocks, rather than by firm specific characteristics or macroeco-nomic conditions. Hence, this approach allows us to rule out alternative explanations driven byomitted variables that could generate patterns in the data consistent with our money creationhypothesis. The remaining control variables in Equation (3.2) are the same as in Equation (3.1).Measuring Firm-level Information Cost Shock using the 9\/11 EventNext, we conduct an event study analysis using the 9\/11 shock to see how it affects firms\u2019 rel-ative cost of loan vs bond financing. Since the 9\/11 uncertainty shock was orthogonal to firmfundamentals, the firm-level equity volatility induced by this shock can be considered an ex-ogenous increase in investors\u2019 cost of acquiring information about the firms\u2019 assets. Formally,we define a firm\u2019s information cost shock as the difference in annualized stock return volatilitybetween the post- and the pre-9\/11 period and estimate the impact of this shock on the loan-bond spread using the following equation:(loan\u2212bond)i ,post \u2212 (loan\u2212bond)i ,pre =\u03b21+\u03b22(\u03c3i ,post \u2212\u03c3i ,pre)+Xi ,t +\u03c6 j +\u03c8t +\u03f5i ,t , (3.3)where (loan\u2212bond)i ,post \u2212 (loan\u2212bond)i ,pre is the difference in the loan-bond spreads be-tween the post- and the pre-event. \u03c3i ,post \u2212\u03c3i ,pre is the difference in annualized stock returnvolatility between the post- and the pre-9\/11 period. Xi ,t is the set of controls and includesbond ratings and differences in maturity of the loan-bond pairs. We restrict the sample to loan-bond pairs within [-365, -90] and [90, 365] days around the 9\/11 event and define these periodsas the pre- and post-shock periods respectively.Alternative Measures of Information Cost: Opacity Index and Rating DisagreementWe consider two alternative measures of firm information cost. First, we follow Anderson et al.(2009) to construct a firm-level opacity index, which ranks the relative opacity of firms in oursample based on four proxies for opacity: bid-ask spread, trading volume, analyst coverage, and78analyst forecast errors. We rank each firm based on quintiles of each of these four variables.The opacity index is the sum of the rankings based on these four variables, normalized by 20.A higher opacity index indicates larger information asymmetry which should make it more dif-ficult for outside investors to evaluate the firm\u2019s assets. Second, we follow Morgan (2002) anddefine opacity as the difference in ratings assigned to a firm by Standard & Poor\u2019s and Moody\u2019s.The idea is that if a firm\u2019s assets are harder to evaluate, there will be more disagreement amongrating agencies about the true value of the firm\u2019s assets. Following this argument, we constructtwo additional proxies for firm opacity shocks. Rating gap measures the absolute difference inratings assigned to a firm by Standard & Poor\u2019s and Moody\u2019s. Rating disagreement, an indicatorvariable which equals one if the rating gap is equal or greater than two and is zero otherwise.This indicator variable captures large bond rating disagreements at issuance. We re-estimateEquation (3.1) using these three variables that measure firm information cost.3.4 Main Results3.4.1 Information Cost and the Loan-bond SpreadFirm-level Uncertainty Shock: OLS Estimation ResultsTable 3.2 presents estimates for the effect of information cost shocks on the loan-bond spread.Panel A shows results for the OLS estimation based on Equation (3.1). The dependent variableis a firm\u2019s relative cost of bank credit, measured as the difference between the loan spread andbond spread on new loans and bonds issued by the firm with the same maturity and at the sametime. A negative loan-bond spread indicates that the bank loan is cheaper than public debt. InPanel A, the main independent variable \u2013 the information cost shock \u2013 is proxied by the firm-level uncertainty shock, measured as year-on-year change in the annualized equity volatilitylagged by one quarter from the start of the loan facility.Columns (1)-(3) report the OLS results controlling for lagged stock returns and the level of79firm opacity, measured as the annualized stock return volatility of the firm prior to the informa-tion cost shock. Columns (4)-(6) control for additional loan- and bond-level characteristics thatcan affect the loan-bond spread as described in Section 3.3.3. We include year-by-quarter fixedeffects in all specifications. We also include bank holding company fixed effects and lenderfixed effects in columns (2) and (5), and columns (3) and (6), respectively, to control for anyvariation in the loan-bond spread driven by lender\/underwriter time-invariant characteristics.The coefficient of the information cost shock is negative and statistically significant acrossall specifications we consider, suggesting that firms whose assets become harder to evaluatereceive a discount when borrowing from banks. In terms of magnitudes, results in column (4)imply that a one-standard deviation increase in the information cost shock leads to a discountof 22 bps, which is about 17.8% of the sample average loan-bond spread.6Firm-level Uncertainty Shock: IV Estimation ResultsTo address the concern that part of the decline in the loan-bond spread in our baseline estima-tion could be explained by unobservable firm-level factors that simultaneously affect firm-levelinformation cost and the relative pricing of bank and public debt, we present results from theIV estimation.Panel B of Table 3.2 shows results from the second stage of the IV estimation described inSection 3.3.3. Columns (1)-(3) report the results based on instruments constructed using real-ized volatility shocks, and columns (4)-(6) report the results based on instruments constructedusing aggregate implied volatility shocks. Year and quarter fixed effects are included in all spec-ifications. We include bank holding company fixed effects in columns (2) and (5), and lenderfixed effects in columns (3) and (6), respectively, to control for any variation in the loan-bondspread driven by lender\/underwriter time-invariant characteristics.6Table 3.2 Panel A also suggests that firms with higher level of opacity pay relatively less for bank credit com-pared to debt in the public market. Firms with high stock returns secure cheaper funding in the public debt mar-ket relative to bank credit. Loans with higher residual maturity than the matched bonds have a higher loan-bondspread. Term loans and secured loans have significantly higher spreads. The positive coefficient of secured loansindicator could reflect the endogenous choice of more risky firms asking for secured credit.80Across all specifications we consider, the IV results are consistent with our main hypothesis.Firms experiencing positive information cost shocks, which are induced by aggregate volatilityshocks, receive larger discounts from banks. P-values of the LM underidentification tests andthe Sargan-Hansen overidentification tests support the validity of our identification strategy.Results in column (1) and (4) imply that a one-standard deviation increase in the firm informa-tion cost shock leads to a discount of 46 bps and 30 bps, respectively, which is larger than theestimate of opacity discount from the baseline analysis.7 Larger magnitude of the opacity dis-count in the IV estimation reflects the ability of this estimation strategy to address endogeneityissues that were biasing the estimates downward in our baseline results.8Firm-level Uncertainty Shock: The 9\/11 Event Study ResultsPanel C of Table 3.2 reports results for the 9\/11 event study. We consider the period [-365, -90] days before the 9\/11 event as the pre-shock period and the period [90, 365] days after theshock as the post-shock period. The dependent variable is the difference between the loan-bond spread for each firm in the post- and pre-shock period. The main independent variable isthe difference between the realized stock return volatility in the post- and the pre-shock period.Realized volatility is calculated using stock returns in the year preceding the loan originationdate.Column (1) does not include any control variables. Columns (2), (3), and (4) include residualdifference in maturity of the loan-bond pairs and bond ratings as additional control variables.All columns include year fixed effects. Column (3) includes bank holding company fixed effectsand column (4) includes lender fixed effects. We find that, across all specifications we consider,7Magnitudes of the coefficients are smaller in columns (4) to (6) when instruments are constructed using im-plied volatilities of aggregate variables. Data for daily implied volatility for treasuries (TYVIX) starts in 2003. Theinstruments are lagged by three years and require one additional year of data to calculate annual average impliedvolatility. Thus, instruments in columns (4) to (6) are only available after 2007. The magnitudes drop becauseglobal systemically important banks offer significantly smaller opacity discounts after the global financial crisis.We discuss this fact in detail in Section 3.4.2.8Coefficients of the control variables are similar to those in the baseline analysis. Loan-bond spreads are lowerfor firms with low stock returns and high level of opacity. Loan-bond spreads are higher for term loans and securedloans, and bonds with higher ratings.81the coefficient of the post-shock change in firm-level volatility is negative and statistically sig-nificant. These results suggest that firms that experienced a larger increase in volatility due tothe 9\/11 shock had lower loan-bond spreads in the post-shock period relative to the pre-shockperiod. This finding again lends support to the hypothesis that an increase in the informationacquisition cost after the 9\/11 event made it harder for outside investors to learn about thefirms\u2019 fundamentals, which in turn made it relatively easier for banks to fund projects of thesefirms.Results using Alternative Measures of Information CostTable 3.3 reports results for the OLS regressions of the loan-bond spread on firm informationcost proxied using alternative measures. The dependent variable is a firm\u2019s relative cost of bankcredit, measured as the difference between the loan spread and bond spread on new loans andbonds issued by the firm with the same maturity and at the same time. In columns (1) and (2),we measure information cost using opacity index constructed following Anderson et al. (2009).In columns (3) and (4), information cost is measured as the rating gap between bond ratingsby Standard & Poor\u2019s and Moody\u2019s. In columns (5) and (6), we measure information cost usingan indicator variable that equals one if the rating gap is greater or equal to two. All columnsinclude year-by-quarter fixed effects. Columns (1), (3), and (5) include bank holding companyfixed effects. Columns (2), (4), and (6) include lender fixed effects. All columns include bondand loan-level control variables as described in Section 3.3.3. Irrespective of how we define thecost of information acquisition, we find that higher information cost are associated with lowerloan-bond spread.3.4.2 Evidence on the Money Creation ChannelSo far, our results suggest that firms experiencing information cost shocks are able to obtaincheaper credit from banks relative to raising funds in the public debt market. We argue that thisreduction in the cost of bank credit relative to that of public debt is driven by the banks\u2019 need82to create money in the form of safe and liquid deposits. In this section, we implement threetests to provide support on this channel by exploiting heterogeneity in banks\u2019 need to maintainopacity of their assets that arises due to this channel.Uninsured Deposits OutflowsAccording to the financial intermediation theory, the need for banks to maintain opacity shouldbe larger when private money creation is not backed by the government. Specifically, depositinsurance provided by the government on demandable debt produced by banks makes such in-sured deposits insensitive to information, reducing banks\u2019 need to maintain opacity of their as-sets though lending decisions. Therefore, a higher opacity discount should be offered by banksthat create relatively more liquidity in the form of uninsured deposits and when the value ofthose deposits is under threat. To test this hypothesis, we exploit variation in the ratio of unin-sured deposits to total assets across banks and test whether banks with a higher ratio of unin-sured deposits offer a larger opacity discount to firms with information cost shocks when theyexperience deposits outflows. Specifically, we estimate equation:(loan\u2212bond)i ,t =\u03b21+\u03b22\u2206\u03c3i ,t\u22121\u00d7Udep j ,t\u22125\u00d7Out f low j ,t\u22121+\u03b23 \u00b7\u2206\u03c3i ,t\u22121+\u03b24 \u00b7\u2206\u03c3i ,t\u22121\u00d7Udep j ,t\u22125+\u03b25\u2206\u03c3i ,t\u22121\u00d7Out f low j ,t\u22121+\u03b26Udep j ,t\u22125+\u03b27Out f low j ,t\u22121+\u03b28Udep j ,t\u22125\u00d7Out f low j ,t\u22121+\u03b29\u03c3i ,t\u22125+\u03b210ri ,t\u22121+\u03b211Bondi ,t +\u03b212Loani ,t+\u03c6 j +\u03c8t +\u03f5i , j ,t ,(3.4)where Out f low j ,t\u22121 is an indicator variable which equals one if the bank holding companyexperiences large uninsured deposits outflow in the past year, measured at the quarter beforeloan origination. We define banks with large uninsured deposits outflow as those in the bottom5 percentile of the sample, representing an annual decrease in uninsured deposits of more than25%. Udep j ,t\u22125 is the ratio of uninsured deposits to total assets of the bank holding companybefore the uninsured deposits outflow.83Panel A of Table 3.4 reports the results. In columns (1) and (2), we measure information costshock using firm-level changes in equity volatility, in columns (3) and (4) information cost isproxied using opacity index following Anderson et al. (2009), and in columns (5) and (6) infor-mation cost is measured using the rating gap between bond ratings by Standard & Poor\u2019s andMoody\u2019s. All columns include year-by-quarter fixed effects. Columns (1), (3), and (5) includebank holding company fixed effects while columns (2), (4), and (6) include lender fixed effects.Consistent with the baseline results, we find that the coefficient of the information cost isnegative and statistically significant across all specifications we consider. Firms receive largeropacity discounts from banks when they experience positive information cost shocks. Further-more, the coefficient on the interaction term between reliance on uninsured deposits, depositoutflows, and firm information cost shock is negative and statistically significant. This resultimplies that banks that rely more on uninsured deposits offer larger opacity discount to firmsthat experience positive information cost shock when the value of those deposits is under threatas proxied by deposit outflows. These results support the view that the economic mechanismdriving the baseline results is the money creation channel. Since uninsured deposits are moreresponsive to negative information about banks\u2019 assets, banks that rely more heavily on unin-sured deposits have a greater need to maintain opacity, and, hence, they offer larger opacitydiscounts when they see depositors worrying about the value of their deposits.Uninsured Deposit Outflows: Evidence using the Money Market Funding ShockIn this section, we refine our evidence on the money creation channel by focusing on depositsoutflows that are induced by a shock to investors\u2019 confidence in the asset quality of banks. Tothis end, we use the money market dollar funding shock of April 2011. European banks activein the U.S. raise most of their dollar funding from uninsured sources, such as the commercialpaper market while the dollar funding of U.S. banks is mostly sourced from insured retail de-posits (Ivashina et al., 2015). In April 2011, money market funds started becoming concernedabout European banks\u2019 exposure to Greek sovereign debt and they reduced their exposure to84the Eurozone banks in the U.S., which led to uninsured deposit outflows from these banks. Wehypothesize that after the money market funding shock, the European banks should offer largeropacity discounts to firms affected by larger information cost shocks since those banks have agreater need to keep information about their assets secret to prevent further withdrawals. Weestimate the following equation to test our hypothesis:(loan\u2212bond)i ,t =\u03b21+\u03b22\u2206\u03c3i ,t\u22121\u00d7EuropeanBank j \u00d7Postt +\u03b23 \u00b7\u2206\u03c3i ,t\u22121+\u03b24 \u00b7\u2206\u03c3i ,t\u22121\u00d7EuropeanBank j +\u03b25\u2206\u03c3i ,t\u22121\u00d7Postt +\u03b26EuropeanBank j +\u03b27Postt+\u03b28EuropeanBank j \u00d7Postt +\u03b29\u03c3i ,t\u22125+\u03b210ri ,t\u22121+\u03b211Bondi ,t +\u03b212Loani ,t+\u03c6 j +\u03c8t +\u03f5i , j ,t ,(3.5)where EuropeanBank j is the fraction of lead European banks in the loan syndicate j andPostt is an indicator variable that takes a value of one after April 2011. Firm information cost ismeasured as the year-on-year change in annualized stock return volatility lagged by one quarterbefore loan origination. All other control variables are the same as in Equation (3.1). We esti-mate Equation (3.5) on a sample of matched loan facilities and investment-grade bond pairsissued by U.S. non-financial public firms within 60 days between July 2004 to June 2007, andbetween May 2011 to April 2014. The main coefficient of interest is \u03b22 which measures thedifferential impact of firm information cost shock on European banks loan pricing in the post-shock period. Since European banks had a greater need to keep information about their assetssecret after the uninsured deposit outflows following the 2011 money market funding shock, weexpect European banks to offer larger opacity discounts, which implies a negative \u03b22.Panel B of Table 3.4 presents the results. All columns include year-by-quarter fixed effects.Column (1) controls for stock return and volatility while columns (2) and (3) include additionalcontrol variables as in Equation (3.1). Column (2) adds bank holding company fixed effectswhile column (3) includes lender fixed effects. We find that the estimate of \u03b22 is negative andstatistically significant in all specifications we consider. This result suggests that larger opacitydiscount was offered by European banks to firms whose assets became harder to evaluate in85the post-funding-shock period. This result further suggest that the money creation channel isdriving our results.Difference-in-Differences Analysis of G-SIBsSimilar to the deposit insurance argument above, banks that are deemed to be systemically im-portant should offer lower opacity discounts as the value of their deposit contracts is implicitlybacked by the government which makes these contracts information insensitive. To test this hy-pothesis, we conduct a difference-in-differences analysis employing the classification of globalsystemically important banks (G-SIBs) following the 2008-09 global financial crisis. To maintainglobal financial stability, the Financial Stability Board started to publish a list of G-SIBs after thefinancial crisis. The global systemically important banks receive implicit \u201ctoo-big-to-fail\u201d guar-antees, which should diminish their need to maintain opacity on the asset side of their balancesheet. Hence, these banks should offer lower opacity discounts to firms after they are classifiedas G-SIBs, compared to non-G-SIBs. We estimate the differential opacity discount offered byG-SIBs vs non-G-SIBs after the global financial crisis using the following regression:(loan\u2212bond)i ,t =\u03b21+\u03b22\u2206\u03c3i ,t\u22121\u00d7Non\u2212GSIBs j \u00d7Postt +\u03b23 \u00b7\u2206\u03c3i ,t\u22121+\u03b24 \u00b7\u2206\u03c3i ,t\u22121\u00d7Non\u2212GSIBs j +\u03b25\u2206\u03c3i ,t\u22121\u00d7Postt +\u03b26Non\u2212GSIBs j +\u03b27Postt+\u03b28Non\u2212GSIBs j \u00d7Postt +\u03b29\u03c3i ,t\u22125+\u03b210ri ,t\u22121+\u03b211Bondi ,t +\u03b212Loani ,t+\u03c6 j +\u03c8t +\u03f5i , j ,t ,(3.6)where Non \u2212GSIBs j is the fraction of non-global systemically important banks among leadbanks in the loan syndicate. Postt is an indicator variable that equals one if the loan facilitystarts after November 2009.9 We exclude the period 2007-08 from the analysis.Panel C of Table 3.4 reports estimates from Equation (3.6). All columns include year-by-quarter fixed effects. Column (1) controls for stock return and volatility while columns (2)9Although the list of G-SIBs was officially published in 2011, an \u201cunofficial\u201d list of G-SIBs was leaked in Novem-ber 2009.86and (3) include additional control variables as in Equation (3.1). Column (2) has bank hold-ing company fixed effects while column (3) includes lender fixed effects. Consistent with thebaseline results, we find that the coefficient of the information cost is negative and statisti-cally significant across all specifications we consider. The estimates of difference-in-differencescoefficients \u03b22 are negative in all specifications we consider and are statistically significant incolumns (1) and (3). This result suggests that Non-G-SIBs offer larger opacity discounts com-pared to G-SIBS after 2009. In terms of economic magnitude, after the Financial Stability Boardpublished the list of globally systemically important banks, non-G-SIBs offer 75 bps larger opac-ity discounts than G-SIBs when there is one standard deviation increase in firm informationcost shock. The results are consistent with the bank opacity channel of money creation: the im-plicit government guarantee on deposits offered to G-SIBs after the 2008-09 crisis implied thatthese banks have less incentives to maintain opacity on their asset side after the global financialcrisis. In other words, G-SIBs offer lower discounts when lending to opaque firms.3.5 Alternative Explanations and Robustness Tests3.5.1 Alternative ExplanationsIn this section, we discuss and rule out multiple other channels that could explain our mainresults.Relationship LendingA large literature has documented that relationship lending can affect the quantity and cost ofbank financing for firms (Petersen and Rajan, 1994). Firms with longer bank relationships paylower interest rates (Berger and Udell, 1995) and, in crisis times, firms are able to receive cheaperbank credit from relationship banks compared to banks with which firms have only transactionlending relationships (Bolton et al., 2016). Following these arguments, one can argue that thelower loan-bond spread in response to firm information cost shock is driven by relationship87lending and not by the money creation channel.To test this hypothesis, we construct two measures of relationship lending at the firm leveland examine whether the opacity discount offered by banks differs by the strength of lendingrelationship. Our first measure of lending relationship is an indicator variable that takes a valueof one if a firm has received a loan from the lead bank in the past two years and is zero otherwise.The second measure is a continuous measure of the strength of lending relationship and isconstructed as the natural logarithm of the number of years since the start of the relationshipbetween the firm and the lead bank.We re-estimate Equation (3.1) interacting the information cost with our measures of lendingrelationship. Results are presented in Table 3.5. In columns (1) and (2), we measure informationcost shocks using firm-level changes in equity volatility, in columns (3) and (4) informationcost is proxied using opacity index following Anderson et al. (2009), and in columns (5) and (6)information cost is measured as the rating gap between bond ratings by Standard & Poor\u2019s andMoody\u2019s. All columns include year-by-quarter fixed effects. Columns (1), (3), and (5) includebank holding company fixed effects while columns (2), (4), and (6) include lender fixed effects.Panel A of Table 3.5 shows results for the discrete measure of lending relationship and PanelB shows the results for the continuous measure. We find that across all specifications we con-sider, the coefficient of the information cost is negative and statistically significant, consistentwith the baseline results. The coefficients of the interaction term between relationship lendingand the information cost are not statistically different from zero in most specifications. For theopacity index measure in Panel A, the coefficients of the interaction term are positive and statis-tically significant, which is inconsistent with relationship lending leading to opacity discount.Overall, the evidence suggests that there is no difference in the opacity discount received byfirms with and without prior bank relationships. This finding, therefore, alleviates the concernthat our results are driven by relationship lending.88Changes in Demand for Bank CreditCan active switching between bank credit and bonds by firms over business cycles explain ourresults? Becker and Ivashina (2014) finds that firms substitute bonds for loans at times withdepressed aggregate lending and tight lending criteria. We argue that this substitution betweenbank loans and bonds over the business cycle does not affect our results since we control foryear-by-quarter fixed effects. Our main finding is that in the cross-section of firms, a higherinformation cost shock is associated with relatively lower cost of bank credit.Can the negative relationship between the loan-bond spread and firm information cost inthe cross-section of firms be driven by changes in firm demand for credit when firms are hit byinformation cost shocks? For example, it is possible that firms that experience larger volatilityin their stock returns lower their demand for bank credit and increase their demand for publicdebt. This relative fall in demand for bank credit may translate into a lower cost of bank creditand a higher cost of public debt and may thereby generate a negative relationship between firmopacity and the loan-bond spread. We argue and show evidence suggesting that this channelcannot explain our results.The argument that the impact of lower demand for bank credit translates into a lower costof bank credit is more likely to be valid for the aggregate cost of bank credit in the economy andnot for individual firms, which take the price of loans as given. A change in the demand for bankcredit by one firm should not significantly affect the cost of credit for that firm. However, onecan still argue that a lower volume of bank credit demanded by a firm might be associated witha lower loan spread. If this is true, and if firms actively switch to public debt when they are hit byan information cost shock, we should see a fall in the share of bank loans to the total credit byfirms that experience positive information cost shocks. We test this hypothesis formally usingthe regression in Equation (3.1) with LoanSharei ,t as the dependent variable. LoanSharei ,t isdefined as the share of bank borrowing in the total amount borrowed by a firm from