Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Essays on labour market fluctuations in emerging markets Kabaca, Serdar 2013

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

Item Metadata


24-ubc_2013_fall_kabaca_serdar.pdf [ 912.78kB ]
JSON: 24-1.0166817.json
JSON-LD: 24-1.0166817-ld.json
RDF/XML (Pretty): 24-1.0166817-rdf.xml
RDF/JSON: 24-1.0166817-rdf.json
Turtle: 24-1.0166817-turtle.txt
N-Triples: 24-1.0166817-rdf-ntriples.txt
Original Record: 24-1.0166817-source.json
Full Text

Full Text

Essays on Labour Market Fluctuationsin Emerging MarketsbySerdar KabacaB.A., Bog?azic?i University, 2006A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFDOCTOR OF PHILOSOPHYinThe Faculty of Graduate and Postdoctoral Studies(Economics)THE UNIVERSITY OF BRITISH COLUMBIA(Vancouver)October 2013c? Serdar Kabaca 2013AbstractThe goal of this dissertation is to compare and contrast labour market fluc-tuations in emerging and developed markets, and to explore the sources ofdifferences in these fluctuations across country groups. Chapter 2 documentscyclical properties of labour share over the cycle for various countries andshow that there is a close relationship between labour share and the costof borrowing. Labour share tends to be more volatile and procyclical withoutput especially in countries with highly volatile and countercyclical inter-est rates. The results are driven neither by sectoral shifts over the cycle norby the measurement errors in the labour compensation data. In Chapter 3,we show that working capital requirements can predict the right sign of thelabour share comovement with output and can partly account for the volatil-ity of the labour share. It is also shown that imperfect financial markets inthe form of credit restrictions not only amplify the results for the variabilityof labour share but also helps better explain some of the striking businesscycle regularities in emerging markets, such as highly volatile consumption,strongly procyclical investment and consumption, and countercyclical netexports.Fluctuations in real wages are mostly responsible for the highly volatilelabour share in emerging markets. Previous literature showed that searchfrictions with countercyclical interest rates can explain movements in wagesin these economies. Chapter 4 shows that when agents are allowed to choosethe amount of hours worked (intensive margin of the labour input), the ef-fects of search frictions on wages are mitigated. Our motivation of intro-ducing intensive margin comes from the fact that variations in hours perworker are at least as significant as those in the employment in emergingiiAbstractmarkets. They are also more cyclical with output in these economies thanin developed ones. Search frictions fail to explain these cyclical properties ofthe intensive margin. On the other hand, by introducing financial frictions,the model can predict them together with movements in real wages. Thissuggests that frictions in both labour and financial markets go further inexplaining emerging market business cycles.iiiPrefaceChapter 4 of this dissertation, titled ?Search Frictions, Financial Frictions,and Labour Market Fluctuations in Emerging Markets?, is a manuscriptco-authored with Sumru G. Altug. The identification and design of theresearch program for this paper were carried out by Serdar Kabaca, withcomments by Sumru G. Altug. Background research, the data analysis, andthe preparation of the manuscript were performed jointly. All other chaptersof the dissertation is accomplished solely by Serdar Kabaca.?Search Frictions, Financial Frictions, and Labour Market Fluctuationsin Emerging Markets? Altug and Kabaca (2013), has been submitted forpublication.ivTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . vList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiiList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xAcknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . xiiDedication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiv1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 The Empirics of Labour Share Fluctuations in EmergingMarket Economies . . . . . . . . . . . . . . . . . . . . . . . . 42.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 42.2 Stylized Facts on the Fluctuations of Labour Share . . . . 102.2.1 The Measure of Labour Share . . . . . . . . . . . . 102.2.2 Data . . . . . . . . . . . . . . . . . . . . . . . . . . 112.2.3 Observations . . . . . . . . . . . . . . . . . . . . . . 122.3 Robustness Analyses . . . . . . . . . . . . . . . . . . . . . 292.3.1 Self-employment Adjustments . . . . . . . . . . . . 302.3.2 Informal Sector Considerations . . . . . . . . . . . . 332.3.3 Sectoral Shifts and Government Expenditure over theCycle . . . . . . . . . . . . . . . . . . . . . . . . . . 38vTable of Contents2.4 Interest Rates and Financial Environment . . . . . . . . . 412.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . 443 The Role of the Cost of Borrowing on Labour Share Fluc-tuations in Emerging Markets . . . . . . . . . . . . . . . . . 453.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 453.2 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 503.2.1 Optimization Problem . . . . . . . . . . . . . . . . 503.2.2 Competitive Equilibrium and Labour Share . . . . 543.3 Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . 583.3.1 Shocks . . . . . . . . . . . . . . . . . . . . . . . . . 583.3.2 Other Model Parameters . . . . . . . . . . . . . . . 623.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 663.5 Implications for Developed Economies . . . . . . . . . . . . 723.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . 754 Search Frictions, Financial Frictions, and Labour MarketFluctuations in Emerging Markets . . . . . . . . . . . . . . 774.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 774.2 Fluctuations in the Intensive and Extensive Margins of LabourInput in Emerging Markets . . . . . . . . . . . . . . . . . . 814.3 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 874.3.1 The Firm?s Problem . . . . . . . . . . . . . . . . . . 874.3.2 The Household?s Problem . . . . . . . . . . . . . . . 894.3.3 Nash Bargaining . . . . . . . . . . . . . . . . . . . . 944.3.4 Equilibrium Prices and Allocation . . . . . . . . . . 954.4 Quantitative Analysis . . . . . . . . . . . . . . . . . . . . . 954.4.1 Calibration of Parameter Values . . . . . . . . . . . 964.4.2 Equilibrium and Impulse Responses . . . . . . . . . 1004.4.3 Quantitative Results . . . . . . . . . . . . . . . . . 1074.4.4 Sensitivity Analysis . . . . . . . . . . . . . . . . . . 1144.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . 1165 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117viTable of ContentsBibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119AppendicesA Appendices to Chapter 2 . . . . . . . . . . . . . . . . . . . . 127A.1 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127B Appendices to Chapter 3 . . . . . . . . . . . . . . . . . . . . 132B.1 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132B.2 Parameters for Mexico and Canada . . . . . . . . . . . . . 133C Appendices to Chapter 4 . . . . . . . . . . . . . . . . . . . . 135C.1 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135viiList of Tables2.1 The Cyclical Properties of Labour Share in EMs . . . . . . 132.2 The Cyclical Properties of Labour Share in DMs . . . . . . 142.3 Evidence from Quarterly Data . . . . . . . . . . . . . . . . 182.4 Labour Share Components ? Cyclical Wages . . . . . . . . . 222.5 Labour Share Components ? Cyclical Employment . . . . . 232.6 Adjustments for Self-employment . . . . . . . . . . . . . . . 322.7 Cyclical Variation of Self-Employment and Total Employ-ment in EMs . . . . . . . . . . . . . . . . . . . . . . . . . . 332.8 Sectoral Shifts over the Cycle in EMs . . . . . . . . . . . . . 403.1 Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . 603.2 Model Implications for Mexico . . . . . . . . . . . . . . . . 703.3 Model Implications for Mexico and Canada . . . . . . . . . 744.1 Movements in Employment and Hours Per Worker . . . . . 834.2 Correlation between Spread and Labour in Emerging Mar-kets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 864.3 Parameter Values . . . . . . . . . . . . . . . . . . . . . . . . 1014.4 Results with Separable Preferences . . . . . . . . . . . . . . 1094.5 Results with JR Preferences . . . . . . . . . . . . . . . . . . 1114.6 More Sensitivity Analyses . . . . . . . . . . . . . . . . . . . 112A.1 Data Sources for Labor Compensation and Interest Rates . 129B.1 Parameter Values for Mexico and Canada . . . . . . . . . . 134B.2 Shock Processes . . . . . . . . . . . . . . . . . . . . . . . . . 134viiiList of TablesC.1 Data Sources . . . . . . . . . . . . . . . . . . . . . . . . . . 136ixList of Figures1.1 Labour Share Fluctuations across Income Level . . . . . . . 21.2 Labour Share vs. Interest Rate Fluctuations . . . . . . . . . 32.1 Labour Share Fluctuations in Mexico, Korea, Turkey and theUS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62.2 Correlation between Labour Share and Interest Rate . . . . 192.3 Labour Share in EMs during the Global Financial Crisis . . 252.4 Advanced Economies? Labour Share during the Debt Crisis 272.5 Interest Rates in Selected Advanced Economies during theDebt Crisis . . . . . . . . . . . . . . . . . . . . . . . . . . . 282.6 Informality and the Cyclical Properties of Labour Share . . 352.7 Informal Sector as a Share of GDP in Mexico . . . . . . . . 362.8 Labour Share Adjusted for Informality and Self-Employmentin Mexico . . . . . . . . . . . . . . . . . . . . . . . . . . . . 372.9 Labour Share across Industries in Mexico and Korea . . . . 423.1 Timing Line . . . . . . . . . . . . . . . . . . . . . . . . . . . 513.2 Labour Share in Mexico . . . . . . . . . . . . . . . . . . . . 593.3 The Cyclical Component of the Credit-to-GDP Ratio in Mex-ico . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 633.4 Adjusted Labour Share and Sectoral Shares in Mexico . . . 653.5 Impulse Responses . . . . . . . . . . . . . . . . . . . . . . . 684.1 Employment and Hours Per Worker during Crises . . . . . 794.2 Spread during Crises . . . . . . . . . . . . . . . . . . . . . . 854.3 Impulse Responses . . . . . . . . . . . . . . . . . . . . . . . 103xList of FiguresA.1 Correlation of Labor Share and EMBI Interest Rates withOutput . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130A.2 Labor Share (Self-Employment-Adjusted) vs. Interest RateFluctuations . . . . . . . . . . . . . . . . . . . . . . . . . . . 131C.1 EPL Index across EMs ad DMs . . . . . . . . . . . . . . . . 137xiAcknowledgementsSeveral people contribute to this dissertation and help me during my gradu-ate work. First of all, I am grateful to my supervisors Michael B. Devereuxand Henry E. Siu for their invaluable advice and guidance that helped meunderstand the subject and research standards in the study of economics.I would also like to thank Viktoria Hnatkovska for spending considerabletime to discuss with me on my research ideas especially at the early stagesof my dissertation. You vastly enhanced my understanding of business cy-cle work. I also benefitted from very helpful comments and suggestionsfrom Paul Beaudry, Yaniv Yedid-Levi, and other faculty at UBC Economicsthroughout my Ph.D study. Okan Y?lankaya and Go?rkem C?elik were alwaysready to offer friendly support, advice, and long conversations over variousaspects of life. Thank you, both of you. I am also indebted to Sumru G.Altug? for insightful comments and honest advice, and for co-authoring oneof the chapters here.I benefited also from discussions with participants at UBC Macro Lunch,UBC Macro Seminars, Canadian Economic Association Conference 2010 and2011, Computing in Economics and Finance Conference 2010, and Euro-Conference 2010 organized by Society for the Study of Emerging Markets.I would also like to thank my professors at Bogazici University, especiallyC. Emre Alper, Go?khan O?zertan, and I?smail Sag?lam for helping me startthe graduate program well-prepared and for inspiring me to pursue a Ph.Ddegree. I also acknowledge the financial support from Paul Beaudry andHenry E. Siu.My special thanks go to my classmates for discussions on my researchand those beautiful five years together, particularly Mat??as Corte?s for un-conditional friendship and willingness to help with my questions. GoodxiiAcknowledgementsfriends left so many memories and kept me productive during the gradu-ate work. Thank you, C?ag?la Suzan Alt?ntas?, Adem Aygu?n, Amelia Bain,Arturo Belleste, Mathieu Caissy, Caner Eks?iog?lu, O?zge Go?ktepe, TayfunGu?rdal, Dilek Kayaalp, Dave Narayan, Gerc?ek Ug?ur O?zcan, S?ahende Peker,Alexander Pluke, Mustafa Tug?an, and other great people in my life. Mygreatest debt go to my parents and my brother for their constant support,encouragement, and love. I do not think I can ever repay my parents forcreating such a priceless family environment for me and my brother. That iswhy this dissertation is dedicated to them. And lastly, but surely not least,I am grateful to my love, Su?ndu?s, for entering my life during the final stageof my studies, and for her unconditional support and love, which kept mehappy and ultimately more productive.xiiiDedicationTo my parents Ayten and Og?uz Kabaca.None of this would have been possible without their love and sacrifices.xivChapter 1IntroductionLabour markets in emerging markets (EMs) are characterized by highlyvolatile and procyclical real wages in the literature. Real wages in EMs arealmost twice as volatile as in developed markets (DMs). Additionally, theyare more procyclical with output in EMs than in DMs. This dissertationshows that the high volatility in real wages in these economies are not off-setby small variations in labour input. Thus, total labour income as well tendsto be more volatile than output