UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Essays on politically connected firms Duo, Yi 2017

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2017_november_duo_yi.pdf [ 1.82MB ]
Metadata
JSON: 24-1.0357147.json
JSON-LD: 24-1.0357147-ld.json
RDF/XML (Pretty): 24-1.0357147-rdf.xml
RDF/JSON: 24-1.0357147-rdf.json
Turtle: 24-1.0357147-turtle.txt
N-Triples: 24-1.0357147-rdf-ntriples.txt
Original Record: 24-1.0357147-source.json
Full Text
24-1.0357147-fulltext.txt
Citation
24-1.0357147.ris

Full Text

 ESSAYS ON POLITICALLY CONNECTED FIRMS  by  Yi Duo B.S. in Business Administration, American University, Washington, DC, 2010 M.S. in Accounting, American University, Washington, DC, 2011   A DISSERTATION SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Business Administration)   THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)   October 2017 © Yi Duo, 2017   ii  Abstract This thesis explores a number of issues related to politically connected firms in two separate chapters. I follow Goldman et al. (2009), defining politically connected firms as those with at least one former politician serving as member of its board of directors, and construct a sample containing the S&P 500 firms between 2004 and 2013. The first chapter explores why firms seek these political connections, and how they benefit from two direct value extraction channels: government procurement and subsidies. I find that firms that aim for government contracts seek executive branch connections, while those that face heavy regulations target congressional connections. Next, I show that politically connected firms do get more government contracts and subsidies. Firm performance (using accounting based measures) suffers with government contracts and subsidies, and political connections fail to increase or decrease this negative relation, which suggests effective safeguards against overpricing and cronyism. However, politically connected firms do seem to enjoy a temporary increase in future ROA, when government contracts are taken into consideration. The second chapter asks if politically connected firms pay higher audit fees, and explores the underlying reasons. Prior studies have mixed implications on how risky these clients are for auditors. On the one hand, some studies suggest politically connected firms have lower accounting quality and face higher political risk, hence incur higher audit fees. On the other hand, less investor pressure and lower litigation and bankruptcy risks would decrease audit fees for firms with political connections. I find that politically connected firms do pay higher audit fees, and the effect is stronger for those with executive branch connections. Neither lower accounting quality nor higher political risk is found to be the underlying reason. The fact that many politically connected firms are government contractors, who are subject to additional regulations and government audit, is found to be the main factor for this difference in audit fee.   iii  Lay Summary This thesis examines how corporate political connections in the United States are associated with firms’ business operations and audit fees, and how they differ from those of their peers. I find that if the goal is to obtain government contracts, firms seek connections to the executive branch; if the goal is to manage and to influence regulations, they seek connections to the legislative branch. Politically connected firms do receive more government contracts and subsidies. After taking the negative effect of government contracts on accounting performance into consideration, politically connected firms enjoy temporarily higher future profits. Meanwhile, auditors charge higher fees to audit politically connected firms. This increase in audit fees is likely a result of these firms’ involvement in government contracts, which is subject to increased audit risks due to additional regulations and government audit.              iv  Preface My committee members provided helpful discussions on prior and existing research in various fields. Both chapters are my own work.     v  Table of Contents Abstract ..................................................................................................................................... ii Lay Summary ............................................................................................................................ iii Preface ...................................................................................................................................... iv Table of Contents ........................................................................................................................ v List of Tables ............................................................................................................................ vii List of Figures..........................................................................................................................viii Acknowledgements .................................................................................................................... ix Introduction ................................................................................................................................ 1 Chapter 1: Political Connections, Government Procurements, and Subsidies ................................. 4 1.1 Introduction .................................................................................................................. 4 1.2 Literature review and hypotheses development ................................................................. 9 1.3 Research design ...........................................................................................................15 1.3.1 Research method ......................................................................................................15 1.3.2 The 2006 and 2008 Elections .....................................................................................20 1.4 Sample and data ...........................................................................................................22 1.5 Results and interpretations ............................................................................................26 1.6 Conclusion ..................................................................................................................31 Chapter 2: Audit Fees of Politically Connected Firms ................................................................. 62 2.1 Introduction .................................................................................................................62 2.2 Literature review and hypothesis development .................................................................66 2.3 Research design ...........................................................................................................70 2.4 Sample and data ...........................................................................................................75 2.4.1 Sources of data ........................................................................................................75 2.4.2 Choice of accounting quality measures .......................................................................78 2.5 Results and interpretations ............................................................................................80   vi  2.6 Additional analyses ......................................................................................................84 2.7 Conclusion ..................................................................................................................85 Conclusion ...............................................................................................................................108 Bibliography ............................................................................................................................109 Appendices ..............................................................................................................................114 Appendix A Appendix for Chapter 1 ........................................................................................ 114 A.1 Variable Definitions ............................................................................................... 114 Appendix B Appendix for Chapter 2 ........................................................................................ 117 B.1 Variable Definitions ............................................................................................... 117    vii  List of Tables Table 1.1 Number of Politically Connected Firms by Year. ..............................................................36 Table 1.2 Industry Composition ....................................................................................................37 Table 1.3 Summary Statistics ........................................................................................................39 Table 1.4 Choices of Political Connections .....................................................................................42 Table 1.5 Government Contracts and Political Connections ..............................................................45 Table 1.6 Government Subsidies and Political Connections ..............................................................47 Table 1.7 Government Procurement Post 2006 and 2008 Election .....................................................51 Table 1.8 Government Subsidies Post 2006 and 2008 Election ..........................................................54 Table 1.9  Current Firm Performance .............................................................................................58 Table 1.10 Future Firm Performance – Single Factor .......................................................................59 Table 1.11 Future Firm Performance – With Interactions..................................................................60 Table 2.1 Number of Politically Connected Firms by Year ...............................................................88 Table 2.2 Summary of Industries ...................................................................................................89 Table 2.3 Summary Statistics ........................................................................................................91 Table 2.4 Political Connections and Audit Fees ...............................................................................94 Table 2.5 Accounting Quality and Political Connections ..................................................................96 Table 2.6 Audit Fees and Political Risk ..........................................................................................97 Table 2.7 Audit Fees Path Analysis ...............................................................................................99 Table 2.8 Audit Fees and Paths Analysis: Isolate Government Contract ........................................... 100 Table 2.9 Audit Report Lag ........................................................................................................ 103 Table 2.10 Internal Control System .............................................................................................. 106    viii  List of Figures Figure 1.1 Discretionary Spending of Federal Government for 1990 – 1998 (in Billion $) ....................33 Figure 1.2 Timeline of Political Power in Control of the United States Government.............................34 Figure 1.3 Number of Politically Connected Firms by Year. .............................................................35 Figure 2.1 Politically Connected Firms by Year ..............................................................................87     ix  Acknowledgements I would like to express my sincere gratitude to my supervisory committee members – Dr. Dan Simunic, Dr. Kin Lo, Dr. Jenny Li Zhang, and Dr. Ralph Winter. Useful inputs were also provided by the other Accounting department faculty members as well as my fellow PhD students during my presentations. This dissertation would not have come about had it not been for the gentle guidance and warm encouragement of the chair of my supervisory committee, Dan Simunic. Dan’s expertise in audit literature and general curiosity in interesting research topics kept me motivated, even when it seemed like my research had come to a dead-end. I learned to be more open-minded and persistent in research from working with him. The commitment and generosity that he demonstrated to both his work and his students has inspired me both as a person and as an educator.  I would also like to give my special thanks to Jenny Li Zhang, who generously contributed her time and energy in shaping my dissertation. With Jenny, I can drop by un-announced and pick her brain whenever I am in doubt. She has challenged me and pushed me to move forward.  My PhD life would not have gone so smoothly without the help of our PhD administrator Elaine Cho, and our divisional assistant Debra Harris. I also owe my parents for their generous financial and emotional support through this journey in pursuit of a PhD, which they never planned nor envisioned for me.  Last but not least, this work is dedicated to my maternal grandparents and the late Ms. Sue Marcum, who was my first accounting professor at American University. The loving memories of these people have motivated and propelled me through all of the most difficult moments in my life.  1  Introduction Why do firms seek political connections? How do connections affect firms and other stakeholders? This thesis examines multiple research questions regarding politically connected firms in the United States in separate chapters. There are various ways of measuring and identifying political connections in the United States: lobbying expenses (Hill et al., 2014), corporate political campaign contributions (Claessens et al., 2008), politician equity ownership (Baloria, 2014; Tahoun, 2014),  and former politicians as board members (Goldman et al., 2009). There are advantages and disadvantages to each measure. Although former politicians on the board are not as influential as sitting public servants, I choose the politically connected board of director because it is easily identified1 and this relationship is clearly bilateral – not only does a firm want to establish connections, but also the former politician is willing to be on the board.  In the first chapter, I examine whether firms seek political connections according to their specific needs. Two possible reasons are identified: obtaining government contracts, and managing regulatory risk. If obtaining government contract is the goal, firms should target connections to the executive branch; if influencing regulation and law making is the goal, then firms should seek connections to the legislative branch. Next, I hypothesize two channels where a politically connected board can contribute to a firm’s operation, namely government contracts and subsidies. I find that politically connected firms do receive more government contracts and subsidies, and establish a directional link between political connections and government contracts. At the end of the first chapter, I analyze current and future accounting performances of politically connected firms, especially in the presence of government contracts and subsidies. The results suggest that                                                  1 Available from public proxy statement filings from the Securities and Exchange Commission, almost annually.   2  politically connected firms have some temporary advantage in future accounting performance, in the presence of government contracts. In the second chapter, I examine a two-part research question: do politically connected firms pay higher or lower audit fees, and why? Existing literature on political connections is not clear as to which direction the relation between political connections and audit fees will go. Some researchers find politically connected firms have lower financial reporting quality (Chaney et al., 2011) and are more aggressive in their accounting (Baloria, 2014; Kim and Zhang, 2016). In this case, auditors should charge higher audit fees because of higher risk. On the other hand, other researchers find politically connected firms face less pressure from shareholders (Cooper et al., 2010;  Goldman et al., 2009), creditors (Houston et al., 2014), and regulators (Correia, 2014); all  these factors help decrease audit risk, thus lowering audit fees. Empirical results indicate that politically connected firms do have higher audit fees, suggesting an increase in audit risk. However, upon closer investigation, lower accounting quality suggested by prior literature does not seem to be the reason behind this difference in audit fees. Neither does political risk (political and policy uncertainty) seem to be the reason. Instead, my findings suggest government contract involvement of politically connected firms is the main driver behind the higher audit fees. Government contractors are subject to additional regulations like Federal Acquisition Regulation (FAR) aside from Generally Accepted Accounting Principles (GAAP). Pressure from politically connected board members to protect their reputation and the possibility of government audits both increase audit risk for external auditors. Hence, politically connected firms pay higher audit fees. This thesis sheds light on the existence and effects of corporate political connections in the United States. By hiring former politicians, firms are actively adapting and influencing the competitive environment in which they operate.  The results do not support the assertion that   3  political connections are associated with collusion and nefarious motives. It is definitely not suggesting that the current system is perfect. However, it reminds us to stay hopeful that, with checks and balances, the system can perform as intended and safeguard against corruption and collusion.  4  Chapter 1: Political Connections, Government Procurements, and Subsidies 1.1 Introduction Every election brings an end to the public service career of many government officials. Where do these former politicians go after public office? According to the findings of Palmer and Schneer (2015), approximately half of former congressional members will get a seat on the board of at least one publically traded firm. As the authors point out, a board directorship “allows a public servant to cash out on political connections and credentials” through helping their companies “enter the political arena, navigate regulations and bureaucracies, and improve governmental relations and engagement”. But do firms target political connections specifically according to their own needs or seek whatever connections that are available? Are these connections effective in their intended purposes and do they ultimately contribute to firm value? This paper seeks to shed light on these questions. First, I identify two main reasons why firms want to be politically connected: obtaining government contracts and managing regulations. To obtain government contracts, connections to the executive branch of government is vital, as government agencies, like the Department of Defense, Department of Energy, and Department of Health and Human Services, award the most in government procurement to outside contractors. Meanwhile, the US Constitution grants Congress the authority to make laws. Firms that intend to manage regulations and exert influence in changes of laws, which is identified by how much they lobby, should look for political connections with legislative branch experience. Next, to assess whether political connections are effective for their intended purposes, I identify two channels to which these connections can contribute: government contracts and subsidies. The first channel is a directly observable result of an intended purpose mentioned   5  above. Government procurement is an important source of revenue for many public firms. For example, big contractors like Boeing Co. and Lockheed Martin Corp. received $4.68 and $13.5 billion in total awarded obligation respectively during the 2013 Federal Fiscal Year2. These two companies’ annual revenues were $86.82 and $45.36 billion for fiscal year 2013, respectively. This means government procurements account for roughly 5% and 30% of their annual revenue. To win the bids of these procurements, firms need to go through the bidding process, be aware of government product standards, navigate bureaucracies, and comply with expense disclosure and reporting requirements of government contractors. Having political connections helps in gaining an information advantage in all of these aspects, and even possibly discovering potential bidding opportunities before they are publically announced. The second channel, government subsidy, is an indirect result of managing regulation but also affects firm value. Not all regulation changes result in subsidies, but subsidies are proposed and written into laws by Congress, typically to relieve economic burden or to promote economic vitality. Subsidies are commonly given in the form of tax credits, government loans, and government grants, which lower firms’ taxes as well as operating expenses. In their research, Duchin and Sosyura (2012) find politically connected firms are more likely to be funded under the Troubled Asset Relief Program (TARP), which can be seen as a government subsidies to those with “toxic assets” during the subprime mortgage crisis. According to Mattera and Tarczynska (2015), firms like Goldman Sachs, JPMorgan Chase, Dow Chemical, NRG Energy, Sempra Energy, Solar City, and United Technologies are among the top recipients of state and federal subsidies. These firms are in highly regulated industries, where political connections can                                                  2 Information from http://www.insidegov.com/   6  help them weather the political landscape and government policies, and hopefully promote law changes that affect competition and the vitality of the industry. A major concern among economists is that government procurements and subsidies are assigned inefficiently, and collusion between politicians and corporations will make it worse. In that case, having political connections effectively turns these two channels into ways of extracting public funds for private gains. Recent research from other countries like China (Li et al., 2008) , Korea (Schoenherr, 2015), and Malaysia (Johnson and Mitton, 2003) provides empirical evidence that such cases exist, and that the economic magnitude, at least in government procurement rent, is significant. According to Schoenherr (2015), the inefficiency in government contract allocations result in a cost of 0.21-0.32% of the Korean annual GDP. However, most of these countries are known for being relation-focused and their legal systems are not considered the most stringent. For politicians in the United States, the collusion may not be as easy as one thinks, when both firms and politicians are facing public scrutiny and risking reputations and litigations. In this paper, using a detailed hand collected sample of the S&P 500 firms’ politically connected board information between 2004 and 2013, I find that firms do seek political connections according to specific needs. They establish connections with former executive branch politicians if obtaining government contracts is the goal, measured as the ratio of government contracts to annual sales. When regulatory risk management is the goal, measured using the ratio of lobbying expense to annual sales, firms hire former congressional members as directors. Next, I explore whether there are differences in the amount of government contracts and government subsidies that politically connected firms receive compared to their non-connected   7  peers. As expected, politically connected firms do receive more in government contracts; they also have more access to total subsidies in general. However, political connections do not seem to help firms to receive more government loans nor tax credits. To address endogeneity, I explore the difference in importance of Democratic and Republican political connections pre- and post- the 2006 midterm election and 2008 Presidential election. In the 2006 midterm election, the Republican Party lost their majority to the Democratic Party in both houses of Congress. Nevertheless, the White House was still under Republican President George W. Bush. If indeed executive branch connections are more important for procurement allocation, there should not be a significant change in the influence of Democratic connections to the executive branch. In the 2008 Presidential election, the Democratic Party took over the White House while maintaining their majority in both houses of Congress. With all major government agencies and departments changing leadership, government procurement influence of the Democratic connection should increase post-2008 election. The result suggests that, indeed, the winning party connections (those with the Democratic Party) are more important in obtaining government procurement only after the Presidential election, rather than the midterm election. These findings indicate that political connections do increase the government contracts that firms obtain. I also use the election setting to test the impact of political connections on government subsidies, but there is little evidence found in both elections. In the end, current and future accounting performance are analyzed. Though political connections may help firms to obtain these benefits from the public sector, they do not necessarily translate into higher profitability. I find that politically connected firms do not enjoy any particular advantage in general. Meanwhile, the two value extraction channels, government contracts and subsidies, are negatively associated with firm performance. However, when   8  government contracts are taken into consideration, politically connected firms are found to have higher return on assets 3 and 4 years in the future. This finding is interesting as it is consistent with the assertion that government contractors benefit from investments made to enable production to satisfy procurement contracts turning into productive assets in the future.  