the bankingsector and the public debt market combined.89Table 3.6 presents the results from the regression of bank loan share on information cost.In columns (1) and (2), we measure information cost shock using firm-level changes in eq-uity volatility, in columns (3) and (4) information cost is proxied using opacity index followingAnderson et al. (2009), and in columns (5) and (6) information cost is measured as the ratinggap between bond ratings by Standard & Poor\u2019s and Moody\u2019s. All columns include year-by-quarter fixed effects. Columns (1), (3), and (5) include bank holding company fixed effects whilecolumns (2), (4), and (6) include lender fixed effects. Table 3.6 shows that the impact of informa-tion cost on the loan share is close to zero and statistically insignificant for firm-level changes inequity volatility and rating gap and is positive and statistically significant for the opacity indexmeasure. This evidence suggests that there is no loan-bond switching when firms experienceinformation cost shocks or that the share of bank loans to the total credit increases with in-formation cost in our sample, which is inconsistent with changes in firm demand for creditexplaining our results. Overall, we conclude that the fall in the loan-bond spread is unlikely tobe driven by compositional changes in demand or supply of bank credit.Bank Allocation in Loan SyndicatesAnother alternative explanation of our results is that banks may participate less in loan syn-dicates when borrowers are opaque, which may lower their exposure and thereby lower theircost of lending to such borrowers. To investigate this alternative hypothesis, we examine theeffects of firm information cost on bank participation in loan syndicates using the regressionin Equation (3.1) with Bank Fractioni ,t or Bank Allocationi ,t as the dependent variables.Bank Fractioni ,t is the number of bank lenders in a loan syndicate divided by total numberof lenders in the syndicate. Bank Allocationi ,t is the percentage of total loan amount banklenders have committed to a loan facility.The results are reported in Table 3.7. The dependent variable in columns (1), (3), and (5) isBank Allocationi ,t , and the dependent variable in column (2), (4), and (6) isBank Fractioni ,t .In columns (1) and (2), we measure information cost shock using firm-level changes in equity90volatility, in columns (3) and (4) information cost is proxied using opacity index following An-derson et al. (2009), and in columns (5) and (6) information cost is measured as the rating gapbetween bond ratings by Standard & Poor\u2019s and Moody\u2019s. All columns include year-by-quarterfixed effects. We find no or positive relationship between information cost and the measures ofbank participation in loan syndicates, which is inconsistent with the possibility that our resultscan be due to changes in bank allocation in loan syndicates.3.5.2 Robustness TestsIn this section, we discuss the results of a series of robustness tests. The results are presentedin Appendix B.3. First, we re-estimate Equation (3.1) using alternative samples. In Panel A ofTable B.3.3, we exclude the global financial crisis (years 2007-09) from the sample. In Panels Band C, we restrict our sample to the loan-bond pairs issued by the same firm within 30 days and10 days, respectively. In Panel D, we relax the restriction on maturity when matching loan-bondpairs. In Panel E, we match the loan-bond pairs based on effective maturity instead of maturity.In Panel F, we re-stimate the results from Panels A through E using the opacity index, rating gap,and the rating disagreement measures. Overall, the evidence presented in Table B.3.3 showsthat our results are robust to all these alternative ways we construct the sample and suggeststhat our findings are thus unlikely to be driven by any specific sample selection procedure.Second, we re-estimate Equation (3.1) with alternative sets of control variables. Panel A ofTable B.3.4 reports the results controlling for a full set of firm, loan and bond characteristics,such as firm profitability, distance to default, and various bond terms. These results suggeststhat our baseline findings are robust to including a wide variety of characteristics that couldaffect the loan-bond spread. Panel B excludes stock return as a control variable from the re-gressions. Panel C excludes stock return volatility from the regressions with alternative infor-mation cost measures. The results in Panel B and C suggests that our results are not driven bythe correlation between information costs, stock return, and stock return volatility.Lastly, we re-estimate Equation (3.4) using alternative regression specifications. Uninsured91deposits may be correlated with bank size. To ensure our results are not driven by the com-parative advantage of big banks lending to large firms, we control for size of the bank holdingcompany or the borrower in Table B.3.5 Panel A and B, respectively. In Table B.3.5 Panel C, weexclude stock return volatility as a control variable to address the concern that the opacity indexor rating disagreement measures may be correlated with stock return volatility. Results in TableB.3.5 suggest that our results are robust to all these the alternative regression specifications.3.6 ConclusionThis chapter shows that the need for banks to be opaque to support their primary function ofprivate money creation translates into a lower cost of bank credit compared to public debt forfirms that experience information cost shocks. By using a unique sample design that facilitateswithin firm comparison of the cost of bank credit and public bonds at issuance, we are able torule out firm-specific factors, such as credit risk, that could affect the relative cost of bank creditand public bonds.The chapter contributes to the literature by not only providing empirical evidence on oneof the core theories of financial intermediation but also documenting the potential benefits ofendogenous bank opacity, which could guide the ongoing debate on disclosure of stress testresults. Our results show that the cost advantage for banks to finance opaque borrowers ispassed on to opaque firms through a lower cost of bank credit relative to that of public debt.The results from our analyses can also guide government intervention policies during periodsof heightened economic uncertainty, such as the ongoing pandemic. Higher uncertainty in-creases firms\u2019 overall borrowing cost, but leads to a smaller increase in the cost of bank credit.Support for corporate sector in times of economic uncertainty may be more effective if an ad-ditional dollar of government support is intermediated through the banking sector rather thanthrough the public capital market.92Figure 3.1 The Loan-Bond Spread: 1995-2020The figure plots the average cost of bank credit, public debt, and the relative cost of bank creditfor firms in our sample over the period 1995-2020. The relative cost of bank credit \u2013 the loan-bondspread \u2013 is measured as the difference between the loan rate and the bond yield on new loansand bonds with the same maturity issued by the same firm at the same time. The shaded verticalbars represent NBER recessions.93Figure 3.2 The Loan-Bond Spread across Firm Information Cost Shock QuartilesThe figure illustrates the average cost of bank credit, public debt, and the relative cost of bankcredit for firms in different quartiles of the information cost shock, where the information costshock is measured using changes in firm-level stock return volatility. The loan rate and the bondyield are residuals from a regression of loan rate and bond yield on maturity, firm equity returnvolatility, and year fixed effects. The relative cost of bank credit \u2013 the loan-bond spread \u2013 ismeasured as the difference between the loan rate residual and the bond yield residual on newloans and bonds with the same maturity issued by the same firm at the same time.94Table 3.1 Summary StatisticsThe table reports summary statistics for the main variables used in the empirical anal-ysis. The sample includes new loan facilities and investment-grade bonds issued by U.S.non-financial public firms from 1995 to 2019. The sample is restricted to senior loans and bondsdenominated in USD. Each loan is paired with the closest bond issued by the same firm within60 days. We further restrict the loan-bond pairs to those with the same maturity category, i.e.short-term, mid-term or long-term in maturity. See Appendix B.1 for variable definitions.Mean SD p10 p50 p90 ObservationsLoan-Bond Pair CharacteristicsLoan-bond Spread (bps) -123 152 -314 -112 35 1,597Total Borrowing ($MM) 1,980 1,785 500 1,425 4,000 1,597\u00a2Maturity (years) -9.18 10.42 -25.07 -5.03 -0.01 1,567Loan Share 0.61 0.21 0.29 0.65 0.85 1,597Information CostUncertainty Shock -0.03 0.29 -0.42 -0.02 0.32 1,597Opacity Index 0.61 0.16 0.4 0.6 0.8 1565Rating Gap 0.61 0.71 0 0 2 1,217Rating Disagreement 0.10 0.31 0 0 1 1,217Firm CharacteristicsVolatility 0.28 0.12 0.16 0.25 0.43 1,597Stock Return 0.16 0.29 -0.16 0.14 0.48 1,597Total Assets ($B) 1,413 889 242 1,458 2,466 1,072Profitability 0.04 0.02 0.02 0.03 0.06 1,558Implied Prob. Default 0.00 0.03 0.00 0.00 0.00 1,525No. of Loan-Bond Pairs per Firm 3.86 3.52 1 3 9 1,597Loan CharacteristicsFacility Amount ($MM) 1,340 1,391 150 1,000 3,000 1,597All-in-drawn Spread (bps) 117 90 27 110 200 1,597Syndicated Loan 0.99 0.08 1 1 1 1,597Term Loan 0.17 0.37 0 0 1 1,597Secured Loan 0.15 0.36 0 0 1 972Bond CharacteristicsFace Value ($MM) 639 628 200 500 1,250 1,597Bond Spread (bps) 497 233 259 475 740 1,597Bond Rating BBB+ 1.87 BBB- BBB A 1,597Secured Bond 0.03 0.17 0 0 0 1,597Redeemable Bond 0.89 0.32 0 1 1 1,597Embedded Investor Option 0.27 0.44 0 0 1 1,597Bondholder Protective Covenant 0.64 0.48 0 1 1 1,597Negative Pledge Covenant 0.66 0.48 0 1 1 1,59795Table 3.2 Firm Information Cost Shock and the Loan-Bond SpreadPanel A. OLS EstimationThe table reports results from the OLS regression of the loan-bond spread on firm uncer-tainty shock, estimated using Equation (3.1). The sample includes matched loan facilities andinvestment-grade bond pairs issued by U.S. non-financial public firms within 60 days from 1995to 2019. The dependent variable is the difference between loan rate and bond yield for eachmatched loan-bond pair in the sample. Uncertainty shock is the year-on-year change in theannualized stock return volatility lagged by one quarter before the loan origination. All columnsinclude year by quarter fixed effects. Column (2) and (5) include bank holding company fixedeffects. Column (3) and (6) include lender fixed effects. Columns (4)-(6) include contract-levelcontrol variables. See Appendix B.1 for variable definitions. The standard errors (in parentheses)are clustered at firm level. *, **, and *** denote significance at the 10%, 5%, and 1% levels,respectively.(1) (2) (3) (4) (5) (6)Uncertainty Shock -0.62** -0.63*** -0.52** -0.83*** -0.70*** -0.76***(0.27) (0.23) (0.20) (0.30) (0.25) (0.23)Volatility -2.29*** -1.57*** -1.60*** -2.65*** -2.16*** -2.31***(0.66) (0.53) (0.52) (0.59) (0.58) (0.56)Stock Return 0.53*** 0.42** 0.47*** 0.71*** 0.58*** 0.45***(0.20) (0.17) (0.16) (0.24) (0.18) (0.16)Log Total Borrowing -0.03 -0.21 -0.24(0.17) (0.20) (0.20)\u00a2Maturity (years) 0.06*** 0.06*** 0.06***(0.00) (0.00) (0.00)Term Loan 0.34*** 0.22** 0.23***(0.12) (0.10) (0.08)Secured Loan 1.07*** 0.67*** 0.55***(0.27) (0.19) (0.17)Log Facility Amount -0.16 0.06 0.13(0.11) (0.14) (0.15)Bond Rating 0.19*** 0.19*** 0.17***(0.04) (0.04) (0.04)Secured Bond -0.60** -0.25 -0.11(0.26) (0.26) (0.22)Redeemable Bond 0.12 0.15 0.29(0.23) (0.23) (0.20)Embedded Investor Option -0.14 -0.00 -0.07(0.11) (0.10) (0.11)Bondholder Protective Covenant -0.03 -0.08 -0.06(0.12) (0.12) (0.11)Negative Pledge Covenant -0.17 -0.18* -0.23**(0.11) (0.10) (0.10)Year\u00a3Quarter FE Yes Yes Yes Yes Yes YesBank Holding Company FE No Yes No No Yes NoLender FE No No Yes No No YesAdjusted R2 0.36 0.43 0.52 0.63 0.68 0.73Observations 1,597 1,338 1,588 963 838 96096Panel B. Second Stage of the IV EstimationThe table reports the results from the IV regression of the loan-bond spread on firm un-certainty shocks. The sample includes matched loan facilities and investment-grade bond pairsissued by U.S. non-financial public firms within 60 days from 1995 to 2019. The dependentvariable is the difference between loan rate and bond yield for each matched loan-bond pairin the sample. Uncertainty shock is the year-on-year change in the annualized stock returnvolatility lagged by one quarter before loan origination, instrumented using volatility shocksfor macro variables. In columns (1)-(3), the instruments are calculated using realized volatilityshocks. In columns (4)-(6), the instruments are calculated using implied volatility shocks. Allcolumns include year and quarter fixed effects. Column (2) and (5) include bank holding companyfixed effects. Column (3) and (6) include lender fixed effects. See Appendix B.1 for variabledefinitions. The standard errors (in parentheses) are clustered at 3-digit SIC level. *, **, and ***denote significance at the 10%, 5%, and 1% levels, respectively.Realized Volatility Instruments Implied Volatility Instruments(1) (2) (3) (4) (5) (6)Uncertainty Shock -2.72** -3.79*** -2.83** -1.30** -1.58** -1.57**(1.17) (1.10) (1.28) (0.66) (0.69) (0.76)Volatility -3.91*** -4.51*** -3.51*** -3.03*** -2.14*** -2.29***(1.30) (1.18) (1.34) (0.74) (0.70) (0.75)Stock Return 0.79*** 0.87*** 0.66*** 0.50* 0.31 0.27(0.25) (0.27) (0.23) (0.28) (0.26) (0.20)Control Variables Yes Yes Yes Yes Yes YesYear, Quarter FE Yes Yes Yes Yes Yes YesBank Holding Company FE No Yes No No Yes NoLender FE No No Yes No No YesAdjusted R2 0.55 0.51 0.65 0.70 0.76 0.81Observations 797 750 794 515 482 513P-value LM underidentification 0.04 0.01 0.04 0.01 0.01 0.02F-statistic CD 3.21 4.56 2.50 4.18 4.09 4.38P-value-SarganHJ 0.19997Panel C. 9\/11 Event StudyThe table reports the results of the 9\/11 event study. The sample includes loan facilityand investment-grade bond pairs issued by U.S. non-financial public firms in the pre- [-365d,-90d] and the post-9\/11 period [90d, 365d]. The dependent variable is the difference betweenthe loan-bond spread in the post 9\/11 period and that in the pre 9\/11 period. The independentvariable is the difference between realized firm stock return volatility in the post-9\/11 periodand that in the pre-9\/11 period. Realized volatility is calculated using stock returns in the yearpreceding the loan staring date. All columns include year fixed effects. Column (3) includes bankholding company fixed effects, and column (4) includes lender fixed effects. See Appendix B.1 forvariable definitions. The standard errors (in parentheses) are clustered at firm level. *, **, and*** denote significance at the 10%, 5%, and 1% levels, respectively.(1) (2) (3) (4)VolatilityPost-9\/11 - Pre-9\/11 -3.67* -4.11** -5.61*** -5.15***(1.75) (1.78) (1.35) (0.21)\u00a2Maturity (Pre-9\/11 Loan-bond Pair) 0.00 -0.02 -0.03***(0.04) (0.02) (0.00)\u00a2Maturity (Post-9\/11 Loan-bond Pair) -0.02 -0.07 -0.09(0.05) (0.04) (0.08)Pre-9\/11 Bond Rating -0.80 1.19 1.74(1.31) (0.83) (1.24)Post-9\/11 Bond Rating 0.86 -1.62 -2.27(1.54) (1.02) (1.59)Year FE Yes Yes Yes YesBank Holding Company FE No No Yes NoLender FE No No No YesAdjusted R2 0.61 0.59 0.83 0.95Observations 47 47 47 4798Table 3.3 Firm Information Cost and the Loan-Bond Spread: Alternative MeasuresThe table reports results from the OLS regressions of the loan-bond spread on firm infor-mation cost, where firm information cost is estimated using alternative proxies. The sampleincludes matched loan facilities and investment-grade bond pairs issued by U.S. non-financialpublic firms within 60 days from 1995 to 2019. The dependent variable is the difference betweenloan rate and bond yield for each matched loan-bond pair in the sample. In columns (1) and(2), firm information cost is measured using the opacity index following Anderson et al. (2009)In columns (3) and (4), information cost is measured as the rating gap, defined as the absoluterating gap between a firm\u2019s bond ratings by Standard & Poor\u2019s and Moody\u2019s. In columns (5) and(6), information cost is measured using a rating disagreement dummy variable that equals oneif the rating gap is greater or equal to two. All columns include year by quarter fixed effects.Column (1), (3) and (5) include bank holding company fixed effects. Column (2), (4) and (6)include lender fixed effects. Contract level control variables are included in all columns. SeeAppendix B.1 for variable definitions. The standard errors (in parentheses) are clustered at firmlevel. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.Opacity Index Rating Gap Rating Disagreement(1) (2) (3) (4) (5) (6)Information Cost -1.17*** -1.14*** -0.17*** -0.18*** -0.52*** -0.49***(0.36) (0.36) (0.06) (0.06) (0.17) (0.16)Volatility -1.05* -1.25** -1.85*** -1.95*** -2.10*** -2.18***(0.60) (0.58) (0.62) (0.63) (0.64) (0.64)Stock Return 0.56*** 0.52*** 0.47** 0.41** 0.49** 0.43**(0.18) (0.17) (0.20) (0.19) (0.20) (0.19)Control Variables Yes Yes Yes Yes Yes YesYear\u00a3Quarter FE Yes Yes Yes Yes Yes YesBank Holding Company FE Yes No Yes No Yes NoLender FE No Yes No Yes No YesAdjusted R2 0.69 0.73 0.68 0.73 0.69 0.74Observations 823 945 675 736 675 73699Table 3.4 Money Creation MechanismThe table presents evidence of the money creation mechanism. The dependent variable is thedifference between loan rate and bond yield for each matched loan-bond pair in the sample. Allcolumns include year by quarter fixed effects. Control variables are included in all columns. SeeAppendix B.1 for variable definitions. The standard errors (in parentheses) are clustered at firmlevel. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.Panel A. Uninsured Deposits Outflow and the Loan-Bond SpreadThe table presents evidence of the economic mechanism using heterogeneity in banks\u2019 relianceon uninsured deposits. The sample includes matched loan facilities and investment-grade bondpairs issued by U.S. non-financial public firms within 60 days from 1995 to 2019. Uncertaintyshock is the year-on-year change in the annualized stock return volatility lagged by one quarterbefore the loan origination. Opacity index is constructed following Anderson et al.(2009). Ratinggap is the absolute rating gap between bond ratings by Standard & Poor\u2019s and Moody\u2019s. Outflowis a dummy variable which equals to one if the bank experiences large uninsured deposits outflowin the past year (banks in the bottom 5 percentile of the sample in terms of changes in uninsureddeposits, representing a decrease in uninsured deposits of more than 25%), lagged by one quarterbefore loan origination. Udep is the ratio of uninsured deposits to total assets of the bank holdingcompany before the uninsured deposits outflow. All columns include year by quarter fixed effects.Column (1), (3) and (5) include bank holding company fixed effects. Column (2), (4) and (6) includelender fixed effects.Uncertainty Shock Opacity Index Rating Gap(1) (2) (3) (4) (5) (6)Info. Cost \u00a3 Udep \u00a3 Outflow -2.82** -2.54*** -3.29*** -3.37*** -1.13*** -1.06***(1.21) (0.95) (1.20) (0.97) (0.26) (0.21)Info. Cost -0.89*** -1.03*** -1.38*** -1.51*** -0.20*** -0.17***(0.25) (0.25) (0.38) (0.42) (0.06) (0.06)Info. Cost \u00a3 Udep -0.35* -0.63*** -0.71** -0.60* 0.11* 0.23***(0.20) (0.20) (0.35) (0.31) (0.06) (0.08)Info. Cost \u00a3 Outflow -0.09 0.69 0.48 1.44 -0.33 -0.18(0.96) (0.85) (1.18) (1.17) (0.29) (0.32)Udep 0.30* 0.45*** 0.72** 0.82*** 0.27** 0.25*(0.16) (0.16) (0.28) (0.26) (0.13) (0.15)Outflow 1.29** 2.14*** -0.22 -0.49 1.71* 2.37**(0.55) (0.45) (0.85) (0.84) (0.98) (0.95)Udep \u00a3 Outflow -0.51 -0.41 1.92*** 2.09*** 1.36*** 1.33***(0.34) (0.26) (0.72) (0.56) (0.23) (0.22)Volatility -2.08*** -2.38*** -0.61 -0.91 -1.73*** -2.07***(0.56) (0.55) (0.58) (0.57) (0.58) (0.59)Stock Return 0.66*** 0.54*** 0.65*** 0.56*** 0.59*** 0.55***(0.19) (0.19) (0.18) (0.18) (0.19) (0.19)Control Variables Yes Yes Yes Yes Yes YesYear\u00a3Quarter FE Yes Yes Yes Yes Yes YesBank Holding Company FE Yes No Yes No Yes NoLender FE No Yes No Yes No YesAdjusted R2 0.65 0.68 0.65 0.67 0.65 0.67Observations 760 761 749 750 607 607100Panel B. Uninsured Deposits Outflow and the Loan-Bond Spread: European Banksafter the Money Market Funding ShockThe table reports the opacity discount offered by European banks after the dollar fundingshock in April 2011. The sample includes matched loan facilities and investment-grade bondpairs issued by U.S. non-financial public firms within 60 days between July 2004 to June 2007,and between May 2011 to April 2014. Uncertainty shock is the year-on-year change in theannualized stock return volatility lagged by one quarter before loan origination. Post is a dummyvariable that equals to one if the loan facility starts between May 2011 to April 2014, andzero otherwise. European Bank is the fraction of European banks among lead banks in a loansyndicate. Column (2) includes bank holding company fixed effects, and column (3) includeslender fixed effects.(1) (2) (3)Uncertainty Shock \u00a3 European Bank \u00a3 Post -5.77** -9.78*** -7.69*(2.22) (3.17) (3.90)Uncertainty Shock -0.79 -0.51 -0.84(0.66) (0.74) (0.93)Uncertainty Shock \u00a3 European Bank 4.09** 5.86*** 4.30(1.80) (2.11) (2.66)Uncertainty Shock \u00a3 Post -0.59 0.53 0.67(0.99) (0.87) (1.11)European Bank 0.90 0.75 1.10(0.58) (0.70) (0.78)Post 0.56 0.16 0.26(0.64) (0.48) (0.74)European Bank \u00a3 Post -1.13 -1.26 -2.10**(0.78) (0.85) (1.00)Volatility -1.51 0.15 -0.92(1.39) (1.17) (1.32)Stock Return 0.68 0.42 0.66*(0.43) (0.32) (0.36)Control Variables Yes Yes YesYear\u00a3Quarter FE Yes Yes YesBank Holding Company FE No Yes NoLender FE No No YesAdjusted R2 0.59 0.62 0.68Observations 296 285 295101Panel C. Loan-Bond Spread by Non-Global Systemically Important Banks (non-GSIBs)The table reports the opacity discount offered by non-global systemically important banks(non-GSIBs) after November 2009. The sample includes matched loan facilities and investment-grade bond pairs issued by U.S. non-financial public firms within 60 days from 1995 to 2019,excluding 2007-08. Uncertainty shock is the year-on-year change in the annualized stock returnvolatility lagged by one quarter before loan origination. Post is a dummy variable that equals toone if a loan facility starts after November 2009, and zero otherwise. Non-GSIBs is the fractionof non-global systemically important banks among lead banks in the loan syndicate. Column (2)includes bank holding company fixed effects, and column (3) includes lender fixed effects.(1) (2) (3)Uncertainty Shock \u00a3 Non-GSIBs \u00a3 Post -4.38*** -1.79 -4.48**(1.10) (1.90) (2.14)Uncertainty Shock -1.37*** -1.46*** -1.86***(0.43) (0.51) (0.49)Uncertainty Shock \u00a3 Non-GSIBs 0.97 -0.87 0.99(0.76) (1.39) (1.33)Uncertainty Shock \u00a3 Post 1.43** 1.66** 1.91***(0.56) (0.70) (0.65)Non-GSIBs 1.06*** 0.67 0.52(0.28) (0.45) (0.45)Post 1.91*** 0.09 1.84***(0.48) (0.75) (0.55)Non-GSIBs \u00a3 Post -2.12*** -0.96* -0.81*(0.33) (0.50) (0.47)Volatility -2.55*** -2.52*** -3.02***(0.74) (0.84) (0.83)Stock Return 0.32 0.52* 0.41(0.22) (0.28) (0.26)Control Variables Yes Yes YesYear\u00a3Quarter FE Yes Yes YesBank Holding Company FE No Yes NoLender FE No No YesAdjusted R2 0.68 0.64 0.72Observations 551 480 551102Table 3.5 Relationship Lending and the Loan-Bond SpreadThe table reports the results examining the alternative hypothesis of relationship lending and theopacity discount. The sample includes matched loan facilities and investment-grade bond pairsissued by U.S. non-financial public firms within 60 days from 1995 to 2019. Uncertainty shock isthe year-on-year change in the annualized stock return volatility lagged by one quarter before theloan origination. Opacity index is constructed following Anderson et al.(2009). Rating gap is theabsolute rating gap between bond ratings by Standard & Poor\u2019s and Moody\u2019s. All columns includeyear by quarter fixed effects. Column (1), (3) and (5) include bank holding company fixed effects.Column (2), (4) and (6) include lender fixed effects. Control variables are included in all columns.See Appendix B.1 for variable definitions. The standard errors (in parentheses) are clustered atfirm level. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.Panel A. Relationship Lending Measured by Previous Lending RelationshipRelationship in Panel A is a dummy variable which equals one if a firm had received a loan fromthe lead bank in the past two years, and zero otherwise.Uncertainty Shock Opacity Index Rating Gap(1) (2) (3) (4) (5) (6)Info. Cost -0.82*** -0.79*** -1.71*** -1.45*** -0.16* -0.17*(0.29) (0.27) (0.46) (0.45) (0.09) (0.09)Info. Cost \u00a3 Relationship 0.17 -0.01 1.36** 1.02* -0.03 0.01(0.31) (0.29) (0.60) (0.61) (0.13) (0.12)Relationship -0.21** -0.23** -1.03*** -0.82** -0.09 -0.16(0.09) (0.09) (0.40) (0.40) (0.13) (0.12)Volatility -2.20*** -2.35*** -1.21** -1.38** -1.84*** -1.92***(0.57) (0.55) (0.59) (0.57) (0.62) (0.63)Stock Return 0.60*** 0.49*** 0.61*** 0.55*** 0.49** 0.43**(0.18) (0.16) (0.17) (0.17) (0.20) (0.19)Control Variables Yes Yes Yes Yes Yes YesYear\u00a3Quarter FE Yes Yes Yes Yes Yes YesBank Holding Company FE Yes No Yes No Yes NoLender FE No Yes No Yes No YesAdjusted R2 0.68 0.74 0.69 0.74 0.68 0.74Observations 838 960 823 945 675 736103Panel B. Relationship Lending Measured by Log Years since the First LoanRelationship in Panel B is the natural log of one plus the number of years since the firstloan with the lead bank.Uncertainty Shock Opacity Index Rating Gap(1) (2) (3) (4) (5) (6)Info. Cost -0.96*** -0.85*** -1.06** -1.04* -0.21* -0.19*(0.32) (0.33) (0.51) (0.54) (0.11) (0.11)Info. Cost \u00a3 Relationship 0.23 0.07 -0.10 -0.09 0.03 0.01(0.16) (0.15) (0.30) (0.30) (0.06) (0.06)Relationship 0.00 -0.05 0.04 0.01 -0.03 -0.07(0.06) (0.05) (0.20) (0.20) (0.07) (0.07)Volatility -2.22*** -2.38*** -1.07* -1.27** -1.83*** -1.98***(0.59) (0.57) (0.60) (0.58) (0.64) (0.64)Stock Return 0.59*** 0.45*** 0.55*** 0.50*** 0.48** 0.40**(0.18) (0.17) (0.18) (0.17) (0.20) (0.19)Control Variables Yes Yes Yes Yes Yes YesYear\u00a3Quarter FE Yes Yes Yes Yes Yes YesBank Holding Company FE Yes No Yes No Yes NoLender FE No Yes No Yes No YesAdjusted R2 0.68 0.73 0.68 0.73 0.68 0.73Observations 838 960 823 945 675 736104Table 3.6 Information Cost and Loan ShareThe table reports the results of the OLS regression of the proportion of bank loans to thetotal amount of borrowing on firm information cost. The dependent variable is loan facilityamount as a proportion of total borrowing, measured as the sum of the loan volume and theface value of the bond. Uncertainty shock is the year-on-year change in the annualized stockreturn volatility lagged by one quarter before the loan origination. Opacity index is constructedfollowing Anderson et al.