in EMs. This is in contrast to developedmarkets, where labour markets tend to be more sluggish and lag behind out-put. Our measure of labour market responsiveness is the cyclical propertyof labour share over the cycle.Figure 1.1 illustrates the characteristics of labour share fluctuations inboth emerging and developed markets, the details of which are studied inChapter 2. Labour share tends to be procyclical with output on average inemerging markets whereas it is slightly countercyclical in developed markets.In addition, labour share is much more volatile in low income countries thanin developed markets.1 However, there is a large variation across countriesin terms of characteristics of labour share fluctuations. India, for instance,having the lowest income per capita in our sample does not show a procycli-cal labour share. Yet Korea has a strongly procyclical one although it is oneof the richest emerging economies. Figure 1.2, on the other hand, gives usa more clear picture. It shows that labour share is procyclical with outputespecially in the countries with countercyclical interest rates, i.e., a decreasein the cost of borrowing during booms tends to increase labour share. The1Despite small variations and weak cyclicality in the labour share of developed markets,there is a declining trend in recent years in these economies. This dissertation studies thedifferent business cycle movements of this variable across country groups, rather thanchanges in the trend.1Chapter 1. IntroductionFigure 1.1: Labour Share Fluctuations across Income LevelARGBRACHLCRIKORMEXPERPOLPHLZAFTURCOLCZHEGYHUNINDISRRUSAUSAUTCANNLDNZLESPSWEUKUSDNKFINFRAGERGRCISLIRLITANOR?.50.5corr(s,y)0 10000 20000 30000 40000GDP per capitaARGBRACHLCRIKORMEXPERPOLPHLZAFTURCOL CZHEGY HUNINDISRRUSAUSAUTCANNLDNZLESPSWEUKUSDNKFINFRAGERGRCISLIRLITANOR0246sigma(s)0 10000 20000 30000 40000GDP per capitaNote: corr(s,y) and sigma(s) denote the correlation of labour share with output and thestandard deviation of labour share, respectively. The variables are detrended and logged.Labour share is annual and covers the period after 1980 for most countries. GDP per capita(PPP adjusted in US dollars) in 2000 is taken for the income level. See the Appendix Afor data sources.more countercyclical interest rates are, the more procyclical labour share is.Moreover, the countries that face more volatile interest rate tend to havemore volatile labour share, as well. We also show below that negative slopecoefficients in Figure 1.1 disappear when we take fluctuations in the interestrate into account. Finally, these results are robust to adjustments of thelabour share that controls self-employment and the informal sector.After discussing the volatility and the cyclicality of labour share in bothEMs and DMs, we build a model in Chapter 3, where wages have to befinanced through working capital loans, and show that the variation in thecost of borrowing can account for the movements of the labour share over thecycle. The premise of this dissertation is that financing matters to labourmarket fluctuations and that emerging markets serve as a good natural ex-periment due to the financial problems and the different features of theinterest rate that they face. On the other hand, the contribution of thesefrictions might be smaller in developed markets as they face a more stableinterest rate over the cycle.2Chapter 1. IntroductionFigure 1.2: Labour Share vs. Interest Rate FluctuationsARGBRACHLCRIKORMEXPERPOLPHLZAFTURCOLCZHEGYHUNINDISRRUSAUSAUTCANNLDNZLESPSWEUKUSDNKFIN FRAGERGRCISLIRLITANOR?.50.5corr(s,y)?.5 0 .5corr(r,y)ARGBRACHLCRIKORMEXPERPOLPHLZAFTURCOLCZHEGYHUNINDISRRUSAUSAUTCANNLDNZLESPSWEUKUSDNKFINFRAGERGRCISLIRLITANOR0246sigma(s)0 2 4 6 8 10sigma(r)Note: corr(s,y) and sigma(s) denote the correlation of labour share with output and thestandard deviation of labour share, respectively. The variables are detrended and logged.Labour share is annual and covers the period after 1980 for most countries. r denotes realannualized interest rates in those countries. See the Appendix A for data sources.Chapter 4 reveals two more distinguishable features of labour marketfluctuations in EMs. (1) The contribution of hours per worker to the move-ments of total labour input ? total hours ? are higher in EMs than in DMs.(2) Hours per worker is procyclical with output and positively correlatedwith employment in EMs, while they are less cyclical with output and notcorrelated with employment in DMs. This chapter attempts to answer thesedifferences through the behaviour of interest rate in EMs in an environmentwith search and financial frictions.3Chapter 2The Empirics of LabourShare Fluctuations inEmerging Market Economies2.1 IntroductionOne of the stylized facts of growth in Kaldor (1961) is that factor incomeshares are stable over time.2 This fact has justified many economic modelsusing constant income shares. Unlike its stability over the long run, though,business cycle literature has shown that labour (income) share has short-term dynamics moving slightly against output in major economies. However,the literature has focused on developed markets (DMs), mostly on the US,and is silent on labour share fluctuations in emerging markets (EMs).In this chapter, we document the volatility and the cyclicality of labourshare in emerging markets and compare them to ones in developed markets.The contribution of this chapter is threefold. First, it shows that the cyclicalproperties of labour share in EMs, on average, significantly differ from thosein DMs. Second, the behaviour of interest rate and financing conditionsover the cycle account more for these differences than the lower incomelevel in these economies. Therefore, we will claim that there is a closerelationship between labour share and the interest rate that these economiesface in international markets. This chapter also provides more evidencesupporting this claim from both the emerging and developed world, along2Recetly, there has been a declining trend in labour share in many advanced economies,particularly the US.42.1. Introductionwith robustness checks for different definitions of labour share. we willthen, in Chapter 3, build a model where wages have to be financed throughworking capital loans and show that the variation in the cost of borrowingcan account for the movements of the labour share over the cycle. Thepremise of this dissertation is that financing matters to labour share and thatemerging markets serve as a good natural experiment due to the financialproblems and the different features of the interest rate that they face.Figure 2.1 shows the annual movements of labour share (deviations fromtrend) over the cycle in selected emerging market economies such as Korea,Mexico, and Turkey. For comparison, we include the cyclical properties oflabour share in the US, as well. There are significant variations for thecyclical component of labour share over the cycle in these selected EMs.More importantly, these variations are systematic; they move positively withoutput. Labour share rises during boom times, and it falls during outputcontractions. In contrast, the variations are much smaller in a developedmarket, such as the US, and they tend to move in the opposite direction tooutput.3We document labour share fluctuations for other developing countries aswell. In Section 2.2, we first state our definition of labour share and calculateit across countries and over time using data on compensation of employeesfrom official accounts compatible with the 1993 System of National Accounts(SNA). These observations are plotted in Figure 1.1, which illustrates thatlabour share tends, on average, to be procyclical with output in emergingmarkets. On the contrary, it is slightly countercyclical in developed markets.In addition, labour share is much more volatile in low income countries.Despite these average cyclical properties of labor share in EMs, there isa large variation across countries.4 India, for instance, having the lowestincome per capita in our sample, does not show a procyclical labour share.Yet Korea has a strongly procyclical one although it is one of the richest3The correlation coefficient for the US is -0.36, whereas these coefficients are 0.52, 0.55,and 0.25 for Korea, Mexico, and Turkey.4Gollin (2002) shows that, after adjustments, the level of labour share across countriesdoes not depend on income level. Here, we show that the volatility of labour share doesnot necessarily depend on it either.52.1. IntroductionFigure 2.1: Labour Share Fluctuations in Mexico, Korea, Turkey and the USMexico?.1?.050.05.1deviations1970 1980 1990 2000 2010Korea?.1?.050.05.1deviations1970 1980 1990 2000 2010Turkey?.15?.1?.050.05.1deviations1970 1980 1990 2000 2010US?.1?.050.05.1deviaitons1970 1980 1990 2000 2010Labor Share OutputNote: The variables are detrended using the HP-filter. The y-axis shows percentage deviationsfrom the trend.62.1. Introductionemerging economies.Figure 1.2, on the other hand, gives us a more clear picture. It showsthat labour share is procyclical with output, especially in the countries withcountercyclical interest rates, i.e., a decrease in the cost of borrowing duringbooms tends to increase labour share. The more countercyclical interestrates are, the more procyclical labour share is. Additionally, the countriesthat face a more volatile interest rate tend to have a more volatile labourshare, as well. We also show below that the negative slope coefficients in Fig-ure 1.1 disappear when we take fluctuations in the interest rate into account.These results are proven robust to different measures of and adjustments onlabour share.In Section 2.2, we provide more evidence using quarterly data. In doingso, we are also interested in the lead-lag relationship; labour share is shownto lag behind output in emerging economies. This is perhaps due to thesluggishness in labour markets addressed in the literature, which we willdescribe below. In this section, we also document the volatility statisticsof the components of labour share, namely wages and employment. Weshow that the high volatility of labour share is mainly driven by highlyvolatile (real) wages. But, importantly, this comes not at the expense of lowvolatility of employment; labour input is still nearly as volatile as output inmany EMs.Despite the limited data availability for recent years, we discuss thecyclical component of labour share in the 2008-2009 global financial crisis.We show that labour share did not significantly change in the recent financialcrisis although capital flew out of EMs and output collapsed. However,we also show that the real cost of borrowing for these economies did notchange, either. In other words, interest rates were not countercyclical during2008-2009 in these economies despite the capital flight. On the other hand,countries such as, Greece, Spain, and Iceland, that saw increases in the costof borrowing, tend to have a declining labour share. Therefore, our resultsfrom emerging markets above have the potential to shed some lights on thedebt-troubled European economies.We perform robustness analyses on labour share measures in Section IntroductionFirstly, we adjust labour share by including the labour portion of the self-employment income, which does not change the main empirical findings.Secondly, we discuss if the lack of labour income data from the informalsector in EMs could be a reason for the procyclical share. We show that largeinformal employment does not necessarily associate with more procyclicallabour share. Moreover, when labour income in the informal sector is takeninto account using official data on labour income in Mexico, the variationsbecome even more apparent and more procyclical. Thirdly, we show thatsectoral shifts cannot be a reason for the movements in labour share sincecontributions of these sectors to the value added are not significantly cyclicalwith output. Lastly, the results on labour share are shown to hold even whenwe exclude the public sector. Thus, a procyclical government sector cannotexplain the procyclicality of labour share in EMs.Movements in labour share in developed economies have already beenaddressed in the literature. The countercyclicality of labour share is well-know in the US (see Kydland and Prescott (1982), Gomme and Green-wood (1995), Boldrin and Horvath (1995), and Rotemberg and Woodford(1999), among others). In addition, Gomme and Greenwood (1995) andGiammarioli et al. (2002) find similar patterns for labour share in otherOECD countries. There are many labour-market related explanations forthis behaviour of labour share.5 Gomme and Greenwood (1995), Boldrin andHorvath (1995), and Danthine et al. (2005) include an insurance componentin the wage that makes total labour income less sensitive to output and,therefore, generates countercyclical labour share. Adjustment costs, such ashiring and firing costs and/or vacancy costs as in search models, have thepotential to explain the sluggish labour market and countercyclical labourshare. Andolfatto (1996), Merz (1995), and R??os-Rull and Choi (2009) cangenerate this behaviour of labour share in the US using search theory. Ina modified version of the firing-cost model of Bentolila and Bertola (1990),Vermeulen (2007) shows that adjustment costs, such as firing costs, can ex-plain an important part of the variation in labour share in France. In anempirical work, Bentolila and Saint-Paul (1998) find that adjustment costs,5These mechanisms will be further explained in Chapter 3.82.1. Introductionwhich result in a gap between marginal product of labour and real wage,have a significant explanatory power for the movements of labour share,along with shifts in the capital-output ratio as a result of the price of oil orcapital-augmenting technological progress. Another strand of literature usescapacity constraints (see Hansen and Prescott (2005)) or a framework