This paper contributes to the literature in a number of ways. To begin with, this paper is the first to show that firms indeed seek different political connections for different purposes. Prior research does not detail the backgrounds of the politically connected board members that this paper uncovers. This study enables us to see that firms do indeed target specific political connections for their own needs. Second, instead of a positive association, by exploiting the difference between 2006 midterm and 2008 presidential elections, this paper links political connections with increased government procurements. Being politically connected to a winning party does increase the amount of government contracts that a firm obtains after the 2008 presidential election, when the heads of major departments and government agencies were replaced. However, the same cannot be said for the 2006 midterm election, because government procurement processes are mostly controlled by the executive branch and their agencies, not Congress. Last but not least, this paper casts a different light on the profitability impact of having political connections. While prior research in many other countries indicates that government contracts and subsidies are ways to extract political rents from the public sector to the private sector, my findings suggest this may not apply to cases in the United States. Although care should be taken in extrapolating to other jurisdictions, these results caution us against holding a negative bias towards the role of political connections in all countries.   9  The rest of this chapter is organized in the following way: I review the existing related literature and develop my hypotheses in Section 1.2; Section 1.3 is on research design; sample selection and data are described in Section 1.4; Section 1.5 provides test results and interpretation; and Section 1.6 ends the paper with a brief conclusion. 1.2 Literature review and hypotheses development  A growing number of papers discuss the economic effects of political connections. Shaffer (1995) summarizes that corporate political activities, such as seeking and establishing political connections, is a firm’s response to how public policies and regulations affect the competitive environment of the firm. Meanwhile, in Faccio (2006), perhaps one of the most cited papers on corporate political connections, the author points out the clustering of politically connected firms in more corrupted countries where political and economic rents can be easily extracted. Using the same sample, follow-up papers find that politically connected firms are more likely to be bailed out during financial hardship (Faccio et al., 2006), have higher leverage (Faccio, 2010), and lower financial information quality (Chaney et al., 2011). But cross-country studies have been criticized by many, as the lack of control for country-specific effects may drive the results. Some researchers do a more in-depth analysis on a specific country, where legal institution and political environment are relatively homogenous. Most of these studies have been done on developing countries, where rent is easily extracted through corruption and cronyism. For example, Fishman (2001) estimates the value of political connection in Indonesia. Johnson and Mitton (2003) find that it pays off to have political connections in Malaysia, especially in the aftermath of the Asian financial crisis. Brazilian firms who contribute to important elections are shown in Claessens et al. (2008) to have better access to financing through banks.   10  Recently, more researchers have turned their interests to functioning democracies like the United States, where the benefits of political connections may be more ambiguous, and arguably more interesting. Cooper et al. (2010) find firms that contribute to U.S. political campaigns have higher future stock returns. Goldman et al. (2009) find that politically connected boards are associated with positive returns. Duchin and Sosyura, (2012) show that when applying for TARP, politically connected firms are more likely to be funded, consistent with the story in Faccio et al. (2006).  Following Faccio (2010)’s idea of easy access to credit, Houston et al. (2014) also find the cost of bank loans is lower for politically connected firms in the United States, and identify this as a possible value extraction channel. Nevertheless, there is a lack of systematic research in the composition of politically connected boards and how else a politically connected board may contribute to firm value. But perhaps the research most relevant to this paper is Goldman et al. (2013). In their paper, the authors take advantage of the change in control of both houses of the US Congress from the Democratic Party to the Republican Party following the 1994 midterm election in the United States, and find that firms with winning (losing) party connections experienced an increase (decrease) in government procurement contracts. Yet, there are concerns with their setting and research design. I argue that instead of midterm elections, Presidential elections, which can change the executive branch leadership, should be the one that matters. Although the legislative branch controls the overall budget, it does not control the procurement allocation process. Interestingly, the sample period of Goldman et al. (2013) falls in a period where executive branch leadership change took place (i.e. the 1992 Presidential election) and then significant budgetary changes happened (i.e. after the 1994 midterm election). As shown in Figure 1.1, after the 1994 midterm   11  election, the federal government’s discretionary spending3 was dramatically decreased. The authors use a stationary connection identifier in 1994 as independent variable, and the 4-year difference before and after the 1994 election in government contracts as the dependent variable (i.e. sum of contract value from 1995 to 1998 minus sum of contract value from 1990 to 1993). This sample period and research design may have had unintended consequence on reliability of the empirical finding. Overall, I believe the authors looked at the wrong election. In this paper, I argue that the majority of government procurement allocations are not directly controlled by the US Congress. Instead, connections to the executive branch are more important in obtaining government contracts. The US Constitution sets up a system where the legislative branch can constrain the executive branch through legislations and budget appropriations. Yet the majority of the appropriations are actually spent by the executive branch through different government agencies. Government agencies, like the Department of Defense, Department of Education, Department of Health and Human Services, are essential parts of the executive branch, and enjoy freedom and authority to designate specific contractors and specify contract requirements. As a result, if firms want to obtain government contracts, they should seek executive branch connections. I use the weight of government contracts in a firm’s revenue to proxy for its need to obtain government contracts. At the same time, corporations are not only competing for government contracts. Government legislation means that, for some firms, political and regulatory risks run high in their operation.  Managing legislative risk is then crucial to the vitality – even the existence – of the industry. In this case I assume that firms facing high regulatory risk spend more in lobbying                                                  3 Discretionary spending is where government procurement dollars mostly come from. It includes various spending items like national defense, education, transportation, etc.   12  in hopes of managing regulatory environment and policy issues that the firms are interested in. These firms with a need to manage regulatory risk will seek related political connections in order to exert their influence in law-making, such that they are granted favorable treatment. Hypothesis 1.1 When obtaining government contracts is the goal, a firm seeks executive branch political connection; when managing regulatory risk is the goal, a firm seeks legislative branch political connection. Next, I identify two possible channels through which political connections can exert influence to contribute to firm value: government contracts and subsidies. As mentioned, obtaining government contracts is a direct result of the first part of Hypothesis 1. Because government contracts are part of revenues, this is also a straight-forward channel through which political connections can contribute to firm value. Receiving government subsidies is an indirect benefit from managing regulatory risk. However, not all regulatory changes are about subsidies. Subsidies are authorized by Congress, and their funding comes from the annual budgets, which is also controlled by the legislative branch. For example, right after the recession of 2010, lawmakers looked for ways to cut the budget. The 2012 fiscal year budget proposed a 22% cut to farm subsidies, including $5 billion direct payment program to farmers, whereas Congress allowed the Deepwater Royalty Waiver Program to stay, permitting oil companies to drill on Federal property without paying royalties. It is easy to see that these subsidies help improve firms’ bottom lines by directly paying in, or lowering operation expenses.  The first set of tests examines the link between political connections and these two channels. Because government contracts and subsidies contribute positively to profits, the   13  relation between political connections and these two channels should also be positive. As a result, I phrase the second hypothesis as followed. Hypothesis 1.2 Firms with political connections receive more government contracts compared with their non-connected peers; they also receive more government subsidies compared with their non-connected peers. In order to address endogeneity, I employ a different set of tests. If Hypothesis 1 is true, that firms seek specific types of political connections according to their needs, then the connection’s power to influence target benefits should be affected only after political control changes in the target government branch. This means, post-2006 midterm election, when the control of both houses of the US Congress changed from the Republican Party to the Democratic Party, the ability of Democratic connections to influence subsidies should have increased as their influence in legislation increased. Yet the executive branch did not change leadership, so we should not expect a significant change in political connections’ influence on government contracts regardless of which party firms are connected to.  A change in executive branch leadership can lead to a change in the effectiveness of political connections in obtaining information advantage. After the 2008 presidential election, leadership of the executive branch changed hands. This means that Democratic connections can exert more influence in government procurement allocations, as the executive branch has autonomy in deciding the specifics of the government contracts they award – project requirements, project size, and how a bid should be structured. Although a large proportion of government contracts are multi-year, new contracts are more likely to be granted to firms that have a closer connection to the current administration. Firms may also seek to establish new   14  political connections with the winning party who controls the executive branch. In summary, this means that when there is a change in control of the White House, it should have a pronounced impact on the government procurement allocation from the winning political party’s connections. It is worth noting that the argument above does not imply that, following a change in political control, Congress has no impact on the amount of government contracts that firms can obtain. After all, Congress controls the total government spending through their authority to adjust the budget and appropriations. Nevertheless, it is after an executive branch leadership change that the ability of the winning party’s connections in obtaining government contracts changes. In this case, a differentiation between the midterm election and presidential election is useful, and the 2006, 2008 elections are settings to test the aforementioned theory. The third hypothesis examines the profitability difference between politically connected firms and their non-connected counterparts. This problem is particularly interesting. Firstly, politically connected firms should enjoy higher profitability than non-connected firms, if simply having political connections is valuable as former politicians provide their seals of approval as reputation guarantee. Secondly, if government contracts and subsidies are means to extract political rent, then higher profitability should be associated with these two channels. In that case, as political connections are able to improve bottom lines through these two channels, firms with political connections who are receiving government contracts or subsidies should see an additional advantage in profit. Profitability of government contracts is a direct channel to increase firm value. Prior literature has found some theoretical as well as empirical evidence, from countries like China, Malaysia, and Korea, that government contracts are highly profitable to firms with cost shifting, and are considered quid pro quo between politicians and corporations (McGowan and Vendrzyk, 2002; Li et al., 2008; Hung et al., 2012; Schoenherr, 2015). Both   15  reasons can help us better understand why the market values political connections, as documented by studies like Cooper et al., (2010) and Goldman et al. (2009). I thus state the third hypothesis accordingly. Hypothesis 1.3 Politically connected firms have better accounting performance than their non-connected counterparts. On the other hand, if government contracts and subsidies are handed to firms with lower profitability to begin with, as subsidies in particular are intended for, even when political connections help to secure these government benefits, profitability will not be higher. To address this issue, firms that receive at least one form of these government benefits should be compared. In this case, there should be a negative correlation between profitability and these government benefits, and profitability’s association to political connection is unknown. 1.3 Research design 1.3.1 Research method There are various ways of measuring and identifying corporate political connections in the United States: lobbying expenses (Hill et al., 2014), corporate political campaign contributions (Claessens et al., 2008), politician equity ownership (Baloria, 2014),  and former politicians as board members (Goldman et al., 2009). However, not all measures are suited for this study. Lobbying expenses, which are usually issue and industry specific, are not a good identification because an industry association may lobby on behalf of the member firms, and lobbyists do lobby both political parties without public disclosure of specifics. Corporate political campaign contribution to some extent measures which political party a firm wants to be associated with more, yet most firms donate to both parties with similar amounts. Recent decades also witnessed an increase in popularity of utilizing Political Action Committees (PACs). Since   16  2010’s Citizens United v. Federal Election Commission, the use of Super PACs has made it almost impossible to identify the donors of political campaigns. Lastly, on politician equity ownership, I argue that it is a matter of politicians selecting their own stock holdings, rather than firms selecting political connections. In this paper, I follow Goldman et al., (2009)’s definition and classify politically connected firms as those with at least one politically connected board member.4 This measure is taken at the beginning of every fiscal year. Comparatively, this is a clear measure of political connection when it comes to identifying which political party and what government branches firms are targeting. Having a politically connected board member is not only a signal that the firm is seeking political connection, but also that it is successful in doing so, as the politician needs to agree to be on the board of directors, and have his/her name associated with the firm when filing for public disclosure. To test Hypothesis 1.1, I use two proxies for a firm’s intent of seeking political connections. The first one is the amount of government contracts scaled by sales, which measures how important government procurement is to a firm’s operation (𝐺𝑂𝑉_𝑆𝐴𝐿𝐸𝑖,𝑡). A higher ratio indicates greater importance of obtaining government contracts. It proxies for a firm’s need to obtain government contracts. The second measure is the amount of lobbying expense scaled by sales (𝐿𝑂𝐵_𝑆𝐴𝐿𝐸𝑖,𝑡), which measures how important lobbying is to a firm’s operation. Similar to the former proxy, a higher ratio indicates greater importance of managing                                                  4 A company is classified as politically connected if it has at least one board member with one of the following former positions: President, presidential (vice-presidential) candidate, senator, member of the House of Representatives, secretary, assistant secretary, deputy secretary, deputy assistant secretary, undersecretary, director, associate director, deputy director, commissioner of any federal government department or federal government agency (including CIA, FEMA, CIA, OMB, IRS, NRC, SSA, CRC, FDA, and SEC), governor, mayor, treasury of the city, representative to the UN, trade representative for the US, ambassador, staff (White House, president, presidential campaign), chairman of the party caucus, and chairman or staff of the presidential election campaign.   17  regulatory risk. So it proxies how much regulatory risk a firm faces. I then regress the target political connection type one period ahead on these two variables measuring the importance of government contracts and the need to manage regulatory risk, respectively. The regression has the following structure: Pr(𝐶𝑜𝑛𝑛𝑒𝑐𝑡𝑖𝑜𝑛 𝑇𝑦𝑝𝑒𝑖,𝑡+1)= 𝛼0 + 𝛼1𝐺𝑂𝑉_𝑆𝐴𝐿𝐸𝑖,𝑡 + 𝛼2𝐿𝑂𝐵_𝑆𝐴𝐿𝐸𝑖,𝑡 + 𝛼3𝑆𝐼𝑍𝐸𝑖,𝑡 + 𝛼4𝐺𝑅𝑂𝑊𝑇𝐻𝑖,𝑡+ 𝛼5𝐻𝐻𝐼𝑖,𝑡 + 𝛼6𝑅𝑂𝐴𝑖,𝑡 + 𝛼7𝐿𝑂𝑆𝑆𝑖,𝑡                                                         (1.1) where 𝐶𝑜𝑛𝑛𝑒𝑐𝑡𝑖𝑜𝑛 𝑇𝑦𝑝𝑒𝑖,𝑡+1 is an indicator of whether the one period ahead politically connected board has legislative branch experience (LEGISLATIVEt+1) or executive branch experience (EXECUTIVEt+1). This test applies to both a sample limited to those with political connection, such that the ability to get political connection is constant, and to the full sample with both connected and non-connected firms. As mentioned, the variables of interest are 𝐺𝑂𝑉_𝑆𝐴𝐿𝐸𝑖,𝑡 and 𝐿𝑂𝐵_𝑆𝐴𝐿𝐸𝑖,𝑡, each measures a different reason why a firm wants to seek political connections. I also replace 𝐺𝑂𝑉_𝑆𝐴𝐿𝐸𝑖,𝑡 and 𝐿𝑂𝐵_𝑆𝐴𝐿𝐸𝑖,𝑡 to indicator variables 𝐺𝐶𝑖,𝑡 and 𝑅𝐸𝐺𝑈𝐿𝐴𝑇𝐸𝑖,𝑡, which indicates when government contract need is high (over 10% of sales) or when regulatory risk is high (lobbying is over 0.03% of sales, which means approximately 10% of the observations are considered to face high regulatory risk). These cut-off points are somewhat arbitrary, but I believe they provide good approximations of which firms face higher pressure on either front.  I control for market capitalization (𝑆𝐼𝑍𝐸𝑖,𝑡) and sales growth (𝐺𝑅𝑂𝑊𝑇𝐻𝑖,𝑡) as larger firms and growth firms tend to attract more attention, and have more resources and ability to recruit a board of directors. I include industry competition, measured as the Herfindahl Index (𝐻𝐻𝐼𝑖,𝑡) calculated using the 2-digit SIC industry in COMPUSTAT, as the   18  level of competition affects incentives to seek political connections. Control of firm performance using return on assets (𝑅𝑂𝐴𝑖,𝑡) and a loss indicator (𝐿𝑂𝑆𝑆𝑖,𝑡) is also included, as high profitability of a firm can make it easier to attract politicians as board members. To address for selection issues, I also include a Heckman two-step model, where the selection model includes the aforementioned control variables of firm characteristics and industry. Following the argument in Hypothesis 1.1, when the dependent variable is EXECUTIVEt+1, α1 should be positive and significant; when the dependent variable is LEGISLATIVEt+1, α2 should be positive and significant. To test Hypothesis 1.2, I employ the following regression: ln(𝐵𝑒𝑛𝑒𝑓𝑖𝑡𝑠𝑖,𝑡+1)= 𝛽0 + 𝛽1𝑃𝑜𝑙𝑖𝑡𝑖𝑐𝑎𝑙 𝐶𝑜𝑛𝑛𝑒𝑐𝑡𝑖𝑜𝑛𝑖,𝑡 + 𝛽2𝑆𝐼𝑍𝐸𝑖,𝑡+1 + 𝛽3𝑀𝐵𝑖,𝑡+1 + 𝛽4𝐻𝐻𝐼𝑖,𝑡+1+ 𝛽5𝑅𝑂𝐴𝑖,𝑡+1 + 𝛽6𝐿𝑂𝑆𝑆𝑖,𝑡+1 + 𝛽7𝐿𝑁𝐴𝐺𝐸𝑖,𝑡+1 + 𝛽8𝐺𝑅𝑂𝑊𝑇𝐻𝑖,𝑡+1+ 𝛽9𝐶𝑂𝐺𝑆𝑖,𝑡+1 + 𝛽10𝐶𝐴𝑃𝐸𝑋𝑖,𝑡+1                                                                         (1.2.1) where 𝐵𝑒𝑛𝑒𝑓𝑖𝑡𝑠𝑖,𝑡+1 can be total government contract amount or subsidy amount for a firm in a year. As mentioned, variables of political connection are measured at the beginning of a fiscal year. Political connection measures include indicator variable PCB, which takes on the value of 1 when at least one member of the board is politically connected, percentage variable PPCB, which is the proportion of politically connected board members, and log number of political connections LN_NPC. Equation (1.2.1) controls for a firm’s current year market capitalization (𝑆𝐼𝑍𝐸𝑖,𝑡+1), market-to-book ratio (𝑀𝐵𝑖,𝑡+1), and industry competition (𝐻𝐻𝐼𝑖,𝑡+1) in order to capture the general characteristics of the firm. To control for profitability of the firm, two additional variables are included: return on assets (𝑅𝑂𝐴𝑖,𝑡+1) and a loss indicator (𝐿𝑂𝑆𝑆𝑖,𝑡+1).   19  Age of the firm (𝐿𝑁𝐴𝐺𝐸𝑖,𝑡+1) and sales growth (𝐺𝑅𝑂𝑊𝑇𝐻𝑖,𝑡+1) are controls of the business cycle. 𝐶𝑂𝐺𝑆𝑖,𝑡+1 is the ratio of cost of goods sold to sale, which captures the cost structure of the firm, and a measure of cost efficiency. This could be an important factor when competing for contract assignments and subsidies. The last control is the ratio of capital expenditure to sales (𝐶𝐴𝑃𝐸𝑋𝑖,𝑡+1), which controls for possible investment into increase production, especially in expectation of government contract needs. Hypothesis 2 predicts 𝛽1to be positive and significant.  To examine Hypothesis 1.3, I first regress current accounting performance measures on political connections, government contracts, subsidies, and lobbying, and the interaction terms between political connection and the latter three measures, in order to assess the impact of political connection on current firm performance. Next I regress future accounting performance, ranging from 1 to 5 years ahead, in the same format in order to assess the impact of political connection on future firm performance. 