(2009). Rating gap is the absolute rating gap between bond ratings byStandard & Poor\u2019s and Moody\u2019s. All columns include firm fixed effects, and year by quarter fixedeffects. Column (1), (3) and (5) include bank holding company fixed effects. Column (2), (4) and(6) include lender fixed effects. Control variables are included in all columns. See Appendix B.1for variable definitions. The standard errors (in parentheses) are clustered at firm level. *, **,and *** denote significance at the 10%, 5%, and 1% levels, respectively.Uncertainty Shock Opacity Index Rating Gap(1) (2) (3) (4) (5) (6)Information Cost 0.02 0.01 0.20** 0.14* 0.01 -0.00(0.04) (0.03) (0.08) (0.08) (0.02) (0.02)Volatility 0.34** 0.16* 0.27* 0.02 0.11 0.12(0.14) (0.09) (0.15) (0.11) (0.26) (0.24)Stock Return -0.06* -0.05 -0.07** -0.05* -0.05 -0.01(0.03) (0.03) (0.03) (0.03) (0.05) (0.06)Control Variables Yes Yes Yes Yes Yes YesYear\u00a3Quarter FE Yes Yes Yes Yes Yes YesFirm FE Yes Yes Yes Yes Yes YesBank Holding Company FE Yes No Yes No Yes NoLender FE No Yes No Yes No YesAdjusted R2 0.92 0.92 0.92 0.92 0.93 0.93Observations 956 1,118 936 1,098 757 836105Table 3.7 Information Cost and Bank Participation in the Loan SyndicateThe table reports the results for the OLS regression of bank participation in the loan syn-dicate on the firm information cost. The dependent variable in column (1), (3) and (5) is thefraction of loan amount syndication allocated to a bank in a syndicate. The dependent variablein column (2), (4) and (6) is the fraction of bank counts in a syndicate. Uncertainty shock is theyear-on-year change in the annualized stock return volatility lagged by one quarter before theloan origination. Opacity index is constructed following Anderson et al.(2009). Rating gap isthe absolute rating gap between bond ratings by Standard & Poor\u2019s and Moody\u2019s. All columnsinclude firm fixed effects, and year by quarter fixed effects. Control variables are included inall columns. See Appendix B.1 for variable definitions. The standard errors (in parentheses)are clustered at firm level. *, **, and *** denote significance at the 10%, 5%, and 1% levels,respectively.Uncertainty Shock Opacity Index Rating GapBank Bank Bank Bank Bank BankAllocation Count Allocation Count Allocation Count(1) (2) (3) (4) (5) (6)Information Cost 0.91* 0.26 0.30*** 0.07 -0.05 0.01(0.50) (0.33) (0.02) (0.10) (0.12) (0.03)Volatility 3.90 -0.58 0.61 0.24 0.16 0.27(2.90) (0.89) (0.57) (0.20) (0.18) (0.39)Stock Return -0.21** -0.25** 0.93*** 0.10* 0.10 0.07(0.10) (0.10) (0.10) (0.06) (0.07) (0.11)Control Variables Yes Yes Yes Yes Yes YesYear\u00a3Quarter FE Yes Yes Yes Yes Yes YesFirm FE Yes Yes Yes Yes Yes YesAdjusted R2 0.88 0.88 0.95 0.83 0.82 0.84Observations 570 567 441 1,166 1,147 884106Chapter 4Earnout: Managing Valuation Risks inMergers and Acquisitions underUncertainty4.1 IntroductionUncertainty plays a vital role in economic outcomes, especially during economic downturns.Many studies have documented the negative impact of uncertainty on investment.1 In mergersand acquisitions, the impact of uncertainty has also received growing attention. Studies showthat higher uncertainty leads to less M&A activities. Uncertainty can hinder M&A activities formany reasons, one of which is the elevated target valuation risks following uncertainty shocks.Steve Baronoff, Chairman of Global Mergers & Acquisitions at Bank of America Merrill Lynch,describes the impact of uncertainty on M&A transactions as follows: \u201cSeveral transactions hitthe \u2018pause\u2019 button. With the current market volatility, it can be difficult to price and executedeals.\u201d2Uncertainty can make it difficult to predict future cash flows of the target company and theexpected synergies. It can also aggravate information asymmetry (Nagar et al., 2019), whichleads to significant disagreements on the target valuation between the two parties. Even afterthe buyer and seller form a consent on the initial valuation, it can change substantially in theperiod between deal announcement and completion. Bhagwat et al. (2016) show that target1Bloom (2009); Mian and Sufi (2010); Pastor and Veronesi (2012); Alfaro et al. (2021)2Deals Fall by the Wayside, the Wall Street Journal, Oct 3, 2011.107stand-alone valuation changes more than 20% within the deal completion window more than50% of the time. Because of the valuation risks, many acquirers postpone M&A activities duringperiods of high uncertainty. Bhagwat et al. (2016) find that a one standard deviation increase inthe VIX reduces public deal activity by 6% in the subsequent month.In this chapter, I focus on the earnout agreement, a contingent payment contract primarilyused to manage valuation risks in mergers and acquisitions. Payment in an earnout contractconsists of two parts, an upfront payment and the earnout payment. The earnout paymentis contingent on the post-transaction performance of the acquired business. Conditions to re-ceive the earnout payment are usually specified in the M&A contract. The criteria can be achiev-ing some earnings or sales target for manufacturing companies, or obtaining FDA approvals forpharmaceutical companies. As described by the term \u201cearnout\u201d, the sellers will earn the secondpart of the M&A payment out.Earnout agreement is primarily used to bridge the valuation gap between the buyer and theseller during the negotiation process. In addition, it facilitates the post-transaction transitionperiod by aligning the seller\u2019s incentives to the acquired company\u2019s performance. Because ofthese benefits, earnout agreement has been increasingly used in recent years. Earnout is mainlyemployed in deals with private targets with high information asymmetry. Figure 4.1 Panel Ashows that the fraction of earnout transactions increased from almost 0 in 1991 to 21.5% in thefull sample, and 32.8% in the private target sample in 2019. Earnout payment accounts for 33%of total transaction value on average in the M&A transactions with earnout. Figure 4.2 illustratesthe fraction of earnout transactions within each industry, indicating that earnout is mostly usedwhen the target company operates in the healthcare industry.While earnout is helpful to reduce information asymmetry and resolve moral hazard prob-lems of target manager\/owner shirking, it can cause potential problems. It is almost impossibleto design a complete earnout contract. Many issues need to be addressed in the negotiationprocess. For example, how long should the earnout period last? What size should the earnoutamount be? Who controls the business during the earnout period? Other issues include the108metrics and accounting standards to calculate the earnout payment. Failure to incorporatethese covenants in the M&A agreement can lead to legal disputes in the post-transaction period.Resolution of earnout disputes involves third parties as arbitrators and is fact specific. As a con-sequence, the outcome of an earnout dispute is usually beyond the control of the buyer and theseller. Even though earnout is used to resolve valuation uncertainty, it can introduce more un-certainty to both parties in the post-transaction period. As one court commented: \u201cAn earnoutoften converts today\u2019s disagreement over price into tomorrow\u2019s litigation over outcome.\u201d3More importantly, the contingent payment scheme may cause additional moral hazard prob-lems after the transaction. Buyers (sellers) can engage in value-destroying activities to minimize(maximize) earnout payments. Such activities can be earning manipulation by the seller, or un-willingness to provide resources by the buyer. If the seller continues to manage the companyafter the transaction, he(she) can reduce R&D expenses to achieve the earnout objective in theshort run. However, such activities can be detrimental to the acquired business in the long run.On the other hand, if the buyer controls the business, the earnout agreement can generate anopposite incentive. The moral hazard problems can be severe when the earnout targets are notobjectively verifiable. Even for the verifiable targets such as obtaining FDA approval, the timingmay still be manipulated. While the earnout contracts imply obligations of the buyers to ex-ert reasonable efforts to facilitate achievement of the earnout targets, many sellers argue in thelegal disputes that they fail to achieve the objectives due to a lack of buyer cooperation.4To understand the trade-off of earnout agreements, I collect a sample of 23,304 M&A trans-actions announced by U.S. public acquirers from 1991 to 2019. Among these transactions, 1,971transactions involve earnouts. I find that employing earnout in a M&A transaction facilitatesdeal completion. Including an earnout agreement in the transaction increases the deal com-pletion rate by 3.2%, which is 14.2% of the standard deviation. The number is economically sig-nificant considering an average deal failure rate of 9.4%. The positive effect of earnout on dealcompletion rate suggests the vital role earnout plays in bridging the valuation gap. However,3Aveta, Inc. v. Bengoa, 984 A 2d 126, 132 (Del. Ch. 2009).4Wolf and Fox (2012)109the probability of deal completion decreases as the earnout fraction increases. A large fractionof earnout payment may suggest a valuation gap which is too large to fill, or other moral hazardproblems discussed above that can lead to deal failure.I apply a standard event study methodology to investigate the effect of earnout agreementson acquirer wealth gains. The results suggest that acquirer announcement returns for earnouttransactions are insignificantly different from those of the transactions without earnout when asmall fraction of earnout payment is used. When earnout payments constitute only a small frac-tion of the total deal value, the incentive distortion problems are modest because the amountof contingent payment is low. The benefits of bridging a small valuation gap can be offset bythe costs of setting up an earnout agreement. As a result, public investors react indifferentlybetween such transactions and the transactions without earnout. When an earnout represents8% of the total transaction value (the 10th percentile), the acquirer\u2019s announcement return isinsignificantly (0.05%) lower than that of a transaction without earnout. However, acquirers re-ceive significantly lower CARs when large fractions of earnout are included in the transaction.An average earnout fraction of 33.38% decreases acquirer wealth gains at announcement by0.44%. When the earnout fraction increases to 66.67% (the 90th percentile), the acquirer expe-riences a 1.17% lower announcement return. In such deals, the incentive distortion problemsare expected to be higher because of the large amount of contingent payment. Disputes mayalso arise in the future, destroying the value of the combined business. As a result, acquiresexperience lower announcement returns.To investigate the reasons for earnout usage, I study the impact of target uncertainty on theprobability of earnout in mergers and acquisitions. It is challenging to measure target uncer-tainty directly because most earnout targets are private companies. I use the value-weightedaverage of uncertainty shocks faced by public companies in the target industry as a proxy. Pre-cisely, I measure uncertainty shocks as the annual change in equity volatility of the public com-panies. To address the endogeneity concerns, I focus on the component in equity volatilitychanges which are induced by macro uncertainty shocks. Results suggest that earnouts are110more likely to be used when target industry uncertainty is high. A one standard deviationincrease in target industry uncertainty increases the probability of earnout usage by 1%, rep-resenting a 12% increase given an average earnout rate of 8.5% in the sample. The fractionof earnout payment also increases with target industry uncertainty. The results suggest thatbridging the valuation gap between the acquirer and the target company induced by target un-certainty shocks is one of the reasons for acquirers to use earnouts.If target uncertainty is low, using an earnout agreement may not be optimal because the po-tential costs can outweigh the benefits. To further understand market perceptions on earnouts,I investigate the effect of earnout misuse on acquirer announcement returns. I employ a logisticmodel to predict the probability of earnout usage based on target uncertainty and other charac-teristics that are correlated with information asymmetry between the two parties. The earnouttransactions with predicted probabilities lower than the median are categorized as improperlyused earnouts. The results suggest that acquirers experience lower CARs when earnouts are im-properly used. The market perceives an earnout agreement as detrimental to acquirer value ifearnouts are not used to manage the valuation risks of the target company.Comparison between the earnout and non-earnout transactions suggests significant differ-ences in deal, target, and acquirer characteristics between the two groups. I conduct a match-ing analysis to ensure the results are not driven by fundamental differences between the twogroups. The matched control sample includes transactions that happen in the same year withsimilar deal values as the earnout transactions. Additional matching conditions are includedto ensure the target company has the same status as the earnout target firm, and the acquirersshare similar characteristics. The results are robust to the matching analysis. Earnout increasesdeal completion rate significantly, while a large fraction of earnout decreases the probabilityof deal completion. Acquirers in deals with a larger fraction of earnout payment receive lowerCARs. Earnouts are more likely to be used when target industry uncertainty is high. A largerfraction of earnout payment is employed as target industry uncertainty increases.The chapter is related to the literature on earnouts in mergers and acquisitions. Kohers and111Ang (2000) is the seminal paper that studies the earnout agreements. They show that earnoutsare more likely to be used in deals with high information asymmetry, e.g. in deals with privatetargets and targets operating in hi-tech or service industries. They also find that earnouts aremore likely to be used when the acquirer and the target operate in different industries. In addi-tion, they highlight the benefits of employing earnout to retain key talents of the target companyafter the transaction. Reuer et al. (2004) show that earnouts are more likely to be used in interna-tional M&A where information asymmetry is high. Viarengo et al. (2018) find that earnouts aremore likely to be used in countries with strong legal enforcement. Cain et al. (2011) conduct adetailed analysis of the earnout contracts. Barbopoulos and Sudarsanam (2012) and Barbopou-los and Adra (2016) argue that earnout structure matters for takeover premia and acquirer gains.Bates et al. (2018) find earnouts provide a source of financing for financially constrained acquir-ers. Different from previous studies, this chapter documents a negative impact of earnout onacquirer wealth gains when earnout accounts for a large fraction of the total deal value. Thefindings highlight the potential issues with earnouts and suggest that such contracts should beemployed with caution.The chapter also relates to the literature on uncertainty and M&A activities. Mitchell andMulherin (1996), Harford (2005), Ahern and Harford (2014) find that M&A activities are affectedby economic, technological, and regulatory shocks, and are clustered by industry. On the otherhand, Shleifer and Vishny (2003) and Rhodes-Kropf et al. (2005) argue that mispricing in thestock market drives M&A activities. Empirically, studies document a negative correlation be-tween economic uncertainty and M&A activities. Bhagwat et al. (2016) find that firms delayM&A transactions because of the interim uncertainty of target valuation. Bonaime et al. (2018)and Hao et al. (2022) show that policy uncertainty reduces M&A activities through the real op-tions framework. Nguyen and Phan (2017) document that uncertainty lengthens the deal com-pletion time. They also find that stock payments are more likely to be used in M&A when policyuncertainty is high. This chapter contributes to the literature by highlighting the advantages ofcontingent payment contracts in managing the valuation risks under uncertainty. In addition,112the chapter discusses the potential problems of including such agreements in the transactions.The chapter is organized as follows. Section 4.2 describes the sample and variables used inthe empirical analysis. Section 4.3 discusses the empirical strategies. Section 4.4 summarizesthe results, and section 4.5 concludes.4.2 DataThis section describes the sample construction process and discusses the variables used in theempirical analysis. The section also provides summary statistics of the sample.4.2.1 SampleThe sample consists of acquisitions in the Thomson Reuters SDC M&A database announced byU.S. public companies between January 1, 1991 and December 31, 2019. The sample periodstarts from 1991 because earnout agreements are seldomly used in the 1980s. The sample in-cludes both completed and withdrawn deals. The following criteria are applied to construct thefinal sample: (1) Acquirers are U.S. public companies listed on NYSE, AMEX, or Nasdaq witha market valuation no less than $1 million four weeks prior to announcement. (2) Acquirersconduct more than one deal throughout the sample period from 1991 to 2019. (2) Status ofthe target company is public, private, or subsidiary. (3) Deal value is no less than $1 million.(4) Bidders own less than 50% before the acquisition and are seeking a transfer of control (ownmore than 50% after the acquisition). Deals with missing acquirer ownership information areexcluded from the sample. (5) Deals announced on the same day by the same acquirer areexcluded from the sample. (6) Deals with targets from the financial and utility industries areexcluded from the sample. (7) Repurchases and recapitalizations are excluded from the sam-ple. The selection criteria yield a full sample of 23,304 deals by 6,502 acquirers. Information onstock returns are from CRSP.Figure 4.1 and Table C.3.1 Panel A illustrate the annual distribution of the earnout agree-113ments. Figure 4.1 Panel A shows the relative number of earnout transactions to the total numberof M&A deals by year. The dotted lines represent the fraction of earnout transactions with dif-ferent subsamples when the target is a public, private, or subsidiary company. Figure 4.1 PanelA illustrates a counter-cyclical pattern of the earnout agreements. They are more likely to beused during periods of economic contraction. During economic expansion periods, earnoutsare less likely to be used because the M&A market is competitive. Panel B illustrates the fractionof earnout transactions in terms of deal volume. Despite the increase in the number of earnouttransactions, total deal volume with earnouts stables after the financial crisis. The low earnoutdeal volume after 2009 may be attributed to the regulatory reforms in the post-recession pe-riod, which led to a more careful selection of the earnout deals. Panel C illustrates the fractionof earnout value to total deal value. It shows that the fraction of earnout payments increasesdramatically during recessions. Figure 4.1 also suggests that earnouts are most likely to be usedwhen the target is a private company, and least likely when the target is public. Panel B andC suggest that the increase in earnout amount is primarily driven by the rise in earnout agree-ments with private targets.Figure 4.2 shows the fraction of earnout transactions within each industry.5 Panel A demon-strates the fraction in the number of deals, and Panel B shows it in terms of deal volume. Figure4.2 Panel C illustrates the ratio of earnout value to the total M&A transaction value within eachindustry. Figure 4.2 Panel A suggests that the number of earnout transactions surges throughoutthe sample period within each industry. Among the five sectors, the healthcare sector experi-ences the most significant increase. The popularity of earnout agreements in the healthcaresector may be attributed to the objectively verifiable earnout targets, e.g., the FDA approvalof a drug. Figure 4.3 illustrates the distribution of earnout transactions by industry. Panel Aand B demonstrate the fraction in deal number and deal value respectively. 39.1% of the targetcompanies of earnout transactions operate in the hi-tech industry. 21.4% target companies arefrom the healthcare sector. 12% and 9.3% operate in consumer and manufacturing industries5Industries are classified based Fama-French five industry classifications.114respectively. Table C.3.1 Panel B offers the industry distribution of earnout transactions basedon the Fama-French twelve industry classifications.A standard event study methodology is applied to calculate acquirer announcement re-turns of the M&A transactions in the sample. Acquirer cumulative abnormal returns (CARs)are estimated in the 5-day event window centered around the M&A deal announcement date.The benchmark returns are estimated using the CAPM model from 300 to 46 days prior to an-nouncement. This allows a 45-day gap between the estimation window and the event window.Acquirers must have at least 70 valid returns during the estimation window to be included inthe analysis. Stock returns and other financial variables are winsorized at the 1st and 99th per-centiles. Appendix C.1 provides definitions of the variables used in the analysis.4.2.2 Measuring UncertaintyIt is very challenging to measure target uncertainty directly since most targets in earnout trans-actions are private. To capture uncertainty shocks to the target company, I calculate the value-weighted average of the uncertainty shocks to the public companies operating in the targetindustry. One way to measure uncertainty shocks is to use the changes in stock return volatility.However, changes in equity volatility can be endogenous. For example, stock return volatility iscorrelated with company performance. In the meanwhile, performance can affect the valuationof a company, which further affects the decision of whether an earnout should be employed.To address the endogeneity concerns, I follow Alfaro et al. (2021) and calculate the exoge-nous changes in stock return volatility due to macroeconomic uncertainty shocks for each listedcompany operating in the target SIC 3-digit industry:\u2206\u03c3p,t =\u03b20+\u2211c\u03b2c \u00b7 IV cj ,t +\u03f5p,t . (4.1)\u2206\u03c3p,t is the year-on-year change in annualized equity volatility for the public company poperating in the target\u2019s SIC 3-digit industry j. IV cj ,t is the uncertainty shock for each macro115factor c faced by industry j, taking industry j\u2019s exposure into account. Macro factor c includesoil, interest rate, 7 major currency exchange rate6 and economic policy uncertainty. To gener-ate the industry variations, the macro uncertainty variables IV cj ,t are constructed by exploitingcompanies\u2019 differential exposures to the aggregate volatility shocks. The idea is that companiesoperating in different industries can experience various levels of uncertainty shocks becauseof the difference in exposure. For instance, a mining company will experience a higher levelof uncertainty compared to companies operating in the public sector when oil uncertainty ishigh. On the other hand, companies in the public sector may face a higher uncertainty levelwhen economic policy uncertainty is high. To construct the variables, I first estimate industryexposure to each aggregated variable c as the sensitivity of regressing stock returns to the pricechanges of c for each industry j. IV cj ,t is constructed as the product of the estimated sensitivityand the year-on-year change in the standard deviation of daily price changes for c. A detaileddescription of how IV cj ,t is constructed is described in Appendix C.2.The predicted value \u0083\u2206\u03c3p,t from equation (4.1) is treated as the exogenous part in changesin equity volatility that is induced by macroeconomic uncertainty. The aggregate uncertaintyshocks, by construction, do not correlate with any observable firm characteristics that may af-fect the usage of earnout. The value-weighted average of \u0083\u2206\u03c3p,t of the public companies withineach SIC 3-digit industry, \u2206\u03c3 j ,t , is used as the proxy for target industry uncertainty.4.2.3 Descriptive StatisticsTable 4.1 Panel A reports summary statistics for the main variables used in the empirical anal-ysis. The sample consists of 23,304 M&A transactions from 1991 to 2019. 1,971 transactions,representing 8.5% of the full sample, employ earnout agreements. The average earnout valueis $44 million, accounting for 33% of the total earnout transaction value on average. Both theamount and fraction of earnout payment vary significantly across transactions. The medianearnout value is $8 million, which is significantly lower than the average. The distribution of6These include: Australian Dollar, British pound, Canadian Dollar, the Euro, Japanese Yen, and Swedish Krona.116earnout value suggests that including only an indicator variable for earnout is not enough forthe earnout analyses. Acquirer CARs are 1.35% on average, consistent with Betton et al. (2008)that acquirers receive positive but modest CARs at announcement. The majority of M&A trans-actions are completed with an average deal completion rate of 90.6%.50.3% of the targets are private companies. 32.5% are subsidiaries of other companies, and17.1% are public companies. 22.8% of the target companies operate in hi-tech industries. Themean (median) deal value is $356.7 ($42) million. The acquirer and target company operatein the same SIC 2-digit industry 62.6% of the cases. 17.1% of the transactions are cross-borderacquisitions. In terms of acquirer characteristics, the acquirer\u2019s mean (median) total assets is$5,043.8 ($560) million. The average log market to book ratio is 4.96. Acquirers have a leverageratio of 46.9% on average.Table 4.1 Panel B compares deal, target, and acquirer characteristics between M&A trans-actions with and without earnouts. Transactions without earnout demonstrate slightly higheracquirer announcement returns, but the difference is statistically insignificant. Earnout trans-actions demonstrate a significantly higher deal completion rate compared to the control group.Target industry uncertainty is slightly higher for earnout transactions compared to the dealswithout earnout. There are significantly more transactions with private targets, and less trans-actions with subsidiary or public targets in the earnout sample. 