withmonopolistic competition and mark-ups (see Hornstein (1993) and Amblerand Cardia (1998)).To our knowledge, Diwan (2002) is the only work that discusses labourshare movements in developing countries, albeit not over the business cycle.He concentrates on financial crises in developing economies and shows thatlabour share tends to decline during these crises. He then discusses poten-tial reasons behind this decline, and emphasizes the relative immobility oflabour. Maarek and Orgiazzi (2010) take a similar stance and show thatlow bargaining power stemming from capital flights can explain the declineof the labour share following financial crises. However the bargaining powerargument cannot explain recent increase in the US labour share and/or inemerging markets after the 2008 financial crisis. The difference in this paperis that first, we not only consider the crisis period but also extend the sampleto expansions and mild recessions. Second, our analysis that shows a strongrelationship between interest rates and labour share in both emerging anddeveloped markets is also novel. Third, the model we present in Chapter3 is consistent with observations not only during the past crises but alsoduring the 2008-2009 global financial crisis.In Section 2.4, we discuss the financial environment in emerging marketsand summarize the literature on EM business cycles related to financiallinkages. Emerging markets, indeed, have large capital inflows during boomtimes and difficulties (limits) in financing during economic downturns. Theyalso face high interest rates during recessions due to default risk. Therefore,during recessions, borrowing is costly not only in that it is limited but alsothat it requires high pay-back. We will show in Chapter 3 that when labouris financed through borrowing, these hikes in the effective cost of borrowingdrive the labour share down.92.2. Stylized Facts on the Fluctuations of Labour Share2.2 Stylized Facts on the Fluctuations of LabourShare2.2.1 The Measure of Labour ShareLabour share is computed using the total compensation of employees fromGDP income accounts. In the income approach, gross value-added GDP isthe sum of labour compensation, capital income (corporate profits, interestincome, rental income and depreciation), mixed income of the self-employed(unincorporated income) and indirect taxes excluding subsidies. Most coun-tries officially announce the total compensation of employees and indirecttaxes net of subsidies. Using this information, we measure labour share asfollows:Labour Share =Labour CompensationGDP-net indirect taxes(2.1)Since we are interested in the incomes earned by the factors of production,we exclude government income from gross value-added. In doing so, weassume that net indirect taxes go to both capital and labour income.The above-mentioned formula is the widely-used measure for labourshare in macroeconomics. However, data on labour compensation of em-ployees suffer from measurement problems, particularly because they do notinclude labour income of the self-employed or of workers in the informal sec-tor. Gollin (2002), for instance, shows that the absence of self-employment ispartly responsible for significantly lower shares of labour income in develop-ing countries due to the high fraction of self-employment in total employmentin these countries.6 Ignoring self-employment in labour cost will only mis-lead in terms of the cyclical component of labour share if self-employmentis cyclical with output over the cycle. Below, we show that self-employmentdoes not have any systematic comovement with output even though totalemployment is highly procyclical. In addition, Section 2.3 corrects labourshare using approaches following Gollin (2002) and shows that the resultsremain unchanged. In that section, we also take into account informal sec-tor, and conclude that these measurement errors do not play a role in our6The ratio of self-employment goes up to 30% in many low-income countries.102.2. Stylized Facts on the Fluctuations of Labour Shareresults.2.2.2 DataWe choose countries that report income accounts compiled with the 1993System of National Accounts. Income accounts data have annual frequencyand come from the OECD and the UN.7 We include countries that have atleast 10 annual observations to make sure that each country has recessionsand expansions over the sample period. This leaves us with 18 emergingmarket economies. These economies cover most of the countries definedas emerging markets by institutions providing investment analysis.8 9 Inaddition to emerging markets, 18 developed markets are included in thesample for comparison. These countries are listed in Appendix A. Data formost of the emerging markets start in the 1980s. We take labour share datafor developed countries after 1980 as well. For self-employment adjustment,we use ratios of self-employment from either the OECD or ILO statistics.The details on data sources can also be found in Appendix A.As for real interest rates, Uribe and Yue (2006) have a dataset on quar-terly interest rates (annualized) for emerging markets. They construct in-terest rates for each country using their corresponding JP-Morgan EMBI+spread over US T-bills. Since these bonds are denominated by the US dol-lar, real yields are calculated using a proxy for the expected inflation in theUS, which is equal to the average of the current and three preceding periodsof annual inflation in the US, based on the GDP deflator.10 One drawbackof using these interest rates is the limited coverage over our sample period.For most countries, EMBI data start at 1994Q1 or 1999Q3, which gives us7The OECD has longer labour compensation data for some developed countries, Mex-ico, and Turkey. We check that data in the OECD is consistent with data reported to theUN such that results are relatively fixed. Therefore, we choose the longer dataset fromthe OECD to do individual country analysis over time for these countries as well.8Korea has been recently classified as a developed market in these institutions. How-ever, during our sample period, 1980-2008, the country was mostly in the category ofemerging markets, in that the GDP per capita in Korea was below $20,000 until 2004.9We also include Costa Rica, given its relatively high per capita income and longtime-series data, although it is not listed as an emerging market in FTSE or MSCI lists.10Using future inflation as expected inflation does not change the results very muchsince inflation is more or less stable in the US over this period.112.2. Stylized Facts on the Fluctuations of Labour Sharea small number of observations at an annual level. Another drawback isthat these rates are the cost of borrowing in US dollars. Firms with localcurrency-denominated assets in their balance sheet would face an extra costbecause of exchange rate movements. In addition, a varying intermediarycost (to access the international credit market) over the cycle might makefinancing more difficult, too. For these reasons, we mainly use domesticinterest rates, as the dataset is longer and more representative of borrowingcosts that firms face. Domestic rates come from IFS and represent the costof financing short-term needs of the private sector.11 The GDP deflator isused to obtain the real interest rate for each economy. In Appendix A, wealso re-plot the figures above with available EMBI rates from the Uribe andYue (2006) dataset, in order to check the sensitivity of the results to differentmeasures of interest rates in emerging markets. For developed economies,interest rates come from OECD financial indicators. These are domesticshort-term (treasury bill) interest rates on bonds that are denominated inlocal currency. As in the case of emerging markets, these rates are assumedto be representative cost of borrowing that an economic agent faces in theseeconomies. Details by country can be found in Appendix A.2.2.3 ObservationsWe document the statistics of annual labour share fluctuations in Tables 2.1and 2.2 for emerging and developed markets, respectively. HP filtering isused to detrend the data together with a smoothing parameter of 6.25. Wealso add p-values for correlation coefficients in parentheses.Volatile Labour Share. These statistics show that labour share, on av-erage, is almost twice more volatile in emerging markets than in developedmarkets. Although the standard deviation over time is higher in emergingmarkets, there is also large variation in terms of this statistic within emerg-ing economies. The median of standard deviation of the labour share is 2.0511Either lending or treasury bill rates are used, depending on availability. If neither isavailable, money market or deposit rates are used. We confirm that all types of interestrate series move together within a country.122.2. Stylized Facts on the Fluctuations of Labour ShareTable 2.1: The Cyclical Properties of Labour Share in Emerging Markets?(s) ?(s,y) ?(r) ?(r,y) ?(r,s)Argentina 5.02 0.56 7.8 -0.57 -0.57(0.03) (0.03) (0.03)Brazil 2.38 0.03 4.08 -0.36 -0.08(0.92) (0.22) (0.39)Chile 2.36 -0.12 4.67 -0.02 0.19(0.54) (0.89) (0.32)Colombia 1.22 -0.06 5.21 0.42 -0.25(0.88) (0.02) (0.36)Costa Rica 2.92 -0.22 4.53 -0.03 -0.05(0.26) (0.88) (0.82)Czech Rep. 1.37 0.27 2.30 -0.16 0.03(0.30) (0.55) (0.92)Egypt 1.29 0.18 1.93 -0.40 -0.60(0.58) (0.27) (0.02)Hungary 1.14 0.12 2.17 0.06 -0.11(0.67) (0.83) (0.71)India 1.17 -0.44 2.89 0.22 -0.52(0.04) (0.33) (0.01)Israel 1.71 0.17 1.94 0.10 -0.31(0.57) (0.74) (0.31)Korea 1.36 0.45 1.83 -0.57 -0.20(0.02) (0.01) (0.31)Mexico 3.72 0.60 9.10 -0.56 -0.66(0.0) (0.0) (0.0)Peru 2.03 0.36 7.01 -0.16 0.02(0.07) (0.52) (0.92)Philippines 2.85 0.02 2.33 -0.03 -0.09(0.95) (0.92) (0.74)Poland 1.97 -0.19 3.45 0.44 -0.35(0.46) (0.08) (0.17)Russia 4.8 -0.12 3.22 0.20 -0.70(0.68) (0.49) (0.02)South Africa 1.47 -0.06 2.18 0.16 0.11(0.76) (0.41) (0.59)Turkey 5.45 0.25 9.66 -0.49 -0.24(0.30) (0.02) (0.31)Mean 2.45 0.11 4.27 -0.10 -0.24Mean* 2.78 0.34 5.49 -0.41 -0.29Note: P-values are in parenthesis. The symbols ? and ? denote standard deviation and cor-relation. Mean* represents the average for countries facing countercyclical interest rates:Argentina, Brazil, Czech Republic, Egypt, Korea, Mexico, Peru and Turkey. Interest ratesare net annual domestic rates from the IFS. See Appendix A for data sources.132.2. Stylized Facts on the Fluctuations of Labour ShareTable 2.2: The Cyclical Properties of Labour Share in Developed Markets?(s) ?(s,y) ?(r) ?(r,y) ?(r,s)Australia 1.05 -0.41 1.45 -0.03 0.62(0.02) (0.50) (0.0)Austria 0.65 -0.29 0.65 0.40 -0.06(0.08) (0.08) (0.80)Canada 1.06 -0.58 1.30 0.37 -0.22(0.0) (0.06) (0.26)Denmark 1.18 -0.38 1.16 -0.03 -0.05(0.05) (0.0) (0.81)Finland 1.74 -0.29 1.79 0.15 0.32(0.13) (0.49) (0.14)France 0.50 -0.32 0.70 0.29 0.13(0.09) (0.13) (0.50)Germany 0.79 -0.04 0.63 0.45 0.09(0.88) (0.01) (0.64)Greece 1.96 0.14 1.31 -0.12 -0.23(0.49) (0.54) (0.64)Iceland 2.8 0.39 2.01 -0.22 -0.03(0.03) (0.0) (0.89)Ireland 1.4 -0.54 1.73 0.05 -0.06(0.01) (0.87) (0.75)Italy 0.8 -0.31 1.11 -0.07 0.23(0.10) (0.72) (0.25)Netherlands 1.07 -0.35 0.76 0.11 -0.01(0.10) (0.63) (0.97)New Zealand 1.86 -0.09 1.62 0.01 0.53(0.65) (0.95) (0.01)Norway 3.17 0.04 3.4 -0.05 -0.44(0.82) (0.81) (0.02)Spain 1.06 0.30 1.59 0.09 0.34(0.13) (0.63) (0.08)Sweden 1.48 0.07 1.49 -0.12 0.21(0.70) (0.56) (0.30)UK 0.99 -0.31 1.19 -0.02 0.25(0.10) (0.83) (0.18)US 0.69 -0.36 1.00 0.50 -0.42(0.05) (0.01) (0.03)Mean 1.34 -0.19 1.38 0.10 0.07Note: P-values are shown in parenthesis. The symbols ? and ? denote standard deviationand correlation, respectively. Interest rates are the annual average of short-term interestrates on local denominated treasury bills over the period taken for labour share. SeeAppendix A for details on data.142.2. Stylized Facts on the Fluctuations of Labour Shareversus the mean of 2.48.One notable fact here is that the mean is significantly higher in countriesfacing high interest rate volatility, such as Argentina, Brazil, Mexico, Peru,Russia, and Turkey. Statistically, countries with higher volatility in their in-terest rates tend to have higher volatility in their labour share, as well. Wedivide the emerging market sample into two groups depending on whetherthey have lower or higher volatility in interest rates than the median12. Themeans of standard deviations of labour share for these two groups are: 1.62versus 3.28, the higher being for the second group. The sample mean testrejects the equality of labour share volatilities across these two groups at a2% significant level.Procyclical Labour Share. In addition to higher volatility, comovementof labour share with output differs in emerging markets compared to devel-oped ones. The correlation between the cyclical component of labour shareand output tends to be procyclical (0.10) in emerging markets, whereas it isnegative (-0.19) in developed markets. Since there is variation among coun-tries in each group, we apply a sample mean test where the null hypothesisis that there is no difference in these correlations across different countrygroups. T-statistic from the mean test is 3.08, which falls in the region ofrejection for the significance level of 1%. This indicates that labour sharefluctuations are, indeed, statistically different in emerging economies thanin developed ones.Furthermore, it can be seen in Table 2.1 that average procyclical labourshare is mostly driven by countries with countercyclical interest rates. Figure1.2 also suggests that procyclicality of labour share becomes more apparentas the country faces more countercyclical interest rates. In a statisticalanalysis, we take Argentina, Brazil, Czech Republic, Egypt, Korea, Mexico,Peru and Turkey as countries having a countercyclical cost of borrowing.13Then, the average correlation of labour share with output goes up to 0.34,12The median for the standard deviation of interest rate is 3.00513When EMBI rates are considered, interest