𝐴𝑐𝑐𝑜𝑢𝑛𝑡𝑖𝑛𝑔 𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒𝑡= 𝛾0 + 𝛾1𝑃𝐶𝐵𝑡−1 + 𝛾2𝐿𝑁𝐺𝐶𝑡−1 + 𝛾3𝐿𝑁𝑆𝑈𝐵𝑆𝐼𝐷𝐼𝐸𝑆𝑡−1 +  𝛾4𝑃𝐶𝐵𝑡−1 × 𝐿𝑁𝐺𝐶𝑡−1+ 𝛾5𝑃𝐶𝐵𝑡−1  × 𝐿𝑁𝑆𝑈𝐵𝑆𝐼𝐷𝐼𝐸𝑆𝑡−1 + 𝛾6𝑆𝐼𝑍𝐸𝑡−1 + 𝛾7𝐿𝑁𝐴𝐺𝐸𝑡−1 + 𝛾8𝐺𝑅𝑂𝑊𝑇𝐻𝑡−1+ 𝛾9𝑀𝐵𝑡−1 + 𝛾10𝐻𝐻𝐼𝑡−1 + 𝛾11𝐿𝑂𝑆𝑆𝑡−1                                                                 (1.3) Equation (1.3) regresses current period accounting performance measures, namely return on assets (ROA) and profit margin (PM) on last period’s PCB, LN_GC, LN_SUBSIDIES, and their interactions, when controlling for other firm characteristics. In this regression, γ1 should pick up the general profitability difference between politically connected firms and their non-connected counterparts; γ2 and γ3 are general association between firm performance and government contract and subsidy; γ4 and γ5 represent the marginal effects of having political connection on   20  top of having possible value extraction channels and lobbying. According to Hypothesis 1.3, the predicted sign for γ1 should be positive. If indeed, government contracts and lobbying are channels to extract political rents, then γ2 and γ3 are expected to be positive and significant as well. If having political connection does help to improve firm profitability when there are government contracts and subsidies, then γ4, and γ5 should be positive. 1.3.2 The 2006 and 2008 Elections As noted above, this paper explores the difference between 2006 midterm and 2008 presidential elections in order to address endogeneity. Endogeneity arises from reverse causality – firms that wants certain benefits will seek political connections, and those connections help them obtain these benefits. In this case, I look to the 2006 and 2008 election outcomes as shocks that have an impact on the ability of political connections to obtain government benefits, when it does not directly affect a firm’s ability to have certain political connections. The 2006 midterm election is the first time that the Democratic Party regained control of the US Congress since 1994. The Republican Party lost both houses of Congress and the majority of state governorships. The Democratic Party takeover was a complete reversal of the 1994 election, partly due to the declining public imagine of George W. Bush. This election is comparable to the 1994 election in that in both midterm elections, both houses of the Congress changed control. The incumbent political party’s defeat was complete and clear. Nevertheless, the midterm election did not alter the color of the White House. In this case, all of the executive branch government departments remained under the control of the Republican Party. Therefore, the 2006 midterm election serves as a bench mark, where Democratic connections should not have an increased influence in government allocation, even though they control the federal budget.   21  The 2008 presidential election brought Democrats back to power, with Barack Obama winning the largest percentage of the popular vote for a Democrat since 1964. As Obama nominated members of his own cabinet, the Democratic Party effectively took control of all the government departments, like the Department of Defense, the Department of Commerce, and the Department of Energy, where government procurements are common and substantial. Because of the aforementioned argument, political connections with the Democratic Party should be able to exert more influence after the 2008 election.  I utilize these two elections in my sample period, and rewrite Equation (1.2.1) into the following format: ln(𝐵𝑒𝑛𝑒𝑓𝑖𝑡𝑠𝑖,𝑡+1)= 𝛿0 + 𝛿1𝐷𝑒𝑚𝑜𝑐𝑟𝑎𝑡𝑖𝑐 𝐶𝑜𝑛𝑛𝑒𝑐𝑡𝑖𝑜𝑛𝑖,𝑡 + 𝛿2𝑅𝑒𝑝𝑢𝑏𝑙𝑖𝑐𝑎𝑛 𝐶𝑜𝑛𝑛𝑒𝑐𝑡𝑖𝑜𝑛𝑖,𝑡+ 𝛿3𝐸𝑙𝑒𝑐𝑡𝑖𝑜𝑛 𝐼𝑛𝑑𝑖𝑐𝑎𝑡𝑜𝑟+ 𝛿4𝐷𝑒𝑚𝑜𝑐𝑟𝑎𝑡𝑖𝑐 𝐶𝑜𝑛𝑛𝑒𝑐𝑡𝑖𝑜𝑛𝑖,𝑡  × 𝐸𝑙𝑒𝑐𝑡𝑖𝑜𝑛 𝐼𝑛𝑑𝑖𝑐𝑎𝑡𝑜𝑟+ 𝛿5𝑅𝑒𝑝𝑢𝑏𝑙𝑖𝑐𝑎𝑛 𝐶𝑜𝑛𝑛𝑒𝑐𝑡𝑖𝑜𝑛𝑖,𝑡  × 𝐸𝑙𝑒𝑐𝑡𝑖𝑜𝑛 𝐼𝑛𝑑𝑖𝑐𝑎𝑡𝑜𝑟+ 𝛿6𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠                                                                                (1.2.2) where the dependent variable is again either LN_GC or LN_SUBSIDIES, Election Indicator is the two years following the 2006 midterm election (fiscal years 2007 and 2008) or 2008 Presidential election (fiscal years 2009 and 2010). 𝐷𝑒𝑚𝑜𝑐𝑟𝑎𝑡𝑖𝑐 𝐶𝑜𝑛𝑛𝑒𝑐𝑡𝑖𝑜𝑛𝑖,𝑡 and 𝑅𝑒𝑝𝑢𝑏𝑙𝑖𝑐𝑎𝑛 𝐶𝑜𝑛𝑛𝑒𝑐𝑡𝑖𝑜𝑛𝑖,𝑡 take one of three forms: binary indicators (DDEM, DREP), log  number of connections (LN_DEM, LN_REP), or ratio of connections to board size (PDEM, PREP). In this regression, 𝛿3 picks up the budgetary difference before and after election. Meanwhile, because Republican Connection and Democratic Connection are available each year,   22  subject to a firm’s decision of adding or dropping connections, the measure of influence of different political connections on benefits will be represented by 𝛿4 and 𝛿5. When government contract is the dependent variable, the expected sign for 𝛿4 (𝛿5) is positive (negative) if the Election Indicator is for the 2008 presidential election. When government subsidy is the dependent variable, the expected sign for 𝛿4 (𝛿5) is positive (negative) if the Election Indicator is for the 2006 mid-term election. Controls of this regression include post-election indicators5, in order to separate the change of pre- and post- election periods, in addition to variables included in Equation (1.2.1). 1.4 Sample and data I start my sample with the Monthly S&P Constituents from CRSP from 2004 to 2013. Information on politically connected boards is collected from each firm’s annual proxy statement (DEF 14A) from EDGAR on the website of the Securities and Exchange Commission (SEC).6 At the beginning of the fiscal year, approximately 9-11 months before the issuance of the annual report, a firm usually files a proxy statement to nominate and confirm the board of directors, and to disclose additional information such as last fiscal year’s committee meetings attendance. For each nominated director, a short description of prior experience is included. I manually read through this part of the proxy statements in order to identify former politicians who are sitting on the board.  My classification of politically connected board members mostly follows Goldman et al. (2009). Note that for this paper, those who have worked for or are related to the judicial branch of government, Federal Reserve System, and those who only have military rankings are                                                  5 For 2006 election, post-election indicator equals to 1 for fiscal years 2008 to 2013; for 2008 election, post-election indicator equals to 1 for fiscal years of 2011, 2012, and 2013. 6 http://www.sec.gov/edgar/searchedgar/companysearch.html   23  not classified as politically connected. Also, those whose description mentioned a qualifying position but did not serve on the position (e.g. “served as a senior advisor to the Secretary of US Department of Defense”) are not considered political connections. For firm-years without proxy statements, I use information from the subsequent proxy statements for up to two years. If neither of the subsequent year information is available, then the observation is dropped. The result is a sample of 5,012 firm-year observations. The resulting sample is as shown in Table 1.1. Approximately 32% of the sample firm-years are politically connected firms, which is very similar to Goldman et al. (2009)’s sample of 31%, but considerably lower than Houston et al. (2014)’s 43%. As shown in Figure 3, the proportion of political connected firms amongst the S&P 500 is fairly stable over the sample period. Summary of industries in the sample, according to the 2-digit SIC code, is presented in Table 1.2. We see that there are indeed industry clusters of politically connected firms in the sample. Most noticeably industries like oil and gas, chemical and allied products, electric, gas, and sanitary services, insurance, and business services have the most politically connected observations. However, the proportion of political connected observations are all under 50%. If a politically connected board member is identified, then I performed a Google search on this person to identify his or her political party affiliation. Republicans and Democrats are classified accordingly, and any other political party association is classified as other. If this information is not readily available, then I look into the period when that person held office, and identify them as being part of the same party as the then-serving President of the United States. Because most of the positions, especially those that belong to the executive branch of government, are appointed directly by the United States President, I believe assuming that person’s political connection is from the same party as the President is reasonable. If, however,   24  that person served under both Democratic and Republican Presidents, then they are classified as other. For those who served in either house of the United States Congress, they are classified as legislatives; those who served under the executive branch are classified as executive. Politicians who are female or racial minorities are coded according to their names, pictures, and other online information7. For government contract information, I use the Federal Procurements Data System-Next Generation (FPDS-NG), and search each firm-year’s specific government contract information. The FPDS-NG has all federal procurements whose estimated value is $3,000 or more. The search is performed in the following way: for each firm, I search the main part of the firm name for a firm’s specific fiscal year, matching the fiscal year-end date. If there is more than one entity in the resulting list, I only include those that are confirmed to be wholly-owned subsidiaries, or venture businesses. The total dollar amount of action obligations8 and the total number of actions are collected. The resulting government contract amount is downward biased, as the amount omits contracts signed by the subsidiaries of the sample firms, whose names do not include the parent companies at all. This biases my sample towards not finding any results for politically connected firms. For lobbying expenses, I downloaded federal lobbying data provided by OpenSecrets.org.9 For all lobbying transactions that are required to be publically disclosed (even if the actual amount is zero) since 1998 to 2014, the transaction year, amount, registrant (lobbyist), and client names, and client parent name (interest groups, firms, and individuals) are recorded in the lobbying spreadsheet provided by OpenSecrets.org. I manually match these                                                  7 The information sources are the proxy statements, official government websites, Wikipedia.org, and nndb.com. 8 Action obligation represents the value of the contract. 9 For non-registered users, the source is http://data.influenceexplorer.com/bulk/.   25  transactions using client name and year into my own sample10, and generate the sum of lobbying expenses for each firm-year. Government subsidies information is compiled by and obtained from Good Jobs First, a policy resource center on subsidy data11. Good Jobs First’s Subsidy Tracker maintains entries for grants, loans, and other subsidies distributed by the federal, state, and local government as early as 1976. However, more comprehensive data is available since 2000. For each subsidy award entry, the Subsidy Tracker records the company as well as its parent firm (if applicable). If the parent firm belongs to one of the 2,782 parent companies covered, then a unique parent identification number (PARENT_ID) is given. Using this unique PARENT_ID, for each year, I calculate how many total subsidies and loans were granted by all levels of government, and then manually match this information to my sample. This means, when a subsidiary and a parent firm are both in my sample, the subsidies-related information is only for the parent firm. This creates issues for the accuracy of the matched information. However, it should only bias the sample in not finding statistical differences, as the amount of variation in government subsidies between firms dramatically decreased. Because subsidies can be in many different forms, I also look into some sub-categories of subsidies in the forms of government loans (separately listed in the dataset obtained from Good Jobs First), and tax credits (classified when subsidy type is listed as “federal allocated tax credit” or “tax credit/rebate”).                                                  10 I would like to thank Ting Xu for sharing his Stata name matching score code with me. With this program, I can generate a score between 0 and 1 for each of the name matches between the client parent names listed in the lobbying spreadsheet and firm names in my own sample, where 1 is a non-case sensitive perfect match. I then assume that matches with a score higher than 0.9 are all successful matches, and anything below 0.6 is an unsuccessful match. For those between 0.6 and 0.9, I conducted a manual check to ensure the match is correct.  11 http://www.goodjobsfirst.org/   26  Financial statement and market variables for each observation are collected from COMPUSTAT and CRSP. The summary statistics are as shown in Table 1.3. All variables are defined in the Appendix. I exclude financial firms (i.e., those with SIC 6000-6999) for all tests; however, other than firm performance comparisons, all results remain unchanged when financial industries are included. For hypothesis 1.3, tests for a sample that only includes observations receiving at least one form of government benefits are run but not tabulated, as the results remain essentially the same. Continuous variables are all winsorized at 1%. All standard errors are corrected for heteroscedasticity and are clustered by firm. 1.5 Results and interpretations Hypothesis 1.1 tests the validity of my assumption that firms target specific types of political connections according to their needs. The logit regression results are presented in Table 1.4. Panel A presents the results with the conditional subsample, where Panel B presents the results with full sample and the Heckman two-step model. In Panel A, the loading on government contract is positive and significant when the one period ahead connection type is with the executive branch, whether using a dummy variable GC or continuous proportion variable GOVT_SALE. Current regulation need is an important predictor for the one period ahead legislative branch connection, whether measured as a binary variable REGULATE or a continuous proportion variable LOB_SALE. In Panel B, with the full sample, we see the coefficient of GOVT_SALE is positive for prediction models regarding executive connection. However, in Heckman two-step model, coefficient on LOB_SALE is also positive and significant, though the magnitude is significantly less. Whereas the full sample for legislative connection, both GOVT_SALE and LOB_SALE’s coefficients are not statistically different from zero. This can be because the number of firms with legislative connection is a lot less than those   27  with executive connections. We also see that for the legislative connection models in Panel B, both pseudo R-square and Wald Chai-square are relatively small, compared to all the other models. Overall, the evidence is somewhat consistent with the prediction of Hypothesis 1.1. When obtaining government contracts is the goal, firms seek connections to the executive branch of government. However, when managing regulatory risks is the goal, they look for former congressional members to help them navigate the legislative landscape. Table 1.5 and 1.6 present the general correlations between different political connections and the two value extraction channels: government procurement and subsidies. In Table 1.5, we see a positive and significant association between various measures of political connections and government contracts in columns (1)-(3). In particular, the coefficient of the binary variable of PCB is 0.735 in column (1), which suggests a 208% difference in total contract amount between firms with and without political connection.12 This translates to an economic magnitude of approximately $72 million raw dollar amount in government contracts, as the sample has an average $36.27 million in government contracts. In column (4)-(5), I compare whether having connections in different political parties makes a difference. Even though, for all columns, Republican connections are loading positively and significantly, an F-test reveals that there is no significant difference between the loadings on Republican connections and Democratic connections. Column (6) examines whether political connections in different government branches have a different impact on government contracts. As expected, the coefficient on EXECUTIVE is positive and significant, suggesting that connections with the executive branch                                                  12 The amount is calculated as e0.735 = 2.085.   28  can influence the amount of government contracts a firm obtains. Taken as a whole, there is a positive relationship between political connections and government contracts. As for government subsidies, the results are tabulated in Table 1.6. For total subsidies, as shown in Panel A columns (1)-(3), the coefficients on political connection measures are all positive and significant. The 0.989 coefficient in Column (1) suggests that on average, politically connected firms get over 2.69 times in total subsidies compared to their non-connected peers13. This difference is both statistically and economically significant, as the sample average in total subsidies is around $17 million, which translates to a difference of over $45 million. The dataset also clearly classifies the indirect subsidy types, most commonly in the form of government loans and tax credits, I decide to take a deeper look into these two forms in Panel B and C. As shown in Panel B and C, even though most of the coefficients of political connection variables are positive, few of them are statistically significant. Column (4) in all panels tests whether connections to executive and legislative branch have a different impact on subsidies. Only government loans are shown to be statistically significant, yet the useful political connection is from the executive branch. Columns (5)-(7) in all panels are various regressions testing whether connections with a different political party have a different impact on subsidies. Positive and significant coefficients are found for firms with Democratic connections when it comes to total subsidies, yet the difference between Democratic and Republican connections are not statistically significant. Political party line does not seem to matter when it comes to loans and tax credits granted. Overall, there is a generally positive association between political connection and total subsidies.                                                  13 The amount is calculated as e0.989 = 2.689.   29  Table 1.7 tabulates the influence of different political connections on government procurement using the two elections within my sample period. In Panel A, Equation (1.2.2) is run for each election individually. This way, it is easier to see the difference in influence between the political party connections. As we see, the interaction between the post 2008 election indicator and Democratic connection always loads positively and significantly. This piece of evidence suggests that after the 2008 election, we see an increase in influence from Democratic connections on the allocation of government procurements. However, this same result does not present itself in the post 2006 election. In fact, there is no significant difference in changes of influence between the Democratic connections and the Republican connections.   In Table 1.7 Panel B, I include both post-election indicators, and tested the changes of influence for the Democratic connections between the two elections. As shown in the table, only the 2008 election results in positive and significant changes for the Democratic connections in their influence of government contract allocations, and this difference is significantly higher than that of the 2006 election.  Table 1.8 presents the same regressions run on subsidies. Panel A, B and C present results with total subsidies, total loan, and total government credit as the dependent variable, respectively. The positive and significant coefficients are for Democratic connections interacting with the 2006 election indicator. However, there is a lack of evidence that the 2006 election made any significant difference for firms with Democratic connections in their power to influence subsidies in general. In Panel B, however, government loans are found to have increased for firms with Democratic political connections, even though the F-test reveals that the difference is not statically different from those with Republican connections. Overall, the two   30  elections did not seem to significantly impact the different political connections differently when it comes to government subsidies.  Hypothesis 1.3 focuses on the firm performance of politically connected firms, whether indeed political rent extraction exists in the form of inefficient government procurement allocations and subsidies granting, and how political connection affects them on the margin. Table 1.9 presents the results of using current return on assets (ROA) and profit margin (PM) as profitability measures. There are some interesting findings that cast doubt on prior perception that politically connected firms are more profitable. In fact, when it comes to profit margin, I find that politically connected firms perform worse than their non-connected counterparts. Also, government contracts and subsidies do not, by themselves, seem to increase a firm’s accounting performance. The association between accounting performance and these two channels is negative and significant in general. This means that firms that are government contractors and subsidies receivers tend to do worse in accounting performance in general. This is consistent with the expectation that subsidies receivers are usually at a competitive disadvantage. But it can also be a result of firms receiving government contracts tend to have excess capacity, which translates to a lower current accounting performance. When political connection is added into the mix, interacting with other variables, it does not seem to increase or decrease the effect of the negative association between firm performance and the two channels. Results using return on equity (ROE) are not tabulated in the table, but they are essentially the same as using ROA as dependent variable. Table 1.10 presents results on future performance with each independent variable on its own, with industry and year fixed effects. Results are similar to that of Table 1.9. Coefficients on   31  government contracts and subsidies remain negative for all future ROA, which seem to imply that government benefits are awarded to firms with lower profitability. Table 1.11 presents results on both short term and long term future performance, focusing on 1 to 5 years ahead ROA as dependent variables. Panel A includes one value extraction channel at a time, whereas Panel B includes both channels together. I find that politically connected firms do enjoy a 3% advantage 3-4 years into the future in terms of ROA, when government contracts are taken into consideration. However, this advantage disappears in the 5th year. Overall, the data points to future benefits of politically connected firms, considering government contracts. This is possibly because current investment made to satisfy government contracts turns into productive assets later on. Table 1.11 also shows that neither interaction terms between PCB and the value extraction channels has positive correlation with future ROA. This suggests that on the margin, having political connections does not significantly improve firms’ profitability, when firms receive government contracts and subsidies. As mentioned, the same set of tests are performed for a subsample that includes observations receiving at least one form of the two government benefits, and the results remain the same. In this case, I argue the evidence implies that the government contract bidding process and subsidy granting is still efficient, and the stricter legal system in the United States makes it difficult to extract political rent. Overall, I find that politically connected firms are not outperforming their non-connected peers. 