25.5% of the target companiesoperate in the hi-tech industries in the earnout sample, which is 3% higher than the controlsample. Transactions with earnout are smaller in size on average. Differences in the fraction ofdiversification deals, i.e., when the acquirer and target company operate in different industries,or cross-border deals are insignificant between the two samples. In terms of acquirer charac-teristics, acquirers that employ earnout transactions are smaller in size, have lower market tobook ratio, and perform worse than those that do not use earnout on average. They also havehigher leverage ratios. Acquirers in earnout transactions can be more financially constrainedand use earnout due to financial considerations.1174.3 Empirical MethodologyIn this section, I discuss the methodology used in the empirical analysis. First, I describe theanalyses of the effect of earnout on deal completion and acquirer wealth gains. Then, I discussthe study to investigate the impact of target uncertainty on the earnout employment. Afterthat, a matching analysis is conducted to address the sample selection bias. The study of howearnout misuse can affect acquirer wealth gains is discussed at the end of this section.4.3.1 Earnout and M&A Deal CompletionIf earnout helps bridge the valuation gap between the acquirer and the target company, employ-ing an earnout agreement in the M&A transaction should facilite deal completion. A binomiallogistic model is applied to estimate the impact of earnout on the probability of deal comple-tion:lnPi1\u2212Pi=\u03b20+\u03b21Earnouti +\u03b22Earnout Pcti +\u03c7i +\u03c5a +\u03b8t +\u03c6k +\u03c8 j +\u03f5i ,t . (4.2)The probability that deal i is completed, Pi = p(Deal Completioni = 1), is given byPi = 11+exp\u2212(\u03b20+\u03b21Earnouti+\u03b22Earnout Pcti+\u03c7i+\u03c5a+\u03b8t+\u03c6k+\u03c8 j ) . (4.3)The log likelihood function is estimated using maximum likelihood techniques. Deal Completioniis an indicator variable which equals to one if deal i is completed, and zero if the deal is with-drawn. The variables Earnouti and Earnout Pcti are the main independent variables. Earnoutiis an indicator variable which equals one if an earnout agreement is involved in the transaction,and zero otherwise. Earnout Pcti is the fraction of earnout value of the total transaction value.For transactions that do not employ earnout agreement, both Earnouti and Earnout Pcti areequal to zero.The hypothesis that earnout bridges the valuation gap implies a positive and statisticallysignificant estimate of \u03b21. The fraction of earnout payment exerts two confronting effects on118deal completion. On one hand, it helps bridge the valuation gap. On the other hand, a largefraction of earnout payment can introduce greater post-transaction moral hazard problems,which may bread down the deal. Depending on the relative size of benefits and costs of theearnout agreement, the sign of \u03b22 becomes an empirical question.Deal and target characteristics \u03c7i are included to control for any deal- and target-specificeffects on deal completion. \u03c5a is a vector of acquirer-level attributes including log assets, logmarket to book ratio, return on assets, and leverage ratio. Year fixed effects, \u03b8t , are included tocontrol for any macroeconomic factors that may affect the deal completion rate. Last, I includeacquirer and target SIC 2-digit industry fixed effects to control for any industry characteristicsaffecting the estimated results.4.3.2 Earnout and Acquirer Wealth GainsNext, I investigate the impact of earnout on acquirer wealth gains using an OLS regression:CAR[-2,+2]i ,t =\u03b20+\u03b21Earnouti +\u03b22Earnout Pcti +\u03c7i +\u03c5a +\u03b8t +\u03c6k +\u03c8 j +\u03f5i ,t . (4.4)The dependent variable, CAR[-2,+2]i ,t , is acquirer cumulative abnormal returns (CARs) es-timated using a five-day event window [-2,+2] centered around the announcement date (day 0).Earnouti is an indicator variable which equals one if an earnout agreement is involved in thetransaction, and zero otherwise. Earnout Pcti is the ratio of earnout value to total deal value.Because of the trade-offs of earnout agreements, signs of the estimated \u03b21 and \u03b22 are empiricalquestions, depending on the relative magnitude of the benefits and expected costs.Similar to Equation (4.2), deal, target, and acquirer characteristics \u03c7i and \u03c5a are included tocontrol for any deal- and acquirer-specific effects on acquirer announcement returns. Year fixedeffects, \u03b8t , are included to control for any macroeconomic factors affecting acquirer wealthgains from the M&A transaction. Acquirer and target SIC 2-digit industry fixed effects are in-119cluded to control for any industry characteristics that might affect acquirer CARs.4.3.3 Earnout and Target Industry UncertaintyIf earnouts are primarily used to manage valuation risks of the target company, the likelihoodof employing an earnout agreement should increase with target uncertainty. To investigate thehypothesis, I estimate the regressions:lnPi1\u2212Pi=\u03b20+\u03b21\u2206\u03c3 j ,t +\u03c7i +\u03c5a +\u03b8t +\u03c6k +\u03c8 j +\u03f5i ,t , (4.5)where Pi = p(Earnouti = 1) is the probability that deal i involves an earnout agreement.Earnout Pcti =\u03b20+\u03b21\u2206\u03c3 j ,t +\u03c7i +\u03c5a +\u03b8t +\u03c6k +\u03c8 j +\u03f5i ,t . (4.6)Earnouti is an indicator variable that equals one if an earnout agreement is included. Earnout Pctiis the ratio of earnout payment to deal value. The main independent variable\u2206\u03c3 j ,t is the value-weighted uncertainty shock of the public companies operating in the target SIC 2-digit industryj described in Section 4.2.2. When the target industry experiences an uncertainty shock, targetvaluation risks are expected to increase. Acquirers may be more likely to employ earnout in thetransaction to resolve the enlarged disagreement between the two parties. As target industryuncertainty increases, the fraction of earnout payment is also expected to increase to reflectthe larger information asymmetry and valuation gap. This implies that the estimates of \u03b21 inEquations (4.5) and (4.6) should be positive and statistically significant.Similar to Equation (4.2), deal, target, and acquirer characteristics \u03c7i and \u03c5a are includedto control for any deal, target, and acquirer effects on the usage of earnout. Year, acquirer, andtarget industry fixed effects, \u03b8t , \u03c6k , and \u03c8 j are included to control for any macroeconomic andindustry factors that might affect the employment of earnout agreements.1204.3.4 Matching AnalysisA comparison between the earnout transactions and the M&A transactions without earnoutin Table 4.1 Panel B suggests that deal, target, and acquirer characteristics differ significantlybetween the two groups. To address the concern that the results of the above analyses maybe driven by the fundamental differences between the earnout and non-earnout transactions,I conduct a matching analysis using various matching criteria. In this section, I discuss thedetails of the matching process.The control group constitutes M&A transactions without earnout agreements announcedby U.S. acquirers between 1991 to 2019. The following criteria are applied to construct thematched sample: (1) The deal is announced in the same year as the earnout transaction. (2)The target company shares the same status as the earnout target company. (3) Deal value is +\/-20% of the earnout transaction value. (4) Acquirer total assets are +\/- 20% of the acquirer totalassets of the earnout transaction. I also conduct a one-to-two propensity score matching withno replacement using criteria (1)-(3) and acquirer characteristics including total assets, mar-ket to book ratio, return on assets, and leverage ratio. The propensity score matching yields asample of 1,836 earnout transactions and a control sample of 2,835 non-earnout transactions.Table 4.5 Panel A compares the earnout and non-earnout transactions after the propensityscore matching. The summary statistics indicate that most differences in target characteristicsbecome insignificant after the matching. The two samples have similar compositions in termsof target status. Acquirer characteristics are still significantly different, but the differences be-come much smaller compared to Table 4.1 Panel B. One thing worth noticing in Table 4.5 PanelA is acquirer CARs. Table 4.1 Panel B shows an insignificant difference in acquirer CARs be-tween the two groups. However, earnout transactions demonstrate significantly lower acquirerwealth gains compared to the non-earnout group in the matching sample.1214.3.5 Earnout MisuseLastly, I investigate market responses on acquirer wealth gains when earnouts are misused. Thebenefits of employing earnout agreements are limited when valuation risks are low, e.g., whentarget industry uncertainty is low. On the other hand, acquirers bear the expected costs suchas the settlement costs for future legal disputes, or destruction in firm value due to incentivemisalignment in the post-transaction period. Acquirers are expected to receive lower wealthgains if the expected costs outweigh the benefits of the earnout agreements.To investigate this, I calculate the predicted probability of earnout usage based on Equa-tion (4.5). Deals with earnouts are categorized into two samples where earnouts are properlyused or misused, i.e., whether earnouts are employed to manage valuation risks. An earnout isidentified as proper if the predicted probability is above the median. Otherwise, the earnout isdeemed as misused. One possible reason when an earnout is not used to manage the valuationrisks might be that acquirers are trying to exploit the target companies with earnouts, whichpossibly causes incentive misalignment and legal disputes in the future.The impact of earnout misuse on acquirer wealth gains is estimated using:CAR[-2,+2]i ,t =\u03b20+\u03b21Earnout Misusei +\u03c7i +\u03c5a +\u03b8t +\u03c6k +\u03c8 j +\u03f5i ,t . (4.7)The dependent variable, CAR[-2,+2]i ,t , is acquirer CARs within the 5-day event window aroundthe announcement date. The main independent variable is an indicator variable which equalsone if the predicted probability of earnout is lower than the median. Deal, target, and acquirercharacteristics are included in the analysis to control for the impacts on CARs. Year, acquirer,and target SIC 2-digit industry fixed effects are included to control for the impact of any macroe-conomics or industry factor. Based on the hypothesis, the estimate of \u03b21 is expected to be neg-ative and statistically significant.1224.4 ResultsThis section summarizes results of the empirical analyses described in Section 4.3.4.4.1 Earnout and M&A Deal Completion: Logistic Estimation ResultsTable 4.2 presents estimates for the effect of earnout agreement on deal completion based onEquation (4.2). The dependent variable is an indicator variable which equals one if the deal iscompleted. The main independent variables are an indicator variable of whether an earnoutagreement is employed, and the ratio of earnout value to deal value. Columns (1) and (2) in-clude the earnout dummy variable as the independent variable. Columns (3) and (4) add theearnout fraction as an additional independent variable. Previous literature (Betton et al., 2008)documents that target status plays an important role in various M&A outcomes. Therefore, tar-get status is included as a control variable in all specifications. Columns (5) and (6) control fordeal and target characteristics that can affect deal completion. Columns (7) and (8) include ad-ditional acquirer characteristics that may affect M&A outcomes. Year fixed effects are includedin all specifications. Columns (2), (4), (6) and (8) include acquirer and target industry fixedeffects.Results from column (2) show that the impact of earnout agreement on deal completion isinsignificantly different from zero when the value of the earnout is neglected in the analysis.Results in columns (3) to (8) indicate that including an earnout agreement in the M&A trans-action increases the deal completion rate significantly controlling for the earnout fraction. Theprobability of deal completion increases by 3.2% when earnout is employed in the transaction.The effect is economically significant given the average deal withdrawal rate is 9.4%. Earnoutfraction, on the other hand, decreases the deal completion rate significantly. A one standarddeviation increase in earnout fraction decreases the probability of deal completion by 1.23%. Acomparison of the results between columns (1)-(2) and columns (3)-(8) suggests that both theearnout usage and the fraction of earnout payment are important factors in explaining M&A123outcomes. They play confronting roles in terms of facilitating deal completion.Consistent with previous studies, a deal is more likely to complete when the target is a pri-vate or subsidiary company. The deal withdrawal rate is higher in cross-border acquisitions.A transaction is more likely to complete when acquirers have better performance and lowerleverage ratio.4.4.2 Earnout and Acquirer Wealth Gains: OLS Estimation ResultsTable 4.3 reports estimates for the impact of earnout agreement on acquirer CARs based onEquation (4.4). The dependent variable is acquirer cumulative abnormal returns in the 5-dayevent window around the announcement date. The key independent variables are the earnoutindicator variable and the fraction of the earnout payment. Specifications in columns (1) and(2) include only the earnout indicator variable as the independent variable. Columns (3) and (4)add the earnout fraction as an additional variable of interest. Dessaint et al. (2021) show that ac-quirer announcement returns are significantly higher when the target is a private or subsidiarycompany. To control for such effects, target status is included in all specifications. Columns (5)to (8) include deal and acquirer characteristics affecting acquirer wealth gains. Year fixed effectsare included in all specifications to control for any macroeconomic factor that may drive M&Awaves and affect acquirer CARs. Specifications in columns (2), (4), (6), and (8) include acquirerand target industry fixed effects to control for any industry characteristics affecting acquirerannouncement returns.The insignificant impact of earnout usage on acquirer CARs can suggest the trade-offs ofan earnout agreement. Because of the benefits and expected costs, acquirer CARs are similarto the transactions without earnouts. However, incorporating a large fraction of earnout pay-ment significantly reduces acquirer wealth gains. The results are consistent with the hypothe-sis that large contingent payments may introduce incentive misalignment problems which aredetrimental to firm value. A one standard deviation increase in the earnout fraction decreasesacquirer CARs by 0.44%. The effect is economically significant given an average acquirer an-124nouncement return of 1.35%. Target characteristics that affect acquirer CARs include targetstatus and whether the target company operates in hi-tech industries. Transactions with pri-vate and subsidiary targets receive significantly higher acquirer announcement returns. TheCARs are lower when target company operates in high-tech industries. Consistent with the lit-erature, smaller acquirers with higher leverage ratios receive higher CARs.4.4.3 Estimation Results on Earnout and Target Industry UncertaintyTable 4.4 reports estimates for the impact of target industry uncertainty on earnout agreement.Table 4.4 Panel A reports the estimated results from the logistic regression of the earnout indica-tor on target industry uncertainty, estimated using Equation (4.5). Table 4.4 Panel B shows theestimated results from the OLS regression of the earnout fraction on target industry uncertainty,estimated using Equation (4.6). The dependent variable in Panel A is an indicator variable thatequals one if an earnout agreement is used. The dependent variable in Panel B is the ratio ofearnout value to deal value. The main independent variable is the target industry uncertainty,which is the value-weighted average of the uncertainty shocks to public companies operatingin the target SIC 3-digit industry. To address the endogeneity concerns, uncertainty shocks topublic companies are estimated as the changes in annualized stock return volatility induced bymacroeconomic uncertainty shocks.Previous studies find that earnout transactions are more likely to be used when there is highinformation asymmetry between the acquirer and target company. Deal characteristics such aswhether the acquirer and the target operate in the same industry, whether the target companyoperates in hi-tech sectors, and whether the transaction is a cross-border deal are included incolumns (3) to (6) to control for the impact of information asymmetry on earnout employment.Additional acquirer characteristics are included to control for any acquirer effect on the adop-tion of an earnout agreement. Year fixed effects are included in all specifications to accountfor the aggregate time trend of earnout adoption. Acquirer and target SIC 2-digit industry fixedeffects are included in columns (2), (4), and (6) to control for any time-invariant industry char-125acteristics affecting the earnout usage.Results in Table 4.4 suggest that earnout agreements are more likely to be used when thetarget industry uncertainty is high. Results in Panel A column (6) indicate that a one standarddeviation increase in targe industry uncertainty increases the likelihood of earnout usage byapproximately 1%, representing a 12% increase given an average earnout rate of 8.5% in thesample. The fraction of earnout payment also increases with target industry uncertainty. Con-sistent with previous studies, earnouts are more likely to be used when the target is a privateor subsidiary company. Smaller acquirers with low leverage ratios are more likely to employearnout agreements.4.4.4 Results on Matching AnalysisTo address the concern that the earnout and non-earnout groups are fundamentally different,I re-estimate Equations (4.2) to (4.6) using various matched control samples. Table 4.5 Panel Acompares the earnout transactions and a matched sample using a one-to-two propensity scorematching on deal announcement year, target status, deal value, acquirer total assets, marketto book ratio, return on assets, and leverage ratio. Results in Panel A indicate that differencesbetween the two groups in most deal and target characteristics become insignificant after thematching. The differences in acquirer characteristics become smaller compared to those inTable 4.1.Table 4.5 Panel B to Panel E report results of re-estimating Equations (4.2) to (4.6) with differ-ent control samples. The control sample in columns (1) and (2) is matched on deal announce-ment year and the target status. The sample in columns (3) and (4) further restricts the dealvalue to be within 20% of the earnout transaction. An additional criterion of within 20% of ac-quirer total assets is imposed in the control sample in columns (5) and (6). Columns (7) and(8) use the control sample based on the propensity score matching mentioned above. Year,acquirer, and target industry fixed effects are included in all specifications.126Table 4.5 suggests that the results in Table 4.2 to Table 4.4 are robust to various matched con-trol samples. Earnout agreement facilitates deal completion, while a larger fraction of earnoutpayment decreases the probability of deal completion. Acquirers receive insignificant CARswhen an earnout agreement is included. However, acquire announcement returns are signif-icantly lower when a large fraction of earnout amount is involved. The magnitude of the im-pact of target industry uncertainty on earnout usage becomes larger with the propensity scorematched sample.4.4.5 Earnout Misuse and Acquirer Wealth Gains: OLS Estimation ResultsTable 4.6 reports estimates of the impact of earnout misuse on acquirer wealth gains. The de-pendent variable is the acquirer cumulative abnormal returns estimated within the 5-day eventwindow around the announcement date. The main independent variable is an indicator vari-able that equals one if the earnout is identified as misused. An earnout is identified as misused ifthe predicted probability of using an earnout agreement based on Equation (4.5) is lower thanthe median. Additional control variables, and year and industry fixed effects are included tocontrol for any characteristic that may affect acquirer wealth gains from the M&A transaction.The results show that acquirers receive significantly lower CARs when earnouts are mis-used. Estimates in column (6) suggest that acquirer receives 1.08% lower announcement re-turns when the earnout is improperly used. A comparison between estimations in Table 4.3and Table 4.6 reveals market perceptions of earnout agreements. In general, acquirer CARs areinsignificantly different between the earnout and non-earnout transactions. However, whenearnouts are used improperly for reasons other than to resolve high valuation risks, acquirersreceive significantly lower announcement returns.1274.4.6 Robustness TestsIn this section, I discuss the results of the robustness tests. The results are presented in Ap-pendix C.3. First, I re-estimate Equations (4.2) and (4.5)using linear probability models:Deal Completioni =\u03b20+\u03b21Earnouti +\u03b22Earnout Pcti +\u03c7i +\u03c5a +\u03b8t +\u03c6k +\u03c8 j +\u03f5i ,t , (4.8)Earnouti =\u03b20+\u03b21\u2206\u03c3 j ,t +\u03c7i +\u03c5a +\u03b8t +\u03c6k +\u03c8 j +\u03f5i ,t . (4.9)The dependent variable in Equation (4.8), Deal Completioni , is an indicator variable whichequals one if the M&A transaction is completed, and zero if the deal is withdrawn. Same asEquation (4.2), the main independent variables are Earnouti and Earnout Pcti . The dependentvariable in Equation (4.9), Earnouti , is an indicator variable of earnout usage. \u2206\u03c3 j ,t is the valueweighted uncertainty shock of all the public companies operating in the target industry as de-scribed in Section 4.2.2. Same as Equations (4.2) and (4.5), deal, target, and acquirer controlvariables are included in the regressions. Year, acquirer, and target industry fixed effects arealso included to control for any macroeconomic conditions or industry characteristics affect-ing the estimates.Table C.3.2 and Table C.3.3 suggest that the results are robust using linear probability mod-els. Earnout significantly increases the probability of deal completion. The economic magni-tudes are similar to Table 4.2. Including an earnout agreement in the M&A transaction increasesthe deal completion rate by approximately 3.1%. The estimated impacts of target uncertaintyon earnout usage are also similar to the main analysis. Target industry uncertainty increasesthe probability of using an earnout significantly. Earnouts are 1.2% more likely to be used witha one standard deviation increase in target industry uncertainty.Second, I include additional variables to control for the impact of the target company\u2019sgrowth prospective on earnout usage. An earnout agreement may be more likely to be em-ployed when the target company demonstrates high growth potential. I construct two variablesto measure the target company\u2019s future growth: the median of the target industry\u2019s sales growth128and the medium age of the public companies operating in the target industry. Target companiesoperate in less mature industries with higher sales growth are expected to have better growthperspectives.Third, I include additional variables to control for the impacts of M&A advisors on earnoutusage. The likelihood of earnout usage may depend on whether the acquirer and the targetcompany hire boutique banks as M&A advisors. On one hand, an earnout agreement may notbe necessary if the financial advisor has specific knowledge about an industry. On the otherhand, the boutique financial advisor may be more likely to suggest an earnout agreement whenit is necessary. I include two indicator variables of whether the acquirer and the target hireboutique financial advisors to control for the impacts. I also include the acquirer and the targetcompany\u2019s financial advisor fixed effects to control for any additional impact of the financialadvisors.Table C.3.4 Panel A reports the results controlling for the impacts of the target company\u2019sgrowth prospective. Panel B reports the results controlling for the impact of M&A advisors.Columns (1) and (2) in Panel A and columns (1) to (3) in Panel B report the estimated resultsusing Equation (4.5), where the dependent variable is an indicator variable of whether earnoutis employed in the transaction. In columns (3) and (4) in Panel A and columns (4) to (6) in PanelB, the dependent variable is the fraction of the earnout payment, and the results are estimatedbased on Equation (4.6). Table C.3.4 shows that the results are robust controlling for variousalternative control variables. The impacts of target industry uncertainty on earnout usage andearnout percentage remain positive and significant. Earnouts are more likely to be used whenthe target company operates in less mature industries, which may suggest more informationasymmetry or growth potential. Earnouts are less likely to be used when the target companyhires a boutique advisor.1294.5 ConclusionEarnout agreements have been increasingly used in M&A transactions in the past 30 years, es-pecially in the deals with private or subsidiary targets. The chapter finds that earnouts are pri-marily used as a mechanism to bridge the valuation gap between the buyer and the seller. Thelikelihood of earnout usage and the fraction of earnout payment increase significantly with thetarget industry uncertainty. Including an earnout agreement in the M&A transaction increasesthe deal completion rate significantly.Despite the benefits of bridging the valuation gap, industry practitioners have controver-sial opinions on the application of earnout agreement. The process of negotiating an earnoutcontract can be very complicated. Failure to design a complete earnout contract can lead tolegal disputes at the end of the earnout period. In addition, the contingent payment schemeintroduces an incentive misalignment problem in the merged business. The acquirer\u2019s objec-tive is to maximize firm value and minimize earnout payment in some cases, while the target\u2019sobjective is to maximize the earnout payment. When the earnout fraction is low, such conflictof interest is trivial. Acquirer wealth gains from the earnout deal are insignificantly differentfrom those deals without earnout. However, when a large contingent payment amount is in-cluded, acquirer experiences significantly lower announcement returns. The chapter providesa deep understanding of the earnout agreement, and sheds light on the trade-offs to considerwhen employing an earnout agreement. Managers who would like to employ earnouts in M&Atransactions should be aware of the potential problems and use them with caution.130Figure 4.1 Fraction of M&A Transactions with Earnout: 1991-2019Panel A. Relative Number of M&A Transactions with EarnoutThe figure depicts the annual distribution of the number of earnout transactions relative tototal number of M&A transactions in the sample. The sample constitutes of completed M&Atransactions by U.S. public acquirers over the period 1991-2019. The shaded vertical barsrepresent NBER recessions.131Panel B. Relative Deal Volume of M&A Transactions with EarnoutThe figure depicts the annual distribution of deal volume of earnout transactions relative to totaldeal volume of M&A transactions in the sample. The sample constitutes of completed M&A trans-actions by U.S. public acquirers over the period 1991-2019. The shaded vertical bars representNBER recessions.132Panel