rates still show countercyclicality withoutput in these countries.152.2. Stylized Facts on the Fluctuations of Labour Sharewhich implies a stronger procyclical labour share in those countries thanthe average in emerging markets. In Appendix A, we redo this exercise withinterest rates constructed by EMBI spreads and show that the results arenot sensitive to measures of interest rates.Other countries in the emerging-market group do not have significantlydifferent movements in labour share than in developed markets. T-stat andp-value from a sample-mean test between these countries and developedeconomies are equal to 1.36 and 0.19, respectively. This indicates that wheninterest rate variations are taken into account in those emerging economies,movements in labour share are not statistically different from those in de-veloped economies.Among developed markets, Table 2.2 shows that many countries exhibitprocyclical real interest rates. Possible explanations for procyclical ratesare accommodative monetary policy during recessions, i.e., countercyclicalpolicies (see King and Watson (1996)) and cyclical marginal productivityof capital following a productivity shock (see Kydland and Prescott (1982).On the other hand, developed markets such as Greece, Iceland and Swe-den are likely to face countercyclical rates.14 These are also countries thattend to have a labour share moving more positively with output than therest of the developed group. Therefore, similar pattern appears in advancedeconomies, albeit not frequently.15Quarterly Evidence. We also document quarterly fluctuations of labourshare on which we have data for only Brazil, Korea, Mexico and Turkey.For comparison, we add the results from the US as well. Table 2.3 presentsthese results and the starting date for the sample countries. We confirmthat the results are similar to those from annual data. One interesting pointto mention is that the labour share seems to lag behind output in all fourcountries. This is perhaps because of a sluggish response of labour markets14For these countries, the correlation between real interest rates and output is below-0.10.15We also discuss below the labour share movements during the recent debt crisis insome of the advanced economies such as Greece, Portugal and Spain.162.2. Stylized Facts on the Fluctuations of Labour Shareto shocks in the economy through search and matching frictions, hiring andfiring costs, and contractual agreements between firm and worker. In fact,the literature on labour share in developed economies models these frictionsto explain the slightly negative correlation between labour share and outputin these economies (See Boldrin and Horvath (1995), Bentolila and Saint-Paul (1998), and R??os-Rull and Choi (2009) among others).The above-mentioned labour frictions might be important in develop-ing countries as well. For instance, the dataset from Botero et al. (2004)suggests that emerging economies? labour markets are not flexible: they laybetween European and Anglo-Saxon countries in terms of employee protec-tion laws.16 However, a positive correlation between labour share and outputimplies that total labour income responds to economic shocks more than theoutput does. Therefore, those frictions will potentially have a lesser powerto explain labour share movements in economies with procyclical labourshare. This is why, in this dissertation, we try to explore other effects onlabour share movements such as labour financing through working capitalfrictions and the ability to borrow for short-term needs, the details of whichare discussed in the next chapter.The Correlation between Labour Share and Interest Rate. We haveso far discussed the link between labour share and interest rate movementswith output, but not the direct correlation between labour share and interestrates. If there were a strong relationship, then we should also observe a non-zero correlation between these variables over the cycle.Figure 2.2 shows the scatter plot of correlations between labour shareand (real) interest rates for all the countries in the sample. We note that,on average, the correlation between labour share and interest rates is belowzero. More importantly, in regions where interest rates are cyclical with out-put17 ? whether procyclical or countercyclical ? labour share has a strongernegative correlation with interest rates. On the other hand, in countries16We will discuss this in more detail in Chapter 4.17We assume here that the (absolute) correlation between interest rates and output hasto be at least 0.15 for interest rates to have a sufficient comovement with output.172.2. Stylized Facts on the Fluctuations of Labour ShareTable 2.3: Evidence from Quarterly Data?(s) ?(s, y) ?(s+1, y) ?(s+2, y) ?(s+3, y) ?(r) ?(r, y)Brazil.91Q1 4.3 0.09 0.14 0.13 0.16 2.9 -0.33(0.47) (0.24) (0.28) (0.18) (0.07)Korea.89Q1 2.69 0.11 0.27 0.46 0.58 1.72 -0.72(0.39) (0.02) (0.01) (0.01) (0.0)Mexico.87Q1 3.65 0.42 0.56 0.58 0.52 2.1 -0.54(0.01) (0.0) (0.0) (0.0) (0.01)Turkey.87Q1 7.8 0.12 0.23 0.27 0.33 1.72 -0.45(0.39) (0.04) (0.02) (0.01) (0.0)Mean 4.61 0.19 0.30 0.36 0.40 2.33 -0.55US.80Q1 0.72 -0.34 -0.17 -0.11 -0.05 1.25 0.41(0.0) (0.07) (0.24) (0.64) (0.0)The table shows the cyclical properties of labour share in quarterly frequency with currentand lagged output. The correlation between interest rate and output is also presented forthose countries in the sample. See Table 2.1 for the symbol denotation. The startingquarter of the sample for each country follows the country name. Sources: IPEADATA(national source) for Brazil, OECD for Korea and Mexico, and TURKSTAT (nationalsource) for Turkey. See Appendix A for the sources of interest rates.182.2. Stylized Facts on the Fluctuations of Labour ShareFigure 2.2: Correlation between Labour Share and Interest RateARGBRACHLCRIKORMEXPERPOLPHLZAFTUR COLCZHEGYHUNINDISRRUSAUSAUTCANNLDNZLESPSWE UKUSDNKFINFRA GERGRCISL IRLITANOR?1?.50.5corr(r,s)?.5 0 .5corr(r,y)Note: corr(r,s) and corr(r,y) denote the correlation between labour share and (real) in-terest rates and between output and interest rates, respectively. All variables are in logs(except interest rates) and are detrended using the HP filter. The yellow line representsthe benchmark correlation between interest rates and output, 0.15. See Appendix A forsources of these variables.192.2. Stylized Facts on the Fluctuations of Labour Sharewhere interest rates are not cyclical with output, the cyclical properties oflabour share vary dramatically. A model with wage bill financing, presentedlater in Chapter 3, is consistent with this fact observed in the data. Partic-ularly, this model will be powerful to explain the movements of labour sharein countries with considerable changes in interest rates over the businesscycle, such as risky emerging market economies. In contrast, this effect willbe small in countries with relatively stable and acyclical interest rates, as inmany advanced economies.Labour Share Components ? Wages and Employment: Total labourincome (or total labour compensation) in a country is, in the end, the mul-tiplication of wages (compensation per employee) and total employment. Anatural question to ask here is: how much this high volatility comes fromeach of those two components? Is it wages or the labour input that makestotal labour income more responsive than output?Table 2.4 and Table 2.5 document the cyclical properties of wages andemployment, respectively.18 First of all, real wages are more volatile thanoutput for all developing countries, with the exception of Chile. In fact,on average, real wage is almost twice as volatile as output over the cyclein emerging economies. This result is quite different in developed markets,where wages are less volatile than output (see Krugman (1999), for exam-ple). Secondly, Table 2.5 suggests that employment is less volatile thanoutput. Lastly, both wages and employment tend to be procyclical withoutput in emerging economies. These results show that the cyclical proper-ties of labour share we discuss above is driven more by the movements inwages than those in employment.Recently, Li (2011) and Boz et al. (2009) have also documented wageand employment regularities in some emerging markets and found similarempirical results. Our contribution here is to show that total labour income18Total hours would be more relevant for the labour input. However, hours data areavailable only for a smaller set of countries with a shorter time horizon. Therefore, wechoose to present employment movements here. With a shorter horizon, we will alsoanalyze hours per worker and employment dynamics in Chapter 4.202.2. Stylized Facts on the Fluctuations of Labour Sharein these economies tends to be more volatile than output, generating a pro-cyclical labour share over the cycle. Thus, high volatility of real wages inthese economies is hardly at the expense of low employment volatility. Weshow in Table 2.5 that some of the low volatility in employment is due to thehigh self-employment ratio that is weakly cyclical with output (see Section2.3 for details on self-employment and its cyclical properties). In fact, whenwe only take into account the number of employees (wage earners), bothvolatility and comovement of labour input with output increase. Therefore,employment is responsive in these economies over the cycle, albeit not asvolatile as real wages. To sum up, overall labour market is more volatilethan output to shocks in the economy, and this extra volatility in the labourincome is mostly attributable to the price component (real wages) ratherthan the quantity component (labour input).Individual-Country Analyses. Mexico and Korea have long horizon datawhich allow us to examine changes in the cyclical properties of labour shareover different decades. Figure 2.1 shows that in Mexico, where data onlabour share go back to the 1970s, the volatility and procyclicality of labourshare appear especially in the 1980s and 1990s. This period is accompaniedby a highly unstable financial environment and highly volatile capital flowsin and out of the country. The comovement of labour share and outputdisappears when the economy stabilizes after 2001; so does the relationshipbetween interest rates and output. In addition, we do not observe the pro-cyclical labour share in the 1970s, when financial liberalization had not yetoccurred.The results are similar in Korea, except that we do observe procycli-cal labour share in the 1970s as well as in other decades, but we verifythat Korean (real) interest rates were already countercyclical with outputin the 1970s, a phenomenon that started earlier than in Mexico. Anotherremarkable point in the figure is that labour share peaked just before thefinancial crises in all countries ? the financial crises are Mexico (1995), Korea(1998), Turkey (1994 and 2001) ? and saw a large drop during those crises.These anecdotal pieces from individual-country analyses also support the re-212.2. Stylized Facts on the Fluctuations of Labour ShareTable 2.4: Labour Share Components ? Cyclical WagesObs ?(y) ?(w)?(y) ?(w, y)?(wman)?(y) ?(wman, y)Argentina 27 3.87 1.58 0.55Brazil 18 1.87 2.87 0.54Chile 28 2.35 0.89 0.22Costa Rica 27 2.05 2.44 0.63Korea 28 2.06 1.2 0.7Mexico 28 2.1 1.99 0.49Peru 28 4.14 2.39 0.42Philippines 28 2.50 1.16 0.44Turkey 28 2.55 2.61 0.39Average 2.61 1.82 0.50 2.06 0.48Note: Real wage applies to the overall economy. For countries such as Argentina, Mexico,and Turkey, we have real wage data for only manufacturing, represented as Wman here.Wages are total gross earnings per worker except for Argentina, Chile, and Turkey inwhich they are hourly earnings. The source for wages is ECLAC (Economic Commissionfor Latin America and the Caribbean) for Brazil, Chile, Costa Rica, and Mexico; ILO forthe Philippines; and INEC and TURKSTAT (National sources) for Argentina and Turkey,respectively. The first column shows the number of annual observations we have over time,untill 2008.222.2. Stylized Facts on the Fluctuations of Labour ShareTable 2.5: Labour Share Components ? Cyclical EmploymentObs ?(y) ?(e)?(y) ?(e, y) Obs?(ee)?(y) ?