1.6 Conclusion This paper examines the existence and effects of corporate political connections, namely using politically connected boards. First, I examine if firms target different political connections by considering different needs. I find evidence that government contractors seek to connect with   32  the executive branch, where the majority of contracts are from, and firms that need to manage regulatory risks align themselves with former congressional members. Next, I identify two direct channels where political connections can contribute to firm value: government contracts and subsidies. I find that politically connected firms do receive more government contracts and subsidies. To address endogeneity, I explore the effects of political connections on government procurements and subsidies before and after the 2006 and 2008 elections. Results demonstrate that the 2008 Presidential election affected firms with different political connections differently in regards to the amount of government contracts they receive. In particular, firms with winning party connections saw an increase in the amount of government contracts after the 2008 general election, but not the 2006 mid-term election. Lastly, this paper examines current and future firm performance in relation to political connections. Overall, the findings cast doubt on prior research’s conclusions based on evidence from other countries, as I failed to find evidence that politically connected firms enjoy higher profits. Meanwhile, the two value extraction channels that political connections can contribute to do not benefit firm profitability. In my sample, government contracts and subsidies are shown to negatively correlate with accounting performances, and having political connections does not decrease this negative impact. Interestingly, in 3-4 years, firms with political connections seem to have higher ROA temporarily when their involvement in government contracting is taken into consideration. This is consistent with the assertion that government contractors can benefit in the future from current investments, possibly to fulfill the demand for current government contracts. Overall, my findings suggest that, at least in the United States, the government contract bidding process and legal system are effective in safeguarding contract overpricing and political cronyism.    33  Figure 1.1 Discretionary Spending of Federal Government for 1990 – 1998 (in Billion $) This graph plots annual discretionary spending of the United States Federal Government for the period of 1990-1998.        6406606807007207407607808001990 1991 1992 1993 1994 1995 1996 1997 1998  34  Figure 1.2 Timeline of Political Power in Control of the United States Government. This figure demonstrates the time line of the political party that controls the United States White House and both houses of the United States Congress. The 2006 midterm election saw both houses changes from a Republican majority to a Democratic majority. During the 2008 presidential election, not only did the Democratic Party maintain control of both houses of Congress, but also Democrat Barack Obama was elected the 44th President of the United States, replacing Republican George W. Bush.           35  Figure 1.3 Number of Politically Connected Firms by Year. Figure 1.3 is a graphic demonstration of the breakdown of politically connected firms (PCB=1) and non-connected firms (PCB=0) in the sample by year.         01002003004005006002004 2005 2006 2007 2008 2009 2010 2011 2012 2013PCB=0 PCB=1  36  Table 1.1 Number of Politically Connected Firms by Year. This table presents the number of observations in the sample by year, and lists a breakdown of politically connected firms (PCB=1) and non-connected firms (PCB=0) in the sample by year.      PCB Year 0 1 Total 2004 339 154 493 2005 329 158 487 2006 322 171 493 2007 336 168 504 2008 340 168 508 2009 339 172 511 2010 334 167 501 2011 337 169 506 2012 341 160 501 2013 348 160 508 Total 3365 1647 5012     37  Table 1.2 Industry Composition This table presents the breakdown of industry according to 2-digit SIC in the S&P 500 index from 2004 to 2013. 2-digit SIC Industry Full Sample PCB=1 Percentage of PCB=1 Frequency Percentage Frequency 1 Agricultural Production - Crops 10 0.20% 10 100.00% 10 Metal, Mining 25 0.49% 11 44.00% 12 Coal Mining 21 0.41% 8 38.10% 13 Oil & Gas Extraction 225 4.43% 90 40.00% 14 Nonmetallic Minerals, Except Fuels 10 0.20% 7 70.00% 15 General Building Contractors 39 0.77% 8 20.51% 16 Heavy Construction, Except Building 17 0.33% 15 88.24% 17 Special Trade Contractors 5 0.10% 0 0.00% 20 Food & Kindred Products 229 4.51% 60 26.20% 21 Tobacco Products 36 0.71% 19 52.78% 22 Textile Mill Products 1 0.02% 0 0.00% 23 Apparel & Other Textile Products 40 0.79% 14 35.00% 24 Lumber & Wood Products 23 0.45% 4 17.39% 25 Furniture & Fixtures 20 0.39% 0 0.00% 26 Paper & Allied Products 75 1.48% 31 41.33% 27 Printing & Publishing 41 0.81% 19 46.34% 28 Chemical & Allied Products 401 7.90% 118 29.43% 29 Petroleum & Coal Products 69 1.36% 32 46.38% 30 Rubber & Miscellaneous Plastics Products 31 0.61% 18 58.06% 31 Leather & Leather Products 10 0.20% 0 0.00% 32 Stone, Clay, & Glass Products 5 0.10% 0 0.00% 33 Primary Metal Industries 56 1.10% 24 42.86% 34 Fabricated Metal Products 42 0.83% 10 23.81% 35 Industrial Machinery & Equipment 279 5.49% 77 27.60% 36 Electronic & Other Electric Equipment 345 6.79% 77 22.32% 37 Transportation Equipment 124 2.44% 63 50.81% 38 Instruments & Related Products 238 4.69% 77 32.35% 39 Miscellaneous Manufacturing Industries 30 0.59% 0 0.00% 40 Railroad Transportation 37 0.73% 24 64.86% 42 Trucking & Warehousing 10 0.20% 10 100.00% 44 Water Transportation 10 0.20% 6 60.00% 45 Transportation by Air 22 0.43% 14 63.64% 47 Transportation Services 21 0.41% 0 0.00% 48 Communications 176 3.47% 68 38.64% 49 Electric, Gas, & Sanitary Services 391 7.70% 165 42.20% 50 Wholesale Trade - Durable Goods 36 0.71% 5 13.89% 51 Wholesale Trade - Nondurable Goods 45 0.89% 10 22.22%         38  2-digit SIC Industry Full Sample PCB=1 Percentage of PCB=1 Frequency Percentage Frequency       52 Building Materials & Gardening Supplies 26 0.51% 8 30.77% 53 General Merchandise Stores 105 2.07% 27 25.71% 54 Food Stores 40 0.79% 6 15.00% 55 Automotive Dealers & Service Stations 29 0.57% 5 17.24% 56 Apparel & Accessory Stores 56 1.10% 4 7.14% 57 Furniture & Home Furnishings Stores 39 0.77% 4 10.26% 58 Eating & Drinking Places 47 0.93% 28 59.57% 59 Miscellaneous Retail 81 1.60% 11 13.58% 60 Depository Institutions 256 5.04% 40 15.63% 61 Non-Depository Institutions 66 1.30% 44 66.67% 62 Security & Commodity Brokers 153 3.01% 44 28.76% 63 Insurance Carriers 255 5.02% 117 45.88% 64 Insurance Agents, Brokers, & Service 20 0.39% 10 50.00% 65 Real Estate 15 0.30% 8 53.33% 67 Holding & Other Investment Offices 133 2.62% 19 14.29% 70 Hotels & Other Lodging Places 23 0.45% 20 86.96% 72 Personal Services 10 0.20% 6 60.00% 73 Business Services 392 7.72% 94 23.98% 75 Auto Repair, Services, & Parking 13 0.26% 8 61.54% 78 Motion Pictures 4 0.08% 0 0.00% 79 Amusement & Recreation Services 20 0.39% 16 80.00% 80 Health Services 45 0.89% 18 40.00% 82 Educational Services 21 0.41% 3 14.29% 87 Engineering & Management Services 20 0.39% 0 0.00% 99 Non-Classifiable Establishments 14 0.28% 10 71.43% Total 5,078 100% 1,646 32.41%            39  Table 1.3 Summary Statistics This table presents the summary statistics for different variables in the sample. Detailed variable definitions are in Appendix A.1. Panel A tabulates the summary statistics of political connections. Panel B presents the summary statistics for other variables used in all tests in non-politically connected observations (PCB=0), politically connected observations (PCB=1), and the full sample.     Panel A: Summary Statistics for Political Connections  N mean standard deviation p25 p50 p75 PCB 5012 0.329 0.470 0 0 1 NUM_PCB 5012 0.446 0.744 0 0 1 PPCB 5012 0.040 0.067 0 0 0.083 LN_NPC 5012 0.271 0.411 0 0 0.693 DREP 5012 0.197 0.398 0 0 0 DDEM 5012 0.186 0.389 0 0 0 LN_DEM 5012 0.128 0.287 0 0 0 LN_REP 5012 0.141 0.307 0 0 0 PREP 5012 0.020 0.046 0 0 0 PDEM 5012 0.018 0.041 0 0 0 LEGISLATIVE 5012 0.124 0.330 0 0 0 EXECUTIVE 5012 0.269 0.444 0 0 1     40  Panel B: Summary Statistics for Other Variables     PCB=0        N mean standard deviation p25 p50 p75 LN_GC 2081 14.728 3.291 12.267 15.025 17.107 GOVT_SALE 3363 0.005 0.027 0 0.000 0.001 LN_SUBSIDIES 3365 8.128 7.113 0 11.484 14.473 LN_LOAN 3365 1.646 5.118 0 0 0 LN_TAX_CREDIT 3365 4.389 6.356 0 0 11.753 LOB_SALE 3363 0.134 0.432 0.000 0.028 0.132 COGS 3363 0.560 0.224 0.405 0.583 0.729 CAPEX 3363 0.075 0.140 0.020 0.036 0.065 HHI 3363 0.226 0.200 0.084 0.169 0.296 ROA 2707 0.055 0.064 0.018 0.050 0.090 SIZE 3359 9.236 0.995 8.560 9.165 9.793 LNAGE 3347 3.262 0.752 2.789 3.391 3.773 GROWTH 3362 0.080 0.172 -0.002 0.068 0.150 MB 3359 3.047 3.304 1.549 2.457 3.864 LOSS 3365 0.091 0.287 0 0 0        PCB=1        N mean standard deviation p25 p50 p75 LN_GC 1146 16.395 3.600 14.183 16.548 18.683 GOVT_SALE 1646 0.026 0.113 0 0.000 0.003 LN_SUBSIDIES 1647 10.137 7.152 0 13.355 15.763 LN_LOAN 1647 2.614 6.351 0 0 0 LN_TAX_CREDIT 1647 5.222 6.818 0 0 13.007 LOB_SALE 1646 0.188 0.254 0.016 0.099 0.249 COGS 1646 0.620 0.333 0.453 0.672 0.783 CAPEX 1646 0.075 0.122 0.019 0.038 0.081 HHI 1646 0.239 0.206 0.092 0.189 0.312 ROA 1399 0.054 0.064 0.022 0.046 0.087 SIZE 1645 9.770 1.141 8.961 9.716 10.496 LNAGE 1643 3.442 0.787 2.914 3.607 4.110 GROWTH 1646 0.067 0.167 -0.015 0.059 0.126 MB 1645 3.527 4.314 1.559 2.424 4.147 LOSS 1647 0.090 0.287 0 0 0                        41         Full Sample        N mean standard deviation p25 p50 p75 LN_GC 3227 15.320 3.496 12.806 15.603 17.609 GOVT_SALE 5009 0.012 0.069 0 0.000 0.002 LN_SUBSIDIES 5012 8.788 7.187 0 12.247 14.926 LN_LOAN 5012 1.964 5.571 0 0 0 LN_TAX_CREDIT 5012 4.663 6.522 0 0 12.136 LOB_SALE 5009 0.151 0.384 0 0.049 0.174 COGS 5009 0.580 0.266 0.416 0.608 0.748 CAPEX 5009 0.075 0.134 0.020 0.036 0.070 HHI 5009 0.230 0.202 0.085 0.178 0.303 ROA 4106 0.055 0.064 0.019 0.048 0.089 SIZE 5004 9.411 1.075 8.676 9.323 10.034 LNAGE 4990 3.321 0.768 2.820 3.500 3.851 GROWTH 5008 0.076 0.170 -0.007 0.065 0.140 MB 5004 3.205 3.673 1.553 2.450 3.913 LOSS 5012 0.091 0.287 0 0 0     42  Table 1.4 Choices of Political Connections This table examines whether a politically connected firm targets political connection according to various needs. Panel A presents the logit regression results with the sample contains only firms with political connections (PCB=1). Panel B presents the logit regression and the Heckman two-step results with the full sample. Political connection variables are measured at the beginning of the year. Dependent variables are one period ahead of the executive branch connection indicator for the first two columns, and one period ahead of the legislative branch connection indicator for the latter two columns. EXECUTIVE (LEGISLATIVE) is equal to 1 when a firm has at least one politically connected board member that was identified to have executive (legislative) branch experience. GOVT_SALE is the ratio of government contracts to sales, whereas LOB_SALE is 1000 times the ratio of lobbying expense to sales, GC and REGULATE are dummy variables, which is equal to 1 when GOVT_SALE is greater or equal to 0.1 and LOB_SALE is greater or equal to 0.3, respectively. Control variables are return on assets (ROA), Herfindahl index (HHI), market size (SIZE), firm age (LNAGE), sales growth (GROWTH), and loss indicator (LOSS). In the Heckman model, these control variables and industry fixed effects are included in the selection model. Detailed variable definitions are in Appendix A.1. Standard errors are corrected for heteroscedasticity and clustered at firm level. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.     43  Panel A.      EXECUTIVEt+1 LEGISLATIVEt+1 GOVT_SALE 11.811***   -3.028   (4.407)   (4.026)  LOB_SALE 0.029   1.142*   (0.710)   (0.679)  GC  1.806*  -1.328   (0.935)  (1.345) REGULATE  0.088  0.826**   (0.415)  (0.382) ROA 0.400 0.300 -2.539 -2.811  (2.999) (2.992) (3.099) (3.132) SIZE 0.408*** 0.420*** 0.166 0.161  (0.157) (0.154) (0.190) (0.192) GROWTH -0.258 -0.283 0.942* 0.952*  (0.480) (0.481) (0.497) (0.502) HHI 2.633** 2.859** -2.080 -2.059  (1.294) (1.277) (1.956) (1.988) LOSS -0.101 -0.167 0.201 0.126  (0.427) (0.433) (0.434) (0.433) Industry Fixed Effects Yes Yes Yes Yes Year Fixed Effects Yes Yes Yes Yes N 1116 1116 960 960 pseudo R-sq 0.139 0.133 0.153 0.159     44  Panel B.          EXECUTIVEt+1 LEGISLATIVEt+1 GOVT_SALE 9.790*** 1.797*** 0.801 0.202  (2.517) (0.184) (3.271) (0.130) LOB_SALE 0.020 0.069*** 0.216 0.020  (0.208) (0.024) (0.222) (0.017) ROA -2.019   -2.337   (1.653)   (2.176)  SIZE 0.467***   0.280**   (0.092)   (0.123)  GROWTH -0.441   0.080   (0.281)   (0.349)  HHI 0.656   -0.631   (0.523)   (0.784)  LOSS -0.109   0.475   (0.247)   (0.298)  Industry Fixed Effects Yes Yes Yes  Yes Year Fixed Effects Yes No Yes No N 3343 4118 3103 4118 pseudo R-sq 0.112   0.087  Wald Chi-sq   109.92   4.22   45  Table 1.5 Government Contracts and Political Connections This table examines the relation between government contracts and political connections. The dependent variable is LN_GC, which is the natural log of government contract amount. Political connection variables are measured at the beginning of the year. PCB is an indicator of political connection, which takes on the value 1 when at least one board member of the firm-year is identified as politically connected. PPCB is the proportion of politically connected board, which is calculated by the number of politically connected board members over the size of the board. LN_NPC is the log number of politically connected board members. PREP (PDEM) are percentage of directors that are identified as being connected to the Republican (Democratic) Party. LN_REP (LN_DEM) is log number of political connections that are identified as being connected to the Republican (Democratic) Party. LEGISLATIVE (EXECUTIVE) are indicators of having at least one political connection with legislative (executive) branch experience in the United States government. Detailed variable definitions are in Appendix A.1. Standard errors are corrected for heteroscedasticity and clustered at firm level. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.                  46   Expected Sign LN_GC  (1) (2) (3) (4) (5) (6) PCB + 0.735***        (0.282)      PPCB +  5.024***        (1.914)     LN_NPC +   0.937***        (0.317)    PREP     7.110***        (2.665)   PDEM     1.499        (2.751)   LN_REP      1.143***        (0.390)  LN_DEM      0.374        (0.409)  LEGISLATIVE       0.177        (0.406) EXECUTIVE +      0.834***        (0.292) COGS  3.806*** 3.789*** 3.784*** 3.768*** 3.742*** 3.697***   (1.000) (0.985) (0.988) (0.989) (0.987) (0.995) CAPEX  -1.099 -1.069 -1.001 -1.118 -1.039 -1.238   (1.351) (1.358) (1.351) (1.346) (1.346) (1.352) HHI  -0.373 -0.505 -0.459 -0.531 -0.509 -0.380   (0.970) (0.968) (0.963) (0.962) (0.957) (0.961) ROA  -3.521 -3.716 -3.510 -3.688 -3.414 -3.649   (2.527) (2.555) (2.543) (2.614) (2.574) (2.536) SIZE  0.942*** 0.956*** 0.924*** 0.966*** 0.935*** 0.920***   (0.157) (0.156) (0.157) (0.157) (0.157) (0.159) LNAGE  0.589*** 0.605*** 0.582*** 0.613*** 0.598*** 0.562***   (0.206) (0.209) (0.207) (0.209) (0.208) (0.205) GROWTH  -1.129** -1.171** -1.137** -1.225** -1.199** -1.136**   (0.489) (0.493) (0.490) (0.498) (0.496) (0.489) MB  -0.034 -0.034 -0.034 -0.034 -0.034 -0.030   (0.024) (0.024) (0.024) (0.024) (0.024) (0.025) LOSS  -0.805** -0.830** -0.824** -0.804** -0.804** -0.757**   (0.354) (0.354) (0.354) (0.362) (0.359) (0.354) Industry Fixed Effects  Yes Yes Yes Yes Yes Yes Year Fixed Effects  Yes Yes Yes Yes Yes Yes N  2230 2230 2230 2230 2230 2230 adj. R-sq  0.475 0.476 0.478 0.477 0.478 0.477 Dem = Rep     2.15 1.92  Legislative = Executive             1.59     47  Table 1.6 Government Subsidies and Political Connections This table presents the relation between government subsidies and political connections. Panel A presents the relation between total government subsidies and political connections. The dependent variable is LN_SUBSIDIES, which is the natural log of the amount of government subsidies (includes government loans, tax credits, government grants, tax rebates, and other forms of subsidies) of a firm year at the end of the year. Panel B and C presents the same relation in the sub-category of subsidies, government loan, and tax credit. In Panel B, dependent variable is LN_LOAN, which is the log amount of government loans granted to a firm at the end of the year. In Panel C, dependent variable is LN_TAX_CREDIT, the log amount of total tax credits available to the firm at the end of the year. Political connection variables are measured at the beginning of the year. PCB is an indicator of political connection, which takes on the value 1 when at least one board member of the firm-year is identified as politically connected. PPCB is the proportion of politically connected board, which is calculated by the number of politically connected board members over size of the board. LN_NPC is the log number of politically connected board members. PREP (PDEM) are percentage of directors that are identified as being connected to the Republican (Democratic) Party. LN_REP (LN_DEM) is the log number of political connections that are identified as being connected to the Republican (Democratic) Party. LEGISLATIVE (EXECUTIVE) are indicators of having at least one political connection with legislative (executive) branch experience in the United States government. Control variables are cost of goods sold (COGS), capital expenditure (CAPEX), Herfindahl index (HHI), return on assets (ROA), market size (SIZE), firm age (LNAGE), sales growth (GROWTH), market-to-book ratio (MB), and loss indicator (LOSS). Details on the variable definitions are provided in Appendix A.1. Standard errors are corrected for heteroscedasticity and clustered at firm level.*, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.       48  Panel A. Total Subsidies Expected Sign LN_SUBSIDIES  (1) (2) (3) (4) (5) (6) (7) PCB + 0.989**         (0.475)       PPCB +  7.887***         (2.972)      LN_NPC +   1.307**         (0.533)     LEGISLATIVE +    -0.384         (0.530)    EXECUTIVE     0.662         (0.458)    DDEM      1.173**         (0.576)   DREP      0.514         (0.528)   PDEM       12.926***         (4.932)  PREP       4.386         (4.111)  LN_DEM        1.710**         (0.761) LN_REP        0.742         (0.645) Controls  Yes Yes Yes Yes Yes Yes Yes Industry Fixed Effects  Yes Yes Yes Yes Yes Yes Yes Year Fixed Effects  Yes Yes Yes Yes Yes Yes Yes N  3265 3265 3265 3265 3265 3265 3265 adj. R-sq   0.326 0.327 0.327 0.323 0.326 0.328 0.327 Legislative=Executive     1.92    Dem = Rep           0.69 1.60 0.92    49  Panel B. Loans Expected Sign LN_LOAN  (1) (2) (3) (4) (5) (6) (7) PCB + 0.489*         (0.266)       PPCB +  2.341         (1.938)      LN_NPC +   0.516         (0.328)     LEGISLATIVE +    -0.455         (0.357)    EXECUTIVE     0.532*         (0.271)    DDEM      0.365         (0.365)   DREP      0.509         (0.345)   PDEM       2.010         (3.407)  PREP       3.603         (2.862)  LN_DEM        0.446         (0.505) LN_REP        0.670         (0.449) Controls  Yes Yes Yes Yes Yes Yes Yes Industry Fixed Effects  Yes Yes Yes Yes Yes Yes Yes Year Fixed Effects  Yes Yes Yes Yes Yes Yes Yes N  3265 3265 3265 3265 3265 3265 3265 adj. R-sq   0.202 0.201 0.201 0.202 0.202 0.201 0.202 Legislative=Executive     4.09**    Dem = Rep           0.07 0.12 0.10    50  Panel C. Tax Credits Expected Sign LN_TAX_CREDIT  (1) (2) (3) (4) (5) (6) (7) PCB + 0.053         (0.359)       PPCB +  0.184         (2.439)      LN_NPC +   -0.012         (0.414)     LEGISLATIVE +    -0.384         (0.462)    EXECUTIVE     0.360         (0.370)    DDEM      -0.165         (0.464)   DREP      0.288         (0.393)   PDEM       -0.052         (4.343)  PREP       0.871         (3.022)  LN_DEM        -0.193         (0.633) LN_REP        0.173         (0.482) Controls  Yes Yes Yes Yes Yes Yes Yes Industry Fixed Effects  Yes Yes Yes Yes Yes Yes Yes Year Fixed Effects  Yes Yes Yes Yes Yes Yes Yes N  3265 3265 3265 3265 3265 3265 3265 adj. R-sq   0.335 0.335 0.335 0.335 0.335 0.335 0.335 Legislative=Executive     1.33    Dem = Rep           0.59 0.03 0.22   51  Table 1.7 Government Procurement Post 2006 and 2008 Election This table presents the results of how the 2006 and 2008 elections affected firms with different political connections’ ability to obtain government contracts. Panel A presents the results for each election separately. Panel B presents results with two elections together. The dependent variable is LN_GC, which is the log amount of government contracts a firm gets for a year. Control variables include appropriate Democratic connection measures (DDEM, PDEM, LN_DEM), matching Republican connection measures (DREP, PREP, LN_REP), and 2006 and 2008 election indicators. Other controls include cost of goods sold (COGS), capital expenditure (CAPEX), Herfindahl index (HHI), return on assets (ROA), market size (SIZE), firm age (LNAGE), sales growth (GROWTH), market-to-book ratio (MB), and loss indicator (LOSS). Details on the variable definitions are provided in Appendix A.1. Standard errors are corrected for heteroscedasticity and clustered at firm level. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.     52  Panel A. Individual Elections   LN_GC  Expected Sign 2008 Election 2006 Election 08 ELECTION×DDEM + 0.786***         (0.302)       08 ELECTION×DREP  -0.186         (0.315)       08 ELECTION×LN_DEM +  1.013**         (0.419)      08 ELECTION×LN_REP   -0.294         (0.342)      08 ELECTION×PDEM +   7.512**        (2.967)    08 ELECTION×PREP    -1.816        (2.057)    06 ELECTION×DDEM ?     0.031         (0.279)   06 ELECTION×DREP      -0.283         (0.260)   06 ELECTION×LN_DEM ?      0.067         (0.351)  06 ELECTION×LN_REP       -0.436         (0.306)  06 ELECTION×PDEM ?       0.538         (2.364) 06 ELECTION×PREP        -3.251*         (1.946) Election×Dem = Election×Rep  4.54** 5.35** 5.95** 0.61 1.08 1.34 Controls  Yes Yes Yes Yes Yes Yes Industry Fixed Effects  Yes Yes Yes Yes Yes Yes Year Fixed Effects  No No No No No No N  2230 2230 2230 2230 2230 2230 adj. R-sq   0.474 0.475 0.474 0.473 0.475 0.473     53  Panel B. Both Elections      Expected Sign LN_GC 08 ELECTION×DDEM + 0.791**     (0.341)   08 ELECTION×DREP  -0.296     (0.372)   06 ELECTION×DDEM ? 0.035     (0.279)   06 ELECTION×DREP  -0.286     (0.260)   08 ELECTION×LN_DEM +  1.033**     (0.473)  08 ELECTION×LN_REP   -0.465     (0.410)  06 ELECTION×LN_DEM ?  0.070     (0.351)  06 ELECTION×LN_REP   -0.441     (0.306)  08 ELECTION×PDEM +   7.699**     (3.430) 08 ELECTION×PREP    -3.088     (2.487) 06 ELECTION×PDEM ?   0.564     (2.366) 06 ELECTION×PREP    -3.287*     (1.953) 08 Election×Dem = 06 Election×Dem  5.73** 5.31** 6.46** Controls  Yes Yes Yes Industry Fixed Effects  Yes Yes Yes Year Fixed Effects  No No No N  2230 2230 2230 adj. R-sq   0.474 0.476 0.474      54  Table 1.8 Government Subsidies Post 2006 and 2008 Election This table presents the results of how the 2006 and 2008 elections affected firms with different political connections’ ability in receiving government subsidies. Panel A presents the results for total subsidies, where the dependent variable is the log amount of total subsidies (LN_SUBSIDIES). Panel B and C present results for loan and tax credits, where the dependent variable is the log amount of government loan (LN_LOAN) and the log amount of tax credit (LN_TAX_CREDIT), respectively. Control variables include appropriate Democratic connection measures (DDEM, PDEM, LN_DEM), matching Republican connection measures (DREP, PREP, LN_REP), and 2006 and 2008 election indicators. Other controls include cost of goods sold (COGS), capital expenditure (CAPEX), Herfindahl index (HHI), return on assets (ROA), market size (SIZE), firm age (LNAGE), sales growth (GROWTH), market-to-book ratio (MB), and loss indicator (LOSS). Details on the variable definitions are provided in Appendix A.1. Standard errors are corrected for heteroscedasticity and clustered at firm level. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.       55  Panel A. Total Subsidies   LN_SUBSIDIES  Expected Sign 2006 Election 2008 Election DDEM×2006 ELECTION + -0.026         (0.836)       DREP×2006 ELECTION  -0.831         (0.697)       LN_DEM×2006 ELECTION +  0.177         (1.075)      LN_REP×2006 ELECTION   -1.208         (0.908)      PDEM×2006 ELECTION +   1.287        (6.950)    PREP×2006 ELECTION    -7.188        (6.420)    DDEM×2008 ELECTION ?     