C. Ratio of Earnout value to Deal Volume of M&A Transactions: 1991-2019The figure plots the annual distribution of the earnout value relative to total deal volume of M&Atransactions in the sample. Earnout value is calculated as the sum of earnout payment of theearnout transactions each year. The sample constitutes of completed M&A transactions by U.S.public acquirers over the period 1991-2019. The shaded vertical bars represent NBER recessions.133Figure 4.2 Fraction of M&A Transactions with Earnout: Within IndustryPanel A. Relative Number of M&A Transactions with EarnoutThe figure depicts the number of earnout transactions relative to total number of M&Atransactions in the sample within each industry. The sample constitutes of completed M&Atransactions by U.S. public acquirers over the period 1991-2019. Industries are classifiedbased on Fama-French five industry classifications. The shaded vertical bars represent NBERrecessions.134Panel B. Relative Deal Volume of M&A Transactions with Earnout: Within IndustryThe figure depicts the deal volume of earnout transactions relative to total deal volume of M&Atransactions in the sample within each industry. The sample constitutes of completed M&Atransactions by U.S. public acquirers over the period 1991-2019. Industries are classified based onFama-French five industry classifications. The shaded vertical bars represent NBER recessions.135Panel C. Ratio of Earnout Value to Deal Volume of M&A Transactions: Within IndustryThe figure plots the ratio of earnout value to total deal volume of M&A transactions within eachindustry in the sample. Earnout value is calculated as the sum of earnout payment of the earnouttransactions within each industry every year. The sample constitutes of completed M&A trans-actions by U.S. public acquirers over the period 1991-2019. The shaded vertical bars representNBER recessions.136Figure 4.3 Fraction of Earnout Transactions by IndustryPanel A. Number of Earnout Transactions by IndustryThe figure plots the number of earnout transactions across industries in the sample. The sampleconstitutes of completed earnout transactions by U.S. public acquirers over the period 1991-2019.The industries are classified based on Fama-French five industry classifications.137Panel B. Deal Volume of Earnout Transactions by IndustryThe figure plots the deal volume of earnout transactions across industries in the sample. Thesample constitutes of completed earnout transactions by U.S. public acquirers over the period1991-2019. The industries are classified based on Fama-French five industry classifications.138Table 4.1 Descriptive Statistics: Full SamplePanel A. Summary Statistics of M&A TransactionsThe table reports summary statistics for the main variables used in the empirical analysis. Thesample includes acquisitions in the Thomson Reuters SDC M&A database announced betweenJanuary 1, 1991 and December 31, 2019 by U.S. public companies with market capitalizationgreater than $1 million four weeks prior to announcement. Only deals that worth at least $1million are included in the sample. The sample is further restricted to deals with a transfer ofcontrol, i.e. bidders own less than 50% before the acquisition and own more than 50% after theacquisition. Deals with target companies from the financial and utility industries are excludedfrom the sample. See Appendix C.1 for variable definitions.Obs. Mean SD P10 P50 P90Deal CharacteristicsEarnout Usage 23,304 0.085 0.278 0 0 0Earnout Pct (%) 1,971 33.380 22.559 8.065 28.571 66.667Earnout Value ($MM) 1,971 44.157 115.689 1.302 8.000 100.000CAR [-2,+2] (%) 21,103 1.348 8.705 -7.993 0.662 11.463Deal Completion 23,304 0.906 0.291 1 1 1Target CharacteristicsTarget Industry Uncertainty Shock 17,074 -0.002 0.057 -0.048 0.000 0.007Public Target 23,304 0.171 0.377 0 0 1Private Target 23,304 0.503 0.500 0 1 1Subsidiary Target 23,304 0.325 0.468 0 0 1Hi-tech Target 23,304 0.228 0.419 0 0 1Log Deal Value ($MM) 23,304 3.915 1.930 1.500 3.738 6.526Same Industry 23,304 0.626 0.484 0 1 1Cross Border 23,304 0.171 0.376 0 0 1Acquirer CharacteristicsAcquirer Log Assets ($MM) 22,027 6.423 2.056 3.795 6.330 9.210Acquirer Log MB 20,967 4.955 1.999 2.562 4.861 7.565Acquirer ROA 22,020 0.004 0.178 -0.141 0.045 0.125Acquirer Leverage Ratio 21,771 0.469 0.241 0.155 0.467 0.767139Panel B. Comparison of M&A Transactions with and without EarnoutThe table compares deal, target, and acquirer characteristics between M&A transactionswith and without earnout in the sample. The sample includes acquisitions in the ThomsonReuters SDC M&A database announced between January 1, 1991 and December 31, 2019 byU.S. public companies with market capitalization greater than $1 million four weeks prior toannouncement. Only deals that worth at least $1 million are included in the sample. The sampleis further restricted to deals with a transfer of control, i.e. bidders own less than 50% before theacquisition and own more than 50% after the acquisition. Deals with target companies from thefinancial and utility industries are excluded from the sample. *, **, and *** indicate statisticalsignificance at the 10%, 5%, and 1% levels, respectively. Variables are defined in Appendix C.1.With Earnout Without EarnoutMean SD Mean SD DifferenceDeal CharacteristicsCAR [-2,+2] (%) 1.289 8.699 1.353 8.706 -0.065Deal Completion 0.938 0.242 0.903 0.295 0.034***Target CharacteristicsTarget Industry Uncertainty Shock -0.001 0.064 -0.003 0.056 0.002Public Target 0.022 0.146 0.185 0.389 -0.163***Private Target 0.751 0.432 0.481 0.500 0.271***Subsidiary Target 0.227 0.419 0.334 0.472 -0.107***Hi-tech Target 0.255 0.436 0.225 0.418 0.030**Log Deal Value ($MM) 3.625 1.605 3.942 1.955 -0.317***Same Industry 0.612 0.487 0.628 0.483 -0.016Cross Border 0.187 0.390 0.169 0.375 0.017Acquirer CharacteristicsAcquirer Log Assets ($MM) 5.852 1.930 6.476 2.060 -0.624***Acquirer Log MB 4.548 1.871 4.993 2.007 -0.445***Acquirer ROA -0.016 0.198 0.006 0.176 -0.022***Acquirer Leverage Ratio 0.400 0.234 0.476 0.241 -0.075***Observations 1,971 21,333 23,304140Table 4.2 Earnout and M&A Deal CompletionThe table reports results from the logistic regression of deal completion on the usage and fraction of earnout payment, estimated using Equa-tion (4.2). The sample includes M&A transactions announced by U.S. public acquirers from 1991 to 2019. The dependent variable is an indicatorvariable that equals one if a deal is completed, and zero if a deal is withdrawn. Earnout usage is an indicator variable which equals one if an earnoutagreement is included in the M&A transaction. Earnout pct is the ratio of earnout value to deal value. Columns (5) to (8) include deal-specific controlvariables. Columns (7) and (8) include additional acquirer-specific control variables. All columns include year fixed effects. Columns (2), (4), (6) and(8) include acquirer and target SIC 2-digit industry fixed effects. See Appendix C.1 for variable definitions. The standard errors (in parentheses) areclustered at acquirer industry (2-digit SIC) level. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.(1) (2) (3) (4) (5) (6) (7) (8)Earnout Usage 0.24* 0.16 0.76*** 0.65*** 0.75*** 0.65*** 0.66*** 0.60***(0.13) (0.11) (0.17) (0.14) (0.16) (0.14) (0.13) (0.12)Earnout Pct (%) -0.01*** -0.01*** -0.01*** -0.01*** -0.01*** -0.01***(0.00) (0.00) (0.00) (0.00) (0.00) (0.00)Private Target 0.90*** 0.83*** 0.90*** 0.84*** 1.15*** 1.16*** 1.20*** 1.21***(0.09) (0.11) (0.09) (0.11) (0.09) (0.10) (0.09) (0.09)Subsidiary Target 0.46*** 0.55*** 0.46*** 0.55*** 0.71*** 0.82*** 0.73*** 0.82***(0.12) (0.10) (0.12) (0.10) (0.10) (0.08) (0.08) (0.08)Hi-tech Target 0.47*** 0.14 0.42*** 0.06(0.11) (0.12) (0.11) (0.14)Log Deal Value ($MM) 0.13*** 0.15*** 0.11** 0.12**(0.02) (0.02) (0.05) (0.05)Same Industry -0.16 -0.05 -0.20 -0.09(0.14) (0.06) (0.15) (0.07)Cross Border -0.21** -0.26*** -0.25** -0.31***(0.08) (0.06) (0.10) (0.07)Acquirer Log Assets ($MM) 0.01 0.05(0.06) (0.04)Acquirer Log MB 0.03* 0.02(0.02) (0.02)141Table 4.2 Continued(1) (2) (3) (4) (5) (6) (7) (8)Acquirer ROA 0.64*** 0.49***(0.16) (0.13)Acquirer Leverage Ratio -0.80*** -0.58***(0.15) (0.18)Deal Control Variables No No No No Yes Yes Yes YesAcquirer Control Variables No No No No No No Yes YesYear FE Yes Yes Yes Yes Yes Yes Yes YesAcquirer, Target Industry FE No Yes No Yes No Yes No YesPseudo R2 0.03 0.06 0.03 0.06 0.04 0.07 0.05 0.07Observations 23,304 23,258 23,304 23,258 23,304 23,258 20,966 20,892Observations with Earnout 1,971 1,968 1,971 1,968 1,971 1,968 1,810 1,803142Table 4.3 Earnout and Acquirer Announcement Returns of M&A TransactionsThe table reports results from the OLS regression of acquirer announcement returns on the usage and fraction of earnout payment, esti-mated using Equation (4.2). The sample includes M&A transactions announced by U.S. public acquirers from 1991 to 2019. The dependent variableis acquirer CAR [-2,+2] (%), which is the cumulative abnormal return of the acquirer in the 5-day event window centered around the announcementdate. Earnout usage is an indicator variable which equals one if an earnout agreement is included in the M&A transaction. Earnout pct is the ratioof earnout value to deal value. Columns (5) to (8) include deal-specific control variables. Columns (7) and (8) include additional acquirer-specificcontrol variables. All columns include year fixed effects. Columns (2), (4), (6) and (8) include acquirer and target SIC 2-digit industry fixed effects.See Appendix C.1 for variable definitions. The standard errors (in parentheses) are clustered at acquirer industry (2-digit SIC) level. *, **, and ***indicate statistical significance at the 10%, 5%, and 1% levels, respectively.(1) (2) (3) (4) (5) (6) (7) (8)Earnout Usage -0.33 -0.37 0.40 0.33 0.43 0.33 0.15 0.10(0.22) (0.22) (0.28) (0.29) (0.28) (0.29) (0.26) (0.27)Earnout Pct (%) -0.02*** -0.02*** -0.02*** -0.02*** -0.02*** -0.02***(0.01) (0.01) (0.01) (0.01) (0.01) (0.01)Private Target 2.45*** 2.44*** 2.45*** 2.45*** 2.50*** 2.49*** 2.54*** 2.52***(0.19) (0.21) (0.19) (0.21) (0.21) (0.22) (0.21) (0.21)Subsidiary Target 2.83*** 2.84*** 2.83*** 2.84*** 2.80*** 2.88*** 2.95*** 3.05***(0.22) (0.21) (0.22) (0.21) (0.22) (0.21) (0.21) (0.20)Hi-tech Target -0.62*** -0.68** -0.52*** -0.51*(0.13) (0.26) (0.17) (0.27)Log Deal Value ($MM) 0.00 0.02 0.39*** 0.40***(0.04) (0.04) (0.06) (0.06)Same Industry 0.01 0.10 -0.07 0.07(0.10) (0.10) (0.14) (0.09)Cross Border -0.38 -0.38 -0.10 -0.13(0.25) (0.25) (0.24) (0.25)Acquirer Log Assets ($MM) -0.62*** -0.62***(0.10) (0.09)Acquirer Log MB -0.08** -0.06143Table 4.3 Continued(1) (2) (3) (4) (5) (6) (7) (8)(0.04) (0.04)Acquirer ROA -0.47 -0.76(0.61) (0.62)Acquirer Leverage Ratio 1.81*** 1.88***(0.30) (0.26)Deal Control Variables No No No No Yes Yes Yes YesAcquirer Control Variables No No No No No No Yes YesYear FE Yes Yes Yes Yes Yes Yes Yes YesAcquirer, Target Industry FE No Yes No Yes No Yes No YesAdjusted R2 0.01 0.02 0.02 0.02 0.02 0.02 0.03 0.03Observations 21,103 21,103 21,103 21,103 21,103 21,103 19,339 19,339Observations with Earnout 1,809 1,809 1,809 1,809 1,809 1,809 1,695 1,695144Table 4.4 Earnout and Target Industry Uncertainty ShockThe table reports the impact of target industry uncertainty on earnout agreements. Panel A reports results from the logistic regression of earnoutusage on target industry uncertainty, estimated using Equation (4.5). Panel B reports results from the OLS regression of earnout fraction on targetindustry uncertainty, estimated using Equation (4.6).The sample includes M&A transactions announced by U.S. public acquirers from 1991 to 2019.The dependent variable in Panel A is an indicator variable that equals one if an earnout agreement is included. The dependent variable in PanelB is the ratio of earnout value to deal value. The independent variable is the normalized target industry uncertainty shock described in Section4.2.2. Columns (3) to (6) include deal-specific control variables. Columns (5) and (6) include additional acquirer-specific control variables. All columnsinclude year fixed effects. Columns (2), (4), and (6) include acquirer and target SIC 2-digit industry fixed effects. See Appendix C.1 for variabledefinitions. The standard errors (in parentheses) are clustered at acquirer industry (2-digit SIC) level. *, **, and *** indicate statistical significanceat the 10%, 5%, and 1% levels, respectively.Panel A. Earnout Usage and Target Industry Uncertainty Shock(1) (2) (3) (4) (5) (6)Target Industry Uncertainty Shock 0.13*** 0.13** 0.13*** 0.13** 0.15*** 0.15**(0.05) (0.06) (0.05) (0.06) (0.05) (0.06)Private Target 2.59*** 2.65*** 2.53*** 2.65*** 2.86*** 2.98***(0.45) (0.41) (0.39) (0.37) (0.46) (0.45)Subsidiary Target 1.72*** 1.90*** 1.67*** 1.91*** 2.09*** 2.28***(0.42) (0.40) (0.38) (0.36) (0.45) (0.45)Hi-tech Target -0.02 -0.21*** -0.08 -0.19*(0.14) (0.08) (0.14) (0.10)Log Deal Value ($MM) -0.03 -0.00 0.17*** 0.18***(0.04) (0.03) (0.03) (0.03)Same Industry 0.06 0.06 0.01 0.02(0.14) (0.07) (0.14) (0.07)Cross Border -0.05 -0.19* 0.07 -0.05(0.10) (0.10) (0.08) (0.09)Acquirer Log Assets ($MM) -0.25*** -0.21***(0.04) (0.04)Acquirer Log MB -0.01 -0.06(0.05) (0.05)145Panel A. Continued(1) (2) (3) (4) (5) (6)Acquirer ROA 0.04 0.23(0.28) (0.21)Acquirer Leverage Ratio -0.99*** -0.47***(0.23) (0.17)Deal Control Variables No No Yes Yes Yes YesAcquirer Control Variables No No No No Yes YesYear FE Yes Yes Yes Yes Yes YesAcquirer, Target Industry FE No Yes No Yes No YesPseudo R2 0.08 0.12 0.08 0.13 0.11 0.15Observations 17,074 16,799 17,074 16,799 15,865 15,438Observations with Earnout 1,565 1,565 1,565 1,565 1,470 1,470146Panel B. Earnout Percentage and Target Industry Uncertainty Shock(1) (2) (3) (4) (5) (6)Target Industry Uncertainty Shock 0.54*** 0.40** 0.55*** 0.42** 0.56*** 0.47**(0.19) (0.17) (0.20) (0.17) (0.21) (0.19)Private Target 4.11*** 4.18*** 3.74*** 3.95*** 3.92*** 4.13***(0.41) (0.58) (0.66) (0.75) (0.74) (0.81)Subsidiary Target 1.53*** 1.97*** 1.18*** 1.78*** 1.72*** 2.15***(0.34) (0.36) (0.44) (0.51) (0.56) (0.58)Hi-tech Target -0.42 -0.80** -0.88 -0.90**(0.52) (0.38) (0.61) (0.41)Log Deal Value ($MM) -0.19 -0.11 0.23* 0.25**(0.14) (0.12) (0.14) (0.11)Same Industry 0.47 0.49** 0.21 0.32(0.51) (0.21) (0.48) (0.22)Cross Border -0.16 -0.49 0.10 -0.21(0.29) (0.31) (0.27) (0.29)Acquirer Log Assets ($MM) -0.61*** -0.49***(0.08) (0.09)Acquirer Log MB 0.21 0.03(0.18) (0.15)Acquirer ROA -2.09 -1.42(1.76) (1.36)Acquirer Leverage Ratio -3.10*** -1.53**(0.97) (0.71)Deal Control Variables No No Yes Yes Yes YesAcquirer Control Variables No No No No Yes YesYear FE Yes Yes Yes Yes Yes YesAcquirer, Target Industry FE No Yes No Yes No Yes147Panel B. Continued(1) (2) (3) (4) (5) (6)Adjusted R2 0.03 0.05 0.03 0.05 0.04 0.06Observations 17,074 17,074 17,074 17,074 15,865 15,865Observations with Earnout 1,565 1,565 1,565 1,565 1,470 1,470148Table 4.5 Matching AnalysisPanel A. Comparison of M&A Deal Characteristics with and without Earnout:Matching SampleThe table compares deal, target, and acquirer characteristics between M&A transactions withand without earnout in the sample. The earnout sample constitutes of earnout acquisitionsannounced by U.S. public acquirers from 1991 to 2019. The control sample is constructed basedon a one-to-two propensity score matching on deal announcement year, target status, deal value,and acquirer characteristics including total assets, market to book ratio, return on assets, andleverage ratio. Deals with target companies from the financial and utility industries are excludedfrom the sample. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels,respectively. Variables are defined in Appendix C.1.With Earnout Without EarnoutMean SD Mean SD DifferenceDeal CharacteristicsCAR [-2,+2] (%) 1.268 8.700 2.180 9.231 -0.912**Deal Completion 0.938 0.240 0.939 0.239 -0.001Target CharacteristicsTarget Industry Uncertainty Shock -0.001 0.064 -0.005 0.064 0.004Public Target 0.020 0.141 0.025 0.156 -0.005Private Target 0.753 0.432 0.720 0.449 0.033*Subsidiary Target 0.227 0.419 0.255 0.436 -0.028*Hi-tech Target 0.259 0.438 0.250 0.433 0.009Log Deal Value ($MM) 3.666 1.579 3.667 1.611 -0.000Same Industry 0.613 0.487 0.629 0.483 -0.016Cross Border 0.188 0.391 0.162 0.369 0.026*Acquirer CharacteristicsAcquirer Log Assets ($MM) 5.857 1.905 5.696 1.964 0.161**Acquirer Log MB 4.527 1.857 4.246 1.886 0.282***Acquirer ROA -0.017 0.199 -0.023 0.212 0.007Acquirer Leverage Ratio 0.399 0.234 0.434 0.255 -0.035***Observations 1,836 2,835 4,671149Panel B. Earnout and M&A Deal Completion Rate: Matching SampleThe table reports results from the logistic regression of deal completion on the usage and fraction of earnout payment, estimated using Equa-tion (4.2). The sample includes earnout transactions announced by U.S. public acquirers from 1991 to 2019, and a matched sample of non-earnouttransactions. The control sample in columns (1) and (2) is matched on deal announcement year and target status. The control sample in columns (3)and (4) is matched on deal announcement year, target status, and +\/-20% of deal value. The control sample in columns (5) and (6) is matched on dealannouncement year, target status, and +\/-20% of deal value and acquirer total assets. The control sample in columns (7) and (8) is constructed basedon a one-to-two propensity score matching on deal announcement year, target status, deal value, and acquirer characteristics including total assets,market to book ratio, return on assets, and leverage ratio. The dependent variable is an indicator variable that equals one if the deal is completed.Earnout usage is an indicator variable which equals one if an earnout agreement is employed. Earnout pct is the ratio of earnout value to deal value.Deal- and acquirer-specific control variables are included in all specifications. All columns include year, acquirer and target SIC 2-digit industry fixedeffects. See Appendix C.1 for variable definitions. The standard errors (in parentheses) are clustered at acquirer industry (2-digit SIC) level. *, **,and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.Year & Status Year, Status & DV Year, Status, DV & Assets Propensity Score(1) (2) (3) (4) (5) (6) (7) (8)Earnout Usage 0.65*** 0.59*** 0.61*** 0.54*** 0.60*** 0.53*** 0.40*** 0.30**(0.13) (0.13) (0.13) (0.14) (0.12) (0.13) (0.12) (0.14)Earnout Pct (%) -0.01*** -0.01*** -0.01*** -0.01*** -0.01*** -0.01*** -0.01*** -0.01***(0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)Private Target 1.20*** 1.22*** 1.14*** 1.17*** 1.27*** 1.34*** 1.15*** 1.11***(0.09) (0.10) (0.15) (0.15) (0.12) (0.13) (0.41) (0.39)Subsidiary Target 0.72*** 0.81*** 0.66*** 0.78*** 0.80*** 0.94*** 0.80** 0.77**(0.09) (0.09) (0.12) (0.12) (0.12) (0.13) (0.35) (0.35)Deal & Acquirer Control Variables Yes Yes Yes Yes Yes Yes Yes YesYear, Acquirer, Target Industry FE Yes Yes Yes Yes Yes Yes Yes YesPseudo R2 0.05 0.08 0.05 0.08 0.04 0.08 0.05 0.11Observations 20,565 20,463 17,707 17,611 13,951 13,824 4,484 4,201Observations with Earnout 1,810 1,801 1,809 1,800 1,768 1,758 1,783 1,684150Panel C. Earnout and Acquirer Announcement Returns of M&A Transactions: Matching SampleThe table reports results from the OLS regression of the acquirer announcement returns on the usage and fraction of earnout payment, esti-mated using Equation (4.4). The sample includes earnout transactions announced by U.S. public acquirers from 1991 to 2019, and a matched sampleof non-earnout transactions. The dependent variable is acquirer CAR [-2,+2] (%), which is the cumulative abnormal return of the acquirer in the5-day event window centered around the announcement date. Earnout usage is an indicator variable which equals one if an earnout agreement isincluded in the M&A transaction. Earnout pct is the ratio of earnout value to deal value. Deal- and acquirer-specific control variables are includedin all specifications. All columns include year, acquirer and target SIC 2-digit industry fixed effects. See Appendix C.1 for variable definitions. Thestandard errors (in parentheses) are clustered at acquirer industry (2-digit SIC) level. *, **, and *** indicate statistical significance at the 10%, 5%,and 1% levels, respectively.Year & Status Year, Status & DV Year, Status, DV & Assets Propensity Score(1) (2) (3) (4) (5) (6) (7) (8)Earnout Usage 0.16 0.11 0.10 0.07 0.01 -0.04 0.05 -0.03(0.26) (0.27) (0.26) (0.27) (0.27) (0.28) (0.33) (0.37)Earnout Pct (%) -0.02*** -0.02*** -0.02*** -0.02*** -0.02*** -0.02*** -0.02*** -0.02**(0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01)Private Target 2.58*** 2.58*** 2.34*** 2.32*** 1.93** 1.92** 3.07*** 2.97***(0.21) (0.21) (0.25) (0.24) (0.83) (0.83) (0.72) (0.75)Subsidiary Target 3.02*** 3.12*** 2.79*** 2.88*** 2.54*** 2.64*** 3.82*** 3.81***(0.21) (0.20) (0.28) (0.27) (0.84) (0.86) (0.71) (0.72)Deal & Acquirer Control Variables Yes Yes Yes Yes Yes Yes Yes YesYear, Acquirer, Target Industry FE Yes Yes Yes Yes Yes Yes Yes YesAdjusted R2 0.03 0.03 0.03 0.03 0.03 0.03 0.04 0.04Observations 18,962 18,962 16,281 16,281 12,821 12,821 4,165 4,165Observations with Earnout 1,695 1,695 1,695 1,695 1,663 1,663 1,673 1,673151Panel D. Earnout Usage and Target Industry Uncertainty Shock: Matching SampleThe table reports results from the logistic regression of earnout usage on target industry uncertainty, estimated using Equation (4.5). Thesample includes earnout transactions announced by U.S. public acquirers from 1991 to 2019, and a matched sample of non-earnout transactions.The dependent variable is an indicator variable that equals one if an earnout agreement is included in the M&A transaction. The independentvariable is the normalized target industry uncertainty shock described in Section 4.2.2. Deal- and acquirer-specific control variables are includedin all specifications. All columns include year, acquirer and target SIC 2-digit industry fixed effects. See Appendix C.1 for variable definitions. Thestandard errors (in parentheses) are clustered at acquirer industry (2-digit SIC) level. *, **, and *** indicate statistical significance at the 10%, 5%,and 1% levels, respectively.Year & Status Year, Status & DV Year, Status, DV & Assets Propensity Score(1) (2) (3) (4) (5) (6) (7) (8)Target Industry Uncertainty Shock 0.14*** 0.15** 0.15*** 0.15** 0.15*** 0.14** 0.25*** 0.22***(0.05) (0.06) (0.05) (0.06) (0.05) (0.06) (0.07) (0.07)Private Target 2.73*** 2.85*** 1.79*** 1.91*** 0.60 0.73 0.44 0.60(0.46) (0.45) (0.50) (0.50) (0.57) (0.56) (0.50) (0.48)Subsidiary Target 1.96*** 2.15*** 1.05** 1.23** -0.01 0.19 0.30 0.56(0.46) (0.45) (0.50) (0.50) (0.57) (0.56) (0.47) (0.46)Deal & Acquirer Control Variables Yes Yes Yes Yes Yes Yes Yes YesYear, Acquirer, Target Industry FE Yes Yes Yes Yes Yes Yes Yes YesPseudo R2 0.11 0.15 0.08 0.12 0.06 0.10 0.01 0.07Observations 15,696 15,272 13,754 13,392 11,002 10,738 3,664 3,599Observations with Earnout 1,470 1,470 1,470 1,470 1,451 1,451 1,458 1,452152Panel E. Earnout Percentage and Target Industry Uncertainty Shock: Matching SampleThe table reports results from the OLS regression of earnout fraction on target industry uncertainty, estimated using Equation (4.6). Thesample includes earnout transactions announced by U.S. public acquirers from 1991 to 2019, and a matched sample of non-earnout transactions.The dependent variable is the ratio of earnout value to deal value. The independent variable is the normalized target industry uncertainty shockdescribed in Section 4.2.2. Deal- and acquirer-specific control variables are included in all specifications. All columns include year, acquirer and targetSIC 2-digit industry fixed effects. See Appendix C.1 for variable definitions. The standard errors (in parentheses) are clustered at acquirer industry(2-digit SIC) level. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.Year & Status Year, Status & DV Year, Status, DV & Assets Propensity Score(1) (2) (3) (4) (5) (6) (7) (8)Target Industry Uncertainty Shock 0.56** 0.47** 0.63** 0.51** 0.71** 0.50 2.16** 1.50**(0.21) (0.20) (0.24) (0.22) (0.33) (0.30) (0.84) (0.73)Private Target 3.71*** 3.94*** 3.41*** 3.63*** 1.51 1.96 1.90 4.14(0.69) (0.77) (0.49) (0.54) (1.90) (1.69) (3.37) (2.64)Subsidiary Target 1.51*** 1.96*** 1.26** 1.74*** -0.64 0.14 0.36 3.50(0.52) (0.55) (0.48) (0.48) (2.10) (1.91) (3.51) (2.82)Deal & Acquirer Control Variables Yes Yes Yes Yes Yes Yes Yes YesYear, Acquirer, Target Industry FE Yes Yes Yes Yes Yes Yes Yes YesAdjusted R2 0.04 0.06 0.04 0.06 0.03 0.06 0.02 0.08Observations 15,696 15,696 13,754 13,754 11,002 11,002 3,664 3,664Observations with Earnout 1,470 1,470 1,470 1,470 1,451 1,451 1,458 1,458153Table 4.6 Earnout Misuse and Acquirer Announcement Returns of M&A TransactionsThe table reports results from the OLS regression of the acquirer announcement returns on the earnout misuse, estimated using Equation(4.7). The sample includes earnout transactions announced by U.S. public acquirers from 1991 to 2019. The dependent variable is acquirer CAR[-2,+2] (%), which is the cumulative abnormal return of the acquirer in the 5-day event window centered around the announcement date. Earnoutmisusage is an indicator variable which equals one if an earnout agreement is improperly used. See Section 4.3.5 for a detailed definition on Earnoutmisuse. Columns (3) to (6) include deal-specific control variables. Columns (5) and (6) include additional acquirer-specific control variables. Allcolumns include year fixed effects. Columns (2), (4), and (6) include acquirer and target SIC 2-digit industry fixed effects. See Appendix C.1 forvariable definitions. The standard errors (in parentheses) are clustered at acquirer industry (2-digit SIC) level. *, **, and *** indicate statisticalsignificance at the 10%, 5%, and 1% levels, respectively.