(e, y)Argentina 27 3.87 0.50 0.71 12 0.75 0.78Brazil 27 1.87 1.02 0.49 27 1.18 0.53Chile 26 2.35 0.98 0.42Costa Rica 24 2.05 1.10 -0.02 22 1.13 0.24Korea 28 2.06 0.64 0.78 28 0.85 0.82Mexico 28 2.10 0.41 0.62 16 0.50 0.61Peru 28 4.14 0.38 0.44Philippines 28 2.50 0.51 0.14Turkey 28 2.55 0.65 -0.05 21 0.59 0.55Average 2.61 0.69 0.38 0.84 0.59Note: Employment, e, is defined as total annual civilian employment. Employees, ee, arethe salaried workers (wage earners). The first and fourth column show the number ofannual observations we have over time untill 2008. The source is ILO for Chile, CostaRica, Peru and Philippines. OECD is the source for Korea, Mexico, and Turkey. ForBrazil, we obtain the data from IPEA (national source). For Argentina, employment datacomes from the World Bank Development Indicators and the time series of employeescomes from ILO.232.2. Stylized Facts on the Fluctuations of Labour Sharelationship between labour share and the cost of borrowing across countriespresented above. For comparison purposes, we add the cyclical componentof the US labour share to the figure as well. It is clear that labour share ismuch more stable in the US and does not reflect such a pattern we observein Mexico, Korea, and Turkey.Emerging Markets and the Global Financial Crisis. We also wantto discuss labour share movements in emerging economies during the recentglobal financial crisis. The 2008-2009 great recession not only hit advancedeconomies but also caused abrupt slowdown in emerging markets, especiallyin 2009. GDP per capita (in constant US dollars) dropped by 1.78% in2009, on average, in our sample emerging markets. This reflects a 5.26%lower growth rate than that reached over 2000-2008. This slowdown wasalso homogeneous in the sense that all emerging countries except India facedlower rates than over the benchmark period.How did labour share move during this recession as opposed to output?Figure 2.3 presents changes in labour share for the countries with avail-able data from the UN statistics, i.e, for Brazil, Columbia, Czech Republic,Korea, Mexico, Peru. As previously discussed, these particular economiesexhibited procyclical labour share until 2008. By contrast, Figure 2.3 showsno indication of a drop in labour share during their recent GDP slowdownin 2009. In fact, in all countries, labour shares were at a higher level in2009 than their pre-crisis levels, such as in 2007. These results are surpris-ing given the remarkable slowdown in GDP and the results discussed over1980-2008. However, we note above that labour share procyclicality appearsin countries or during times with countercyclical interest rates. Figure 2.3shows that, contrary to their previous crisis, the recent crisis did not lead toa significantly higher cost of borrowing in emerging markets. Our analysisshows that the spread these economies faced increased during this period,but lower base rates (US treasury bills) offset this effect. In addition, mon-etary authorities in EMs were able to undertake expansionary policies in2009 whereas they were constrained by the large capital flight in their own242.2. Stylized Facts on the Fluctuations of Labour ShareFigure 2.3: Labour Share in EMs during the Global Financial CrisisNote: Labour shares are indexed so that 2007 value is equal to one. Interest rates are theannual aggregation of quarterly real cost of borrowing for the EM countries constructedusing quarterly JP-Morgan EMBI Global spreads plus US T-bill 3-month rate constantmaturity. The rate is deflated using the average inflation rate over the current and threepreceding quarters as the expected inflation. Source: UN data and JP-Morgan EMBIdatabase from the Haver Analytics.past financial crises.19Other evidence showing the contrast during the recent and earlier domes-tic crises comes from the Mexican labour compensation index in manufactur-ing. The index shows an 8.5% drop in total compensation in manufacturingfor 2009. In the same year, total GDP and manufacturing GDP dropped by7% in Mexico. This implies no significant change in labour share, especiallywhen it is considered that the manufacturing sector is highly responsive tochanges in GDP. On the other hand, the same compensation index droppedby approximately 20% in the crisis year of 1995 when GDP dropped by19See IMF (2009) for the policy rates during the great recession in EM economies.252.2. Stylized Facts on the Fluctuations of Labour Share6.5%, and the country was experiencing difficulties financing its expenses,either because of the high cost of borrowing or the quantity restrictions onloans. The Mexican cost of borrowing, constructed using Mexican spreadsin the Uribe and Yue (2006) data-set, went up to around 20% from 10%levels, whereas Financial Times Emerging Market reports reveal that bidyields on US-dollar-denominated Mexican bonds were around 5-6% duringthe 2008-2009 recession, levels similar to ones in January 2008. Therefore,these observations from both across and individual country analyses moti-vate us to explore the effects of financing labour cost on the share of incomegoing to labour.Developed Economies and the Euro Debt Crisis. For developedeconomies as well, there is variation in terms of the dynamics of labourshare. (see Table 2.2). As mentioned, the developed countries that tend tohave countercyclical interest rates, such as Greece, Iceland and Sweden, alsotend to have procyclical labour share. These countries are famous for theirfinancial crises (Sweden in 1992 and Iceland in 2008) and had increasesin their risk premia in the past. In addition, since 2010 some Europeaneconomies have faced debt crises and consequently increases in their cost ofborrowing. When this dissertation was started, the debt crisis in Europeaneconomies had not yet begun. However, we have recently obtained datafor some of the European countries to track the movements of labour share.Figures 2.4 and 2.5 present movements in the labour share and interest ratesfor those economies.Figure 2.4 shows labour share had a tendency to drop during the timeswhen developed economies started to have debt problems. First of all, thereis a sharp drop in labour share for Iceland during 2008-2009. In addition,Figure 2.5 reveals that Icelandic corporations faced large (real) cost of bor-rowing during these years. Secondly, labour share fell during 2010 and 2011in countries such as Greece, Hungary, Portugal and Spain. Note that thesecountries had increases in their risk premia during the years following theGreek debt problems (see, for example, Figure 1.2 in IMF (2012) for thewidening spreads in 2010 and 2011). Figure 2.5 also suggests higher costs of262.2. Stylized Facts on the Fluctuations of Labour ShareFigure 2.4: Advanced Economies? Labour Share during the Debt CrisisNote: The figure presents the movements of labour share in European countries whichfaced increases in risk premium since 2009. Source: OECD.272.2. Stylized Facts on the Fluctuations of Labour ShareFigure 2.5: Interest Rates in Selected Advanced Economies during the DebtCrisisNote: Interest rates are the monthly yields on government bonds (2-years or more) for allcountries except Iceland, for which they are the lending rate from financial institutions forIceland. Rates are deflated using inflation rates in the current period and then aggregatedto annual levels. Source: OECD and IFS.282.3. Robustness Analysesborrowing for those economies during 2010-2011 relative to 2007-2008. Wenote an increase in real interest rates in 2009 as well; however, this increasewas mainly driven by deflation in these economies in 2009 since we assumecurrent inflation as the expected inflation in calculations of real interestrates. The inflation rate in all those years except 2009 was in the range of2%-5% but it dipped below zero in 2009. Since this phenomenon is rareover the cycle, the assumption for the expected inflation might not hold;therefore, the real cost for 2009 might be distorted. Thus, a comparisonbetween 2007-2008 and 2010-2011 is preferred here, as inflation since 2010has returned to patterns observed earlier (2%-5%) since 2010.In short, even countries with more developed financial markets, as op-posed to lower income countries, had drops in labour share during periodswith a high cost of borrowing ? a phenomenon not observed during theirprevious recessions. Therefore, we claim that there is a strong relationshipbetween the dynamics of labour share and the risk component in the costof borrowing, independent of income level.2.3 Robustness AnalysesAs mentioned above, data on compensation of employees do not includeincome from the self-employed. In addition, labour compensation estimatesmostly originate from establishment surveys in the formal sector whereas thevalue-added output, the denominator in equation (1), takes into accountthe output from the informal sector. If these parts of the economy arecountercyclical with output, the results on the procyclicality of labour sharewould not be reliable. Moreover, sectoral shifts and changes in governmentexpenditures over the cycle might alter labour income shares in the overalleconomy. In this section, we take all these possibilities into considerationand show that the results discussed above are robust to different measuresof labour income shares.292.3. Robustness Analyses2.3.1 Self-employment AdjustmentsDeveloping countries have high ratios of self-employment. In our sample,self-employment as a ratio of total employment varies between 30%-40%in developing countries. This is in contrast to the ratio in developed mar-kets, around 10%-15% of total employment. One reason for the large num-ber of self-employed people in emerging markets is that agriculture is stillwidespread in these economies and farmers are counted as self-employed innational accounts. The fact that data on compensation of employees lackthe labour income of the self-employed would distort the cyclical behaviourof labour income shares as calculated above if there were significant shiftsacross different occupations. For example, if a laid-off manufacturing workermoves into farming during a recession, labour compensation will still fall bythe amount of his gross salary even though he contributes to the total valueadded in the country as a farmer.In order to take into account the labour income of self-employed peoplein the total labour compensation, we apply two adjustment methods here.The first is to take only incorporated businesses when computing labourshare. This requires a deduction in value-added output by an amount equalto the self-employed income (mixed income):Adj-1 : Labour Share =Labour CompensationValue Added GDP - Mixed IncomeThe adjustment shown above assumes that labour share is the same acrossincorporated and unincorporated (self-employed) enterprises and can be ap-plied only to countries that report mixed income in their national accounts.For those that do not report it, we compute a proxy for the labour incomeof a self-employed person and then adjust the overall economy labour shareusing self-employment data for those countries:Adj-2 : Labour Share =Labour Comp. + Labour Income per Self-Emp. ? Self-EmpValue Added GDPLabour compensation per employee is calculated by the total labour com-pensation divided by the number of employees, and this is used as a proxy302.3. Robustness Analysesfor the labour income of a self-employed person, as in Gollin (2002). How-ever, the assumption that the labour cost of the self-employed is equal tothe labour compensation per employee might not be valid for some coun-tries. This method lifts the level of labour share up to a very high fractionof income in countries where the self-employment ratio is very high, such asKorea and Turkey.20 The Korean labour share, for example, rises to 80-90%levels after this correction. We then check household surveys in Korea andTurkey, and verify that the total gross income of a self-employed person isaround 60% of the average gross wage level. One reason for this is because abig part of self-employment comes from rural areas where the pay is lower.In addition, there might be differences in terms of skills across average work-ers and those self-employed. Furthermore, self-employed people often workin the informal sector for which the administrative cost of labour, such aslabour income tax and social security payments, do not show up on records.Since the total income of a self-employed individual also includes his/hercapital income, one should expect for the self-employed a labour compensa-tion lower than 60% of average labour compensation per employee. In thecalculations, we assume that the labour income of a self-employed person ishalf the labour cost of an employee in Korea and Turkey.The results for adjusted labour share are listed in Table 2.6. Thereare only minor changes after self-employment corrections. Adjusted labourshare still shows a high volatility. In terms of the cyclical comovement oflabour share with output, the changes are so miniscule that the adjustmentdoes not alter the sign of correlation between labour share and output. Whenwe plot these adjusted observations with the cyclical properties of interestrates (see Figure A.2) in order to compare with our initial unadjusted Figure1.2, the main result still holds: the more countercyclical interest rates are,the more procyclical labour share is. This shows that the high procyclicalityof labour share in countries such as Argentina, Korea and Mexico is not ameasurement error from a calculation that ignores self-employment.20Korea and Turkey have the highest self-employment ratio in the sample taken here.Half of employed people are working for themselves. The average is around 30% in thedeveloping group, whereas it is only 12% in developed economies.312.3. Robustness AnalysesTable 2.6: Adjustments for Self-employmentLab. Share Adj-1 Adj-2?(s) ?(s,y) ?(s) ?(s,y) ?(s) ?(s,y)Argentina 5.02 0.56 5.60 0.54(0.03) (0.04)Brazil 2.38 0.03 2.88 0.20(0.92) (0.43)Chile 2.36 -0.12 2.18 -0.44(0.54) (0.13)Colombia 1.22 -0.06 1.15 -0.10(0.88) (0.67)Costa Rica 2.92 -0.22 2.59 -0.46(0.26) (0.03)Czech Republic 1.37 0.27 1.42 0.17(0.30) (0.33)Egypt 1.29 0.18 1.35 0.18(0.58) (0.59)Hungary 1.14 0.12 1.51 -0.02(0.67) (0.95)Korea 1.36 0.44 1.45 0.32(0.02) (0.05)Mexico 3.72 0.60 3.34 0.58(0.0) (0.0)Turkey 6.40 0.25 6.22 0.15(0.30) (0.62)Mean 2.65 0.18Mean (Adj.) 2.46 0.10Note: Adj-1 calculates labour share as the ratio of labour compensation of the incorporatesector over value-added excluding the unincorporated sector. Adj-2 assumes that labourincome per self-employed is equal to compensation of the average worker (except in Ko-rea and Turkey) and recalculates labour share as the multiplication of compensation peremployees and total employment.322.3. Robustness AnalysesTable 2.7: Cyclical Variation of Self-Employment and Total Employment inEMsBrazil Korea Mexico Turkey?(se, y) 0.23 -0.28 0.30 -0.29?(l, y) 0.50 0.78 0.75 0.35Note: ?