0.886         (0.747)   DREP×2008 ELECTION      0.324         (0.698)   LN_DEM×2008 ELECTION ?      1.197         (0.999)  LN_REP×2008 ELECTION       0.345         (0.856)  PDEM×2008 ELECTION ?       7.323         (6.614) PREP×2008 ELECTION        2.070         (5.288) Election×Dem = Election×Rep  0.1 0.00 0.01 0.04 0.05 0.05 Controls  Yes Yes Yes Yes Yes Yes Industry Fixed Effects  Yes Yes Yes Yes Yes Yes Year Fixed Effects  No No No No No No N  3265 3265 3265 3265 3265 3265 adj. R-sq   0.322 0.323 0.324 0.309 0.310 0.311        56  Panel B. Loan LN_LOAN  2006 Election 2008 Election DDEM×2006 ELECTION 0.721        (0.789)       DREP×2006 ELECTION 0.561        (0.831)       LN_DEM×2006 ELECTION  1.190        (1.087)      LN_REP×2006 ELECTION  0.793        (0.980)      PDEM×2006 ELECTION   6.825       (6.718)    PREP×2006 ELECTION   4.215       (6.239)    DDEM×2008 ELECTION     1.612**        (0.734)   DREP×2008 ELECTION     0.635        (0.720)   LN_DEM×2008 ELECTION      1.730*        (1.000)  LN_REP×2008 ELECTION      0.781        (0.911)  PDEM×2008 ELECTION       10.553        (7.142) PREP×2008 ELECTION       3.205        (5.793) Election×Dem = Election×Rep 0.00 0.07 0.15 0.80 0.41 0.44 Controls Yes Yes Yes Yes Yes Yes Industry Fixed Effects Yes Yes Yes Yes Yes Yes Year Fixed Effects No No No No No No N 3265 3265 3265 3265 3265 3265 adj. R-sq 0.202 0.202 0.200 0.177 0.176 0.174      57   Panel C. Tax Credit LN_TAX_CREDIT  2006 Election 2008 Election DDEM×2006 ELECTION -0.316        (0.647)       DREP×2006 ELECTION 0.154        (0.691)       LN_DEM×2006 ELECTION  -0.183        (0.882)      LN_REP×2006 ELECTION  -0.133        (0.871)      PDEM×2006 ELECTION   -0.141       (6.215)    PREP×2006 ELECTION   -1.390       (5.575)    DDEM×2008 ELECTION     0.936        (0.687)   DREP×2008 ELECTION     0.167        (0.674)   LN_DEM×2008 ELECTION      0.969        (0.978)  LN_REP×2008 ELECTION      -0.015        (0.792)  PDEM×2008 ELECTION       2.105        (7.164) PREP×2008 ELECTION       -2.446        (4.856) Election×Dem = Election×Rep 1.28 0.58 0.15 0.69 0.99 0.79 Controls Yes Yes Yes Yes Yes Yes Industry Fixed Effects Yes Yes Yes Yes Yes Yes Year Fixed Effects No No No No No No N 3265 3265 3265 3265 3265 3265 adj. R-sq 0.331 0.331 0.330 0.306 0.306 0.305    58  Table 1.9  Current Firm Performance This table examines the current accounting performance of politically connected firms. Dependent variables for column (1)-(6) are current return on assets (ROA); for column (7)-(14) are current profit margin (PM). PCB is an indicator for political connections, LN_GC is the log amount of total government contracts received by the firm in the year, LN_SUBSIDIES is the log amount of total subsidies received by the firm. Control variables include market size (SIZE), firm age (LNAGE), sales growth (GROWTH), market-to-book ratio (MB), Herfindahl index (HHI), and loss indicator (LOSS). Details on the variable definitions are provided in Appendix A.1. Standard errors are corrected for heteroscedasticity and clustered at firm level. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.   Expected Sign ROAt PMt  (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) PCBt-1 + -0.002   0.010 0.001 0.008 -0.011**   0.020 0.001 0.017   (0.004)   (0.022) (0.007) (0.022) (0.005)   (0.029) (0.009) (0.028) LN_GC t-1   -0.002***  -0.002**  -0.002**  -0.005***  -0.004***  -0.004***    (0.001)  (0.001)  (0.001)  (0.001)  (0.001)  (0.001) LN_SUBSIDIES t-1    -0.001***  -0.001* -0.001***   -0.002***  -0.001*** -0.001**     (0.000)  (0.000) (0.000)   (0.000)  (0.000) (0.000) PCB t-1×LN_GC t-1 +    -0.001  -0.001    -0.001  -0.001      (0.001)  (0.001)    (0.002)  (0.002) PCB t-1×LN_SUBSIDIES t-1 +     -0.000 -0.000     -0.001* -0.000       (0.000) (0.001)     (0.001) (0.001) Controls  Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Industry Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Effects  Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes N  2801 1972 2815 1969 2801 1969 3532 2445 3577 2428 3532 2428 adj. R-sq   0.304 0.329 0.314 0.328 0.307 0.333 0.257 0.303 0.260 0.306 0.264 0.311   59  Table 1.10 Future Firm Performance – Single Factor This table presents the analysis on future accounting performance of politically connected firms with a single factor. Dependent variables are one-year-ahead ROA (FROA) for column (1)-(3), two-year-ahead ROA (FROA2) for column (4)-(6), three-year-ahead ROA (FROA3) for column (7)-(9), four-year-ahead ROA (FROA4) for column (10)-(12), and five-year-ahead ROA (FROA5) for column (13)-(15). Control variables include market size (SIZE), firm age (LNAGE), sales growth (GROWTH), market-to-book ratio (MB), and Herfindahl index (HHI). Details on the variable definitions are provided in Appendix A.1. Standard errors are corrected for heteroscedasticity and clustered at firm level. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.  FROA FROA2 FROA3  (1) (2) (3) (4) (5) (6) (7) (8) (9) PCB -0.005    -0.003    -0.001     (0.004)    (0.004)    (0.004)   LN_GC  -0.002***    -0.002**    -0.002*    (0.001)    (0.001)    (0.001)  LN_SUBSIDIES   -0.001***   -0.001***   -0.001**    (0.000)   (0.000)   (0.000) Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Industry Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes N 4003 2727 4052 3494 2424 3536 3016 2117 3052 adj. R-sq 0.271 0.288 0.277 0.239 0.244 0.245 0.220 0.210 0.222                      FROA4 FROA5     (10) (11) (12) (13) (14) (15)    PCB -0.001    0.002       (0.004)    (0.004)      LN_GC  -0.002**    -0.002**       (0.001)    (0.001)     LN_SUBSIDIES   -0.001*   -0.001**       (0.000)   (0.000)    Controls Yes Yes Yes Yes Yes Yes    Industry Fixed Effects Yes Yes Yes Yes Yes Yes    Year Fixed Effects Yes Yes Yes Yes Yes Yes    N 2567 1826 2599 2137 1536 2165    adj. R-sq 0.208 0.210 0.209 0.240 0.259 0.245        60  Table 1.11 Future Firm Performance – With Interactions This table presents the results from regression analyses of future firm performance, with interaction terms. Dependent variables are one-year-ahead roa (FROA), two-year-ahead roa (FROA2), three-year-ahead roa (FROA3), four-year-ahead roa (FROA4), and five-year-ahead roa (FROA5). Panel A presents the results with LN_GC and LN_SUBSIDIES as independent variables separately, and includes these variables’ interaction with political connection indicators (PCB). Panel B presents the results with both LN_GC and LN_SUBSIDIES in the regression as independent variables, and interact with PCB. Control variables include market size (SIZE), firm age (LNAGE), sales growth (GROWTH), market-to-book ratio (MB), and Herfindahl index (HHI). Details on the variable definitions are provided in Appendix A.1. Standard errors are corrected for heteroscedasticity and clustered at firm level. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.  Panel A. FROA FROA2 FROA3 FROA4 FROA5  (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) PCB 0.010 -0.003 0.021 -0.001 0.037* 0.003 0.039* 0.001 0.028 0.002  (0.020) (0.006) (0.019) (0.007) (0.021) (0.008) (0.022) (0.008) (0.021) (0.007) LN_GC -0.002**   -0.001   -0.001   -0.001   -0.001   (0.001)   (0.001)   (0.001)   (0.001)   (0.001)  LN_SUBSIDIES  -0.001***  -0.001***  -0.001  -0.001  -0.001*   (0.000)  (0.000)  (0.000)  (0.000)  (0.000) PCB×LN_GC -0.001   -0.001   -0.002*   -0.002*   -0.002   (0.001)   (0.001)   (0.001)   (0.001)   (0.001)  PCB×LN_SUBSIDIES  -0.000  -0.000  -0.000  -0.000  0.000   (0.000)  (0.000)  (0.001)  (0.001)  (0.001) Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Industry Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes N 2708 4003 2409 3494 2105 3016 1816 2567 1527 2137 adj. R-sq 0.291 0.275 0.250 0.244 0.219 0.222 0.217 0.210 0.260 0.243    61    Panel B. FROA FROA2 FROA3 FROA4 FROA5 PCB 0.008 0.020 0.037* 0.037* 0.025  (0.020) (0.019) (0.021) (0.022) (0.021) LN_GC -0.002** -0.001 -0.001 -0.001 -0.001  (0.001) (0.001) (0.001) (0.001) (0.001) LN_SUBSIDIES -0.001*** -0.001*** -0.000 -0.001 -0.001**  (0.000) (0.000) (0.000) (0.000) (0.000) PCB×LN_GC -0.001 -0.001 -0.002 -0.002 -0.002  (0.001) (0.001) (0.001) (0.001) (0.001) PCB×LN_SUBSIDIES -0.000 -0.000 -0.001 -0.000 0.000  (0.001) (0.001) (0.001) (0.001) (0.001) Controls Yes Yes Yes Yes Yes Industry Fixed Effects Yes Yes Yes Yes Yes Year Fixed Effects Yes Yes Yes Yes Yes N 2708 2409 2105 1816 1527 adj. R-sq 0.297 0.257 0.222 0.220 0.265   62  Chapter 2: Audit Fees of Politically Connected Firms We weaken those ties when we allow our political dialogue to become so corrosive that people of good character aren't even willing to enter into public service; […] when we write off the whole system as inevitably corrupt... — President Barack Obama’s Farewell Address, 2017 2.1 Introduction Do politically connected firms pay higher or lower audit fees, and why? This paper seeks to answer this seemingly straight-forward two-part research question. Although there are studies that hint at the audit risk of politically connected firms, the relation between audit fees and political connections in the United States is an empirical question that has yet to be answered. Existing research suggests that political connections are associated with bad financial reporting. For example, Chaney et al. (2011) find that political connections are associated with lower accounting quality using a sample of firms in 19 countries; Baloria (2014) and Kim and Zhang (2016) find that U.S. firms with political connections are more aggressive in their accounting and tax planning. One may also argue that politically connected firms are self-selected to seek out political connections because they face higher political risk, thus higher audit risk. As a result, many believe that political connections are associated with higher audit fees due to low reporting quality and a more opaque audit environment, in addition to increased political risks. Indeed, Gul (2006) finds that auditors charge politically connected companies in Malaysia higher audit fees as compensation for bearing higher litigation risk due to political cronyism, and the audit fee gap increased during the 1997 Asian Financial Crisis. On the other hand, the results may not hold for politically connected firms in the United States, which enjoy a lower corruption rate as well as a superior regulation environment. In fact, one can argue that politically connected firms are considered safer clients. For instance, Faccio et   63  al. (2006) find that firms with political connections are more likely to be bailed out when they are in trouble; Correia (2014) suggests that politically connected firms are less likely to be involved in SEC enforcement actions, and when they are, they face lower penalties. If investors value political connections (Goldman et al., 2009; Cooper et al., 2010),creditors are willing to lend at a lower rate (Houston et al., 2014), and regulators provide more lenient oversight, then the audit risk of politically connected firms will be lower. This paper examines the empirical relation between audit fees and political connections, and identifies the reasons for the relation. Using hand-collected data of S&P 500 firms between 2004 and 2013, I find that politically connected firms pay higher audit fees than their non-connected counterparts. This result is stronger for firms with former politicians who worked for the executive branch. To investigate the cause of this difference in audit fees, I explore several channels. First, following the prevailing view in current research that politically connected firms have poor financial reporting, I compare various measures of financial reporting quality between politically connected firms and non-connected ones. I find no significant difference in my sample, suggesting that the audit fee gap found is not the result of lower reporting quality.  Next, I explore whether having a higher risk of being affected by political and policy changes, which is labeled “political risk”, contributes to higher audit fees for politically connected firms. Politically connected firms, I assume, are more likely to be involved in industries that are sensitive to political risk. During election years, they face higher uncertainty, hence higher audit risk. This effect should be stronger when control of one or more government branches changes hands between political parties. Using all of the election years available during my sample period, especially with the 2006 and 2008 elections, when Congress and the White   64  House respectively changed hands, I employ the difference-in-difference method and find no significant changes in the audit fee gap during election years. This suggests that the different exposure to political risk is unlikely to be the main reason for the audit fee gap. Finally, I hypothesize and find that because politically connected firms – many of whom are government contractors in my sample – pay higher audit fees because government contractors are subject to additional regulations, which increases audit risk. Federal government contractors must comply with the Federal Acquisition Regulations (FAR). FAR governs not only the “acquisition process”, which is how contractors are selected for different projects, but also how contracts are carried out. To ensure compliance, all federal government contractors are subject to government audits. Government auditors like the Defense Contract Audit Agency (DCAA) and the Government Accountability Office (GAO) perform audits on behalf of the government or government agencies. This means firms that are federal government contractors have to adhere to additional disclosure requirements and are subject to additional government auditor inspections which do not apply to other companies. As a result, when an audit client is a government contractor, the auditor needs to gather additional evidence, bear higher risk, or even alter the normal audit procedure to cater to the client,14 thus increasing audit fees. As a result, the audit fee gap between politically connected firms and their non-connected counterparts is a manifestation of the demand for higher audit quality or additional audit evidence associated with being in the government contract business. This paper makes several contributions to the literature. This is the first paper to empirically document the relation between political connections and audit fees in the United                                                  14 Information is from my private conversation with auditors involved in auditing hospitals contracted by Veteran Affairs.   65  States. Existing literature has provided mixed arguments regarding this issue. This paper documents that politically connected firms pay higher audit fees in the United States. More importantly, it sheds light on the reason behind the audit fee gap. I put to test several channels through which such an audit fee gap can be explained.  Most of the existing literature paints firms with political connections in an unfavorable light, with lower accounting quality and higher incentive to manipulate records. As a result, it is believed that auditors charge these firms higher fees to cover litigation risks. My findings do not support this argument in the U.S. Next, I find no evidence for the political risk hypothesis. Politically connected firms may be in more political and policy-sensitive environments, but political uncertainty, most acute during election years, is not associated with higher audit fees.  In the end, this paper demonstrates that the additional compliance requirements and audit demands of being government contractors are the main reasons why audit fees are higher for politically connected firms. In spite of findings in prior literature, and in spite of the cynical tones regarding corporate political connections that some politicians and many political pundits hold, this paper finds nothing nefarious. My evidence shows that the higher audit fees paid by politically connected firms are driven by the higher audit demands of being a government contractor. To my knowledge, this is the first paper to study and draw connections between politically connected boards, government contracts, and audit fees using U.S. data.  In addition, this paper adds to the current literature on how board characteristics affect audit fees. Existing literature finds a positive relation between board independence and audit fees. For example, Carcello et al. (2002) argue that independent board members demand higher audit quality, which results in higher audit fees. They point out that independent board members, being separated from firm management, seek to protect their reputations as experts by   66  monitoring, and have strong incentive to protect against damage to shareholder wealth. My paper identifies a specific subset of outside board members, namely former politicians, and finds that they demand higher audit quality not only because of their own self-interest. Their demand for higher audit quality is warranted as part of the effort to bid for government contracts, and likewise if firms already hold such contracts.  The rest of this paper is organized as followed. Section 2.2 reviews related existing literature and develops hypotheses, followed by research design in Section 2.3. I discuss my sample and data selection in Section 2.4. The results and interpretation are presented in Section 2.5. Section 2.6 includes some additional analyses, which is followed by a brief conclusion in Section 2.7. 2.2 Literature review and hypothesis development There is a rich literature in audit fee determinants. Starting with the seminal work of Simunic (1980), the past few decades have seen an increasing number of studies into what drives audit fees. Some researchers study supply side factors like auditor size, industry expertise or auditor brand name, and audit market competition (Palmrose, 1986; Maher et al., 1992; Ferguson et al., 2003; Maher et al., 1992). However, many more studies focus on demand factors, which are client firm characteristics such as business risk (Bell et al., 2001; Lyon and Maher, 2005), corporate governance (Carcello et al., 2002; Gul et al., 2003; Larcker and Richardson, 2004), and litigation risk (Choi et al., 2009; Abbott et al., 2012). The main findings on the demand side literature are that audit fees are a function of client firms’ size, complexity, inherent risks, and leverage (Hay et al., 2006). This chapter is more relevant to the latter stream of audit fee research. Closely related to my paper is Carcello et al. (2002), who find that board independence,   67  diligence, and expertise are associated with increased audit fees, because of a higher demand for accounting and audit quality from the board. Parallel to the long and extensive audit fee literature, there is a new and growing stream of research on corporate political connections in accounting and finance. Most of the existing literature paints politically connected firms’ accounting practices in an unfavorable light. Specifically, Chaney et al. (2011) find that politically connected firms have higher abnormal accruals; Kim and Zhang (2016) find that they are more tax aggressive; and Baloria (2014) finds that they are less conservative. These papers’ findings suggest that audit fees should be higher for politically connected firms. However, politically connected firms are found to be favored by their stakeholders. Faccio et al., (2006) find that political connections help firms get bailouts in financial distress; Claessens et al. (2008) find that banks are willing to lend to them at a better interest rate; Cooper et al. (2010) and Goldman et al. (2009) both find that the market values political connections; Correia (2014) finds that the SEC is less likely to litigate politically connected firms, who pay less penalty even when litigated. These findings suggest that it is safer to audit politically connected firms, resulting in lower audit fees. It is unclear which of these two forces dominates. Therefore, the relation between political connection and audit fee is an empirical question. A paper worth noting here is Gul (2006), which finds that politically connected firms in Malaysia are charged higher audit fees. Using the 1997 Asian Financial Crisis as an exogenous shock, the author comes to the conclusion that such an audit fee gap is in response to auditors bearing higher audit risk when their clients benefit from political cronyism. However, this finding is based on a Malaysian sample, where political cronyism and corruption is more common. Whether the same applies to   68  politically connected firms in the U.S. remains unclear. Following Gul (2006)’s observation, I phrase Hypothesis 2.1 as the following. Hypothesis 2.1  Audit fees of politically connected firms are higher than audit fees of non-connected firms.  Prior research such as Chaney et al. (2011), Baloria (2014), Tahoun, (2014), and Kim and Zhang (2016) suggests that corporate political connections are signals of bad accounting quality. In order to examine whether lower financial reporting quality contributes to the difference in audit fees paid by politically connected firms, as most current literature suggests, I compare several measures of accounting quality. As such, I state the second hypothesis accordingly. Hypothesis 2.2  Politically connected firms are associated with lower financial reporting quality compared with non-connected firms.  Aside from substandard reporting quality, there are other factors that contribute to higher audit fees paid by politically connected firms. One reason is that firms are self-selected to be politically connected because they face higher political risk. Shaffer (1995) summarizes that firm-level responses include strategic adaptation and attempts to influence public policy. Having political connections is a good example of such a response. Hillman (2005) finds firms in heavily regulated industries have more politician directors. In this case, politically connected firms are self-selected because they face significant regulation or are sensitive to government policy changes. I call this heavy exposure to regulations and sensitivity to policy changes “political risk”, which increases audit risk from an auditor’s perspective, resulting in a higher audit fee charged. This effect should be stronger during election years when political uncertainty is higher, and most acute when the election result is a change in political control in government branches, as favorable policies and regulations may not remain in place and unfavorable changes may be   69  enacted. This increases audit risk as profitability of the firm is sensitive to these changes. Thus, I posit the next hypothesis accordingly. Hypothesis 2.3 Higher audit fees for politically connected firms are due to higher political risk faced by these firms. Another reason for politically connected firms having higher audit fees may be due to higher demands for audit services. This increase in demand of audit service can be a consequence of government contracting, which politically connected firms are often associated with (Goldman et al., 2008; Tahoun 2014).  After securing government contracts, firms must also comply with additional reporting and audit standards while carrying out these contracts. Federal government contractors are regulated under the Federal Acquisition Regulation (FAR) and are often subject to government audits. In order to limit wasteful spending and ensure the quality of government contract executions, the Defense Contract Audit Agency (DCAA) and Government Accountability Office (GAO) audit government contractors on behalf of the U.S. Government as well as its agencies. Such audits are done based on a contract basis, are not mandatory periodically, and are usually performed upon the request of the government branches. However, big government contractors, like Boeing Co. or Lockheed Martin Corp., who are awarded a great amount of money from government procurement, can reasonably expect regular attention from government auditors. Knowing this additional compliance need, external auditors for these government contractors are more inclined to perform services catered to their clients in order to facilitate a smoother government audit. Such catered services require altered audit procedures, and are thus more likely to result in higher audit fees. Therefore, my last hypothesis is as follows.   70  Hypothesis 2.4 Higher audit fees for politically connected firms are due to their involvement in government contracts. 