(1) (2) (3) (4) (5) (6)Earnout Misuse -1.32*** -1.83*** -1.23*** -1.81*** -0.71 -1.16**(0.33) (0.39) (0.36) (0.42) (0.49) (0.54)Private Target 0.09 0.45 0.02 0.19 0.85 1.08(3.33) (2.96) (3.55) (3.11) (3.61) (3.16)Subsidiary Target 1.37 1.53 1.27 1.26 1.97 1.99(3.06) (2.61) (3.22) (2.75) (3.22) (2.74)Deal Control Variables No No Yes Yes Yes YesAcquirer Control Variables No No No No Yes YesYear FE Yes Yes Yes Yes Yes YesAcquirer, Target Industry FE No Yes No Yes No YesAdjusted R2 0.01 0.02 0.01 0.02 0.02 0.03Observations 1,385 1,385 1,385 1,385 1,385 1,385154Chapter 5ConclusionThe thesis is a collection of three essays studying the impacts of economic uncertainty on thefinancial markets. Economic uncertainty has been increasingly high in recent years. The thesisseeks to understand the vital role uncertainty plays in economic activities.In Chapter 2, I study the impact of economic uncertainty on going private transactionsthrough the corporate governance channel. The chapter shows that companies are more likelyto go private following uncertainty shocks. The effects are more prominent for companies withsevere conflicts between shareholders, e.g., companies with dual-class structure and less insti-tutional ownership. The effects are also stronger for companies facing large conflicts betweenshareholders and creditors: firms with lower asset redeployability, lower loan-to-bond ratio,and firms in financial distress. The results are consistent with the corporate governance hy-pothesis. Uncertainty exacerbates the agency problems of public companies. Companies goprivate to restructure their capital and align the interests between new shareholders, creditors,and managers. This chapter finds that companies receive a lower cost of debt after the agencyproblems are mitigated through going private.Chapter 3 investigates how economic uncertainty shocks can transform into shocks to in-vestors\u2019 information production costs and affect firms\u2019 relative cost of bank loans vs. corporatebonds. The financial intermediation theories suggest that, in order for banks to create safemoney which are liquid and can be redeemable at par, banks should keep information abouttheir assets secret. The chapter argues that this, in turn, translates into a comparative advan-tage when banks lend to more opaque firms. Using a sample of firms that issue bank loans andcorporate bonds simultaneously, the chapter documents that firms pay a relatively lower cost155of bank loans when they experience uncertainty shocks. In other words, banks offer an opacitydiscount to opaque firms. The chapter finds that the opacity discount is offered more by banksthat rely more on the money creation function, e.g., banks that experience uninsured depositoutflows, or banks that do not receive implicit government guarantees after the financial crisis.In Chapter 4, I investigate how companies use earnout agreement, a contingent paymentcontract, to manage the elevated valuation risks following uncertainty shocks. Earnouts arewidely applied in M&A transactions with private targets, increasing from almost 0 to more than30% in the past twenty years. In this chapter, I investigate the trade-offs of using an earnoutagreement. On one hand, earnout bridges the valuation gap between the buyer and the sellerand helps facilitate deal completion. I show that earnouts are more likely to be used when targetindustry uncertainty is high. The probability of deal completion increases significantly when anearnout agreement is employed. On the other hand, earnout can introduce incentive misalign-ment problems in the post-transaction period. The acquirer\u2019s objective is to maximize firmvalue, while the target\u2019s objective is to maximize earnout payments. Such incentive misalign-ment can destroy firm value. I find the acquirer CARs to be significantly lower when earnoutsare misused, i.e., when they are not used to manage the valuation risks. The chapter suggeststhat the costs of earnout should not be neglected and earnout should be used with caution.To sum up, Chapter 2 and 3 of the thesis study how economic uncertainty affects firms\u2019cost of capital in the equity and debt market respectively. The results suggest that the costsof private equity and private debt are relatively lower compared to the public market in thepresence of uncertainty shocks. Chapter 4 focuses on the impacts of economic uncertainty inmergers and acquisitions. It shows how firms use contingent payment contracts to managethe valuation risks after uncertainty shocks, and highlights the potential problems associatedwith such contracts. The thesis provides a deep understanding of the impacts of uncertaintyon the financial markets, which is an important finance question, especially under the currenteconomic turmoil.156BibliographyAcharya, Viral V, and Stephen G Ryan, 2016, Banks\u2019 financial reporting and financial systemstability, Journal of Accounting Research 54, 277\u2013340.Agarwal, Vikas, Rahul Vashishtha, and Mohan Venkatachalam, 2018, Mutual fund transparencyand corporate myopia, Review of Financial Studies 31, 1966\u20132003.Agrawal, Anup, and Gershon N Mandelker, 1990, Large shareholders and the monitoring ofmanagers: The case of antitakeover charter amendments, Journal of Financial and Quan-titative Analysis 25, 143\u2013161.Ahern, Kenneth R, and Jarrad Harford, 2014, The importance of industry links in merger waves,Journal of Finance 69, 527\u2013576.Alfaro, Ivan, Nicholas Bloom, and Xiaoji Lin, 2021, The finance uncertainty multiplier, NationalBureau of Economic Research .Anderson, Ronald C, Augustine Duru, and David M Reeb, 2009, Founders, heirs, and corporateopacity in the United States, Journal of Financial Economics 92, 205\u2013222.Andolfatto, David, 2010, On the social cost of transparency in monetary economies, FederalReserve Bank of St. Louis Working Paper .Ashraf, Badar Nadeem, and Yinjie Shen, 2019, Economic policy uncertainty and banks\u2019 loanpricing, Journal of Financial Stability 44, 100695.Aslan, Hadiye, and Praveen Kumar, 2012, Strategic ownership structure and the cost of debt,Review of Financial Studies 25, 2257\u20132299.Bachmann, R\u00fcdiger, and Christian Bayer, 2013, \u2018Wait-and-see\u2019 business cycles?, Journal ofMon-etary Economics 60, 704\u2013719.Baker, Scott R, Nicholas Bloom, and Steven J Davis, 2016, Measuring economic policy uncer-tainty, Quarterly Journal of Economics 131, 1593\u20131636.Balakrishnan, Karthik, Mary Brooke Billings, Bryan Kelly, and Alexander Ljungqvist, 2014, Shap-ing liquidity: On the causal effects of voluntary disclosure, Journal of Finance 69, 2237\u20132278.Barbopoulos, Leonidas, and Sudi Sudarsanam, 2012, Determinants of earnout as acquisitionpayment currency and bidder\u2019s value gains, Journal of Banking & Finance 36, 678\u2013694.Barbopoulos, Leonidas G, and Samer Adra, 2016, The earnout structure matters: Takeover pre-mia and acquirer gains in earnout financed M&As, International Review of Financial Analysis45, 283\u2013294.157Bartik, Timothy J, 1991, Who benefits from state and local economic development policies?, WEUpjohn Institute for Employment Research .Bates, Thomas W, Jordan B Neyland, and Yolanda Yulong Wang, 2018, Financing acquisitionswith earnouts, Journal of Accounting and Economics 66, 374\u2013395.Becker, Bo, and Victoria Ivashina, 2014, Cyclicality of credit supply: Firm level evidence, Journalof Monetary Economics 62, 76\u201393.Behn, Markus, Rainer Haselmann, and Paul Wachtel, 2016, Procyclical capital regulation andlending, Journal of Finance 71, 919\u2013956.Benmelech, Efraim, and Nittai K Bergman, 2009, Collateral pricing, Journal of Financial Eco-nomics 91, 339\u2013360.Berger, Allen N, and Gregory F Udell, 1995, Relationship lending and lines of credit in small firmfinance, Journal of Business 351\u2013381.Bernanke, Ben S, 1983, Irreversibility, uncertainty, and cyclical investment, Quarterly Journal ofEconomics 98, 85\u2013106.Bernard, Darren, 2016, Is the risk of product market predation a cost of disclosure?, Journal ofAccounting and Economics 62, 305\u2013325.Betton, Sandra, B Espen Eckbo, and Karin S Thorburn, 2008, Corporate takeovers, Handbook ofEmpirical Corporate Finance 291\u2013429.Bhagwat, Vineet, Robert Dam, and Jarrad Harford, 2016, The real effects of uncertainty onmerger activity, Review of Financial Studies 29, 3000\u20133034.Bharath, Sreedhar T, and Amy K Dittmar, 2010, Why do firms use private equity to opt out ofpublic markets?, Review of Financial Studies 23, 1771\u20131818.Bharath, Sreedhar T, and Tyler Shumway, 2008, Forecasting default with the merton distance todefault model, Review of Financial Studies 21, 1339\u20131369.Bloom, Nicholas, 2009, The impact of uncertainty shocks, Econometrica 77, 623\u2013685.Bolton, Patrick, and Xavier Freixas, 2000, Equity, bonds, and bank debt: Capital structure andfinancial market equilibrium under asymmetric information, Journal of Political Economy108, 324\u2013351.Bolton, Patrick, Xavier Freixas, Leonardo Gambacorta, and Paolo Emilio Mistrulli, 2016, Rela-tionship and transaction lending in a crisis, Review of Financial Studies 29, 2643\u20132676.Bolton, Patrick, and Ernst-Ludwig Von Thadden, 1998, Blocks, liquidity, and corporate control,Journal of Finance 53, 1\u201325.Bonaime, Alice, Huseyin Gulen, and Mihai Ion, 2018, Does policy uncertainty affect mergers158and acquisitions?, Journal of Financial Economics 129, 531\u2013558.Boot, Arnoud WA, Radhakrishnan Gopalan, and Anjan V Thakor, 2006, The entrepreneur\u2019schoice between private and public ownership, Journal of Finance 61, 803\u2013836.Brav, Omer, 2009, Access to capital, capital structure, and the funding of the firm, The Journalof Finance 64, 263\u2013308.Cain, Matthew D, David J Denis, and Diane K Denis, 2011, Earnouts: A study of financial con-tracting in acquisition agreements, Journal of Accounting and Economics 51, 151\u2013170.Calomiris, Charles W, and Charles M Kahn, 1991, The role of demandable debt in structuringoptimal banking arrangements, American Economic Review 497\u2013513.Caprio, Lorenzo, Ettore Croci, and Alfonso Del Giudice, 2011, Ownership structure, family con-trol, and acquisition decisions, Journal of Corporate Finance 17, 1636\u20131657.Carhart, Mark M, 1997, On persistence in mutual fund performance, Journal of Finance 52, 57\u201382.Chava, Sudheer, and Amiyatosh Purnanandam, 2011, The effect of banking crisis on bank-dependent borrowers, Journal of Financial Economics 99, 116\u2013135.Chemmanur, Thomas J, and Paolo Fulghieri, 1994, Reputation, renegotiation, and the choicebetween bank loans and publicly traded debt, Review of Financial Studies 7, 475\u2013506.Chemmanur, Thomas J, and Paolo Fulghieri, 1999, A theory of the going-public decision,Reviewof Financial Studies 12, 249\u2013279.Chen, Guojun, 2016, Corporate savings, financing, and investment with aggregate uncertaintyshocks, Financing, and Investment with Aggregate Uncertainty Shocks .Chen, Hui, Hao Wang, and Hao Zhou, 2014, Stock return volatility and capital structure deci-sions, PBCSF-NIFR Research Paper .Chen, Qi, Itay Goldstein, Zeqiong Huang, and Rahul Vashishtha, 2020, Bank transparency anddeposit flows, Available at SSRN 3212873 .Claessens, Stijn, Simeon Djankov, Joseph PH Fan, and Larry HP Lang, 2002, Disentangling theincentive and entrenchment effects of large shareholdings, Journal of Finance 57, 2741\u20132771.Core, John E, Robert W Holthausen, and David F Larcker, 1999, Corporate governance, chiefexecutive officer compensation, and firm performance, Journal of Financial Economics 51,371\u2013406.Dang, Tri Vi, Gary Gorton, and Bengt Holmstr\u00f6m, 2015, Ignorance, debt and financial crises,Working Paper .Dang, Tri Vi, Gary Gorton, Bengt Holmstr\u00f6m, and Guillermo Ordonez, 2017, Banks as secret159keepers, American Economic Review 107, 1005\u201329.Darrough, Masako N, and Neal M Stoughton, 1990, Financial disclosure policy in an entry game,Journal of Accounting and Economics 12, 219\u2013243.DeAngelo, Harry, Linda DeAngelo, and Edward M Rice, 1984, Going private: Minority freezeoutsand stockholder wealth, Journal of Law and Economics 27, 367\u2013401.Dessaint, Olivier, B Espen Eckbo, and Andrey Golubov, 2021, Bidder-specific synergies and theevolution of acquirer returns, Tuck School of Business Working Paper .Diamond, Douglas W, 1984, Financial intermediation and delegated monitoring, Review of Eco-nomic Studies 51, 393\u2013414.Diamond, Douglas W, 1991, Monitoring and reputation: The choice between bank loans anddirectly placed debt, Journal of Political Economy 99, 689\u2013721.Diamond, Douglas W, and Raghuram G Rajan, 2001, Liquidity risk, liquidity creation, and fi-nancial fragility: A theory of banking, Journal of Political Economy 109, 287\u2013327.Diamond, Douglas W, and Robert E Verrecchia, 1991, Disclosure, liquidity, and the cost of capi-tal, Journal of Finance 46, 1325\u20131359.Dye, Ronald A, 1986, Proprietary and nonproprietary disclosures, Journal of Business 331\u2013366.Easton, Peter D, and Mark E Zmijewski, 1989, Cross-sectional variation in the stock market re-sponse to accounting earnings announcements, Journal of Accounting and Economics 11,117\u2013141.Engel, Ellen, Rachel M Hayes, and Xue Wang, 2007, The Sarbanes\u2013Oxley Act and firms\u2019 going-private decisions, Journal of Accounting and Economics 44, 116\u2013145.Fern\u00e1ndez-Villaverde, Jes\u00fas, Pablo Guerr\u00f3n-Quintana, Keith Kuester, and Juan Rubio-Ram\u00edrez,2015, Fiscal volatility shocks and economic activity, American Economic Review 105, 3352\u201384.Fern\u00e1ndez-Villaverde, Jes\u00fas, Pablo Guerr\u00f3n-Quintana, Juan F Rubio-Ram\u00edrez, and MartinUribe, 2011, Risk matters: The real effects of volatility shocks, American Economic Review101, 2530\u201361.Flannery, Mark J, Simon H Kwan, and Mahendrarajah Nimalendran, 2013, The 2007\u20132009 fi-nancial crisis and bank opaqueness, Journal of Financial Intermediation 22, 55\u201384.Fraisse, Henri, Mathias L\u00e9, and David Thesmar, 2020, The real effects of bank capital require-ments, Management Science 66, 5\u201323.Gambacorta, Leonardo, and Paolo Emilio Mistrulli, 2004, Does bank capital affect lending be-havior?, Journal of Financial Intermediation 13, 436\u2013457.Garlappi, Lorenzo, Ron Giammarino, and Ali Lazrak, 2017, Ambiguity and the corporation:160Group disagreement and underinvestment, Journal of Financial Economics 125, 417\u2013433.Garlappi, Lorenzo, Ron Giammarino, and Ali Lazrak, 2021, Group-managed real options,Reviewof Financial Studies, forthcoming .Gigler, Frank, 1994, Self-enforcing voluntary disclosures, Journal of Accounting Research 32,224\u2013240.Gilchrist, Simon, Jae W Sim, and Egon Zakraj\u0161ek, 2014, Uncertainty, financial frictions, and in-vestment dynamics, National Bureau of Economic Research .Goldstein, Itay, and Haresh Sapra, 2014, Should banks\u2019 stress test results be disclosed? An anal-ysis of the costs and benefits, Foundations and Trends in Finance 32, 1\u201354.Gornall, Will, and Ilya A Strebulaev, 2018, Financing as a supply chain: The capital structure ofbanks and borrowers, Journal of Financial Economics 129, 510\u2013530.Gorton, Gary, and George Pennacchi, 1990, Financial intermediaries and liquidity creation,Journal of Finance 45, 49\u201371.Guay, Wayne, Delphine Samuels, and Daniel Taylor, 2016, Guiding through the fog: Finan-cial statement complexity and voluntary disclosure, Journal of Accounting and Economics62, 234\u2013269.Guiso, Luigi, and Giuseppe Parigi, 1999, Investment and demand uncertainty, Quarterly Journalof Economics 114, 185\u2013227.Gulen, Huseyin, and Mihai Ion, 2015, Policy uncertainty and corporate investment, Review ofFinancial Studies 29, 523\u2013564.Hao, Zhiwei, Ahmed Ameya Prapan, Konstantinos Gavriilidis, Dimitris Petmezas, and Evange-los Vagenas-Nanos, 2022, Does geopolitical risk affect acquisitions?, Working Paper .Harford, Jarrad, 2005, What drives merger waves?, Journal of Financial Economics 77, 529\u2013560.Hirshleifer, Jack, 1971, The private and social value of information and the reward, AmericanEconomic Review 61, 561\u2013574.Holmstr\u00f6m, Bengt, 2009, Discussion of G. Gorton, \u201cThe panic of 2007\" in maintaining stabil-ity in a changing financial system, Proceedings of the 2008 Jackson Hole Conference, FederalReserve Bank of Kansas City 39.Holmstr\u00f6m, Bengt, 2015, Understanding the role of debt in the financial system, BIS workingpaper No.479.Im, Hyun Joong, Heungju Park, and Gege Zhao, 2017, Uncertainty and the value of cash hold-ings, Economics Letters 155, 43\u201348.Ivashina, Victoria, David S Scharfstein, and Jeremy C Stein, 2015, Dollar funding and the lending161behavior of global banks, Quarterly Journal of Economics 130, 1241\u20131281.Jayaraman, Sudarshan, and Joanna Shuang Wu, 2019, Is silence golden? Real effects of manda-tory disclosure, Review of Financial Studies 32, 2225\u20132259.Jensen, Michael C, 1986, Agency costs of free cash flow, corporate finance, and takeovers, Amer-ican Economic Review 76, 323\u2013329.Jensen, Michael C, and William H Meckling, 1976, Theory of the firm: Managerial behavior,agency costs and ownership structure, Journal of Financial Economics 3, 305\u2013360.Jones, Jeffrey S, Wayne Y Lee, and Timothy J Yeager, 2012, Opaque banks, price discovery, andfinancial instability, Journal of Financial Intermediation 21, 383\u2013408.Kahle, Kathleen M, and Ren\u00e9 M Stulz, 2013, Access to capital, investment, and the financialcrisis, Journal of Financial Economics 110, 280\u2013299.Kaplan, Todd R, 2006, Why banks should keep secrets, Economic Theory 27, 341\u2013357.Kaviani, Mahsa S, Lawrence Kryzanowski, Hosein Maleki, and Pavel Savor, 2020, Policy uncer-tainty and corporate credit spreads, Journal of Financial Economics 138, 838\u2013865.Khwaja, Asim Ijaz, and Atif Mian, 2008, Tracing the impact of bank liquidity shocks: Evidencefrom an emerging market, American Economic Review 98, 1413\u201342.Kieschnick, Robert L, Jr, 1998, Free cash flow and stockholder gains in going private transactionsrevisited, Journal of Business Finance & Accounting 25, 187\u2013202.Kim, Hyunseob, and Howard Kung, 2017, The asset redeployability channel: How uncertaintyaffects corporate investment, Review of Financial Studies 30, 245\u2013280.Kim, Wi Saeng, and Esmeralda O Lyn, 1991, Going private: Corporate restructuring under in-formation asymmetry and agency problems, Journal of Business Finance & Accounting 18,637\u2013648.Kohers, Ninon, and James Ang, 2000, Earnouts in mergers: Agreeing to disagree and agreeing tostay, Journal of Business 73, 445\u2013476.La Porta, Rafael, Florencio Lopez-de Silanes, Andrei Shleifer, and Robert Vishny, 2002, Investorprotection and corporate valuation, Journal of Finance 57, 1147\u20131170.Leahy, John, and Toni M Whited, 1996, The effects of uncertainty on investment: Some stylizedfacts, Journal of Money, Credit, and Banking 28, 64\u201383.Lehn, Kenneth, and Annette Poulsen, 1989, Free cash flow and stockholder gains in going pri-vate transactions, Journal of Finance 44, 771\u2013787.Lemmon, Michael, and Michael R Roberts, 2010, The response of corporate financing and in-vestment to changes in the supply of credit, Journal of Financial and Quantitative Analysis162555\u2013587.Lemmon, Michael L, and Karl V Lins, 2003, Ownership structure, corporate governance, andfirm value: Evidence from the East Asian financial crisis, Journal of Finance 58, 1445\u20131468.Leuz, Christian, Alexander Triantis, and Tracy Yue Wang, 2008, Why do firms go dark? Causesand economic consequences of voluntary SEC deregistrations, Journal of Accounting andEconomics 45, 181\u2013208.Li, Yinghua, Yupeng Lin, and Liandong Zhang, 2018, Trade secrets law and corporate disclosure:Causal evidence on the proprietary cost hypothesis, Journal of Accounting Research 56, 265\u2013308.Martinez, Isabelle, and St\u00e9phanie Serve, 2011, The delisting decision: The case of buyout offerwith squeeze-out (BOSO), International Review of Law and Economics 31, 228\u2013239.Masulis, Ronald W, Cong Wang, and Fei Xie, 2009, Agency problems at dual-class companies,Journal of Finance 64, 1697\u20131727.Maupin, Rebekah J, Clinton M Bidwell, and Alan K Ortegren, 1984, An empirical investigation ofthe characteristics of publicly-quoted corporations which change to closely-held ownershipthrough management buyouts, Journal of Business Finance & Accounting 11, 435\u2013450.Mehran, Hamid, 1995, Executive compensation structure, ownership, and firm performance,Journal of Financial Economics 38, 163\u2013184.Mehran, Hamid, and Stavros Peristiani, 2009, Financial visibility and the decision to go private,Review of Financial Studies 23, 519\u2013547.Mian, Atif, and Amir Sufi, 2010, The great recession: Lessons from microeconomic data, Ameri-can Economic Review 100, 51\u201356.Mitchell, Mark L, and J Harold Mulherin, 1996, The impact of industry shocks on takeover andrestructuring activity, Journal of Financial Economics 41, 193\u2013229.Monnet, Cyril, and Erwan Quintin, 2017, Limited disclosure and hidden orders in asset markets,Journal of Financial Economics 123, 602\u2013616.Morgan, Donald P, 2002, Rating banks: Risk and uncertainty in an opaque industry, AmericanEconomic Review 92, 874\u2013888.Morris, Stephen, and Hyun Song Shin, 2002, Social value of public information, American Eco-nomic Review 92, 1521\u20131534.Murfin, Justin, 2012, The supply-side determinants of loan contract strictness, Journal of Fi-nance 67, 1565\u20131601.Nagar, Venky, Jordan Schoenfeld, and Laura Wellman, 2019, The effect of economic policy un-certainty on investor information asymmetry and management disclosures, Journal of Ac-163counting and Economics 67, 36\u201357.Nguyen, Nam H, and Hieu V Phan, 2017, Policy uncertainty and mergers and acquisitions, Jour-nal of Financial and Quantitative Analysis 52, 613\u2013644.Opler, Tim, and Sheridan Titman, 1993, The determinants of leveraged buyout activity: Freecash flow vs. financial distress costs, Journal of Finance 48, 1985\u20131999.Pagano, Marco, and Paolo Volpin, 2012, Securitization, transparency, and liquidity, Review ofFinancial Studies 25, 2417\u20132453.Paravisini, Daniel, 2008, Local bank financial constraints and firm access to external finance,Journal of Finance 63, 2161\u20132193.Pastor, Lubos, and Pietro Veronesi, 2012, Uncertainty about government policy and stockprices, Journal of Finance 67, 1219\u20131264.P\u00e1stor, L\u2019ubo\u0161, and Pietro Veronesi, 2013, Political uncertainty and risk premia, Journal of Fi-nancial Economics 110, 520\u2013545.Peek, Joe, and Eric S Rosengren, 1997, The international transmission of financial shocks: Thecase of Japan, American Economic Review 495\u2013505.Peek, Joe, and Eric S Rosengren, 2000, Collateral damage: Effects of the Japanese bank crisis onreal activity in the United States, American Economic Review 90, 30\u201345.Petersen, Mitchell A, and Raghuram G Rajan, 1994, The benefits of lending relationships: Evi-dence from small business data, Journal of Finance 49, 3\u201337.Rajan, Raghuram G, 1992, Insiders and outsiders: The choice between informed and arm\u2019s-length debt, Journal of Finance 47, 1367\u20131400.Rauh, Joshua D, and Amir Sufi, 2010, Capital structure and debt structure, Review of FinancialStudies 23, 4242\u20134280.Reuer, Jeffrey J, Oded Shenkar, and Roberto Ragozzino, 2004, Mitigating risk in internationalmergers and acquisitions: The role of contingent payouts, Journal of International BusinessStudies 35, 19\u201332.Rhodes-Kropf, Matthew, David T Robinson, and Sean Viswanathan, 2005, Valuation waves andmerger activity: The empirical evidence, Journal of Financial Economics 77, 561\u2013603.Schandlbauer, Alexander, 2017, How do financial institutions react to a tax increase?, Journal ofFinancial Intermediation 30, 86\u2013106.Schwert, Michael, 2020, Does borrowing from banks cost more than borrowing from the mar-ket?, Journal of Finance 75, 905\u2013947.Shah, Salman, and Anjan V Thakor, 1988, Private versus public ownership: Investment, owner-164ship distribution, and optimality, Journal of Finance 43, 41\u201359.Shleifer, Andrei, and Robert W Vishny, 1986, Large shareholders and corporate control, Journalof Political Economy 94, 461\u2013488.Shleifer, Andrei, and Robert W Vishny, 1997, A survey of corporate governance, Journal of Fi-nance 52, 737\u2013783.Shleifer, Andrei, and Robert W Vishny, 2003, Stock market driven acquisitions, Journal of Finan-cial Economics 70, 295\u2013311.Shumway, Tyler, 2001, Forecasting bankruptcy more accurately: A simple hazard model, Journalof Business 74, 101\u2013124.Stein, Luke CD, and Elizabeth Stone, 2013, The effect of uncertainty on investment, hiring, andR&D: Causal evidence from equity options, Working Paper .Subrahmanyam, Avanidhar, and Sheridan Titman, 1999, The going-public decision and the de-velopment of financial markets, Journal of Finance 54, 1045\u20131082.Thakor, Anjan V, 1996, Capital requirements, monetary policy, and aggregate bank lending:Theory and empirical evidence, Journal of Finance 51, 279\u2013324.Thomsen, Steen, and Frederik Vinten, 2014, Delistings and the costs of governance: a study ofeuropean stock exchanges 1996\u20132004, Journal of Management & Governance 18, 793\u2013833.Viarengo, Luca, Stefano Gatti, and Annalisa Prencipe, 2018, Enforcement quality and the useof earnouts in M&A transactions: International evidence, Journal of Business Finance & Ac-counting 45, 437\u2013481.Wolf, Daniel, and David Fox, 2012, Earnouts\u2013a siren song?