(x, y) is the correlation between two variables. se denotes self-employmentwhereas l denotes total employment. The data covers the period after 1990.The reason self-employment is not a concern for cyclical properties oflabour share is because this part of employment does not show a significantcomovement with output. Table 2.7 shows the correlations of the cyclicalcomponent of self-employment and total employment with output in dif-ferent countries. Although self-employment is less correlated with outputcompared with total employment, this is not significant enough to reversethe results on overall labour share. Although self-employment constitutesaround 30% of total employment, its contribution to the GDP is only around10-15%.2.3.2 Informal Sector ConsiderationsAnother important issue for low-income countries is the high ratios of em-ployment in the informal sector. Comparable estimates from ILO suggestthat developing countries in our sample have informal employment (bothin formal and informal enterprises) ranging between 40% and 60% of totalemployment.21 Labour compensation in the informal sector is not usuallyrepresented in the national accounts by the income approach for developing21Informal employment refers to the self-employed in their own informal sector enter-prises, contributing family workers (unpaid), members of informal producers? cooperatives(not established as legal entities), employees holding informal jobs (i.e., jobs not subjectto national labour legislation, income taxation, social protection or entitlement to certainemployment benefits, such as sick leave); own-account workers engaged in production ofgoods exclusively for own final use by their household.332.3. Robustness Analysescountries.22 On the other hand, the value added GDP has estimates of theproduction in the informal sector. In these countries, though, an importantpart of employment in the informal sector actually comes from unregulatedself-employment (Thomas (1992) and De Soto (1990)). Therefore, adjust-ments that we have done for the self-employed partially correct informalityproblem, too.Another possibility here is to analyze comparable cross-section data oninformality that we have obtained from ILO, in order to see if differentcyclical patterns exist in different levels of informality. The possible theorybehind here is that a large informal sector would absorb more employmentduring recessions, which would push the labour share down. Figure 2.6plots the cyclical patterns of labour share with respect to informal employ-ment observed in developing countries, with comparable data availability.If informality ? strictly speaking, the lack of informality in the labour com-pensation ? were the main driver of the procyclical labour share, we wouldobserve a higher correlation between labour share and output in countrieswith a larger informal sector. However, we do not observe such a pattern inFigure 2.6. In fact, countries with the highest ratios of informality (Colom-bia and India) have a significantly smaller correlation of labour share withoutput than the others. Therefore, it is hard to say that the absence oflabour compensation from the informal sector accounts for the procyclicallabour share.Additional evidence comes from Mexico.23 Official authorities releaseinformation on the contribution of the informal sector to total value added.We plot them in Figure 2.7. Besides its small contribution to GDP in Mexico(on average, 12.4% of the value added), a crucial takeaway from the figure22Having said that, we should also note that informal employment in the formal sectoris included in the official labour compensation since the data is usually derived fromestimates in the formal sector. It is employment in the informal sector that is missing inthe official compensation of employees.23Throughout the dissertation, we sometimes concentrate on Mexico, especially fromdisaggregated data. The main reason for this is Mexico is a country with ample dataavailable. Moreover, the reader will see that models in Chapter 2 and 3 are calibrated tothe Mexican economy. Therefore, we want to make sure that potential concerns about theMexican labour share are addressed and proven not effective.342.3. Robustness AnalysesFigure 2.6: Informality and the Cyclical Properties of Labour ShareARGBRACRIMEXPERCOLEGYIND?.50.5corr(s,y).4 .45 .5 .55informal employment as a percentage of total employmentNote: Corr(s,y) denotes the correlation between the cyclical components of labour shareand output. Data on the informal employment is the cross-section data in 2000, and coversemployment in the informal sector as well as informal employment in the formal sector(informal jobs in formal enterprises). See the text for details of the coverage. Source: ILOfor the informal employment.352.3. Robustness AnalysesFigure 2.7: Informal Sector as a Share of GDP in MexicoNote: The figure presents the contribution of the informal sector to the value added inMexico. Data period: 1993-2003. Source: INEGI362.3. Robustness AnalysesFigure 2.8: Labour Share Adjusted for Informality and Self-Employment inMexicoNote: Adjusted share is obtained first taking care of the informal sector by adjustingofficial labour compensation data using the contribution of the informal sector to thevalue added. The implicit assumption here is that labour content is the same acrossformal and informal sectors. Then, this is further adjusted for self-employment to accountfor the labour income of the self-employed.372.3. Robustness Analysesis that the informal sector actually moves quite cyclically with output inMexico. Therefore, the absence of labour cost from the informal sector inthe official labour compensation data tends to work in the opposite directionfor the observed positive correlation between labour share and output. Thisis perhaps the reason we do not observe a greater procyclicality for labourshare in countries with larger informality. Thus, the informal sector mightactually shrink during recessions in other developing economies, as in thecase of Mexico.Lastly, using the informal sector?s contribution and the self-employmentratio between 1993-2003, we adjust labour share in Mexico. Figure 2.8 com-pares the adjusted share to the unadjusted one. First of all, the adjusted oneis significantly larger than the unadjusted: the difference between them is24% of the value added. More importantly, the adjusted one still keeps thesame cyclical pattern that the unadjusted one follows. The correlation coef-ficient between the two is 0.94. In addition, the adjusted one becomes evenmore volatile due to the highly cyclical properties of the informal sector.In summary, although the level of labour share suffers from the informalityproblem, the cyclical component does not.2.3.3 Sectoral Shifts and Government Expenditure over theCycleAnother driving force in the change of labour share might be shifts acrosssectors over the cycle. This is because different sectors in the economy havedifferent labour intensities. Manufacturing, particularly, has a lower shareof labour income, compared with the service sectors. For instance, dur-ing 2000-2007, 59% of income generated in US industrial sectors (includingmanufacturing, electricity, gas and water supply, and construction) went tolabour. This is in contrast to 70% in service sectors such as trade, trans-port, and financial services.24 We observe similar differences in intensities24During the same period, the labour share in trade, transport, and communication was0.69, and that in financial services and professional business services such as real estatewas 0.71.382.3. Robustness Analysesin a developing nation such as Mexico, as well. Labour share (adjustedwith self-employment ratios) between 2000 and 2007 was 0.30 in industrialactivities whereas in Mexico it was 0.44 in Mexican service sectors.25These dissimilarities in labour intensities across sectors might generateprocyclical labour share if the contributions of sectors to the economy arecyclical over the cycle. Specifically, if the share of manufacturing in the valueadded increased during recessions, we would observe a decline in labour sharemaking it procyclical with output. We document in Table 2.8 the cyclicalproperties of the contributions of sectors with the overall output. The mainresult from Table 2.8 is that there is no significant patterns for sectoralshifts over the cycle in emerging markets. When we narrow the sampleto the economies with procyclical labour share from Table 2.1, the resultdoes not change. Therefore, the procyclicality of labour share cannot beattributed to sectoral shifts during boom and bust cycles.In addition, we present some more evidence from Mexico and Korea onthe changes of industry-specific labour share. Figure 2.9 shows the paths forlabour share (adjusted using sectoral self-employment ratios similar to thatin the previous section) in major sectors in Mexico and Korea since 1980s,including industry, trade, transport and communication, and financial andbusiness services. The main observation here is that all these sectors followsimilar cyclical patterns. When labour share in the overall economy drops,shares in all sectors tend to drop, too. Indeed, the cyclical component oflabour share in every sector is positively correlated with that in the overalleconomy.26 Therefore, the movements in the overall labour share over thecycle are not driven by sectoral reasons.Another explanation for the procyclical labour share could be changesin the government expenditure over the business cycle in emerging markets.25The source for the industry-specific shares for Mexico is the OECD. Note that, evenwhen adjusted, the labour share seems to be much lower in Mexico than in the US in bothsectors of the economy. This is partially due to the lack of labour compensation fromthe informal sector, discussed above. Another potential reason might be different marketstructures and bargaining powers across countries.26The lowest correlation is 0.33, coming from the trade sector in Korea. All othercorrelation coefficients in both countries are greater than 0.42.392.3. Robustness AnalysesTable 2.8: Sectoral Shifts over the Cycle in EMsCountry ?(ind, y) ?(ser, y) ?(agr, y)Argentina -0.20 0.46 -0.67Brazil -0.42 0.40 -0.13Chile 0.04 -0.10 0.14Colombia -0.28 0.26 -0.01Costa 0.48 -0.54 0.10Czech 0.08 -0.34 0.41Egypt -0.05 0.08 -0.06Hungary 0.21 -0.14 -0.11India 0.39 -0.7 0.24Korea -0.04 0.08 -0.11Mexico -0.21 0.17 0.05Peru 0.26 -0.34 0.32Philippines -0.36 -0.19 0.35Poland 0.41 -0.56 0.17Russia 0.05 -0.26 0.37South 0.05 -0.26 0.42Turkey 0.04 -0.19 0.23Mean 0.03 -0.13 0.10Mean* -0.02 -0.01 0.02Note: The table documents the correlations between detrended shares of three majorsectors ? industry, services, and agriculture ? in the value added with detrended real GDP.A positive correlation denotes an increase in the share of that sector in output duringeconomic expansions. The data period is the same as in Table 2.1. Industry includesthe sectors of mining and quarrying, manufacturing, electricity, gas and water supply,and construction. Agriculture covers agriculture, hunting, forestry, and fishing. Finally,services includes all the other sectors combined. Data is not available for Israel. Mean*denotes the average of correlations for countries with significantly procyclical labour sharefrom Table 2.1: Argentina, Czech Rep., Egypt, Korea, Mexico, Peru, Turkey. Source:World Economic Indicators (World Bank).402.4. Interest Rates and Financial EnvironmentThis is true if the public sector is procyclical and more labour intensivethan the other sectors. It is well-known that the government expenditure isprocyclical in these economies (see Kaminsky et al. (2004) for the cyclicalproperties of policy responses in EMs). However, even when we exclude thegovernment sector (which is mainly services such as health, education, andadministration), we still observe procyclical labour share in countries such asMexico and Korea. The pink lines in Figure 2.9 represent the labour sharein the private sector in these economies. It shows that the labour share inthe private sector follows the labour share in the overall economy. This ismainly because the public sector represents only 10% to 20% of the outputduring the sample period. This observation is consistent with the above-mentioned fact: the sectoral shifts over the cycle are neither systematic norsignificant in emerging markets.2.4 Interest Rates and Financial EnvironmentThe variation in the cost of borrowing that emerging market economiesface in international markets is widely discussed in the macro literature.Neumeyer and Perri (2005), and Uribe and Yue (2006), document the coun-tercyclical behaviour of interest rates for a number of emerging markets.Here, we show that domestic real rates (Table 2.1) also support this be-haviour. As in previous studies, Argentina, Brazil, Korea, Mexico, andTurkey exhibit highly countercyclical interest rates. In contrast, the Philip-pines and South Africa have a cost of borrowing that mildly responds tooutput changes. The countercyclical movement of interest rate is mostlyexplained by default risk variation over the cycle. Arellano (2008) deriveshigh probabilities of default in equilibrium during a recession when there isless incentive for repayment in incomplete markets. This, in turn, leads tohigher interest rates and consequently causes more output contractions (seeNeumeyer and Perri (2005) and Uribe and Yue (2006)).During financial distress, borrowing becomes not only more costly butalso more limited to agents that engage in risky investment activities. Thus,principal-agent problems might be even more apparent during recessions.412.4. Interest Rates and Financial EnvironmentFigure 2.9: Labour Share across Industries in Mexico and KoreaNote: Labour share is adjusted using sector-specific self-employment ratios (the secondmethod in the text). For the Korean adjustment, see the details in the text. Industryrepresents mining and quarrying, manufacturing, electricity, gas and water supply, andconstruction. Private Sector covers all the sectors excluding the public sectors, such ashealth, education, and administration. The sectoral labour share data start at 1988 forMexico, and 1981 for Korea.422.4. Interest Rates and Financial EnvironmentStiglitz and Weiss (1981) explain credit rationing as an equilibrium phe-nomena in an environment where agents differ in terms of their risk, andfinancial markets are monopolistically competitive. Indeed, macroeconomicimplications of financial frictions are heavily touched upon in the literature.Examples in the literature of developed markets include Kiyotaki and Moore(1997), Bernanke et al. (1999), Aiyagari and Gertler (1999), and Holmstromand Tirole (1997). These works address the high cost of recessions when theagents are credit constrained.In emerging markets, these frictions are still important, especially whentheir level of financial development is considered. In an empirical study,Arteta and Hale (2008) find that crises are accompanied by a sharp de-crease in foreign credit when firm-specific and country-specific characteris-tics are controlled. Their study also reveals that the credit level remainslow for a couple of quarters and only recovers after macro-fundamentals im-prove. In a theoretical perspective, the effects of financial frictions on largeoutput drops in emerging markets have been emphasized by Aghion et al.