2.3 Research design There are various ways to measure and to identify corporate political connections: lobbying expenses (Hill et al., 2014), corporate political campaign contributions (Claessens et al., 2008), politician equity ownership (Baloria, 2014; Tahoun, 2014),  and having former politicians as board members (Goldman et al., 2009). However, not all measures are suited for this study. Lobbying expenses, which are usually issue- and industry-specific, are not a good identification strategy because firms may lobby through industry associations, and lobbyists target both political parties without public disclosure of specifics. Corporate political campaign contributions to some extent measure which political party a firm wants to be associated with more, yet most firms donate similar amounts to both major parties. Recent decades have also witnessed an increased popularity in Political Action Committees (PACs). Since 2010’s Citizens United v. Federal Election Commission, the use of Super PACs has made it almost impossible to identify the donors to political campaigns. Lastly, when it comes to politician equity ownership, I argue that this is a measure of politicians selecting their own stock holdings, rather than firms selecting political connections. In this paper, I follow Goldman et al. (2009)’s definition and classify politically connected firms as those with at least one politically connected board members. This variable is measured at the beginning of every fiscal year. Comparatively, this is a clear measure of political connection when it comes to identifying which political party and what government branches firms are targeting. Having a politically connected board member is not only a signal that the firm is seeking political connections, but also that it is successful in doing so, as the politician   71  needs to agree to be on the board of directors and have his/her name associated with the firm when filing for public disclosure. To test the empirical relation between audit fee and political connections, I analyze audit fees with a standard audit fee model used in the literature (Carcello et al., 2002; Lyon and Maher, 2005). The following specification is used to test Hypothesis 2.1. Ln( 𝐴𝑢𝑑𝑖𝑡 𝐹𝑒𝑒𝑖,𝑡)= 𝛼0 + 𝛼1𝑃𝑜𝑙𝑖𝑡𝑖𝑐𝑎𝑙 𝐶𝑜𝑛𝑒𝑐𝑡𝑖𝑜𝑛𝑠𝑖,𝑡−1 + α2 𝐿𝑁𝐴𝑇𝑖,𝑡 + 𝛼3𝑅𝐸𝐶𝐼𝑁𝑉𝑖,𝑡 + 𝛼4𝐶𝐴𝑇𝐴𝑖,𝑡+ 𝛼5𝑄𝑈𝐼𝐶𝐾𝑖,𝑡 + 𝛼6𝑀𝐵𝑖,𝑡 + 𝛼7𝑅𝑂𝐴𝑖,𝑡 + 𝛼8𝐿𝑂𝑆𝑆𝑖,𝑡 + 𝛼9𝐿𝑇𝐷𝑖,𝑡 + 𝛼10𝐿𝑁𝑆𝐸𝐺𝑖,𝑡+ 𝛼11𝑅𝐸𝑆𝑇𝐴𝑇𝐸𝑀𝐸𝑁𝑇𝑖,𝑡 + 𝛼12𝐺𝑂𝐼𝑁𝐺_𝐶𝑂𝑁𝐶𝐸𝑅𝑁𝑖,𝑡 +  𝛼13𝐵𝐼𝐺4𝑖,𝑡               (2.1) In Equation 2.1, the variable of interest is Political Connections, which can take one of three forms. The first is PCB, an indicator variable that takes on the value 1 when there is at least one board member who is politically connected. The second, PPCB, is the percentage of board members that are politically connected. The third is LN_NPC, the natural log of the number of politically connected board members. I also replace the political connection indicator with different measures of political connection characteristics, such as political party associations and various government branch experiences, to examine the impacts of various types of political connections.  The remaining variables in Equation 2.1 are from the standard audit fee model. Log number of client’s total assets, LNAT, measures the size of audit engagement and audit workload. To control for demand side factors like inherent risk, profitability, and complexity, I include RECINV, which is the ratio of accounts receivables and inventory scaled by total assets, CATA, which is the percentage of current assets to total assets, QUICK, which is quick ratio, ROA and LOSS indicator, LTD, which is long-term debt scaled by total assets, LNSEG, which is the log   72  number of business segments. RESTATEMENT and GOING_CONCERN are indicators when a firm-year observation has restatement or the auditors have issued going concern qualifications, which are associated with high audit risk. BIG4 is a Big 4 auditor indicator, as these auditors are associated with the audit fee premium. Hypothesis 2.1 predicts 𝛼1 to load positively, which indicates politically connected firms pay higher audit fees. When Political Connections measures are PCB, PPCB and LN_NPC, a positive 𝛼1suggests that the audit fee increases with the existence or strength of political connections. When the Political Connections is replaced with indicators of different government branch experiences (e.g. legislative and executive), the coefficient difference between executive branch connections and legislative connections indicates whether or not having different types of political connections matters. I expect having legislative branch connections may help firms decrease legislative risk, thus lowering audit fees; whereas having executive branch connections help firms obtain government contracts, thus increasing audit fees. The latter is consistent with Hypothesis 2.4. To test Hypothesis 2.2, I use Equation 2.2 below. The dependent variable is different measures of financial reporting quality, and Political Connections is the variable of interest. 𝐹𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝑅𝑒𝑝𝑜𝑟𝑡𝑖𝑛𝑔 𝑄𝑢𝑎𝑙𝑖𝑡𝑦= 𝛽0 + 𝛽1𝑃𝑜𝑙𝑖𝑡𝑖𝑐𝑎𝑙 𝐶𝑜𝑛𝑛𝑒𝑐𝑡𝑖𝑜𝑛𝑠 + 𝛽2𝑅𝑂𝐴 + 𝛽3𝐿𝑂𝑆𝑆 +  𝛽4𝑆𝐼𝑍𝐸 +  𝛽5𝐿𝑇𝐷+ 𝛽6𝐺𝑅𝑂𝑊𝑇𝐻 + 𝛽7𝐶𝐴𝑃𝐸𝑋                                                                  (2.2) I use five variables to measure financial reporting quality: C score (C_SCORE) from Khan and Watts (2009), Modified Jones Model’s discretionary accruals (DACC) from Dechow et al. (1995), discretionary accruals (DDACCR) from Dechow and Dichev (2002), Accounting and Auditing Enforcement Releases (AAER) issued by the SEC, and whether there is a restatement of   73  the financial statement (RESTATEMENT). I discuss the choice of these accounting measures in detail in the sample and data section. Hypothesis 2.3 is tested using a difference-in-difference variation of Equation 2.1. Specifically, I use Equation 2.3 to assess whether election years, when political risk is higher, have any marginal impact on the audit fees paid by politically connected firms. Moreover, the sample period includes both the 2006 and 2008 elections, when the controls of U.S. Congress and the executive branch changed hands respectively. Political risk would be more acute during these two elections, so I test these two elections separately using the same structure of Equation 2.3. Ln(𝐴𝑢𝑑𝑖𝑡 𝐹𝑒𝑒)=  𝛾0 + 𝛾1𝑃𝑜𝑙𝑖𝑡𝑖𝑐𝑎𝑙 𝐶𝑜𝑛𝑛𝑒𝑐𝑡𝑖𝑜𝑛𝑠 + 𝛾2𝐸𝑙𝑒𝑐𝑡𝑖𝑜𝑛 𝑌𝑒𝑎𝑟 𝐼𝑛𝑑𝑖𝑐𝑎𝑡𝑜𝑟+ 𝛾3𝑃𝑜𝑙𝑖𝑡𝑖𝑐𝑎𝑙 𝐶𝑜𝑛𝑛𝑒𝑐𝑡𝑖𝑜𝑛𝑠 × 𝐸𝑙𝑒𝑐𝑡𝑖𝑜𝑛 𝑌𝑒𝑎𝑟 𝐼𝑛𝑑𝑖𝑐𝑎𝑡𝑜𝑟+ 𝛴𝛾𝑖𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠                                              (2.3) I use three variations of this regression: separating Democratic and Republican connections, using an indicator of political connections (PCB) and an election year indicator (ELECTION), and interacting the political party identifier with an indicator for election years when political control changed. The first variation of separating political connections along party lines will find whether connections to a specific party are more sensitive towards political risks. The second and third variations are the difference-in-difference tests. Hypothesis 3 predicts 𝛾3 to be positive and significant. The result should diverge in variation 3, as the election year indicator is replaced with power-changing election years. Given that in the 2006 and 2008 elections, the Democratic Party took control of both houses of Congress and the White House respectively, the interaction of these election indicators and the Democratic connections should be negative and significant   74  due to lower political risk, whereas the interaction of these election indicators and the Republican connections should be positive and significant as political risk increases. I also included lobbying expense as it may also measure the political sensitivity of the firm. Testing Hypothesis 2.4 requires path analysis, which I execute in two ways. First, I add in the amount of annual government contracts (LN_GC) as a control variable in the audit fee model. Hypothesis 2.4 indicates that the significance level of the political connections variable should decrease (or disappear) when government procurement information is added into the regression.  Next, though orthogonality is not required for independent variables, to try and see the effect of stand-alone government contracts, I utilize another method to demonstrate the isolated effect of government contracts on audit fees. In the first regression, I use government contract information to predict political connection value. The expected political connection value then contains all the information of having government contracts, whereas the residual value of this regression contains the variation in political connections orthogonal to government contracts. The audit fee model is the second regression, with both predicted political connection and residual political connection information as independent variables. In the second step, any effect that the political connection has on audit fees through the government procurement channel is captured in the coefficient of the expected political connection conditional on government contracts. Any additional impact that political connections have on audit fees that are orthogonal to government contracts is reflected in the coefficient on the residual political connection variable. Hypothesis 4 predicts that, in the second stage regression, the coefficient of the predicted political connection should be positive and significant.   75  2.4 Sample and data 2.4.1 Sources of data I begin my sample construction using information from CRSP to identify firms in the S&P 500 from 2004 through 2013. Information on politically connected boards are collected from each firm’s annual proxy statement (DEF 14A) from EDGAR on SEC’s website15. Each year, approximately 9 months before the issuance of the annual report, a firm usually provides a proxy statement to nominate the board of director, and to disclose additional information such as the past fiscal year’s committee meetings and audit fees. For each nominated director, a short description of his/her prior experience is included. I manually read through this part of the proxy statements in order to identify former politicians who are sitting on the board.  My classification of politically connected board members follows Goldman et al. (2009),16 with the exception that I do not count as politically connected those who have worked for or are related to the judicial branch of government, Federal Reserve System, and those who only have military rankings are not included. Also, those whose description mentioned a qualifying position but the individual did not serve on the position (e.g. “served as a senior advisor to the Secretary of US Department of Defense”) are not considered political connections. If there is no proxy statement for a specific year, I try to infer it from both the prior and the following year’s proxy statement. If inference is not possible, then I assume that the missing year’s information is the same as the year prior. If                                                  15 http://www.sec.gov/edgar/searchedgar/companysearch.html 16 A company is classified as politically connected if it has at least one board member with one of the following former positions: President, presidential (vice-presidential) candidate, senator, member of the House of Representatives, secretary, assistant secretary, deputy secretary, deputy assistant secretary, undersecretary, director, associate director, deputy director, commissioner of any federal government department or federal government agency (including CIA, FEMA, CIA, OMB, IRS, NRC, SSA, CRC, FDA, and SEC), governor, mayor, treasury of the city, representative to the UN, trade representative for the US, ambassador, staff (White House, president, presidential campaign), chairman of the party caucus, and chairman or staff of the presidential election campaign.    76  neither is possible, then the observation is dropped. The resulting sample is as shown in Table 2.1. Approximately 32.9% of the sample firm-years are politically connected firms, which is very similar to Goldman et al.(2009)’s sample of 31%, but lower than Houston et al. (2014)’s sample of 43%, which is likely because their sample was machine-read. The proportion of politically connected firms relative to the sample is stable across years, as shown in Figure 2.1.  Table 2.2 shows the industry composition of my sample. As shown in Table 2.2, there is a good range of industries within my sample. There are few industries where political connections are extremely high (over 95%) or low (below 5%), and the number of firms involved is small. This shows that seeking political connection is not a special phenomenon that only exists in specific industries, but a wide-spread, common practice across firms in various industries. We see the sample has some representation in finance industries. Following prior research, I exclude companies with SIC codes between 6000 and 6999 from all my tests. However, the results remain the same even after including these observations.  Once a politically connected board member is identified, then a Google search is performed on this person to identify his/her political party affiliation. Republicans and Democrats are classified accordingly, and any other political party association is classified as other. If this information is not readily available, then I look into the period when that person held office, and identify them as being part of the same party as the then-serving President of the United States. Because most of the positions, especially those that belong to the executive branch of government, are appointed directly by the President, assuming that a person’s political connection is from the same party as the President is reasonable. If, however, that person served under both Democratic and Republican Presidents, then they are classified as other. For those who served in either house of Congress, they are classified as legislatives; those who served   77  under the executive branch are classified as such. Politicians who are female or racial minorities are identified according to their names, pictures, and other online information17. For firm-years without proxy statements, I use information from the subsequent proxy statements for up to two years. For example, if a firm is missing the 2004 proxy, then I assume the board information to be the same as 2005. If its information is also missing for 2005, then I assume that it is the same as 2006. If neither 2005 nor 2006 information is available, then the observation is dropped. The details of political connections are summarized in Table 2.3 Panel A. As Goldman et al.(2009) suggest, there are more Republican connections than Democratic ones, though not by much. I end up with a sample of 5,012 firm-year observations.  For government contract information, I use the Federal Procurements Data System-Next Generation (FPDS-NG)18, and search for each firm-year’s specific government contract information. The FPDS-NG is a government-run public source that contains all federal procurements whose estimated value is $3,000 or more. Modifications to any of the reported contracts, regardless of dollar value, will also be reported to FPDS-NG. A search is done in the “Adhoc Report” section, by adding the main part of the firm name (e.g. “Boeing” is searched instead of “Boeing Co”) as a selected filter. Duration of the “Adhoc Report” is then selected to match exactly the beginning and end date of the reported fiscal period for each firm-year observation. When multiple entities are listed in the Adhoc Report, an additional search on Google is performed to include those that are confirmed to be subsidiaries, or venture businesses. Total dollar amount of action obligations in the report is then collected. This means there is a downward bias in government contract information, as I omit contracts awarded to subsidiaries                                                  17 The information sources are the proxy statements, official government websites, Business week, Forbes, and nndb.com. 18 Information available at https://www.fpds.gov/fpdsng_cms/index.php/en/.   78  whose name do not contain the name of the parent firm. This biases my sample towards not finding result for Hypothesis 2.4.   Lobbying expenses are included as a control variable in some tests, and they are sourced from the Center for Responsive Politics19. Information required to be publically disclosed (even if the actual amount is zero) since 1998 for all lobbying transactions is available in the lobbying spreadsheet provided by the website. I manually match these transactions using client name and year to my sample, and generate the sum of lobbying expenses for each firm-year. Lobbying is added as a control variable in the audit fee models in tests for Hypothesis 2.3 and 2.4 in order to control for political sensitivity and political influence. Financial statement variables are collected and calculated from COMPUSTAT. Audit fees, audit filing dates, restatement, SOX302 material weakness, and audit opinion variables are from Audit Analytics. AAER firms are identified when the firm is involved in an SEC enforcement action20. A firm is identified to be in the AAER if the firm has been litigated by the SEC in at least one case within the current fiscal year. Financial information data as well as control variables are summarized in Panel B and C of Table 2.3. Continuous variables are winsorized at 1%. All variables are as defined in Appendix 2.  2.4.2 Choice of accounting quality measures The literature provides many different measures of accounting quality. In this paper, I use five measures for this purpose. First, I construct the C_SCORE, a measure of conservatism,                                                  19 Information available from http://www.opensecrets.org/. For non-registered users, the source is http://data.influenceexplorer.com/bulk/. 20 Dechow et al., (2011) has collected most of the data needed for my sample years. This information is publically available on SEC’s Accounting and Auditing Enforcement Releases: http://www.sec.gov/divisions/enforce/friactions.shtml. I thank Weili Ge for sharing most of her collected data with me.    79  following Khan and Watts (2009). Using data from CRSP and COMPUSTAT universe, I estimate the following cross-sectional regression:  𝑋𝑖 = 𝛽1 +  𝛽2𝐷𝑖 + 𝑅𝑖(𝜇1 + 𝜇2𝑆𝑖𝑧𝑒𝑖 + 𝜇3𝑀 𝐵𝑖⁄ + 𝜇4𝐿𝑒𝑣𝑖)      + 𝐷𝑖𝑅𝑖(𝜆1 + 𝜆2𝑆𝑖𝑧𝑒𝑖 + 𝜆3𝑀 𝐵𝑖⁄ + 𝜆4𝐿𝑒𝑣𝑖)+ (𝛿1𝑆𝑖𝑧𝑒𝑖 + 𝛿2𝑀 𝐵𝑖⁄ + 𝛿3𝐿𝑒𝑣𝑖 + 𝛿4𝐷𝑖𝑆𝑖𝑧𝑒𝑖 + 𝛿5𝐷𝑖𝑀 𝐵𝑖⁄ + 𝛿6𝐷𝑖𝐿𝑒𝑣𝑖)+ 𝜖𝑖                                                         (2.4)  I then calculate 𝐶_𝑆𝑐𝑜𝑟𝑒𝑖𝑡 =  𝜆1 + 𝜆2𝑆𝑖𝑧𝑒𝑖𝑡 + λ3𝑀 𝐵𝑖𝑡⁄ + 𝜆4𝐿𝑒𝑣𝑖𝑡. C_SCORE is constructed as a firm-year measure to capture the timely recognition of losses, instead of requiring estimation of a time series. It is a more appropriate measure for my sample in order to examine the claims found in Baloria (2014) that politically connected firms (measured as whether congressional member have equity holdings) are less conservative in their financial reporting applies to my sample. A negative and significant coefficient on political connection measure is expected if the finding is consistent with Hypothesis 2.2, which suggests that politically connected firms have lower accounting conservatism. The second and third measures are discretionary accruals from the Modified Jones Model from Dechow et al. (1995) and from Dechow and Dichev (2002). Both are residuals from accrual models that regress accounting accruals on their economic drivers21. They are commonly used measures for earnings management in the accounting literature to capture management discretion or manipulation. While the first measure is signed, the second one is not, which may be a better measure when we do not expect a specific direction of earnings management. Chaney et al.                                                  21 Modified Jones Model uses the difference in growth of revenues and credit sales and property, plants and equipment as economic factors for total accruals, and produces a signed residual as discretionary accruals. Dechow and Dichev (2002) regress the change of working capital on prior, current, and future cash flow from operations, and use the absolute value of the residual as accrual management.   80  (2011) found that accounting accrual quality is lower for politically connected firms in their cross-country sample; I examine whether this association is also present in my sample. If politically connected firms are managing earnings more than other firms, then we should have positive and significant coefficients on political connections.  The fourth measure is whether the firm is identified by the SEC in the Accounting and Auditing Enforcement Release (AAER) that year and the fifth is whether the firm-year observation has a subsequent financial reporting restatement. These are both subsequent event measures of poor reporting quality, rather than statistical predictions of poor quality . They help explore whether politically connected firms in my sample are more frequently identified as having violated reporting standards by the regulator or other stakeholders. Because of the binary nature of these dependent variables, logit models are used for these two tests. Results from Correia (2014) suggest that the loading on political connections of the AAER regression should be negative. Hypothesis 2.2 predicts that the coefficients of political connections should be positive and significant for both of these measures. 2.5 Results and interpretations All standard errors of the results (in parenthesis) are estimated using robust regressions and clustered by firm, and statistical significance is calculated based on two-tail tests. Industry fixed effects based on two-digit SIC code and year fixed effects are added where appropriate. Table 2.4 presents the result for Hypothesis 2.1, which predicts a positive relation between political connections and audit fees. We see that for politically connected firms, there is an audit fee premium associated with the political connected board (measured by PCB). This result persists when the strength of the political connections (measured by PPCB, and LN_NPC) increases. Moreover, we see that audit fee is positively associated with executive branch   81  connections, whereas the association with legislative branch connections is slightly negative and insignificant. This result suggests those with executive branch experiences are contributing to higher audit fees paid. To explore whether the accounting quality difference is behind the audit fee premium paid by politically connected firms, I proceed to test Hypothesis 2.2, and the results are presented in Table 2.5. Control variables in each regression include financial performance (measured as ROA), loss indicator, firm size (measured as market capitalization), leverage, sales growth, and capital expenditure. Overall, across the five measures, there is little significant difference in accounting quality between politically connected and unconnected firms. Only Modified Jones Model’s discretionary accruals are marginally associated with the dichotomous PCB variable at the 10% significance level. Nevertheless, this relation disappears when I use the continuous measures of political connection strength in the next two columns. Interestingly, there is an opposite result coming from the AAER columns. Though it is only significant at 10%, it shows that the likelihood of a firm being the subject of AAER decreases as the strength of the political connection increases. This is consistent with the findings in Correia (2014), and politically connected firms are less likely to be litigated by the SEC. Taken as a whole, results in Table 2.5 suggest that, at least in my sample, accounting quality difference does not seem to be the main reason for politically connected firms being charged higher audit fees.  