\u2014the harvard law school forum oncorporate governance and financial regulation.Zheng, Yi, 2020, Does bank opacity affect lending?, Journal of Banking & Finance 119, 105900.Zingales, Luigi, 1995, Insider ownership and the decision to go public, Review of Economic Stud-ies 62, 425\u2013448.165AppendicesAppendix A. Appendix to Chapter 2A.1 Variable Definition\u2206 Volatility is the change in annualized stock return volatility (\u03c3i ,t \u2212\u03c3i ,t\u22121)\/( 12\u03c3i ,t + 12\u03c3i ,t\u22121) for firm i ata given year t.Analyst Coverage is the number of analysts following the company.Asset Redeployability is the value weighted asset redeployability index from Kim and Kung (2017).Dual Class is an indicator variable which equals to one if the firm has dual class structure.Financial Distress is an indicator variable which equals to one if the Altman Z-score if lower than 1.8.Altman Z-Score = 1.2*(working capital \/ total assets) + 1.4*(retained earnings \/ total assets) + 3.3*(earn-ings before interest and tax \/ total assets) + 0.6*(market value of equity \/ total liabilities) + 1.0*(sales \/total assets)GDPGrowth is the percent change of gross domestic product from FRED.Institutional Ownership is the percentage ownership by institutional blockholders from SEC 13F hold-ings.Intangible Assets is total intangible assets divided by total assets from Compustat.Leverage is total long term debt divided by total assets from Compustat.Loan to Bond Ratio is the ratio of bank loans to corporate bonds from Capital IQ.Log ERC is the log of earnings response coefficient. Earnings response coefficient is estimated as thecoefficient of regressing size-adjusted CAR around the three day window of the earning announcementon unexpected earnings. Unexpected earning is the actual earning per share minus the median earningforecast from I\/B\/E\/S database. ERC is estimated at SIC 3-digit industry level.LogRelative Tobin\u2019sQ is the log of firm Tobin\u2019s Q divided by industry Tobin\u2019s Q, which is the size-weightedaverage of Tobin\u2019s Q for each 3-digit SIC industry.Log Sales is the log of sales from Compustat.Recession is the recession indicators defined by NBER.Return on Assets is net income divided by total assets from Compustat.Sentiment is the investor sentiment sf1 measure from Baker and Wurgler (2006).Stock Return is the compounded return within a fiscal year, using CRSP daily dividend cumulative stockreturns.Tax Ratio is (federal income taxes + foreign income taxes - total interest and related expense + stateincome tax)\/ market capitalization from Crsp\/Compustat Merged Database.Term Premium is the yield spread between 10 years and 1 year Treasury bond.166Tobin\u2019sQ is (stock price * common shares used to calculate earnings per share + preferred stock\/liquidatingvalue + total long term debt + total debt in current liabilities - deferred taxes and investment tax credit)divided by total assets at the beginning of the fiscal year from Compustat.VIX is the CBOE volatility index from Bloomberg.Volatility is the standard deviation of daily dividend cumulative stock returns (from CRSP) within a fiscalyear, multiplied byp252.167A.2 Additional TablesTable A.2.1 Sample DescriptionPanel A. Sample Composition by IndustryThe table reports the industry distribution of the firms that filed for Sched-ule 13E-3 and delisted within two years after the filing from 1994 to2017. Firms in financial and utilities industries are excluded from the sam-ple. Industries are based on Fama-French twelve industry classifications fromhttp:\/\/mba.tuck.dartmouth.edu\/pages\/faculty\/ken.french\/Data_Library\/det_12_ind_port.html.No. Description No. of Going Private Firms Percentage1 Consumer Nondurables 77 8.232 Consumer Durables 27 2.893 Manufacturing 94 10.054 Energy 46 4.925 Chemicals 15 1.606 Business Equipment 190 20.327 Telecom 52 5.569 Shops 148 15.8310 Healthcare 53 5.6712 Other 233 24.92Total 935 100.00168Panel B. Sample Composition by YearThe table reports the time-series distribution of the firms that filed for schedule 13E-3and delisted within two years after filing from 1994 to 2017. Firms in financial and utilitiesindustries are excluded from the sample. Percentage indicates the number of going private firmsin that year out of the total number of going private firms.Year No. of Going Private Firms Percentage1994 4 0.431995 12 1.281996 27 2.891997 42 4.491998 51 5.451999 77 8.242000 54 5.782001 69 7.382002 53 5.672003 66 7.062004 51 5.452005 53 5.672006 43 4.602007 36 3.852008 24 2.572009 38 4.062010 39 4.172011 28 2.992012 24 2.572013 39 4.172014 20 2.142015 24 2.572016 44 4.712017 17 1.82Total 935 100.00169Table A.2.2 Cox Proportional Hazards Models for Time to Go Private: First StageResultsThis table reports first stage results of the Cox proportional hazards models for time to goprivate with control functions. Columns (1)-(4) correspond to the first stage results in Table 2.2columns (3)-(6) respectively. The sample includes going-private firms over the period of 1994-2017 and a group of control firms that remain public. The dependent variable is \u00a2Volatilityi,t\u00b01.Standard errors (in parentheses) are clustered at SIC 3-digit level. *, **, and *** indicatestatistical significance at the 10%, 5%, and 1% levels, respectively. Variables are defined inAppendix A.1.Cox Proportional Hazards Modelwith Control Function: First Stage(1) (2) (3) (4)\u00a2 Vol Exposure CADi,t\u00b01 1.11*** 1.32*** 1.30*** 1.14***(0.38) (0.37) (0.38) (0.35)\u00a2 Vol Exposure EURi,t\u00b01 1.84*** 1.69*** 1.72*** 1.56***(0.61) (0.59) (0.60) (0.39)\u00a2 Vol Exposure JPYi,t\u00b01 1.62*** 0.99* 1.02* 1.97***(0.61) (0.52) (0.52) (0.42)\u00a2 Vol Exposure AUDi,t\u00b01 4.41*** 3.39*** 3.22*** 2.08***(0.89) (0.90) (0.90) (0.47)\u00a2 Vol Exposure SEKi,t\u00b01 3.53*** 4.22*** 4.16*** 2.94***(0.46) (0.53) (0.53) (0.43)\u00a2 Vol Exposure CHFi,t\u00b01 3.85*** 3.51*** 3.54*** 1.98***(0.68) (0.59) (0.61) (0.37)\u00a2 Vol Exposure GBPi,t\u00b01 -0.05 0.43 0.57 0.86(0.89) (0.95) (0.96) (0.62)\u00a2 Vol Exposure Oili,t\u00b01 4.28*** 3.92*** 3.96*** 2.70***(0.28) (0.24) (0.24) (0.25)\u00a2 Vol Exposure Policyi,t\u00b01 415.78** 507.70** 532.83** 418.46***(176.01) (207.15) (208.78) (152.18)\u00a2 Vol Exposure Treasuryi,t\u00b01 57.65*** 61.83*** 62.69*** 36.88***(4.95) (5.46) (5.41) (4.69)Control Variables No Yes Yes Yes1st Moment 10 IVi,t\u00b01 No Yes Yes YesIndustry FE No No Yes YesYear FE No No No YesFirm-year Observations 33,711 26,034 26,034 26,034F statistic Cragg-Donald 287.4 209.7 210.8 75.5F statistic Kleibergen-Paap 130.4 102.6 104.3 33.3p-val Kleib.-P Underidentification Test 0.000 0.000 0.000 0.000p-val Sargan-H J Overidentification Test 0.412 0.342 0.307 0.495170Table A.2.3 Uncertainty Shocks and Going Private Transactions: Impacts of Macroe-conomic FactorsThis table reports results of the Cox proportional hazards models with control functions,controlling for macroeconomic factors. The sample includes going-private firms over the periodof 1994-2017 and a group of control firms that remain public. The dependent variable is thehazard rate of going private. In the Cox proportional hazards models, the firm-year observationsare treated as recurring censored events until the firm goes private or the end of the sampleperiod. Standard errors (in parentheses) are clustered at SIC 3-digit level and bootstrapped with300 replications. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels,respectively. Variables are defined in Appendix A.1.Cox Proportional Hazards Model with Control Function(1) (2) (3) (4) (5)\u00a2Volatilityi,t-1 1.63*** 1.40*** 1.82*** 1.29*** 1.40***(0.25) (0.25) (0.23) (0.26) (0.26)Volatilityi,t-2 1.17*** 0.73*** 1.04*** 0.78*** 0.77***(0.19) (0.18) (0.20) (0.21) (0.19)Stock Returni,t-1 -0.42*** -0.48*** -0.43*** -0.43*** -0.44***(0.10) (0.11) (0.10) (0.10) (0.10)GDP growtht-1 0.16***(0.05)Sentimentt-1 0.36***(0.10)Term Premiumt-1 -0.23**(0.09)VIXt-1 0.02**(0.01)Recessiont-1 -0.13(0.23)Control Variables Yes Yes Yes Yes Yes1st Moment 10 IVi,t\u00b01 Yes Yes Yes Yes YesIndystry FE Yes Yes Yes Yes YesYear FE No No No No NoFirm-year Observations 26,034 21,875 25,896 25,300 21,917No. of Firms 2,620 2,216 2,482 2,466 2,216No. of Going Private Firms 252 206 222 222 206Wald \u00ac2 394.9*** 325.0*** 332.8*** 363.4*** 307.1***171Table A.2.4 Cox Proportional Hazards Models for Time to Go Private: Instrumentswith Alternative Risk ModelsPanel A. First Stage ResultsThis table reports first stage results of the Cox proportional hazards models for time to go privatewith control functions, with the instruments constructed using alternative risk models. Thesample includes going-private firms over the period of 1994-2017 and a group of control firmsthat remain public. The dependent variable is \u00a2Volatilityi,t\u00b01. Standard errors (in parentheses)are clustered at SIC 3-digit level. *, **, and *** indicate statistical significance at the 10%, 5%,and 1% levels, respectively. Variables are defined in Appendix A.1.Cox Proportional Hazards Modelwith Control Function: First StageRawReturnCAPM FF3F FF5F(1) (2) (3) (4)\u00a2 Vol Exposure CADi,t\u00b01 0.78*** 0.68*** 1.36*** 1.37***(0.10) (0.20) (0.39) (0.37)\u00a2 Vol Exposure EURi,t\u00b01 1.77** 1.16*** 1.59*** 1.69***(0.69) (0.33) (0.31) (0.39)\u00a2 Vol Exposure JPYi,t\u00b01 1.19*** 1.37*** 1.60*** 1.86***(0.42) (0.27) (0.47) (0.42)\u00a2 Vol Exposure AUDi,t\u00b01 -0.07 1.76*** 1.83*** 1.96***(0.17) (0.42) (0.38) (0.39)\u00a2 Vol Exposure SEKi,t\u00b01 -0.44** 1.99*** 2.53*** 2.84***(0.17) (0.35) (0.43) (0.41)\u00a2 Vol Exposure CHFi,t\u00b01 0.88*** 1.94*** 1.45*** 1.70***(0.15) (0.30) (0.35) (0.32)\u00a2 Vol Exposure GBPi,t\u00b01 1.32 0.79 1.54*** 1.25*(1.48) (0.75) (0.56) (0.69)\u00a2 Vol Exposure Oili,t\u00b01 3.23*** 2.29*** 2.39*** 2.58***(0.58) (0.20) (0.24) (0.26)\u00a2 Vol Exposure Policyi,t\u00b01 1142.33*** 473.89* 421.46** 617.05***(359.57) (269.49) (183.15) (165.06)\u00a2 Vol Exposure Treasuryi,t\u00b01 12.24*** 29.62*** 37.15*** 33.86***(1.65) (2.12) (3.95) (4.81)Control variables Yes Yes Yes Yes1st Moment 10 IVi,t\u00b01 Yes Yes Yes YesIndustry FE Yes Yes Yes YesYear FE Yes Yes Yes YesFirm-year Observations 25,814 25,214 25,173 25,300F statistic Cragg-Donald 72.9 85.8 72.6 73.5F statistic Kleibergen-Paap 35.9 57.2 44.1 37.5p-val Kleib.-P Underidentification Test 0.000 0.000 0.000 0.000p-val Sargan-H J Overidentification Test 0.791 0.386 0.328 0.307172Panel B. Cox Proportional Hazards Models with Control FunctionsThis table reports results of the Cox proportional hazards models for time to go private,with the instruments constructed using alternative risk models. The sample includes going-private firms over the period of 1994-2017 and a group of control firms that remain public. Thedependent variable is the hazard rate of going private. In the Cox proportional hazards models,the firm-year observations are treated as recurring censored events until the firm goes privateor the end of the sample period. Standard errors (in parentheses) are clustered at SIC 3-digitlevel and bootstrapped with 300 replications. *, **, and *** indicate statistical significance atthe 10%, 5%, and 1% levels, respectively. Variables are defined in Appendix A.1.Cox Proportional Hazards Model with Control FunctionRaw Return CAPM FF3F FF5F(1) (2) (3) (4)\u00a2Volatilityi,t-1 1.07*** 1.06*** 0.93*** 1.13***(0.27) (0.25) (0.26) (0.24)Volatilityi,t-2 0.68*** 0.73*** 0.68*** 0.71***(0.22) (0.20) (0.19) (0.21)Stock Returni,t-1 -0.45*** -0.48*** -0.48*** -0.45***(0.11) (0.11) (0.11) (0.10)Control Variables Yes Yes Yes Yes1st Moment 10 IVi,t\u00b01 Yes Yes Yes YesIndystry FE Yes Yes Yes YesYear FE Yes Yes Yes YesFirm-year Observations 25,814 25,214 25,173 25,300No. of Firms 2,650 2,635 2,609 2,600No. of Going Private Firms 258 248 245 246Wald \u00ac2 2165.9*** 3354.3*** 3228.7*** 2750.9***173Table A.2.5 Cox Proportional Hazards Models for Time to Go Private: MatchingAnalysis on IPO CharacteristicsThis table reports results of the Cox proportional hazards models for time to go private,estimated using Equation (2.1). The sample includes going-private firms over the period of1994-2017 and control firms that matched on firm characteristics one year after IPO. Thedependent variable is the hazard rate of going private. In the Cox proportional hazards models,the firm-year observations are treated as recurring censored events until the firm goes private orthe end of the sample period. The control samples in columns (1)-(4) are matched on SIC 2-digitindustry, log sales, Tobin\u2019s Q, and stock return respectively. The control sample in column (5)is constructed with propensity score matching on Fama-French 12 industry, log sales, Tobin\u2019sQ and stock return. Standard errors (in parentheses) are clustered at SIC 3-digit level andbootstrapped with 300 replications. *, **, and *** indicate statistical significance at the 10%, 5%,and 1% levels, respectively. Variables are defined in Appendix A.1.Cox Proportional Hazards Model with Control FunctionSIC2 Log Sales Tobin\u2019s QStockReturnP-score(1) (2) (3) (4) (5)\u00a2Volatilityi,t-1 0.98*** 1.14*** 1.59*** 1.46** 0.75**(0.38) (0.38) (0.40) (0.67) (0.32)Volatilityi,t-2 0.72** 0.71*** 0.96*** 1.11* 0.46***(0.30) (0.24) (0.31) (0.65) (0.29)Stock Returni,t-1 -0.41*** -0.49*** -0.48*** -0.55*** -0.58***(0.14) (0.13) (0.14) (0.21) (0.14)Log Salesi,t-1 -0.10* -0.19*** -0.13*** -0.13* -0.04(0.06) (0.06) (0.05) (0.08) (0.05)Tobin\u2019s Qi,t-1 -0.19** -0.10 -0.22*** 0.00 -0.08(0.10) (0.07) (0.08) (0.10) (0.06)Taxi,t-1 4.45 4.38 3.68 3.80 4.20(3.29) (2.68) (2.73) (4.52) (3.10)Leveragei,t-1 0.08 0.26 0.10 1.80** -0.03(0.44) (0.43) (0.44) (0.70) (0.47)Return on Assetsi,t-1 0.60* 0.31 0.84** 0.47 0.25(0.35) (0.28) (0.36) (1.11) (0.32)Intangible Assetsi,t-1 0.19 0.49 0.35 1.07 0.64(0.70) (0.65) (0.63) (1.08) (0.65)Control Variables Yes Yes Yes Yes Yes1st Moment 10 IVi,t\u00b01 Yes Yes Yes Yes YesIndystry FE Yes Yes Yes Yes YesYear FE Yes Yes Yes Yes YesFirm-year Observations 13,151 17,127 17,993 6,149 10,132No. of Firms 1,271 1,715 1,651 536 958No. of Going Private Firms 171 198 174 71 171Wald \u00ac2 6436.2*** 1325.3*** 4030.8*** 88220.3*** 136504.5***174Appendix B. Appendix to Chapter 3B.1 Variable DefinitionLoan-Bond Pair CharacteristicsLoan-Bond Spread All-in-drawn loan spread minus bond spread (Dealscan andFISD).Total Borrowing The sum of loan facility amount and bond face value ($MM)(Dealscan and FISD).Log Total Borrowing The logarithm of the sum of loan facility amount and bond facevalue (Dealscan and FISD).\u00a2Maturity Loan maturity minus bond maturity in years (Dealscan andFISD).Loan Share The loan facility amount divided by the sum of loan facilityamount and bond face value (Dealscan and FISD).Information CostUncertainty Shock (\u00a2\u00e6t\u00b01) The difference between annualized stock return volatility esti-mated 5 quarters and 1 quarter before the loan origination date(CRSP).Opacity Index Sum of quintiles based on bid-ask spread, trading volume, an-alyst coverage, and analyst forecast errors, normalized by 20.(CRSP and IBES)Rating Gap The absolute rating gap between bond ratings at issuance byStandard & Poor\u2019s and Moody\u2019s (FISD).Rating Disagreement A dummy variable that equals to one if the rating gap is greateror equal to two (FISD).Firm CharacteristicsVolatility (\u00e6t\u00b05) Annualized stock return volatility, estimated 5 quarters beforethe loan origination date (CRSP).Stock Return (rt\u00b01) Annual stock return, estimated 1 quarter before the loan origi-nation date (CRSP).Total Assets Asset size ($B) (Compustat).Profitability Ratio of operating income before depreciation to book assets(Compustat).Implied Prob. Default Implied probability of default from Bharath and Shumway(2008).Asset Market-to-book Ratio of quasi-market assets to book assets. (Compustat)Quasi-market Leverage Ratio of book debt to quasi-market assets. (Compustat)Loan CharacteristicsFacility Amount Loan facility amount ($MM) (Dealscan).Log Facility Amount The logarithm of the loan facility amount (Dealscan).All-in-drawn Spread The all-in-drawn spread of loan facilities (Dealscan).Syndicated Loan An indicator variable that equals one if a loan is a syndicatedloan (Dealscan).Term Loan An indicator variable that equals one if a loan is a term loan(Dealscan).Secured Loan An indicator variable that equals one if a loan is a secured loan(Dealscan).175Bond CharacteristicsFace Value Face amount of a bond ($MM) (FISD).Bond Spread Yield to maturity of the bond at issuance (bps) (FISD).Bond Rating Average bond rating at issuance by Moody\u2019s and S&P (FISD)Secured Bond An indicator variable that equals one if a bond is a secured bond(FISD).Redeemable Bond An indicator variable that equals one if a bond is a redeemablebond (FISD).Embedded Investor Option An indicator variable that equals one if a bond is putable, con-vertible or exchangeable (FISD).Bondholder protective Covenant An indicator variable that equals one if a bond has cross defaultor cross acceleration covenants (FISD).Negative Pledge Covenant An indicator variable that equals one if a bond has negativepledge covenants (FISD).Bank CharacteristicsUdep Ratio of uninsured deposits to total assets of the bank holdingcompany, estimated five quarters before the loan origination date(FR Y-9C).Outflow A dummy variable which equals to one if the change in uninsureddeposits of the bank is in the bottom 5 percentile of the sample,estimated 1 quarter before the loan origination date (FR Y-9C).European Bank The fraction of European banks among lead banks in a loan syn-dicate.Non-GSIBs The fraction of non-global systemically important banks amonglead banks in the loan syndicate.176B.2 Instrument Variables ConstructionWe construct the instruments for firm uncertainty shocks following Alfaro, Bloom, and Lin (2019)on a quarterly basis. The instruments are constructed exploiting firms\u2019 differential exposures tovolatility shocks of multiple aggregate variables. For each aggregate variable c, we construct aninstrumentIV ck,t =\u00d8\u00d8\u00d8\u00d8ck,t\u00b02\u00d8\u00d8\u00d8 \u00b7\u00a2\u00e6ct , (B.2.1)where c is crude oil, three currencies (Canadian Dollar, British Pound and Australian Dollar),10-year U.S. Treasury note, or economic policy uncertainty. \u00d8ck,t\u00b02 is the sensitivity of stock re-turns to changes in these aggregate variables estimated at for each 3-digit SIC industry k. \u00a2\u00e6ctis aggregate volatility shock for quantity c, which is measured using year-on-year change in theannualized standard deviation of daily price changes of c, or year-on-year change in the averageannual daily implied volatility of c. We use changes in both realized volatility and implied volatil-ity to capture volatility shocks based on past events as well as expected shocks in the future. Foreconomic policy uncertainty, \u00a2\u00e6ct is the year-on-year change in the 365-day average of economicpolicy uncertainty. The idea behind the instruments is that when there is a volatility shock toaggregate quantity c, firms with different levels of exposure to c, captured by sensitivities \u00d8ck,t\u00b02,experience different opacity shocks.Sensitivities \u00d8ck,t\u00b02 are estimated using:rrisk_ad ji,t =\u00c6k,t+Xc\u00d8ck,t\u00b02rct +\u2264i,t, (B.2.2)where rrisk_ad ji,t is firm i\u2019s daily risk adjusted stock return and rct is the daily price change of c.We estimate the sensitivities for each 3-digit SIC industry k using a 10-year rolling window. Theestimated sensitivities are weight adjusted by their statistical significance levels. The adjustedsensitivities are lagged by two years to ensure that they pre-date the opacity shocks, both inthe aggregate and firm level. The sensitivities are unlikely to be correlated with firm specificcharacteristics two years from now. Aggregate volatility shocks,\u00a2\u00e6ct , are also unlikely to be drivenby firm characteristics. For this reason, the instruments, by construction, do not correlate withany unobservable firm characteristics.We include the aggregate first moment effects, Aggck,t to control for the direct impact ofchanges in aggregate quantities on firm opacity. For each aggregate variable of c, Aggck,t =\u00d8ck,t\u00b02 \u00b7 rct , where rct is the average annual return of c, and \u00d8ck,t\u00b02 is the weighted sensitivity ofc estimated in Equation (B.2.2). Controlling for Aggck,t allows us to isolate the second momenteffects of aggregate volatility shocks on firm opacity, from the first moment effects of changes inlevels of aggregate quantities on firm opacity.177B.3 Additional TablesTable B.3.1 Sample Construction ProcessThe table illustrates our sample construction process. The starting point is the intersec-tion of DealScan and Compustat from 1995 to 2019 with the loan denominated in U.S. dollars bynon-financial U.S. public firms. The sample of loans is restricted to senior loans with non-missingall-in-drawn spread. Each loan is paired with the closest senior bond issued by the same firmwithin 60 days. We restrict the loan-bond pairs to those with investment grade bond rating. Wefurther restrict the loan-bond pairs to those with the same maturity category, i.e. short-term,mid-term or long-term in maturity.Process No. of ObservationsSenior USD denominated loan facilities by non-financial U.S. public firms 67,967Senior USD denominated bond issuances by non-financial U.S. public firms 15,422Loan-bond pairs issued by the same firm within 60 days 8,686Keeping the bond with closest starting date for each loan facility 7,825Less loan-bond pairs by non-investment bond grading 4,015Less loan-bond pairs with different maturity 2,379Less loan-bond pairs with missing data 1,597178Table B.3.2 First Stage of the IV EstimationThe table presents the first stage regression results of the IV estimation. The first stageresults are estimated as follows: \u00a2\u00e6i,t\u00b01 = \u00d81 + \u00d82IV ck,t\u00b01 + \u00d83\u00e6i,t\u00b05 + \u00d84ri,t\u00b01 + \u00d85Bondi,t +\u00d86Loani,t+\u00d87Aggck,t+\u00a1 j +\u221at+ \u2264i, j,t. \u00a2\u00e6i,t\u00b01 is the year-on-year change in the annualized stockreturn volatility lagged by one quarter before loan origination. We instrument firm opacity shockusing aggregate volatility shocks to currency, energy, policy and U.S. Treasury notes togetherwith firms\u2019 exposures to these aggregate volatility shocks. Instruments in columns (1)-(3)are constructed using aggregate realized volatility shocks, and instruments in columns (4)-(6)are constructed aggregate implied volatility shocks. Columns (1) to (6) report the first stageregression results for the corresponding second stage results in columns (1) to (6) of Table 3. Thestandard errors (in parentheses) are clustered at firm level. *, **, and *** denote significance atthe 10%, 5%, and 1% levels, respectively.Realized Implied(1) (2) (3) (4) (5) (6)Exposure \u00a2Vol Cad 4.85** 5.27** 4.27* 4.97 9.55** 6.62(2.37) (2.29) (2.35) (4.86) (4.73) (5.22)Exposure \u00a2Vol Gbp 7.34** 15.21*** 7.62* 14.34 19.12* 21.09**(3.31) (5.09) (4.23) (8.89) (10.65) (9.00)Exposure \u00a2Vol Aud 3.98 3.43 4.46 0.61 -1.77 0.72(2.50) (2.16) (2.84) (4.23) (3.39) (4.10)Exposure \u00a2Vol Tbill 33.86 32.55 33.90 342.75*** 349.40*** 316.14***(35.60) (32.57) (30.99) (132.17) (113.99) (122.07)Exposure \u00a2Vol Oil 0.87 0.74 0.47 2.49*** 2.49*** 2.07**(0.79) (0.75) (0.68) (0.89) (0.91) (0.86)Exposure \u00a2Vol EPU870.54*** 1051.22*** 684.92** 1868.49*** 1767.07*** 1407.31**(309.67) (294.08) (331.03) (662.98) (631.90) (711.05)Observations 797 750 794 515 482 513179Table B.3.3 Information Cost and the Loan-Bond Spread: Alternative SamplesThe table reports results from the OLS regressions of the loan-bond spread on firm informationcost with alternative samples. The dependent variable is the difference between loan rateand bond yield. Uncertainty shock is the year-on-year change in the annualized stock returnvolatility lagged by one quarter before loan origination. All columns include year by quarterfixed effects. In Panel A-E, column (2) and (5) include bank holding company fixed effects, andcolumn (3) and (6) include lender fixed effects. See Appendix B.1 for variable definitions. Thestandard errors (in parentheses) are clustered at firm level. *, **, and *** denote significance atthe 10%, 5%, and 1% levels, respectively.Panel A. Excluding the Financial CrisisThe sample includes loan facility and investment-grade bond pairs issued by U.S. non-financialpublic firms within 60 days from 1995 to 2019, excluding 2007-09.(1) (2) (3) (4) (5) (6)Uncertainty Shock -0.58** -0.59** -0.50** -0.80*** -0.69** -0.74***(0.29) (0.24) (0.21) (0.30) (0.27) (0.24)Volatility -2.36*** -1.66*** -1.74*** -2.67*** -2.28*** -2.51***(0.69) (0.57) (0.54) (0.62) (0.60) (0.57)Stock Return 0.49** 0.33** 0.39** 0.67*** 0.53*** 0.38**(0.22) (0.16) (0.16) (0.24) (0.19) (0.17)Control Variables No No No Yes Yes YesYear\u00a3Quarter FE Yes Yes Yes Yes Yes YesBank Holding Company FE No Yes No No Yes NoLender FE No No Yes No No YesAdjusted R2 0.35 0.41 0.51 0.62 0.66 0.73Observations 1,500 1,243 1,491 901 778 898180Panel B. Loan-Bond Pairs within 30 DaysThe sample includes loan facility and investment-grade bond pairs issued by U.S. non-financialpublic firms within 30 days from 1995 to 2019.(1) (2) (3) (4) (5) (6)Uncertainty Shock -0.62 -0.81*** -0.70*** -0.68 -0.90*** -0.69**(0.44) (0.29) (0.26) (0.43) (0.34) (0.32)Volatility -2.76*** -2.42*** -2.34*** -2.88*** -2.80*** -2.85***(0.85) (0.60) (0.60) (0.78) (0.65) (0.60)Stock Return 0.73*** 0.54*** 0.55*** 0.67** 0.30 0.43**(0.26) (0.21) (0.18) (0.30) (0.19) (0.20)Control Variables No No No Yes Yes YesYear\u00a3Quarter FE Yes Yes Yes Yes Yes YesBank Holding Company FE No Yes No No Yes NoLender FE No No Yes No No YesAdjusted R2 0.37 0.45 0.57 0.62 0.73 0.77Observations 1,042 876 1,038 647 559 647Panel C. Loan-bond Pairs within 10 DaysThe sample includes loan facility and investment-grade bond pairs issued by U.S. non-financialpublic firms within 10 days from 1995 to 2019.(1) (2) (3) (4) (5) (6)Uncertainty Shock -0.85* -0.73 -0.51 -2.61*** -1.47** -1.64**(0.51) (0.50) (0.42) (0.90) (0.68) (0.78)Volatility -2.38** -2.74*** -2.00** -2.55 -4.28*** -5.90***(1.18) (0.99) (0.91) (2.02) (1.41) (2.20)Stock Return -0.12 0.25 -0.24 -0.60 -0.97* -0.71(0.44) (0.48) (0.36) (0.98) (0.54) (0.59)Control Variables No No No Yes Yes YesYear\u00a3Quarter FE Yes Yes Yes Yes Yes YesBank Holding Company FE No Yes No No Yes NoLender FE No No Yes No No YesAdjusted R2 0.40 0.52 0.63 0.53 0.70 0.75Observations 395 324 391 241 219 241181Panel D. Loan-bond Pairs without Matching on MaturityThe sample includes loan facility and investment-grade bond pairs issued by U.S. non-financialpublic firms with no restriction on maturity from 1995 to 2019.(1) (2) (3) (4) (5) (6)Uncertainty Shock -0.89*** -0.95*** -0.80*** -0.56*** -0.50** -0.53**(0.18) (0.20) (0.18) (0.21) (0.24) (0.21)Volatility -2.18*** -2.03*** -2.16*** -0.48 -0.40 -0.45(0.31) (0.32) (0.30) (0.35) (0.39) (0.38)Stock Return 0.36*** 0.36*** 0.41*** 0.47*** 0.48*** 0.49***(0.09) (0.11) (0.10) (0.10) (0.12) (0.11)Control Variables No No No Yes Yes YesYear\u00a3Quarter FE Yes Yes Yes Yes Yes YesBank Holding Company FE No Yes No No Yes NoLender FE No No Yes No No YesAdjusted R2 0.31 0.34 0.38 0.48 0.49 0.54Observations 5,522 4,638 5,500 3,107 2,697 3,091Panel E. Loan-bond Pairs Matched on Effective MaturityThe sample includes loan facility and investment-grade bond pairs issued by U.S. non-financialpublic firms matched on effective maturity from 1995 to 2019.(1) (2) (3) (4) (5) (6)Uncertainty Shock -0.61** -0.63*** -0.51** -0.77** -0.71** -0.73***(0.27) (0.23) (0.21) (0.31) (0.28) (0.25)Volatility -2.29*** -1.57*** -1.59*** -2.28*** -1.91*** -2.30***(0.66) (0.53) (0.52) (0.65) (0.65) (0.62)Stock Return 0.53*** 0.42** 0.47*** 0.66*** 0.54** 0.44**(0.20) (0.17) (0.16) (0.25) (0.21) (0.18)Control Variables No No No Yes Yes YesYear\u00a3Quarter FE Yes Yes Yes Yes Yes YesBank Holding Company FE No Yes No No Yes NoLender FE No No Yes No No YesAdjusted R2 0.36 0.43 0.52 0.54 0.57 0.65Observations 1,599 1,339 1,590 956 830 953182Panel F. Firm Information Cost and the Loan-Bond Spread: Alternative Samples withAlternative MeasuresThe table reports results from the OLS regression of the loan-bond spread on firm infor-mation shock with alternative samples, estimated using alternative measures. The sampleincludes loan facilities and investment-grade bond pairs issued by U.S. non-financial publicfirms from 1995 to 2019: a. excluding the financial crisis; b. loan-bond pairs issued within30 days; c. loan-bond pairs with no restriction on maturity; d. loan-bond pairs matched oneffective maturity. The dependent variable is the difference between loan rate and bond yield foreach matched loan-bond pair in the sample. Opacity index is constructed following Anderson etal.