(2001), Caballero and Krishnamurthy (2001), and Calvo (1998). Moreover,Mendoza and Smith (2006) and Mendoza (2010) stress the importance offinancial frictions on the crashes of asset prices in emerging markets.Empirical studies on the leverage ratio (measured as debt liabilities overthe market value of equity or total credit as a percentage of output) alsoshed some light on large credit booms and sharp declines. Gourinchas et al.(2001) and Mendoza and Terrones (2008) show that credit expansions playa significant role for output expansions in emerging countries. In fact, theprivate credit to GDP ratio displays a positive movement with output inthese countries (see, for example, Figure 3.3 in Chapter 3 for the cyclicalpattern of the private credit-to-GDP ratio in Mexico over time). Previousstudies deliver some explanations on volatile credit, such as poor monitoringon banks? lending activities,27 bailout guarantees aggravating moral hazardissues,28 and imperfection in credit markets serving as a financial accelera-27Lorenzoni (2008) points out the need for financial supervision as a second-best option.28Ranciere et al. (2008) and Schneider and Tornell (2004)432.5. Conclusiontor.29Motivated by these facts, we show in Chapter 3 that a model with coun-tercyclical interest rates, as well as labour financing through working capitalloans, can account for some of the variation in the labour share of emergingmarkets. We further show that perfect credit markets with highly volatileand countercyclical interest rates hardly explain the above-mentioned largecredit expansions and procyclical leverage ratios. When we introduce lever-age constraints, not only the movements of the aggregates of the goodsmarket, such as consumption, but also the variation of the labour share isamplified in the model. Thus, both the level of interest rates of external fi-nancing and borrowing constraints will be proven responsible for the cyclicalproperties of the labour share presented in this chapter.2.5 ConclusionIn this chapter, we show that emerging markets tend to have a more volatileand procyclical labour share in contrast to developed markets. Rather thanthe income level, this cyclical pattern is more related to the cyclical prop-erties of the cost of borrowing. Procyclicality increases as the country facesstronger countercyclical interest rates among emerging markets as well asdeveloped markets. Since countercyclical interest rates usually appear inemerging markets, we observe more procyclical labour share, on average,in these economies. These results are shown robust to adjustments in thelabour share and are not caused by sectoral shifts over the cycle.We also show that these patterns are showing signs of reversal in theaftermath of the recent global financial crisis. The procyclicality of labourshare tends to disappear after the 2000s in emerging economies having betterfinancing outlook relative to earlier times. On the other hand, economieswhich have seen increased risk premia in their borrowed funds, such as someof the European countries, tend to have lower shares of income going tolabour, suggesting a procyclical movement for labour share.29See Gourinchas et al. (2001) for a summary of these explanations.44Chapter 3The Role of the Cost ofBorrowing on Labour ShareFluctuations in EmergingMarkets3.1 IntroductionWe documented in Chapter 2 that labour share is relatively volatile andmoves positively with output in emerging markets (EMs), in contrast to itscyclical behaviour in developed markets (DMs). This procyclicality of labourshare appears especially in countries with countercyclical cost of borrowing.This chapter gives a theoretical explanation for those cyclical observations,and compares business cycle implications of the model with data moments.The model presented here implies a varying labour share even with aCobb-Douglas production function, which implies a constant share in stan-dard frictionless RBC models. The key mechanism resulting in those vari-ations in labour share is the wage-in-advance requirement in the model:Firms have to borrow in order to pay workers before the production takesplace and sales are cashed out.30 Even if the firm used its own resources forthe wage-bill financing, instead of borrowing, labour decisions would stillbe affected as this creates an opportunity cost in a world with a positivereturn on bonds. The liquidity need to finance the wage bill makes labour30Barth III and Ramey (2002) discuss the dataset for US firms and show that workingcapital, including the value of inventories and trade receivables, is 17 months of final sales,on average, over the period 1959 to 2000.453.1. Introductiondemand sensitive to interest rate changes. The duration between the timewhen labour income is paid and the time when the goods market clears willcreate an extra cost on the wage bill, namely interest payments to the rest ofthe world.31 During a recession, the share of output that goes to the rest ofthe world increases due to the higher interest rate, which lowers the labourshare of output. Additionally, we study limits on borrowing capacity as op-posed to a perfect-credit scenario. The introduction of these limits togetherwith working capital requirement generates an effective interest rate that ishigher than the observed one, and leads to larger responses in the labourshare relative to those in the perfect-credit scenario.In the quantitative-analysis section, we calibrate the model to Mexico,and show that working capital mechanisms can generate the right comove-ment of labour share with output, as well as explain part of the volatility inlabour share, cyclical properties of which are shown in Figure 1.1 and Figure1.2. In addition to this effect, the results are amplified when the agents arecredit constrained. The baseline model with both working capital and lever-age constraint can account for 60% to 73% of the variation in the Mexicanlabour share depending on the level of the working capital requirement. Thepresence of the binding leverage constraint not only amplifies the responseof labour share but also improves the performance of the model with respectto other business cycle regularities in emerging markets, particularly highlyvolatile consumption, highly procyclical investment and consumption, andcountercyclical net exports with output.This chapter is related to the literature that previously studies the dif-ferent behaviour of interest rate that emerging markets face. As opposed toslightly procyclical interest rates in developed markets, the countercyclicalinterest rates are mostly explained by the country-risk premium (see Arel-lano (2008)). Neumeyer and Perri (2005), and Uribe and Yue (2006) alsoshow that countercyclical interest rates due to default risk can be a prop-31The use of medium or long-run post-dated checks and illiquid assets as the return togoods sold will increase the liquidity burden on wage bill. This is not only due to the timelag between wages are paid and sales are cashed out, but also because of the uncertaintyin drawing checks which might be high during recessions.463.1. Introductionagation mechanism to generate business cycle fluctuations in this group ofcountries; however, they do not address the implications on labour income.Recently, Li (2011) studies that these models along with income effect onlabour supply can explain a significant part of the wage volatility.32 Here,our contribution to these models is twofold. Firstly, we show that these mod-els with working capital requirement have implications also on labour sharemovements. Secondly, we show that under the shock processes calibratedconsistently with the data, the perfect credit used in the above-mentionedmodels is not substantial enough to match fluctuations in these countries,such as highly volatile consumption and countercyclical net exports. Thisis why we emphasize imperfect credit together with working capital require-ments.Interest rates are not the only cost of borrowing developing countriesface. Credit frictions are also significant for these lower income economies.Arteta and Hale (2008), for instance, find additional decline in foreign creditto private sector as high as 20% of country-specific average after debt crises.This decline is not temporary, either. It persists more than two years, whichpotentially aggravates the recession and affects business cycle fluctuations.In terms of the macroeconomic implications of credit frictions, there arenumerous studies in the literature following Kiyotaki and Moore (1997)and Bernanke et al. (1999).33 These studies highlight the importance ofcredit frictions in the fluctuations of developed markets. A significant roleof these frictions can be expected in developing countries as well, especiallywhen their relatively low level of financial development is considered. Calvo(1998) and Caballero and Krishnamurthy (2001), in fact, study the effects offinancial frictions on output drops in emerging markets. Recently, Mendoza(2010) points out that the real cost of borrowing can be amplified in sudden32At the time of writing this dissertation, earlier version of Li (2011) was studyinga contractual model, in contrast to a Walrasian model presented here and in the finalversion. Therefore, the last version makes the model in her paper closer to the model hereexcept that we also investigate the impact of imperfect credit and concentrate on labourshare fluctuations.33See Kocherlakota (2000), Aiyagari and Gertler (1999), Devereux and Yetman (2009),Jermann and Quadrini (2006) for the use of financial constraints.473.1. Introductionstops through credit frictions.34 This paper stands on the same line with thisliterature and claims that when working capital is introduced, these mod-els have important implications on the short-run dynamics of labor sharethanks to volatile cost of borrowing, whether it be the observed one in themarket or the effective one through imperfect credit.35Another recent paper by Boz et al. (2009) studies labour income fluctu-ations in emerging markets. They incorporate search and matching frictionsin a small open economy real business cycle (SOE-RBC) model with coun-tercyclical interest rates to explain wage volatility in EMs. In these models,wages fall in recessions more than in a frictionless world due to the negativeeffect of higher interest rates on the outside option of workers, namely thevalue of unemployment. Chapter 4 will compare this type of frictions tofinancial frictions proposed in this chapter in order to evaluate the relativeimpact of search frictions to financial frictions on labour income fluctuationsin EMs, and conclude that the relative impact is smaller when householdsendogenously choose hours per worker.Besides the fluctuations in emerging economies, we also study in Section3.5 the performance of the model presented here in a developed market,Canada, and conclude that the model implies more or less stable labourshare that is slightly countercyclical with output. This is because interestrates in developed markets display a slight positive correlation with outputperhaps implied by an increase in marginal product of capital as a responseto positive productivity shocks (see Kydland and Prescott, 1982) and/orthe effectiveness of monetary authorities that might have an impact on realvariables in the short-run (see King and Watson, 1996).Labour share movements in developed markets have already been ad-dressed in the literature. R??os-Rull and Santaeulalia-Llopis (2009) showthat when we allow labour share to have a dynamic response, it displaysan overshooting property as a response to a positive productivity shock.34The difference here is that credit frictions are not temporary; rather they affect busi-ness cycles implications.35By saying imperfect credit, I not only consider credit rationing in the financial system,but also mean leverage cycles explained by various types of asymmetric informationalfrictions in the emerging market asset as in Fostel and Geanakoplos (2008).483.1. IntroductionThus, labour share falls down on impact, then starts to increase, and af-ter a couple of quarters it passes beyond the steady-state level and staysabove the mean for a long time.36 These findings show that there are somemechanisms or frictions that prevent labour share from initially respondingas it does in the medium run. Bentolila and Saint-Paul (1998) empiricallyshow that adjustment costs on labour and union-wage bargaining have asignificant effect on the movements of labour share. Bentolila and Bertola(1990) use adjustment costs ? specifically, firing costs ? to explain coun-tercyclical labour share in European countries. R??os-Rull and Choi (2009)emphasize the effect of non-competitive wages and search frictions on thelabour share dynamics in a developed market, namely the US. Boldrin andHorvath (1995) and Gomme and Greenwood (1995) use contractual agree-ments between workers and employers in which real wages deviate from themarginal product of labour. This mechanism also makes total wage bill lessresponsive to output and generates countercyclical labour share. Followingthese papers, when