Table 2.6 presents results of the tests performed for Hypothesis 2.3. As indicated before, I use a difference-in-difference setting in this analysis, where elections years are treatments with higher political risk. In the first two columns, I find no evidence that the political risk of being associated with different parties impacts audit fees. While the coefficients of Democratic Party connections are positive and significant in both column (1) and (2), the magnitude of the   82  coefficients (0.092 and 0.139) are not dramatically different from those of the Republican Party connection (0.072 and 0.089). F-test also finds that the coefficients on a Democratic Party connection are not statistically different from the coefficients on a Republican Party connection. In the third column, all election years (2004, 2006, 2008, 2010, and 2012) are indicated with an election year indicator, and the variable of interest is on the interaction of PCB and Election, which is not significantly different from zero. This is to address the concern that politically connected firms are facing political risk during every election ex ante, without knowing the outcome of the election of which party wins. Columns (4) to (6) pay special attention to the 2006 and 2008 elections, when political risk should be most acute due to a change in party controls in the U.S. Congress and the White House respectively. However, most of the interaction terms are insignificantly different from zero, except in columns (5) and (6) where the coefficient of DREP×2006 ELECTION is positive and significant at 10%. Overall, these results do not support political risk as the reason for higher audit fees paid by politically connected firms. Table 2.7 presents the first test in a simple path analysis to examine Hypothesis 2.4. In this table, I estimate the normal audit fee model with additional measures for the amount of government contracts (LN_GC) and lobbying expenses (LN_LOB). Lobbying expenses are included to control for regulation risk. Column (1) shows audit fees are positively associated with the amount of government contracts a firm receives. Column (2)’s coefficient on PCB is no longer statistically different from zero, and it drops from 0.127 in Table 2.4 to 0.08. Columns (3) to (4) show that after controlling for government contracts, audit fees still have a marginally positive association with the strength of political connections. But overall, the coefficients of political connection strength measures also decrease, compared to those in Table 2.4 when government contracts are added into the regression. This shows that the audit fee premium found   83  in Table 2.4 is from firms with government contracts. It also is consistent with findings in the last column of Table 2.4, where connections to the executive branch are driving the result. Being politically connected, especially with the executive branch, is useful for winning government contracts, as most government contracts are awarded and controlled by the executive branch (e.g. Department of Defense, Department of Energy).  Being government contractors also means additional compliance requirements, hence auditing government contractors may take more work and expertise. Table 2.8 shows the second approach for testing Hypothesis 2.4, which isolates the effect of government contracts on political connections. The first step is to isolate the impact of government contracts by calculating the expected political connection value conditional on government contracts (EX_PCB, etc.). The second stage is to include both expected political connection and residual political connection (RES_PCB, etc.) in the audit fee model. The loading on predicted political connection value reflects how much of the original association of political connection with audit fee is due to government contracts, while the coefficient on  residual political connections shows the association of audit fees with the portion of political connections orthogonal to government contracts. In the first stage, shown in Panel A, I regress political connection on government contract information as well as other firm characteristics. Whether with or without control variables and fixed effects, political connections are associated with the amount of government contracts received by a company. Second stage regresses audit fees against the expected political connections and the residuals of political connections from the prediction model. Table 8 Panel B shows the results of the second stage regressions, where again the expected values of political connection in columns (2), (4), and (6) are calculated using the pure effect of government contracts. All coefficients on predicted political connections are   84  positive and significant, whereas the residuals are shown to be marginally significant or not significant. Table 2.8 also reveals that government contracts explain approximately 30% of the impact of political connections on audit fees. Together with results from Table 2.7, I conclude that the higher audit fees paid by politically connected firms are driven by their involvements in government contracts. All in all, the results show strong support for Hypothesis 2.4. 2.6 Additional analyses As an attempt to see whether there is any sign of additional audit effort, I also perform a test on audit report lag, the length of time needed for auditors to complete their audit. Knechel and Payne (2001) show that audit report lag is highly correlated with audit hours, and has been used as an indirect measure of audit effort in the literature. As shown in Table 2.9, there is no difference in audit report lag (or audit effort) between politically connected firms and the non-connected firms. This suggests that auditing government contractors may be more likely to involve catering services and additional expertise rather than just additional work.  I also perform a marginal test to see if having better internal control system helps to reduce the audit fees paid by politically connected firms. A better internal control system may help decrease the audit work load needed for a government contractor, as the processing of information becomes more efficient and reliable. Better internal controls also ensure better compliance with laws and regulations, including compliance with FAR, thus lowering demands of auditors for this aspect as a result. Empirically measuring the quality of internal control is difficult. So instead of measuring how good a firm’s internal system is, I look at whether there are material weaknesses identified for a firm’s internal control system, MATERIAL_WEAKNESS, the number of identified material weaknesses, COUNT_WEAK, and the log transformation of the number of identified material weaknesses, LN(COUNT_WEAK). These measures are internal   85  control deficiencies. I assume that fewer deficiencies proxies for better the internal control system. Table 2.10 presents the results and shows that with better internal control systems, politically connected firms can decrease their audit fees. 2.7 Conclusion In this paper, I document a positive and significant positive relation between political connections and audit fees. This relationship is stronger for firms with connections who previously worked in the executive branch of the government. To explore the reasons behind such difference in audit fees, three distinct hypotheses are put to the test. The first hypothesis posits the prevailing view that politically connected firms are bad financial reporters. Yet in my sample, I fail to find any significant difference in financial reporting qualities between politically connected firms and their non-connected counterparts. Next, a hypothesis regarding political risk is examined through a difference-in-difference analysis. I exploit higher political uncertainty during elections, especially when political controls of legislative and executive branch change hands, but fail to find any impact on the audit fee gap. Although politically connected firms may be more sensitive to political risk, this does not seem to be the reason for higher audit fees. Finally, through path analysis, I find that politically connected firms may be charged with higher audit fees because of politically connected firms’ involvement in government contracts. Because government contractors are subject to additional regulations in their financial reporting, like FAR, ensuring compliance increases audit risk. Meanwhile, the former politicians, especially those who help firms obtaining government contracts, have strong incentive to protect their reputations, thus putting more pressure on auditors. The evidence also suggests that by improving the internal control system, politically connected firms can lower this risk, thus decreasing audit fees.   86  This paper demonstrates that higher audit fees paid by politically connected firms do not signal nefarious practices. Rather, this difference in audit fees is a manifestation of increased audit demands of government contractors in order to comply with government requirements. This is therefore the nature of being involved in government contracts, and a legitimate practice. I also acknowledge the shortcomings of this paper. This paper cannot directly examine whether audits for government contractors do involve more audit efforts or different audit procedures. The audit production function may be different when auditing government contractors, due to compliance needed for FAR in addition to GAAP. Moreover, the external validity of my results for Hypothesis 2 and 3 is unknown. Even though I fail to reject the null hypotheses in both cases, one should not conclude the same result applies for another sample. Taken as a whole, this paper sheds new light on the effects of corporate political connections. Specifically, how audit fees may be different for those with political connections and why that is the case. We should not automatically assume corporate political connections as synonymous with corruption and corporation manipulation. At the very least, this paper shows that auditors, at least in the United States, are guarding financial reporting qualities and ensuring their clients’ compliance with laws and regulations in addition to GAAP.      87  Figure 2.1 Politically Connected Firms by Year  Figure 2.1 is a graphic demonstration of the breakdown of politically connected firms (PCB=1) and non-connected firms (PCB=0) in the sample by year.        01002003004005006002004 2005 2006 2007 2008 2009 2010 2011 2012 2013PCB=0 PCB=1  88  Table 2.1 Number of Politically Connected Firms by Year Table 2.1 presents the breakdown of politically connected firms (PCB=1) and non-connected firms (PCB=0) in the sample by year.      PCB Year 0 1 Total 2004 339 154 493 2005 329 158 487 2006 322 171 493 2007 336 168 504 2008 340 168 508 2009 339 172 511 2010 334 167 501 2011 337 169 506 2012 341 160 501 2013 348 160 508 Total 3365 1647 5012       89  Table 2.2 Summary of Industries This table presents the industry composition of my sample, based on 2-digit SIC.  2-digit SIC Industry Full Sample PCB=1 Percentage of PCB=1 Frequency Percentage Frequency 1 Agricultural Production - Crops 10 0.20% 10 100.00% 10 Metal, Mining 25 0.49% 11 44.00% 12 Coal Mining 21 0.41% 8 38.10% 13 Oil & Gas Extraction 225 4.43% 90 40.00% 14 Nonmetallic Minerals, Except Fuels 10 0.20% 7 70.00% 15 General Building Contractors 39 0.77% 8 20.51% 16 Heavy Construction, Except Building 17 0.33% 15 88.24% 17 Special Trade Contractors 5 0.10% 0 0.00% 20 Food & Kindred Products 229 4.51% 60 26.20% 21 Tobacco Products 36 0.71% 19 52.78% 22 Textile Mill Products 1 0.02% 0 0.00% 23 Apparel & Other Textile Products 40 0.79% 14 35.00% 24 Lumber & Wood Products 23 0.45% 4 17.39% 25 Furniture & Fixtures 20 0.39% 0 0.00% 26 Paper & Allied Products 75 1.48% 31 41.33% 27 Printing & Publishing 41 0.81% 19 46.34% 28 Chemical & Allied Products 401 7.90% 118 29.43% 29 Petroleum & Coal Products 69 1.36% 32 46.38% 30 Rubber & Miscellaneous Plastics Products 31 0.61% 18 58.06% 31 Leather & Leather Products 10 0.20% 0 0.00% 32 Stone, Clay, & Glass Products 5 0.10% 0 0.00% 33 Primary Metal Industries 56 1.10% 24 42.86% 34 Fabricated Metal Products 42 0.83% 10 23.81% 35 Industrial Machinery & Equipment 279 5.49% 77 27.60% 36 Electronic & Other Electric Equipment 345 6.79% 77 22.32% 37 Transportation Equipment 124 2.44% 63 50.81% 38 Instruments & Related Products 238 4.69% 77 32.35% 39 Miscellaneous Manufacturing Industries 30 0.59% 0 0.00% 40 Railroad Transportation 37 0.73% 24 64.86% 42 Trucking & Warehousing 10 0.20% 10 100.00% 44 Water Transportation 10 0.20% 6 60.00% 45 Transportation by Air 22 0.43% 14 63.64% 47 Transportation Services 21 0.41% 0 0.00% 48 Communications 176 3.47% 68 38.64% 49 Electric, Gas, & Sanitary Services 391 7.70% 165 42.20% 50 Wholesale Trade - Durable Goods 36 0.71% 5 13.89%   90  2-digit SIC Industry Full Sample PCB=1 Percentage of PCB=1 Frequency Percentage Frequency       51 Wholesale Trade - Nondurable Goods 45 0.89% 10 22.22% 52 Building Materials & Gardening Supplies 26 0.51% 8 30.77% 53 General Merchandise Stores 105 2.07% 27 25.71% 54 Food Stores 40 0.79% 6 15.00% 55 Automotive Dealers & Service Stations 29 0.57% 5 17.24% 56 Apparel & Accessory Stores 56 1.10% 4 7.14% 57 Furniture & Home Furnishings Stores 39 0.77% 4 10.26% 58 Eating & Drinking Places 47 0.93% 28 59.57% 59 Miscellaneous Retail 81 1.60% 11 13.58% 60 Depository Institutions 256 5.04% 40 15.63% 61 Non-Depository Institutions 66 1.30% 44 66.67% 62 Security & Commodity Brokers 153 3.01% 44 28.76% 63 Insurance Carriers 255 5.02% 117 45.88% 64 Insurance Agents, Brokers, & Service 20 0.39% 10 50.00% 65 Real Estate 15 0.30% 8 53.33% 67 Holding & Other Investment Offices 133 2.62% 19 14.29% 70 Hotels & Other Lodging Places 23 0.45% 20 86.96% 72 Personal Services 10 0.20% 6 60.00% 73 Business Services 392 7.72% 94 23.98% 75 Auto Repair, Services, & Parking 13 0.26% 8 61.54% 78 Motion Pictures 4 0.08% 0 0.00% 79 Amusement & Recreation Services 20 0.39% 16 80.00% 80 Health Services 45 0.89% 18 40.00% 82 Educational Services 21 0.41% 3 14.29% 87 Engineering & Management Services 20 0.39% 0 0.00% 99 Non-Classifiable Establishments 14 0.28% 10 71.43% Total 5,078 100% 1,646 32.41%    91  Table 2.3 Summary Statistics Table 2.3 presents the summary statistics of different variables. Please refer to appendix for variable definitions. Panel A presents summary statistics of political connections. Panel B and C present summary statistics for variables used for audit fee tests and other tests by separating them into politically connected firms (PCB=1) and non-connected firms (PCB=0).   N mean standard deviation p25 p50 p75 PCB 5012 0.329 0.470 0 0 1 PPCB 5012 0.040 0.067 0 0 0.083 LN_NPC 5012 0.271 0.411 0 0 0.693 DREP 5012 0.197 0.398 0 0 0 DDEM 5012 0.186 0.389 0 0 0 LN_DEM 5012 0.128 0.287 0 0 0 LN_REP 5012 0.141 0.307 0 0 0 PREP 5012 0.020 0.046 0 0 0 PDEM 5012 0.018 0.041 0 0 0 LEGISLATIVE 5012 0.124 0.330 0 0 0 EXECUTIVE 5012 0.269 0.444 0 0 1      92  Panel B: Summary Statistics for Audit Fees Regression Variables     PCB=0        N mean standard deviation p25 p50 p75 AF 2711 15.301 0.850 14.679 15.299 15.793 LNAT 2711 9.439 1.355 8.440 9.232 10.218 RECINV 2618 0.235 0.192 0.084 0.187 0.320 ROA 2707 0.055 0.064 0.018 0.050 0.090 CATA 2155 0.385 0.208 0.204 0.374 0.540 QUICK 2115 1.532 1.042 0.853 1.220 1.842 LOSS 3365 0.091 0.287 0 0 0 MB 3359 3.047 3.304 1.549 2.457 3.864 BIG4 3365 0.799 0.401 1 1 1 LTD 2690 0.202 0.154 0.085 0.176 0.294 LNSEG 2158 0.987 0.569 0.693 0.693 1.386 RESTATEMENT 3365 0.081 0.273 0 0 0 GOING CONCERN 3365 0.000 0.017 0 0 0        PCB=1        N mean standard deviation p25 p50 p75 AF 1400 15.867 0.916 15.149 15.857 16.510 LNAT 1400 10.094 1.354 9.100 10.092 10.765 RECINV 1378 0.207 0.165 0.080 0.168 0.270 ROA 1399 0.054 0.064 0.022 0.046 0.087 CATA 1165 0.325 0.174 0.180 0.318 0.428 QUICK 1159 1.252 0.769 0.810 1.068 1.437 LOSS 1647 0.090 0.287 0 0 0 MB 1645 3.527 4.314 1.559 2.424 4.147 BIG4 1647 0.846 0.361 1 1 1 LTD 1398 0.214 0.142 0.102 0.194 0.306 LNSEG 1075 1.230 0.643 0.693 1.386 1.609 RESTATEMENT 1647 0.090 0.287 0 0 0 GOING CONCERN 1647 0.001 0.025 0 0 0     93  Panel C: Summary Statistics for Other Regression Variables     PCB=0        N mean standard deviation p25 p50 p75 C_SCORE 3338 -0.208 1.079 -0.600 -0.184 0.448 DACC 2668 -0.148 1.059 -0.071 -0.006 0.054 DDACCR 2803 0.029 0.031 0.008 0.018 0.037 AAER 3365 0.019 0.137 0 0 0 RESTATEMENT 3365 0.081 0.273 0 0 0 SIZE 3359 9.236 0.995 8.560 9.165 9.793 GROWTH 3362 0.080 0.172 -0.002 0.068 0.150 CAPEX 3363 0.075 0.140 0.020 0.036 0.065 LN_GC 2081 14.728 3.291 12.267 15.025 17.107 LN_LOB 3365 8.512 6.641 0 12.206 14.057 ARL 2708 4.637 0.555 4.344 4.489 4.644 MATERIAL_WEAKNESS 3365 0.004 0.064 0 0 0 COUNT_WEAK 3365 0.027 0.648 0 0 0         PCB=1        N mean standard deviation p25 p50 p75 C_SCORE 1644 -0.287 1.294 -0.774 -0.241 0.445 DACC 1303 -0.129 0.967 -0.061 -0.005 0.042 DDACCR 1464 0.028 0.031 0.007 0.018 0.037 AAER 1647 0.023 0.150 0 0 0 RESTATEMENT 1647 0.090 0.287 0 0 0 SIZE 1645 9.770 1.141 8.961 9.716 10.496 GROWTH 1646 0.067 0.167 -0.015 0.059 0.126 CAPEX 1646 0.075 0.122 0.019 0.038 0.081 LN_GC 1146 16.395 3.600 14.183 16.548 18.683 LN_LOB 1647 11.534 6.001 11.695 14.228 15.305 ARL 1399 4.621 0.559 4.317 4.466 4.615 MATERIAL_WEAKNESS 1647 0.004 0.060 0 0 0 COUNT_WEAK 1647 0.012 0.266 0 0 0       94  Table 2.4 Political Connections and Audit Fees Table 2.4 presents the results for the relation between audit fees and political connections. The dependent variable is the log of audit fee (AF). PCB is an indicator of political connection, which takes on the value 1 when at least one board member of the firm-year is identified as politically connected. PPCB is the proportion of politically connected board, which is calculated by the number of politically connected board members over the size of the board. LN_NPC is the log number of politically connected board members. LEGISLATIVE (EXECUTIVE) is indicator of having at least one political connection with the legislative (executive) branch experience in the United States government. Detailed variable definitions are in Appendix B.1. Standard errors are corrected for heteroscedasticity and clustered at firm level. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.       95    Expected Sign (1) (2) (3) (4) PCB  0.127***      (0.045)    PPCB   0.734**      (0.294)   LN_NPC    0.165***      (0.051)  LEGISLATIVE     -0.076      (0.060) EXECUTIVE     0.197***      (0.046) LNAT + 0.615*** 0.620*** 0.613*** 0.615***   (0.023) (0.023) (0.023) (0.023) RECINV + 1.827*** 1.845*** 1.839*** 1.851***   (0.324) (0.323) (0.322) (0.318) ROA - -1.068*** -1.065*** -1.090*** -1.100***   (0.303) (0.307) (0.305) (0.304) CATA - 0.141 0.132 0.136 0.107   (0.168) (0.168) (0.168) (0.169) QUICK - -0.046*** -0.045*** -0.045*** -0.046***   (0.017) (0.017) (0.017) (0.017) LOSS + 0.049 0.053 0.047 0.066   (0.055) (0.055) (0.056) (0.053) MB + 0.004 0.005 0.004 0.003   (0.003) (0.003) (0.003) (0.003) BIG4 + 0.216 0.213 0.215 0.160   (0.217) (0.225) (0.219) (0.239) LTD + 0.399*** 0.399*** 0.403*** 0.412***   (0.154) (0.154) (0.154) (0.154) LNSEG + 0.098** 0.100** 0.094** 0.080*   (0.042) (0.042) (0.042) (0.042) RESTATEMENT + 0.084** 0.081** 0.083** 0.087**   (0.038) (0.039) (0.038) (0.036) GOING_CONCERN + 0.645*** 0.655*** 0.662*** 0.605***   (0.101) (0.107) (0.100) (0.089) Year Fixed Effects  Yes Yes Yes Yes Industry Fixed Effects  Yes Yes Yes Yes N  2545 2545 2545 2545 adj. R-sq   0.775 0.773 0.776 0.779   96  Table 2.5 Accounting Quality and Political Connections Table 2.5 presents the results for the relation between accounting quality and political connections. The dependent variables are C_SCORE (Khan and Watts, 2009), DACC calculated as discretionary accruals from Modified Jone's Model (Dechow et al. 1995), DDACCR calculated as discretionary accruals from Dechow and Dichev (2002), AAER which is an indicator that is equal to 1 when the firm-year observation is named in the Accounting and Audit Enforcement Release, and RESTATEMENT, also indicator that is equal to 1 when the firm-year observation has a restatement. Control variables include return on assets (ROA), loss indicator (LOSS), market capitalization (SIZE), long term debt (LTD), sales growth (GROWTH), and capital expenditure (CAPEX). Detailed variable definitions are in Appendix B.1. Standard errors are corrected for heteroscedasticity and clustered at firm level. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.    C_SCORE DACC DDACCR AAER RESTATEMENT PCB 0.030    0.054*    0.001    -0.421    -0.081    (0.021)    (0.029)    (0.002)    (0.413)    (0.219)   PPCB  0.149    0.060    0.017    -5.133*    0.454    (0.171)    (0.188)    (0.015)    (3.059)    (1.403)  LN_NPC   0.026   0.040   0.002   -0.808*   -0.022    (0.022)   (0.030)   (0.002)   (0.465)   (0.257) Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes  Yes Yes Industry fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes  Yes Yes Year fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes  Yes Yes N 3259 3259 3259 3021 3021 3021 2959 2959 2959 2335 2335 2335 2978 2978 2978 adj. R-sq 0.682 0.682 0.682 0.196 0.196 0.196 0.183 0.184 0.184        pseudo R-sq                   0.150 0.156 0.156 0.068 0.068 0.068          97  Table 2.6 Audit Fees and Political Risk Table 2.6 presents the results for testing whether political risk affects the relation between political connection and audit fee. Columns (1) and (2) test whether different political party affiliations affect the relationship differently. Columns (3) - (6) employ the difference-in-difference approach to test whether increased political risk after major elections (when political control of government branches changed from one party to another) affects the relation between political connection and audit fee. The dependent variable is the log of audit fee (AF). Control variables are the log of assets (LNAT), receivable and inventory (RECINV), return on assets (ROA), current assets (CATA), quick ratio (QUICK), loss indicator (LOSS), market-to-book ratio (MB), Big4 auditor indicator (BIG4), long-term debt (LTD), log of number of segments (LNSEG), restatement indicator (RESTATEMENT), and going concern indicator (GOING_CONCERN). Detailed variable definitions are in Appendix B.1. Standard errors are corrected for heteroscedasticity and clustered at firm level. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.   Expected Sign (1) (2) (3) (4) (5) (6) DDEM  0.092*   0.079 0.087 0.086   (0.051)   (0.064) (0.079) (0.079) DREP  0.072   0.070 0.026 0.025   (0.045)   (0.053) (0.067) (0.067) LN_DEM   0.139**        (0.068)     LN_REP   0.089        (0.056)     PCB +   0.123***        (0.044)    ELECTION    0.030***        (0.011)    ELECTION×PCB +     -0.028               (0.017)       2008 ELECTION     -0.025  0.002      (0.028)  (0.024) DDEM×2008 ELECTION -       -0.013   -0.020           (0.054)   (0.040) DREP×2008 ELECTION +       -0.010   -0.026           (0.052)   (0.040)   98   Expected Sign (1) (2) (3) (4) (5) (6) 2006 ELECTION      0.015 0.015       (0.028) (0.028) DDEM×2006 ELECTION -         -0.034 -0.033             (0.063) (0.063) DREP×2006 ELECTION +         0.109* 0.109*             (0.063) (0.063) POST08     -0.107***  -0.081***      (0.029)  (0.028) DEM×POST08     0.031  0.024      (0.060)  (0.054) REP×POST08     -0.000  -0.016      (0.058)  (0.057) POST06      -0.084** -0.030       (0.033) (0.031) DEM×POST06      0.012 -0.001       (0.073) (0.067) REP×POST06      0.045 0.060       (0.074) (0.077) LN_LOB + 0.002 0.001 0.002 0.002 0.002 0.002   (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) Controls  Yes Yes Yes Yes Yes Yes Industry Fixed Effects  Yes Yes Yes Yes Yes Yes Year Fixed Effects  Yes Yes No No No No N  2545 2545 2685 2545 2545 2545 adj. R-sq   0.773 0.773 0.760 0.767 0.766 0.767 dem = rep F-stat  0.09 0.35     Prob > F   0.759 0.553                        99  Table 2.7 Audit Fees Path Analysis Table 2.7 presents the path analysis results of how government contracts affects the relation between political connection and audit fee. The dependent variable is the log of audit fee. Control variables are the log of assets (LNAT), receivable and inventory (RECINV), return on assets (ROA), current assets (CATA), quick ratio (QUICK), loss indicator (LOSS), market-to-book ratio (MB), Big4 auditor indicator (BIG4), long-term debt (LTD), log of number of segments (LNSEG), restatement indicator (RESTATEMENT), and going concern indicator (GOING_CONCERN). Detailed variable definitions are in Appendix B.1. Standard errors are corrected for heteroscedasticity and clustered at firm level. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.    