(2009). Rating gap is the absolute rating gap between bond ratings by Standard & Poor\u2019s andMoody\u2019s. Rating disagreement is a dummy variable that equals to one if the rating gap is greateror equal to two. Column (1), (3) and (5) include bank holding company fixed effects. Column (2),(4) and (6) include lender fixed effects. Control variables are included in all columns.a. Excluding the Financial Crisis 2007-09Opacity Index Rating Gap Rating Disagreement(1) (2) (3) (4) (5) (6)Information Cost -0.82*** -0.79*** -1.71*** -1.45*** -0.16* -0.17*(0.29) (0.27) (0.46) (0.45) (0.09) (0.09)Adjusted R2 0.68 0.74 0.69 0.74 0.68 0.74Observations 838 960 823 945 675 736b. Loan-bond Pairs within 30 Days(1) (2) (3) (4) (5) (6)Information Cost -0.96*** -0.85*** -1.06** -1.04* -0.21* -0.19*(0.32) (0.33) (0.51) (0.54) (0.11) (0.11)Adjusted R2 0.68 0.73 0.68 0.73 0.68 0.73Observations 838 960 823 945 675 736c. Loan-bond Pairs without Matching on Maturity(1) (2) (3) (4) (5) (6)Information Cost -0.96*** -0.85*** -1.06** -1.04* -0.21* -0.19*(0.32) (0.33) (0.51) (0.54) (0.11) (0.11)Adjusted R2 0.68 0.73 0.68 0.73 0.68 0.73Observations 838 960 823 945 675 736d. Loan-bond Pairs Matched on Effective Maturity(1) (2) (3) (4) (5) (6)Information Cost -0.96*** -0.85*** -1.06** -1.04* -0.21* -0.19*(0.32) (0.33) (0.51) (0.54) (0.11) (0.11)Adjusted R2 0.68 0.73 0.68 0.73 0.68 0.73Observations 838 960 823 945 675 736Control Variables Yes Yes Yes Yes Yes YesYear\u00a3Quarter FE Yes Yes Yes Yes Yes YesBank Holding Company FE Yes No Yes No Yes NoLender FE No Yes No Yes No Yes183Table B.3.4 Firm Information Cost and the Loan-Bond Spread: Alternative ControlVariablesThe table reports the robustness results from the OLS regression of the loan-bond spread onfirm information shock, with alternative control variables. The sample includes loan facility andinvestment-grade bond pairs issued by U.S. non-financial public firms within 60 days from 1995to 2019. The dependent variable is the difference between loan rate and bond yield. All columnsinclude year by quarter fixed effects. See Appendix B.1 for variable definitions. The standarderrors (in parentheses) are clustered at firm level. *, **, and *** denote significance at the 10%,5%, and 1% levels, respectively.Panel A. Controlling for Additional Bond and Firm CharacteristicsThe table reports results from the OLS regression of the loan-bond spread on firm uncertaintyshock controlling for a variety of firm and contract characteristics that can affect the loan-bondspread. Uncertainty Shock is the year-on-year change in the annualized stock return volatilitylagged by one quarter before loan origination. Additional variables (quasi-market assets, assettangibility, total number of banks in a loan syndicate, indicators for bonds with asset salerestriction, credit enhancement, or tender offer) are included in the regressions (results notreported). Column (2) and (5) include bank holding company fixed effects. Column (3) and (6)include lender fixed effects.(1) (2) (3) (4) (5) (6)Uncertainty Shock -0.62** -0.63*** -0.52** -1.08*** -0.67** -0.72***(0.27) (0.23) (0.20) (0.29) (0.26) (0.24)Volatility -2.29*** -1.57*** -1.60*** -3.80*** -3.34*** -3.51***(0.66) (0.53) (0.52) (0.72) (0.59) (0.58)Stock Return 0.53*** 0.42** 0.47*** 0.58* 0.60** 0.77***(0.20) (0.17) (0.16) (0.30) (0.27) (0.25)Quasi-market Leverage -1.49** -1.26* -1.37**(0.66) (0.64) (0.65)Implied Prob. Default -5.46 -5.60** -5.62**(4.17) (2.62) (2.67)Asset Market-to-book -0.31*** -0.18* -0.19**(0.09) (0.09) (0.09)Profitability -0.13 -4.68 -4.31(3.18) (3.33) (3.26)Log Borrowing Amount 10.07*** 9.58*** 9.03***(2.08) (2.10) (2.11)\u00a2Maturity (years) 0.08*** 0.08*** 0.08***(0.00) (0.00) (0.00)Secured Loan 0.58*** 0.23 0.30*(0.20) (0.22) (0.18)Bond Rating 0.23*** 0.15*** 0.13***(0.05) (0.05) (0.04)Redeemable Bond -1.88*** -0.71 -0.70(0.52) (0.62) (0.67)Secured Bond -0.36 -0.26 -0.06(0.28) (0.27) (0.24)Putable Bond 5.84*** 5.60*** 5.61***(0.64) (0.54) (0.61)Cross Acceleration Covenant -0.15 -0.19* -0.09(0.13) (0.11) (0.11)184Panel A. Continued(1) (2) (3) (4) (5) (6)Year\u00a3Quarter FE Yes Yes Yes Yes Yes YesBank Holding Company FE No Yes No No Yes NoLender FE No No Yes No No YesAdjusted R2 0.36 0.43 0.52 0.73 0.74 0.80Observations 1,597 1,338 1,588 529 435 527185Panel B. Excluding Stock Return as a Control VariableThe table reports results from the OLS regression of the loan-bond spread on firm uncer-tainty cost, excluding stock return as a control variable.(1) (2) (3) (4) (5) (6)Uncertainty Shock -0.67** -0.65*** -0.56*** -0.85*** -0.71*** -0.78***(0.26) (0.22) (0.20) (0.31) (0.25) (0.24)Volatility -2.31*** -1.55*** -1.61*** -2.62*** -2.12*** -2.28***(0.66) (0.54) (0.52) (0.60) (0.58) (0.57)Control Variables No No No Yes Yes YesYear\u00a3Quarter FE Yes Yes Yes Yes Yes YesBank Holding Company FE No Yes No No Yes NoLender FE No No Yes No No YesAdjusted R2 0.36 0.42 0.51 0.62 0.67 0.73Observations 1,597 1,338 1,588 963 838 960Panel C. Alternative Measures excluding Stock Return Volatility as a Control VariableThe table reports results from the OLS regression of the loan-bond spread on firm infor-mation cost, estimated using alternative measures without controlling for stock return volatility.Opacity index is constructed following Anderson et al.(2009). Rating gap is the absolute ratinggap between bond ratings by Standard & Poor\u2019s and Moody\u2019s. Rating disagreement is a dummyvariable that equals to one if the rating gap is greater or equal to two. Column (1), (3) and (5)include bank holding company fixed effects. Column (2), (4) and (6) include lender fixed effects.Control variables are included in all columns.Opacity Index Rating Gap Rating Disagreement(1) (2) (3) (4) (5) (6)Information Cost -1.49*** -1.31*** -0.11 -0.16** -0.29 -0.40**(0.35) (0.34) (0.07) (0.06) (0.19) (0.16)Stock Return 0.62*** 0.50*** 0.49** 0.36* 0.50** 0.36*(0.21) (0.17) (0.23) (0.20) (0.23) (0.20)Control Variables Yes Yes Yes Yes Yes YesYear\u00a3Quarter FE Yes Yes Yes Yes Yes YesBank Holding Company FE Yes No Yes No Yes NoLender FE No Yes No Yes No YesAdjusted R2 0.67 0.73 0.66 0.74 0.66 0.74Observations 829 959 681 750 681 750186Table B.3.5 Uninsured Deposits Outflow and the Loan-Bond SpreadThe table presents results of the robustness tests of the money creation mechanism usingheterogeneity in banks\u2019 reliance on uninsured deposits. Panel A and Panel B include log asset ofthe bank holding company or the borrower as an additional control variable. Panel C excludesstock return volatility as a control variable. The sample includes matched loan facilities andinvestment-grade bond pairs issued by U.S. non-financial public firms within 60 days from 1995to 2019. The dependent variable is the difference between loan rate and bond yield for eachmatched loan-bond pair in the sample. Uncertainty shock is the year-on-year change in theannualized stock return volatility lagged by one quarter before the loan origination. Opacityindex is constructed following Anderson et al. (2009). Rating gap is the absolute rating gapbetween bond ratings by Standard & Poor\u2019s and Moody\u2019s. Outflow is a dummy variable whichequals to one if the bank experiences large uninsured deposits outflow in the past year (banks inthe bottom 5 percentile of the sample in terms of changes in uninsured deposits, representing adecrease in uninsured deposits of more than 25%), lagged by one quarter before loan origination.Udep is the ratio of uninsured deposits to total assets of the bank holding company before theuninsured deposits outflow. All columns include year by quarter fixed effects. Column (1), (3)and (5) include bank holding company fixed effects. Column (2), (4) and (6) include lender fixedeffects. Control variables are included in all columns. See Appendix B.1 for variable definitions.The standard errors (in parentheses) are clustered at firm level. *, **, and *** denote significanceat the 10%, 5%, and 1% levels, respectively.Panel A. Controlling for Bank SizeUncertainty Shock Opacity Index Rating Gap(1) (2) (3) (4) (5) (6)Info. Cost \u00a3 Udep \u00a3 Outflow -2.78* -2.46** -3.77** -3.88** -1.34*** -1.05***(1.46) (1.23) (1.83) (1.58) (0.28) (0.24)Info. Cost -0.92*** -0.95*** -1.57*** -1.55*** -0.20*** -0.16***(0.30) (0.29) (0.46) (0.48) (0.06) (0.06)Info. Cost \u00a3 Udep -0.40 -0.22 2.16** 2.44*** 1.64*** 1.32***(0.48) (0.33) (0.91) (0.81) (0.28) (0.30)Info. Cost \u00a3 Outflow -0.07 0.13 0.86 1.03 -0.52 -0.77*(1.23) (1.18) (1.66) (1.57) (0.38) (0.41)Udep 0.54* 0.59** 0.83** 0.94*** 0.40* 0.40*(0.28) (0.26) (0.36) (0.33) (0.23) (0.24)Outflow 1.52* 2.27*** -0.48 -0.17 2.49** 4.26***(0.91) (0.69) (1.18) (1.08) (1.22) (1.17)Udep \u00a3 Outflow -0.51 -0.41 1.92*** 2.09*** 1.36*** 1.33***(0.34) (0.26) (0.72) (0.56) (0.23) (0.22)Volatility -2.33*** -2.81*** -0.65 -1.12* -2.03*** -2.25***(0.71) (0.66) (0.69) (0.62) (0.71) (0.76)Stock Return 0.72*** 0.75*** 0.78*** 0.88*** 0.85*** 0.87***(0.25) (0.24) (0.26) (0.24) (0.27) (0.27)Log Assets (Bank) -0.62 -0.22 -0.66 -0.06 -0.18 -0.34(0.41) (0.46) (0.42) (0.42) (0.38) (0.42)Control Variables Yes Yes Yes Yes Yes YesYear\u00a3Quarter FE Yes Yes Yes Yes Yes YesBank Holding Company FE Yes No Yes No Yes NoLender FE No Yes No Yes No YesAdjusted R2 0.63 0.65 0.64 0.65 0.64 0.65Observations 626 626 617 617 503 503187Panel B. Controlling for Firm SizeUncertainty Shock Opacity Index Rating Gap(1) (2) (3) (4) (5) (6)Info. Cost \u00a3 Udep\u00a3 Outflow -2.98** -2.77*** -3.48*** -3.57*** -1.24*** -1.18***(1.26) (0.99) (1.21) (0.96) (0.27) (0.20)Info. Cost -0.86*** -0.98*** -1.30*** -1.38*** -0.20*** -0.17***(0.25) (0.25) (0.40) (0.45) (0.06) (0.06)Info. Cost \u00a3 Udep -0.37* -0.65*** -0.73** -0.64** 0.10 0.21***(0.22) (0.21) (0.35) (0.32) (0.06) (0.08)Info. Cost \u00a3 Outflow -0.13 0.62 0.46 1.35 -0.33 -0.24(0.98) (0.88) (1.19) (1.21) (0.29) (0.32)Udep 0.27 0.41** 0.71** 0.82*** 0.23* 0.20(0.18) (0.18) (0.30) (0.28) (0.14) (0.15)Outflow 1.32** 2.19*** -0.22 -0.39 1.74* 2.60***(0.56) (0.45) (0.86) (0.88) (0.99) (0.93)Udep \u00a3 Outflow -0.59 -0.53* 1.98*** 2.14*** 1.44*** 1.39***(0.36) (0.28) (0.72) (0.56) (0.24) (0.21)Volatility -2.01*** -2.32*** -0.60 -0.94* -1.65*** -2.03***(0.55) (0.55) (0.57) (0.56) (0.56) (0.58)Stock Return 0.66*** 0.55*** 0.67*** 0.57*** 0.61*** 0.56***(0.19) (0.19) (0.18) (0.18) (0.18) (0.19)Log Assets (Borrower) 0.06 0.08 0.07 0.09 0.15** 0.15**(0.08) (0.07) (0.08) (0.08) (0.07) (0.06)Control Variables Yes Yes Yes Yes Yes YesYear\u00a3Quarter FE Yes Yes Yes Yes Yes YesBank Holding Company FE Yes No Yes No Yes NoLender FE No Yes No Yes No YesAdjusted R2 0.65 0.68 0.65 0.67 0.65 0.67Observations 759 760 748 749 606 606188Panel C. Excluding Stock Return Volatility as a Control VariableUncertainty Shock Opacity Index Rating Gap(1) (2) (3) (4) (5) (6)Info. Cost \u00a3 Udep \u00a3 Outflow -4.07*** -3.80*** -3.56*** -3.72*** -1.35*** -1.35***(1.36) (1.18) (1.18) (1.05) (0.23) (0.20)Info. Cost -0.51** -0.59** -1.48*** -1.63*** -0.19*** -0.16**(0.23) (0.24) (0.36) (0.40) (0.06) (0.06)Info. Cost \u00a3 Udep -0.37* -0.67*** -0.74** -0.64** 0.09 0.18**(0.21) (0.21) (0.35) (0.32) (0.06) (0.07)Info. Cost \u00a3 Outflow -1.14 -0.51 0.49 1.38 -0.53* -0.49(1.05) (0.96) (1.20) (1.22) (0.29) (0.34)Udep 0.30* 0.45*** 0.72** 0.82*** 0.27** 0.25*(0.16) (0.16) (0.28) (0.26) (0.13) (0.15)Outflow 1.27** 2.07*** -0.26 -0.50 2.15** 2.94***(0.55) (0.47) (0.86) (0.84) (1.01) (1.09)Udep \u00a3 Outflow 0.32** 0.45*** 0.74*** 0.84*** 0.32** 0.28*(0.15) (0.16) (0.28) (0.27) (0.14) (0.16)Stock Return 0.66*** 0.54*** 0.63*** 0.53*** 0.54*** 0.51***(0.19) (0.19) (0.18) (0.18) (0.19) (0.19)Control Variables Yes Yes Yes Yes Yes YesYear\u00a3Quarter FE Yes Yes Yes Yes Yes YesBank Holding Company FE Yes No Yes No Yes NoLender FE No Yes No Yes No YesAdjusted R2 0.64 0.67 0.65 0.67 0.64 0.66Observations 760 761 752 753 610 610189Appendix C. Appendix to Chapter 4C.1 Variable DefinitionDeal CharacteristicsEarnout Usage An indicator variable that equals one if an earnout agree-ment is employed in the M&A transaction (ThomsonReuters SDC).Earnout Pct Ratio of earnout value to deal value (%) (Thomson ReutersSDC).Earnout Value The logarithm of the value of the earnout in an M&Atransaction ($MM) (Thomson Reuters SDC).CAR [-2,+2] Acquirer cumulative abnormal returns (CARs) estimatedin the 5-day event window centered around the M&A dealannouncement date. (%) (CRSP).Deal Completion An indicator variable that equals one if a deal is com-pleted, and zero if a deal is withdrawn (Thomson ReutersSDC).Target CharacteristicsTarget Industry Uncertainty Shock Value weighted average of the changes in annualized stockreturn volatility induced by macro uncertainty shocks ofthe public companies operating in target SIC 3-digit in-dustry (CRSP).Public Target An indicator variable that equals one if the target is a pub-lic company (Thomson Reuters SDC).Private Target An indicator variable that equals one if the target is a pri-vate company (Thomson Reuters SDC).Subsidiary Target An indicator variable that equals to one if the target is asubsidiary of another company (Thomson Reuters SDC).Hi-tech Target An indicator variable that equals one if the target com-pany operates in the hi-tech industry (Thomson ReutersSDC).Log Deal Value The logarithm of deal value ($MM) (Thomson ReutersSDC).Same Industry An indicator variable that equals one if the target and ac-quirer operate in the same SIC 2-digit industry (ThomsonReuters SDC).Cross Border An indicator variable that equals one if a deal is a crossborder acquisition (Thomson Reuters SDC).Target Industry Sales Growth The median of the target industry\u2019s sales growth.Target Industry Firm Age The medium age of the public companies operating in thetarget industry.Target Boutique Advisor An indicator variable that equals to one if the target com-pany hires a boutique bank as M&A advisor.Acquirer CharacteristicsAcquirer Log Assets The logarithm of acquirer total assets ($MM) (ThomsonReuters SDC).190Acquirer Log MB The logarithm of the ratio of acquirer market capitaliza-tion 4 weeks prior to announcement to the book value(Thomson Reuters SDC).Acquirer ROA Ratio of acquirer net income in the past twelve months tototal assets (Thomson Reuters SDC).Acquirer Leverage Ratio Ratio of acquirer long term debt to total assets (ThomsonReuters SDC).Acquirer Boutique Advisor An indicator variable that equals to one if the acquirerhires a boutique bank as M&A advisor.191C.2 Measuring Macroeconomic UncertaintyI follow Alfaro et al.(2021) to construct the macroeconomic uncertainty shocks faced by each in-dustry, taking industry\u2019s exposure to the uncertainty shocks into account. The variables areconstructed exploiting firms\u2019 differential exposures to volatility shocks of multiple aggregate vari-ables. For each aggregate factor c, I construct the variableIV cj,t =\u00d8\u00d8\u00d8\u00d8ck, j\u00b02\u00d8\u00d8\u00d8 \u00b7\u00a2\u00e6ct , (C.2.1)where c is crude oil, seven currencies (Australian Dollar, British pound, Canadian Dollar, theEuro, Japanese Yen, and Swedish Krona), 10-year U.S. Treasury note, or economic policy uncer-tainty. \u00d8cj,t\u00b02 is the sensitivity of stock returns to changes in these aggregate variables estimatedat for each 3-digit SIC industry j. \u00a2\u00e6ct is aggregate volatility shock for quantity c, which is mea-sured using year-on-year change in the annualized standard deviation of daily price changes ofc, or year-on-year change in the average annual daily implied volatility of c. Both changes in re-alized volatility and implied volatility are used to capture volatility shocks based on past eventsas well as expected shocks in the future. For economic policy uncertainty, \u00a2\u00e6ct is the year-on-yearchange in the 365-day average of economic policy uncertainty. The idea behind the instrumentsis that when there is a volatility shock to aggregate quantity c, firms with different levels ofexposure to c, captured by sensitivities \u00d8cj,t\u00b02, experience different uncertainty shocks.Sensitivities \u00d8cj,t\u00b02 are estimated using:rrisk_ad ja,t =\u00c6 j,t+Xc\u00d8cj,t\u00b02rct +\u2264a,t, (C.2.2)where rrisk_ad ja,t is firm a\u2019s daily risk adjusted stock return and rct is the daily price change of c.I estimate the sensitivities for each 3-digit SIC industry j using a 10-year rolling window. Theestimated sensitivities are weight adjusted by their statistical significance levels. The adjustedsensitivities are lagged by two years to ensure that they pre-date the opacity shocks, both inthe aggregate and firm level. The sensitivities are unlikely to be correlated with firm specificcharacteristics two years from now. Aggregate volatility shocks,\u00a2\u00e6ct , are also unlikely to be drivenby firm characteristics. For this reason, the variables, by construction, do not correlate with anyunobservable firm characteristics.192C.3 Additional TablesTable C.3.1 Sample DescriptionPanel A. Sample Composition by YearThe table reports the annual distribution of earnout transactions in the sample. The sampleincludes acquisitions in the Thomson Reuters SDC M&A database announced between January1, 1991 and December 31, 2019 by U.S. public companies with market capitalization greaterthan $1 million four weeks prior to announcement. Only deals that worth at least $1 million areincluded in the sample. The sample is further restricted to deals with a transfer of control, i.e.bidders own less than 50% before the acquisition and own more than 50% after the acquisition.Deals with target companies from the financial and utility industries are excluded from thesample.Year No. of Earnout Transactions Percentage1991 8 0.411992 2 0.101993 6 0.301994 6 0.301995 40 2.031996 43 2.181997 76 3.861998 105 5.331999 71 3.602000 86 4.362001 66 3.352002 81 4.112003 83 4.212004 98 4.972005 93 4.722006 103 5.232007 123 6.242008 97 4.922009 67 3.402010 81 4.112011 99 5.022012 84 4.262013 66 3.352014 88 4.462015 78 3.962016 54 2.742017 50 2.542018 63 3.202019 54 2.74Total 1,971 100.00193Panel B. Sample Composition by IndustryThe table reports the industry distribution of the earnout transactions in the sample. Thesample includes acquisitions in the Thomson Reuters SDC M&A database announced betweenJanuary 1, 1991 and December 31, 2019 by U.S. public companies with market capitalizationgreater than $1 million four weeks prior to announcement. Only deals that worth at least $1million are included in the sample. The sample is further restricted to deals with a transfer ofcontrol, i.e. bidders own less than 50% before the acquisition and ownmore than 50% after the ac-quisition. Deals with target companies from the financial and utility industries are excluded fromthe sample. Industries are classified based on Fama-French twelve industry classifications fromhttp:\/\/mba.tuck.dartmouth.edu\/pages\/faculty\/ken.french\/Data_Library\/det_12_ind_port.html.No. Description No. of Earnout Transactions Percentage1 Consumer Nondurables 78 3.962 Consumer Durables 38 1.933 Manufacturing 135 6.854 Energy 33 1.675 Chemicals 22 1.126 Business Equipment 685 34.757 Telecom 43 2.189 Shops 121 6.1410 Healthcare 421 21.3612 Other 395 20.04Total 1,971 100.00194Table C.3.2 Earnout and M&A Deal Completion: Linear Probability EstimationThe table reports results from the linear probability regression of deal completion on the usage and fraction of earnout payment, estimatedusing Equation (4.8). The sample includes M&A transactions announced by U.S. public acquirers from 1991 to 2019. The dependent variable is anindicator variable which equals one if the deal is completed, and zero if the deal is withdrawn. Earnout usage is an indicator variable which equalsone if an earnout agreement is included in the M&A transaction. Earnout pct is the ratio of earnout payment to deal value. Columns (5) to (8) includedeal-specific control variables. Columns (7) and (8) include additional acquirer-specific control variables. All columns include year fixed effects.Columns (2), (4), (6) and (8) include acquirer and target SIC 2-digit industry fixed effects. See Appendix C.1 for variable definitions. The standarderrors (in parentheses) are clustered at acquirer industry (2-digit SIC) level. *, **, and *** indicate statistical significance at the 10%, 5%, and 1%levels, respectively.(1) (2) (3) (4) (5) (6) (7) (8)Earnout Usage 0.01* 0.01 0.05*** 0.04*** 0.04*** 0.03*** 0.04*** 0.03***(0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01)Earnout Pct (%) -0.01*** -0.01*** -0.01*** -0.01*** -0.01*** -0.01***(0.00) (0.00) (0.00) (0.00) (0.00) (0.00)Private Target 0.09*** 0.08*** 0.09*** 0.08*** 0.11*** 0.11*** 0.11*** 0.11***(0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01)Subsidiary Target 0.05*** 0.06*** 0.05*** 0.06*** 0.07*** 0.08*** 0.07*** 0.08***(0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01)Deal Control Variables No No No No Yes Yes Yes YesAcquirer Control Variables No No No No No No Yes YesYear FE Yes Yes Yes Yes Yes Yes Yes YesAcquirer, Target Industry FE No Yes No Yes No Yes No YesAdjusted R2 0.02 0.03 0.02 0.03 0.02 0.04 0.03 0.04Observations 23,304 23,304 23,304 23,304 23,304 23,304 20,966 20,966Observations with Earnout 1,971 1,971 1,971 1,971 1,971 1,971 1,810 1,810195Table C.3.3 Earnout and Target Industry Uncertainty Shock: Linear Probability EstimationThe table reports results from the linear probability regression of earnout usage on target industry uncertainty, estimated using Equation(4.9). The sample includes M&A transactions announced by U.S. public acquirers from 1991 to 2019. The dependent variable is an indicator variablethat equals one if an earnout agreement is included. The independent variable is the normalized target industry uncertainty shock described inSection 4.2.2. Columns (3) to (6) include deal-specific control variables. Columns (5) and (6) include additional acquirer-specific control variables.All columns include year fixed effects. Columns (2), (4), and (6) include acquirer and target SIC 2-digit industry fixed effects. See Appendix C.1 forvariable definitions. The standard errors (in parentheses) are clustered at acquirer industry (2-digit SIC) level. *, **, and *** indicate statisticalsignificance at the 10%, 5%, and 1% levels, respectively.(1) (2) (3) (4) (5) (6)Target Industry Uncertainty Shock 0.01*** 0.01** 0.01*** 0.01** 0.01*** 0.01**(0.00) (0.00) (0.00) (0.00) (0.00) (0.01)Private Target 0.12*** 0.12*** 0.12*** 0.12*** 0.12*** 0.13***(0.01) (0.01) (0.01) (0.01) (0.02) (0.02)Subsidiary Target 0.05*** 0.06*** 0.05*** 0.06*** 0.06*** 0.07***(0.01) (0.01) (0.01) (0.01) (0.01) (0.01)Deal Control Variables No No Yes Yes Yes YesAcquirer Control Variables No No No No Yes YesYear FE Yes Yes Yes Yes Yes YesAcquirer, Target Industry FE No Yes No Yes No YesAdjusted R2 0.04 0.06 0.04 0.06 0.06 0.07Observations 17,074 17,074 17,074 17,074 15,865 15,865Observations with Earnout 1,565 1,565 1,565 1,565 1,470 1,470196Table C.3.4 Earnout and Target Industry Uncertainty Shock: Robustness TestsThe table reports the robustness results of the impact of target industry uncertainty on earnoutagreements, with alternative control variables. Panel A reports the results controlling for thetarget company\u2019s growth prospective. Panel B reports the results controlling for the impactsof M&A advisors. The sample includes M&A transactions announced by U.S. public acquirersfrom 1991 to 2019. Columns (1) and (2) in Panel A and columns (1) to (3) in Panel B reportresults from the logistic regression estimated using Equation (4.5). The dependent variable is anindicator variable that equals one if an earnout agreement is included. Columns (3) and (4) inPanel A and columns (4) to (6) in Panel B report the results from the OLS regression estimatedusing Equation (4.6). The dependent variable is the ratio of earnout value to deal value. Theindependent variable is the normalized target industry uncertainty shock described in Section4.2.2. All columns include year fixed effects. Columns (2) and (4) in Panel A and columns (2), (3),(5), and (6) in Panel B include acquirer and target SIC 2-digit industry fixed effects. Columns(3) and (6) in Panel B include acquirer and target M&A advisor fixed effects. See Appendix C.1for variable definitions. The standard errors (in parentheses) are clustered at acquirer industry(2-digit SIC) level. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels,respectively.Panel A. Controlling for the Target Company\u2019s Growth ProspectiveEarnout Usage Earnout Percentage(1) (2) (3) (4)Target Industry Uncertainty Shock 0.15*** 0.16** 0.60*** 0.50**(0.05) (0.07) (0.22) (0.20)Private Target 2.85*** 2.98*** 3.90*** 4.11***(0.46) (0.45) (0.74) (0.80)Subsidiary Target 2.09*** 2.29*** 1.74*** 2.18***(0.45) (0.43) (0.59) (0.62)Hi-tech Target -0.20 -0.27** -1.57* -1.31***(0.20) (0.11) (0.93) (0.46)Log Deal Value ($MM) 0.17*** 0.18*** 0.23* 0.25**(0.03) (0.03) (0.13) (0.11)Same Industry -0.02 0.02 0.08 0.32(0.12) (0.07) (0.41) (0.20)Cross Boarder 0.08 -0.04 0.18 -0.17(0.08) (0.09) (0.27) (0.28)Acquirer Log Asset ($MM) -0.25*** -0.21*** -0.60*** -0.49***(0.04) (0.04) (0.08) (0.09)Acquirer Log MB -0.02 -0.06 0.17 0.02(0.05) (0.04) (0.17) (0.14)Acquirer ROA 0.15 0.29* -1.48 -1.12(0.20) (0.17) (1.33) (1.15)Acquirer Leverage Ratio -0.93*** -0.42*** -2.81*** -1.35**(0.22) (0.14) (0.82) (0.58)Target Industry Sales Growth -0.44 -0.28 -1.80** -1.00(0.31) (0.29) (0.86) (0.86)Target Industry Firm Age -0.05* -0.06*** -0.27* -0.29**(0.03) (0.02) (0.15) (0.14)197Panel A. Continued(1) (2) (3) (4)Control Variables Yes Yes Yes YesYear FE Yes Yes Yes YesAcquirer, Target Industry FE No Yes No YesPseudo R2 0.11 0.15Adjusted R2 0.05 0.06Observations 15,865 15,438 15,865 15,865Observations with Earnout 1,470 1,470 1,470 1,470198Panel B. Controlling for M&A advisors(1) (2) (3) (4) (5) (6)Target Industry Uncertainty Shock 0.15*** 0.16** 0.29 0.56*** 0.47** 0.40(0.05) (0.06) (0.29) (0.21) (0.19) (0.24)Private Target 2.85*** 2.97*** 3.78*** 3.86*** 4.06*** 3.55***(0.46) (0.45) (0.94) (0.74) (0.81) (1.06)Subsidiary Target 2.09*** 2.27*** 2.44*** 1.67*** 2.10*** 0.80***(0.45) (0.44) (0.83) (0.56) (0.58) (0.23)Hi-tech Target -0.08 -0.19* 0.66 -0.87 -0.91** 0.19(0.14) (0.10) (0.58) (0.61) (0.42) (0.48)Log Deal Value ($MM) 0.18*** 0.19*** -0.21 0.26* 0.28** -0.19(0.03) (0.03) (0.16) (0.13) (0.11) (0.16)Same Industry 0.00 0.02 0.96** 0.20 0.31 0.67***(0.14) (0.07) (0.48) (0.48) (0.23) (0.25)Cross Boarder 0.06 -0.06 0.67** 0.08 -0.24 0.20(0.08) (0.09) (0.31) (0.27) (0.29) (0.42)Acquirer Log Asset ($MM) -0.25*** -0.22*** -0.03 -0.61*** -0.49*** 0.09(0.04) (0.04) (0.11) (0.08) (0.09) (0.13)Acquirer Log MB -0.01 -0.06 -0.03 0.20 0.02 -0.09(0.05) (0.05) (0.07) (0.18) (0.15) (0.07)Acquirer ROA 0.04 0.23 1.95 -2.07 -1.40 0.98(0.28) (0.21) (1.22) (1.76) (1.36) (1.24)Acquirer Leverage Ratio -0.99*** -0.47*** -1.69*** -3.11*** -1.54** -0.60(0.23) (0.17) (0.61) (0.97) (0.72) (0.54)Acquirer Boutique Advisor 0.06 -0.02 1.46* -0.15 -0.42* 1.33(0.08) (0.07) (0.78) (0.27) (0.22) (0.82)Target Boutique Advisor -0.31*** -0.38*** -2.54*** -1.08*** -1.21*** -0.24(0.09) (0.11) (0.48) (0.21) (0.21) (0.46)Control Variables Yes Yes Yes Yes Yes Yes199Panel B. ContinuedEarnout Usage Earnout Percentage(1) (2) (3) (4) (5) (6)Year FE Yes Yes Yes Yes Yes YesAcquirer, Target Industry FE No Yes Yes No Yes YesAcquirer, Target Advisor FE No No Yes No No YesPseudo R2 0.11 0.15 0.43Adjusted R2 0.04 0.06 0.19Observations 15,865 15,438 1,545 15,865 15,865 3,790Observations with Earnout 1,470 1,470 156 1,470 1,470 206200","@language":"en"}],"Genre":[{"@value":"Thesis\/Dissertation","@language":"en"}],"GraduationDate":[{"@value":"2022-11","@language":"en"}],"IsShownAt":[{"@value":"10.14288\/1.0417490","@language":"en"}],"Language":[{"@value":"eng","@language":"en"}],"Program":[{"@value":"Business Administration - Finance","@language":"en"}],"Provider":[{"@value":"Vancouver : University of British Columbia Library","@language":"en"}],"Publisher":[{"@value":"University of British Columbia","@language":"en"}],"Rights":[{"@value":"Attribution-NonCommercial-NoDerivatives 4.0 International","@language":"*"}],"RightsURI":[{"@value":"http:\/\/creativecommons.org\/licenses\/by-nc-nd\/4.0\/","@language":"*"}],"ScholarlyLevel":[{"@value":"Graduate","@language":"en"}],"Supervisor":[{"@value":"Bena, Jan, 1976-","@language":"en"},{"@value":"Carlson, Murray","@language":"en"}],"Title":[{"@value":"Essays in empirical corporate finance : the impact of economic uncertainty on the financial markets","@language":"en"}],"Type":[{"@value":"Text","@language":"en"}],"URI":[{"@value":"http:\/\/hdl.handle.net\/2429\/82456","@language":"en"}],"SortDate":[{"@value":"2022-12-31 AD","@language":"en"}],"@id":"doi:10.14288\/1.0417490"}