Expected Sign (1) (2) (3) (4) (5) PCB   0.080   0.350*    (0.050)   (0.195) PPCB    0.585*       (0.329)   LN_NPC     0.123**       (0.056)  LN_GC + 0.034*** 0.032*** 0.032*** 0.032*** 0.040***   (0.008) (0.008) (0.008) (0.008) (0.010) LN_LOB - 0.000 -0.000 -0.000 -0.000 -0.003   (0.004) (0.004) (0.004) (0.004) (0.004) PCB×LN_GC      -0.024*       (0.012) PCB×LN_LOB      0.009       (0.007) Controls  Yes Yes Yes Yes Yes Year Fixed Effects  Yes Yes Yes Yes Yes Industry Fixed Effects  Yes Yes Yes Yes Yes N  1763 1760 1760 1760 1760 adj. R-sq   0.793 0.793 0.794 0.795 0.794             100  Table 2.8 Audit Fees and Paths Analysis: Isolate Government Contract Table 2.8 presents the results of how government contracts affect the relation between political connection and audit fee. Panel A presents the first regression. The dependent variables in the first stage are all political connection variables (PCB, PPCB, and LN_NPC), where expected political connections are then estimated based on the amount of government contracts and other basic firm characteristics. Panel B presents the results for the second regression, where effects of expected political connections (variables that start with “EX_”), which incorporate the government contract information, and the residual information in actual political connections (variables that start with “RES_”) on audit fees are tested. The dependent variable for second step is the log of audit fee. Control variables are the log of assets (LNAT), receivable and inventory (RECINV), return on assets (ROA), current assets (CATA), quick ratio (QUICK), loss indicator (LOSS), market-to-book ratio (MB), Big4 auditor indicator (BIG4), long-term debt (LTD), log of number of segments (LNSEG), restatement indicator (RESTATEMENT), and going concern indicator (GOING_CONCERN). Detailed variable definitions are in Appendix B.1. Standard errors are corrected for heteroscedasticity and clustered at firm level. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.                101  Panel A. First Step    PCB PPCB LN_NPC LN_GC 0.024*** 0.032*** 0.004*** 0.004*** 0.024*** 0.031***  (0.008) (0.006) (0.001) (0.001) (0.007) (0.006) ROA -0.476  -0.027  -0.376   (0.465)  (0.084)  (0.448)  MB 0.011***  0.002***  0.009***   (0.004)  (0.001)  (0.003)  SIZE 0.111***  0.013***  0.102***   (0.024)  (0.004)  (0.023)  HHI -0.049  0.019  0.059   (0.129)  (0.022)  (0.122)  GROWTH -0.174**  -0.015  -0.124*   (0.081)  (0.011)  (0.067)  CAPEX -0.521**  -0.083**  -0.509**   (0.208)  (0.038)  (0.209)  LOSS 0.037  0.010  0.051   (0.072)  (0.012)  (0.069)  Industry FE Yes No Yes No Yes No Year FE Yes No Yes No Yes No N 2232 2769 2232 2769 2232 2769 adj. R-sq 0.219 0.055 0.210 0.051 0.248 0.065           102  Panel B. Second Step  Expected Sign (1) (2) (3) (4) (5) (6) EX_PCB + 0.880***        (0.239)      RES_PCB ? 0.084*        (0.048)      EX_PCB2 +  1.084***        (0.249)     RES_PCB2 ?  0.085*        (0.049)     EX_PPCB +   6.905***        (1.668)    RES_PPCB ?   0.077        (0.048)    EX_PPCB2 +    7.221***        (1.778)   RES_PPCB2 ?    0.085*        (0.049)   EX_LN_NPC +     1.015***        (0.255)  RES_LN_NPC ?     0.081*        (0.048)  EX_LN_NPC2 +      1.127***        (0.260) RES_LN_NPC2 ?      0.085*        (0.049) Controls  Yes Yes Yes Yes Yes Yes Year Fixed Effects  Yes Yes Yes Yes Yes Yes Industry Fixed Effects  Yes Yes Yes Yes Yes Yes N  1760 1760 1760 1760 1760 1760 adj. R-sq   0.792 0.795 0.795 0.795 0.793 0.795     103  Table 2.9 Audit Report Lag Table 2.9 presents the results of testing the relationship between political connection and audit report lag. The dependent variable is the log of audit report lag (ARL). PCB is an indicator of political connection, which takes on the value 1 when at least one board member of the firm-year is identified as politically connected. PPCB is the proportion of politically connected board, which is calculated by the number of politically connected board members over the size of the board. LN_NPC is the log number of politically connected board members. PREP (PDEM) are percentage of directors that are identified as being connected to the Republican (Democratic) Party. LN_REP (LN_DEM) is log number of political connections that are identified as being connected to the Republican (Democratic) Party. LEGISLATIVE (EXECUTIVE) are indicators of having at least one political connection with legislative (executive) branch experience in the United States government. Detailed variable definitions are in Appendix B.1. Standard errors are corrected for heteroscedasticity and clustered at firm level. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.     104   (1) (2) (3) (4) (5) (6) (7) (8) PCB -0.036         (0.034)        PPCB  -0.158         (0.216)       DDEM   -0.039   -0.028      (0.053)   (0.060)   DREP   -0.001   -0.003      (0.056)   (0.062)   LN_DEM    -0.043   -0.037      (0.071)   (0.082)  LN_REP    0.023   0.016      (0.074)   (0.081)  PDEM     -0.056   -0.052      (0.451)   (0.551) PREP     0.355   0.231      (0.495)   (0.508) 2008 ELECTION   -0.108*** -0.100*** -0.093***       (0.028) (0.029) (0.028)    2006 ELECTION      0.019 0.014 0.008       (0.034) (0.034) (0.033) DDEM×2008 ELECTION   0.088         (0.072)      DREP×2008 ELECTION   -0.007         (0.060)      DDEM×2006 ELECTION      -0.029         (0.068)   DREP×2006 ELECTION      0.002         (0.063)   LN_DEM×2008 ELECTION    0.083         (0.098)     LN_REP×2008 ELECTION    -0.025         (0.076)     LN_DEM×2006 ELECTION       -0.016         (0.096)  LN_REP×2006 ELECTION       0.015         (0.082)  PDEM×2008 ELECTION     0.340         (0.705)      105   (1) (2) (3) (4) (5) (6) (7) (8)          PREP×2008 ELECTION     -0.303         (0.487)    PDEM×2006 ELECTION        -0.038         (0.668) PREP×2006 ELECTION        0.297         (0.615) LNAT -0.024* -0.026* -0.026* -0.026* -0.028* -0.025* -0.026* -0.027*  (0.014) (0.015) (0.015) (0.015) (0.015) (0.015) (0.015) (0.015) RECINV -0.244* -0.249* -0.248* -0.245* -0.245* -0.245* -0.242* -0.241*  (0.132) (0.132) (0.132) (0.132) (0.132) (0.132) (0.132) (0.132) ROA -0.354* -0.358* -0.361* -0.366* -0.382** -0.356* -0.360* -0.377**  (0.186) (0.187) (0.187) (0.189) (0.192) (0.187) (0.188) (0.192) LOSS -0.028 -0.029 -0.032 -0.032 -0.033 -0.034 -0.033 -0.035  (0.044) (0.044) (0.043) (0.044) (0.044) (0.043) (0.044) (0.044) BIG4 -0.028 -0.028 -0.030 -0.031 -0.031 -0.030 -0.031 -0.032  (0.056) (0.055) (0.058) (0.058) (0.057) (0.058) (0.058) (0.057) GOING_CONCERN 0.484 0.481 0.473 0.467 0.465 0.480 0.475 0.475  (0.670) (0.665) (0.658) (0.653) (0.646) (0.660) (0.656) (0.652) COUNT_WEAK 0.106*** 0.107*** 0.106*** 0.107*** 0.107*** 0.107*** 0.107*** 0.107***  (0.019) (0.019) (0.019) (0.019) (0.019) (0.019) (0.019) (0.019) Post-election Controls No No Yes Yes Yes Yes Yes Yes Industry Fixed Effect Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Effect Yes Yes No No No No No No N 3218 3218 3218 3218 3218 3218 3218 3218 adj. R-sq 0.077 0.076 0.077 0.077 0.077 0.076 0.076 0.076     106  Table 2.10 Internal Control System This table presents the results of how internal control system affects the relation between political connection and audit fee. The dependent variable is the log of audit fee (AF). PCB is an indicator of political connection, which takes on the value 1 when at least one board member of the firm-year is identified as politically connected. PPCB is the proportion of politically connected board, which is calculated by the number of politically connected board members over the size of the board. Control variables are the log of assets (LNAT), receivable and inventory (RECINV), return on assets (ROA), current assets (CATA), quick ratio (QUICK), loss indicator (LOSS), market-to-book ratio (MB), Big4 auditor indicator (BIG4), long-term debt (LTD), log of number of segments (LNSEG), restatement indicator (RESTATEMENT), and going concern indicator (GOING_CONCERN). Detailed variable definitions are in Appendix B.1. Standard errors are corrected for heteroscedasticity and clustered at firm level. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.     107   (1) (2) (3) (4) (5) (6) PCB 0.125*** 0.124*** 0.124***     (0.044) (0.044) (0.044)    PPCB    0.717** 0.710** 0.711**     (0.290) (0.290) (0.290) MATERIAL_WEAKNESS 0.143   0.120    (0.257)   (0.259)   COUNT_WEAK  -0.013   -0.014    (0.026)   (0.026)  LN(COUNT_WEAK)   -0.040   -0.048    (0.154)   (0.153) PCB×MATERIAL_WEAKNESS 0.713**       (0.307)      PCB×COUNT_WEAK  0.239***       (0.036)     PCB×LN(COUNT_WEAK)   0.640***       (0.154)    PPCB×MATERIAL_WEAKNESS    4.937***       (1.531)   PPCB×COUNT_WEAK     1.319***       (0.210)  PPCB×LN(COUNT_WEAK)      3.677***       (0.901) Controls Yes Yes Yes Yes Yes Yes Industry Fixed Effect Yes Yes Yes Yes Yes Yes Year Fixed Effect Yes Yes Yes Yes Yes Yes N 2545 2545 2545 2545 2545 2545 adj. R-sq 0.775 0.775 0.775 0.774 0.774 0.774    108  Conclusion This thesis examines research questions regarding politically connected firms in the United States. Political connections are established to satisfy different needs of firms: when the goal is to obtain government contracts, they seek connections to the executive branch; when the goal is to manage regulations and law-making, they seek connections to the legislative branch. In particular, empirical data shows that the right political connections do help firms to obtain more government contracts, which is beneficial for a firm’s future performance but not its current performance. On the other hand, though political connection is positively associated with government subsidies, but research design cannot identify causality in the relations. I find that auditors charge higher audit fees to politically connected firms. This finding is not likely due to lower financial reporting quality, as prior literature has suggested, nor is it due to higher exposure to political risk. Instead, the results are consistent with the hypothesis that higher audit fees are due to politically connected firms’ involvement in government contracts. Being a government contractor means that in addition to compliance with US GAAP, a firm’s financial report needs to comply with additional regulations such as FAR, which are subject to additional audits from government auditors like DCAA and GAO. This additional compliance significantly increase politically connected firms’ audit risk, resulting in higher audit fees. Overall, I fail to find any indication that such relation leads to the extraction of political rent in my sample of U.S. firms. This study shows that some of the cynical views toward corporate political connections, supported by research using international data, may not apply to cases in the United States.   109  Bibliography Abbott,L. J., Parker,S., and Peters,G. F. , 2012. Audit Fee Reductions from Internal Audit-Provided Assistance: The Incremental Impact of Internal Audit Characteristics. Contemporary Accounting Research, 29(1), 94–118. Baloria,V. , 2014a. Politicians’ Equity Holdings and Accounting Conservatism, Working paper. Baloria,V. , 2014b. Politicians’ Equity Holdings and Accounting Conservatism. Working Paper, 0–40. Bell,T. B., Landsman,W. R., and Shackelford,D. A. , 2001. Auditors’ perceived business risk and audit fees: Analysis and evidence. Journal of Accounting Research, 39(1), 35–43. Carcello,J. V., Hermanson,D. R., Neal,T. L., and Riley Jr.,R. a. , 2002. Board Characteristics and Audit Fees. Contemporary Accounting Research, 19(3), 365–384. Carcello,J. V, Hermanson,D. R., Neal,T. L., and Riley,R. A. , 2002. Board characteristics and audit fees. Contemporary Accounting Research, 19(3), 365–384. Chaney,P. K., Faccio,M., and Parsley,D. , 2011. The quality of accounting information in politically connected firms. Journal of Accounting and Economics, 51(1–2), 58–76. Choi,J. H., Kim,J. B., Liu,X., and Simunic,D. A. , 2009. Cross-listing audit fee premiums: Theory and evidence. Accounting Review, 84(5), 1429–1463. Claessens,S., Feijen,E., and Laeven,L. , 2008. Political connections and preferential access to finance: The role of campaign contributions. Journal of Financial Economics, 88(3), 554–580. Cooper,M. J., Gulen,H., and Ovtchinnikov,A. V. , 2010. Corporate Political Contributions and Stock Returns. Journal of Finance, 65(2), 687–724. Correia,M. M. , 2014. Political connections and SEC enforcement. Journal of Accounting and   110  Economics, 57(2–3), 241–262. Dechow,P. M., and Dichev,I. D. , 2002. Quality Earnings : The The Accruals Accrual Estimation Errors. The Accounting Review, 77(2002), 35–59. Dechow,P. M., Ge,W., Larson,C. R., and Sloan,R. G. , 2011. Predicting Material Accounting Misstatements. Contemporary Accounting Research, 28(1), 17–82. Dechow,P. M., Sloan,R. G., and Sweeney,A. P. , 1995. Detecting Earnings Management. The Accounting Review, 70(2), 193–225. Duchin,R., and Sosyura,D. , 2012. The politics of government investment. Journal of Financial Economics, 106(1), 24–48. Faccio,M. , 2006. Politically Connected Firms. The American Economic Review, 96(1), 369–386. Faccio,M. , 2010. Differences between Politically Connected and Nonconnected Firms: A Cross-Country Analysis. Financial Management, 39(3), 905–928. Faccio,M., Masulis,R. W., and McConnell,J. J. , 2006. Political Contributions and Corporate Bailouts. Journal of Finance, 61(6), 2597–2635. Ferguson,A., Francis,J. R., and Stokes,D. J. , 2003. The effects of firm-wide and office-level industry expertise on audit pricing. Accounting Review, 78(2), 429–448. Fishman,R. , 2001. Estimating the Value of Political Connections. The American Economic Review, 91(4), 1095–1102. Goldman,E., Rocholl,J., and So,J. , 2008. Political connections and the allocation of procurement contracts. Review of Finance, 1–50. Goldman,E., Rocholl,J., and So,J. , 2009. Do politically connected boards affect firm value. Review of Financial Studies, 22(6), 2331–2360. Goldman,E., Rocholl,J., and So,J. , 2013. Politically Connected Boards of Directors and The   111  Allocation of Procurement Contracts. Review of Finance, 17(5), 1617–1648. Gul,F. A. , 2006. Auditors’ response to political connections and cronyism in Malaysia. Journal of Accounting Research, 44(5), 931–963. Gul,F. A., Chen,C. J. P., and Tsui,J. S. L. , 2003. Discretionary accounting accruals, managers’ incentives, and audit fees. Contemporary Accounting Research, 20(3), 441–464. Hay,D. C., Knechel,W. R., and Won,N. , 2006. Audit Fees : A Meta-Analysis of the Effect of Supply and Demand Attributes Audit Fees : A Meta-Analysis of The Effect of Supply And Demand Attributes. Contemporary Accounting Research, 23(1), 141–191. Hill,M. D., Fuller,K. P., Kelly,G. W., and Washam,J. O. , 2014. Corporate cash holdings and political connections. Review of Quantitative Finance and Accounting, 42(1), 123–142. Hillman,A. , 2005. Politicians on the Board of Directors: Do Connections Affect the Bottom Line? Journal of Management, 31(3), 464–481. Houston,J. F., Jiang,L., Lin,C., and Ma,Y. , 2014. Political connections and the cost of bank loans. Journal of Accounting Research, 52(1), 193–243. Hung,M., Wong,T. J., and Zhang,T. , 2012. Political considerations in the decision of Chinese SOEs to list in Hong Kong. Journal of Accounting and Economics, 53(1–2), 435–449. Johnson,S., and Mitton,T. , 2003. Cronyism and capital controls: Evidence from Malaysia. Journal of Financial Economics, 67(2), 351–382. Khan,M., and Watts,R. L. , 2009. Estimation and empirical properties of a firm-year measure of accounting conservatism. Journal of Accounting and Economics, 48(2–3), 132–150. Kim,C., and Zhang,L. , 2014. Corporate Political Connections and Tax Aggressiveness. Contemporary Accounting Rsearch, forthcoming. Kim,C., and Zhang,L. , 2016. Corporate Political Connections and Tax Aggressiveness.   112  Contemporary Accounting Research, 33(1), 78–114. Larcker,D. F., and Richardson,S. A. , 2004. Fees paid to audit firms, accrual choices, and corporate governance. Journal of Accounting Research, 42(3), 625–658. Li,H., Meng,L., Wang,Q., and Zhou,L. A. , 2008. Political connections, financing and firm performance: Evidence from Chinese private firms. Journal of Development Economics, 87(2), 283–299. Lyon,J. D., and Maher,M. W. , 2005. The importance of business risk in setting audit fees: Evidence from cases of client misconduct. Journal of Accounting Research, 43(1), 133–157. Maher,M. W., Tiessen,P., Colson,R., and Broman,A. J. , 1992. Competition and Audit Fees. Accounting Review, 67(1), 199–211. Mattera,P., and Tarczynska,K. , 2015. Uncle Sam’s Favoriate Corporations: Identifying the Large Companies that Dominate Federal Subsidies. Good Jobs First Report. McGowan,A. S., and Vendrzyk,V. P. , 2002. The Shifting in Relation and between Cost Segment Profitability the Industry. The Accounting Review, 77(4), 949–969. Palmer,M., and Schneer,B. , 2016. Capitol Gains: The Returns to Elected Office from Corporate Board Directorships. The Journal of Politics, 78(1), 181–196. Palmrose,Z.-V. , 1986. Audit Fees and Auditor Size : Further Evidence. Journal of Accounting Research, 24(1), 97–110. Schoenherr,D. , 2015. Political Connections and Allocative Distortions. Job Market Paper. Shaffer,B. , 1995. Firm-Level Responses To Government-Regulation - Theoretical and Research Approaches. Journal of Management, 21(3), 495–514. Simunic,D. A. , 1980. The Pricing of Audit Services: Theory and Evidence. Journal of   113  Accounting Research, 18(1), 161–190. Tahoun,A. , 2014. The role of stock ownership by US members of Congress on the market for political favors. Journal of Financial Economics, 111(1), 86–110.    114  Appendices Appendix A   Appendix for Chapter 1 A.1 Variable Definitions  Political connection variables PCB Indicator of political connections in the board. It takes on the value of 1 when at least one of the board members is identified as politically connected according to Goldman et al (2009), measured at the beginning of the fiscal year. PPCB Percentage of the board that is politically connected, calculated as the total number of politically connected board members scaled by the size of board.  DDEM Indicator of Democratic connected board. It takes on the value of 1 when at least one of the politically connected board member is identified as associated with the Democratic Party. DREP Indicator of Republican connected board. It takes on the value of 1 when at least one of the politically connected board member is identified as associated with the Republican Party. LN_DEM Log number of Democratic connected board members, calculated as the natural log of 1 plus the number of Democratic connected board members. LN_REP Log number of Republican connected board members, calculated as the natural log of 1 plus the number of Republican connected board members. PDEM Percentage of Democratic connected board members, calculated as the total number of Democratic connected board members scaled by the size of the board. PREP Percentage of Republican connected board members, calculated as the total number of Republican connected board members scaled by the size of the board. LEGISLATIVE Indicator of political connections with the legislative branch of the United States of America. EXECUTIVE Indicator of political connections with the executive branch of the United States of America.   Firm Performance Variables ROA Return on assets, calculated as income before extraordinary item over total assets.   115  FROA One-year-ahead return on assets. FROA2 Two-year-ahead return on assets. FROA3 Three-year-ahead return on assets. FROA4 Four-year-ahead return on assets. FROA5 Five-year-ahead return on assets. PM Profit margin, calculated as net income scaled by sales. FPM One-year-ahead profit margin. FPM2 Two-year-ahead profit margin. FPM3 Three-year-ahead profit margin. FPM4 Four-year-ahead profit margin. FPM5 Five-year-ahead profit margin.     Government benefit variables LN_GC Log number of government contract amount for a firm in a specific fiscal year LN_SUBSIDIES Log number of 1 plus total subsidies amount for a firm in a specific fiscal year. LN_LOAN Log number of 1 plus total government loan for a firm in a specific fiscal year. LN_TAX_CREDIT Log number of 1 plus total tax credit, indicated by subsidy types as "tax credit/rebate" or "federal allocated tax credit" for a firm in a specific fiscal year.     Other variables GOVT_SALE Measures the importance of government contracting. Calculated as the ratio of government contract to sales at the end of the year. LOB_SALE Measures regulatory risk. Calculated as 1000 times the ratio of lobbying to sales at the end of the year. GC Indicator of high government contract dependence, which takes on the value of 1 when GOVT_SALE>=0.1. REGULATE Indicator of firms with high regulatory risk, which takes on the value of 1 when LOB_SALE>=0.3. 2006 ELECTION Indicator of post 2006 midterm election effects, which takes on the value of 1 for firm's fiscal year 2007 and 2008. 2008 ELECTION Indicator of post 2008 presidential election effects, which takes on the value of 1 for firm's fiscal year 2009 and 2010. COGS Cost of goods sold, scaled by sales. CAPEX Capital expenditure, scaled by sales. HHI The Herfindahl index. Calculated using the 2-digit SIC industry classification in COMPUSTAT universe of the year.   116  SIZE Natural log of market capital. LNAGE Natural log of firm's age, measured from the first date that the firm show up on CRSP. GROWTH Sales growth. MB Market-to-book ratio. LOSS Loss indicator, which takes on the value of 1 when net income is less than 0.      117  Appendix B  Appendix for Chapter 2 B.1 Variable Definitions  Political connection variables PCB Politically connected board indicator, which takes up the value of 1 when at least one of the board of directors is defined as a politician. PPCB Percentage of politically connected board, calculated as the number of politicians on the board divided by the total number of directors. LN_NPC The natural log number of 1 plus politically connected board members. DREP Republican indicator, which takes up the value of 1 when at least one of the politician is identified as a Republican. DDEM Democrat indicator, which takes up the value of 1 when at least one of the politicians is identified as a Democrat. LN_REP The natural log number of 1 plus politically connected board members who is identified as a Republican. LN_DEM The natural log number of 1 plus politically connected board members who is identified as a Democrat. LEGISLATIVE Congressman indicator, which takes up the value of 1 when the politician served in the United States Congress. EXECUTIVE Executive branch indicator, which takes up the value of 1 when the politician served in the executive branch of the federal government.     Audit fee variables AF Natural log of audit fee. LNAT Natural log of total assets of the firm, calculated as ln(Compustat at). LNSEG Natural log of a firm's operating segments. CATA Current assets scaled by total assets. MB Market-to-book ratio, calculated as the market value of equity divided by book value of equity (Compustat ceq). LTD Leverage, calculated as total long term debt (Compustat dltt) over total assets (Compustat at). RECINV The sum of receivables (Compustat rect) and inventories (Compustat invt) over total assets (Compustat at). LNSEG Natural log of total number of business segments. GOING_CONCERN Indicator of a going-concern opinion from auditors.   118  ROA Return on assets, calculated as income before extraordinary items (Compustat ib) over total assets (Compustat at). LOSS Loss indicator, which takes up the value of 1 when a firm has negative net income (Compustat ni). QUICK Quick ratio. BIG4 Big four auditor indicator.     Accounting quality variables C_SCORE A firm-year conservatism measure, calculated following Khan and Watts (2009), and truncated at 1% of the COMPUSTAT population after 1999. DACC Discretionary accruals, calculated following modified Jone's model, and truncated at 1% of the COMPUSTAT population after 1999. DDACCR Dichow and Dechiev (2002) discretionary accruals. AAER Indicator for AAER firms, which takes on the value of 1 if there is at least one AAER litigation during that year. RESTATEMENT Indicator of restatement, which takes on the value of 1 if there is at least one restatement during that year. MATERIAL_WEAKNESS Indicator of material weakness according to SOX 302. COUNT_WEAK Number of material weakness.     Other variables LN_GC Log amount of government contracts for a firm in a specific fiscal year. LN_LOB Log number of total lobbying expenditure plus 1 for a firm in a specific fiscal year. SIZE Market capital of the firm in a specific fiscal year.  GROWTH Sales growth compared to last year. CAPEX Capital expenditure over sales. ARL Log number of length of time auditors need to complete the audit. Calculated as the difference between the file date and a firm's fiscal year end date.  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0357147/manifest

Comment

Related Items