the model implications of a developed market are com-pared to those of an emerging market in Section 3.5, we include a laboradjustment cost to the model so as to analyze how much sluggishness inlabour market contributes to the cyclicality of labour share. The result isthat working capital channel is the predominant factor explaining labourshare fluctuations in EMs through volatile interest rates. In contrast, otherfactors producing less responsive wage bill explain more the movements inthe labour share in DMs.The plan for the rest of the paper is as follows. Section 3.2 presentsthe model, and section 3.3 describes the calibration method for the param-eters. Section 3.4 discusses the main findings, and compares them to datamoments. Section 3.5 extends the model with an adjustment cost on labourand shows the implications of this model in EMs and DMs. Section 3.6concludes.36Using Mexican quarterly data, we also find evidence on overshooting property inlabour share. However, in this work, we are interested in the immediate response oflabour share in emerging markets which is procyclical.493.2. Model3.2 ModelThe model is in the class of small open economy real business cycle modelswith an internationally traded single good.37 Asset markets are incompletein the sense that there is only one single internationally traded one-periodbond which pays the buyer a predetermined interest. Agents face shocksin the interest rate on bonds and productivity level. These shocks followexogenous processes, the details of which are described below. The differencefrom a standard RBC model is that wages have to be paid in advance andthat the country is credit constrained.3.2.1 Optimization ProblemLet us consider an economy with an infinitely-lived self-employed repre-sentative household.38 The agent derives utility from consumption ct andleisure 1 ? lt, where the total time that he devotes to labour and leisure isnormalized to one. His preferences are described as follows:??t=0 ?tEtu(ct ?N(lt)) (3.1)where 0 < ? < 1 is the discount factor, u(.) is twice-continuously-differentiableand a concave period utility function, and N(.) expresses the disutilityof labour which is twice-continuously-differentiable and a convex function.This utility representation is known as GHH preferences after Greenwoodet al. (1988). These preferences eliminate the wealth effect and make laboursupply decisions independent of consumption. Neumeyer and Perri (2005)show that the standard Cobb-Douglass utility function generates large wealtheffects when interest rates are volatile and countercyclical, and results in37Our main motivation for real business cycle model ?rather than a model with nominalrigidities? is the high CPI and wage inflation observed in developing countries. Thus, weverify that nominal wages are also highly volatile in these economies.38This is similar to the yeoman-farmer model, in which the farmer uses his own labourto produce the good, which is widely used in monetary literature (see Ball and Romer(1990) and Mankiw (1985)). The alternative is to use a decentralized representation whichhas households and firms as separate agents. We choose this type of modeling since itallows us to impose a constraint on the whole nationwide debt, including both householddebt and working capital loans (see Mendoza (2010)).503.2. ModelFigure 3.1: Timing Linecounterfactual employment movements. An alternative to this form wouldbe to use standard preferences with asset market segmentation to lessen theeffect of interest rates on labour supply decisions. However, this specificform is chosen due to its simplicity to deal with the wealth effect.The agent maximizes the life-time expected utility and chooses the op-timal sequences of consumption, ct, labour, lt, investment, xt, and bondholdings, bt, subject to budget and leverage constraints:ct + xt + bt + ?(bt) ? yt ? ?(Rt ? 1)wtlt +Rt?1bt?1 (3.2)bt ? ?Rtwtlt ? ??tyt (3.3)Income, in this economy, is generated by producing a single traded good,yt, using a constant-returns-to-scale technology which has capital, kt, andlabour, lt, as the factors of production:yt = AtF (kt, lt) = Atk?t l1??t (3.4)where ? is the capital elasticity in the production and At is the total factor513.2. Modelproductivity. The agent chooses an investment level, xt, in order to accu-mulate capital by taking into account the fact that capital depreciates at arate, ?. Capital accumulation follows the law of motion:xt = kt+1 ? (1? ?)kt + ?(kt+1, kt) (3.5)where ?(kt+1, kt) is a quadratic convex capital adjustment cost to mitigatethe excessive volatility of investment that might arise in small open economymodels. The agent can also trade an international one-period bond, bt, inthe market that has a gross return, Rt. A quadratic convex cost function,?(bt), is introduced into the model as in Schmitt-Grohe? and Uribe (2003) toensure a stationary path for bond holdings.39The model has a wage bill financing, that is, a fraction ? of wage bill hasto be paid in advance of the production. This can be rationalized either bythe fact that workers want to consume in the beginning of the period butcannot access the financial markets or by having a production line wherefirms use installments or post-dated checks so that sales are cashed out inlater periods. In the model, these working capital loans are borrowed in thebeginning of the period from international markets at the same rate on thebond, Rt, after the shock is realized and generate interest payments in theend of the period to the rest of the world (see Figure 3.1 for a representationof the timing line). That is why the income net of these payments areentered in the right-hand-side of the budget constraint. The labour marketis competitive. Therefore, the wage is taken as given by the representativeagent and is equal to the marginal disutility of labour:wt = ?N(lt)/?lt (3.6)where lt is the market average. This is similar to the optimal labour supplyin a decentralized competitive equilibrium set-up.The economy also faces an external leverage constraint given in equation39This cost is zero in imperfect credit since non-stationarity does not exist in this caseand it is so small in the perfect credit case that it does not affect the long-run businesscycle implications of the model.523.2. Model(3.3). The net foreign asset is constrained by a fraction of output, ??t. Inother words, net debt including working capital loans has to be smaller thana ?t fraction of output. This fraction has a stochastic component and variesover time. Ludvigson (1999) finds that forecastable (ex-ante) credit growthhas a significant influence on consumption that is independent of variationin predictable income growth. Furthermore, she shows that introducing astochastic upper limit on the debt-to-output ratio improves the correlationbetween consumption and income growth in the US. The high correlationbetween consumption and income appears in emerging markets, as well.40In the quantitative exercise with perfect credit markets, we will see belowthat the model with working capital cannot account for highly volatile (andhighly cyclical) consumption and countercyclical net exports even in thepresence of countercyclical interest rates. A stochastic leverage constraint,however, can indeed improve these results.41Another issue is how frequently these leverage constraints bind over thecycle, which cannot be directly observed from data. Stiglitz and Weiss (1981)show that financial imperfection, as in the form of quantity restrictions, canbe an optimal equilibrium outcome in every state of nature when there areasymmetric informational costs in the environment.42 Moreover, consider-ing the low levels of financial development in emerging markets, these kindsof constraints are more likely to bind. In this dissertation work, we nar-row the scope to see how the presence of binding constraints interacts withworking capital by assuming that the constraint binds permanently. Sincethe constraint binds at every state, ?t represents the leverage ratio of theeconomy as the net debt over GDP. Motivated by interest rates being animportant driving force in emerging markets, the leverage ratio is assumed40See Aguiar and Gopinath (2007) for the documentation of business cycle regularitiesin emerging markets.41Sarquis (2008) and Guajardo (2004) also explore the effects of these types of creditshocks in emerging market business cycles. The difference here is that we introduceworking capital channel in order to explore the effect of changes in the effective cost ofborrowing on labour market variables.42See Mandelman (2011) for more on monopolistic competition among emerging-marketfinancial institutions.533.2. Modelto move over the cycle in the following way:??t = ??R?t (3.7)where ? > 0. This tells us that the economy faces credit restrictions (such aslosing access to financial markets) during financial crises when spreads andinterest rates are high. Our motivation comes from the empirical evidenceon foreign credit supplied to private sector in emerging markets. Arteta andHale (2008), for instance, find debt crises are accompanied with a signif-icant and persistent drop in foreign credit after controlling fundamentals.In addition, Fostel and Geanakoplos (2008) document that during closuresemerging market bond issuance drops even though its spread has increasedminimally relative to other assets ? during mild recessions. They show thatthis phenomenon can be explained in a financial environment where lenderscannot distinguish bad credit from good credit. Therefore, the exogenousstructure of leverage ratio can be endogenized through the mechanism inFostel and Geanakoplos (2008), which imply positive correlation betweentightening and interest rates.43 Furthermore, the increase in the asymmet-ric information cost ? driving spreads up for these economies ? can causemonopolistically competitive banks to not only charge high interest ratesbut also to impose tighter restrictions on credit in the framework describedby Stiglitz and Weiss (1981). This specification would imply a one-to-onerelation between interest rates and leverage ratio. The observed relationshipbetween these two variables in the data are consistent with these explana-tions. Private sector credit-to-GDP ratio for non-financial firms shows ahigh correlation of -0.60 with interest rates in Mexico (see Figure 3.3).3.2.2 Competitive Equilibrium and Labour ShareA competitive equilibrium for this economy consists of sequences of optimalallocations {ct, lt, kt+1, bt, xt, yt} and wages {wt} such that43The external borrowing is crucial here as a model with heterogenous agents and bind-ing collateral constraints in a closed economy might imply a negative correlation betweenthose two variables under endogenous interest rates.543.2. Model1. the representative agent solves the maximization problem subject tobudget and collateral constraints in (2), taking wages, interest rate,and initial states k0 and b0 as given,2. wage equals the marginal disutility of labour wt = ?N(lt)/?lt,3. labour decisions satisfy lt = lt, and4. goods market clear, meaning that the goods that are not spent onconsumption, investment, and the cost of bond holdings, represent thenet export for the economy:ct + xt + nxt + ?(bt) = yt (3.8)Given the problem described above, the optimal condition for bond holdingsand capital accumulation can be expressed as:?t[1 + ??(bt)] = ?t + Et?t+1?Rt (3.9)?t[1 + ?2,t] = ?Et[?t+1(1 +At+1F1,t+1 ? ? ??1,t+1 + ?t+1?t+1)]where the subscript in the functions denotes the partial derivative of thefunction with respect to its argument numbered. These conditions tell usthat bond holdings and capital accumulation are at their optimal level whenthe cost of an additional bond/capital accumulation is equal to the dis-counted benefit to the households. The expression ?t is the Lagrange mul-tiplier on leverage constraint at period t representing the marginal value ofrelaxing the leverage constraint. The other Lagrange multiplier on budgetconstraint, ?t stands for the marginal utility of consumption:?t = uc(ct, lt) (3.10)Finally, the optimal condition for labour demand in this economy can bewritten as follows:?u2(ct, lt) = ?t[AtF2(kt, lt)? ?(Rt ? 1)wt] + ?t[?tAtF2(kt, lt)? ?Rtwt]553.2. ModelThis condition suggests that the marginal cost of increasing labour inputhas to be equal to the marginal benefit of labour to the household. Com-bining this equation with the wage rate formula described above and therelationship ?N(lt)?lt = ?u2(ct,lt)u1(ct,lt), we can express the (inverse) labour demandin this economy as follows:wt =1 + ?t?t?t1 + ?(Rt ? 1) +?t?t?RtAtF2(kt, lt) (3.11)Now, we consider the effects of working capital and credit constraint onlabour share. In order to see the contribution of each friction, we firstlyexamine the case in which the upper limit on borrowing is infinitely high,i.e., the agent is not credit-constrained implying that ?t = 0 for every t. Inthis case of perfect credit, the expression for labour share follows:st =wtltyt=1? ?1 + ?(Rt ? 1)(perfect credit) (3.12)where st is the labour share at period t and ? is the capital exponent in theproduction function. Equation (14) tells us that labour share would still bemoving even when the credit market is frictionless since wages deviate frommarginal product of labour.44 An increase (decrease) in interest rates drivesthe wages to a lower (higher) level than the marginal product of labour whichreduces (increases) the labour share of income and increases (decreases) theshare of the interest payments in output.44In a decentralized set-up, a similar implication can be derived from firm maximizationas in the following:maxkt,ltF (At, kt, lt)? rkt kt ? (1 + ?(Rt ? 1))wtltThis maximization problem produces the same labour share as in the equation (6) whenCobb-Douglass production function is taken.563.2. ModelWhen the credit constraint is introduced, the effect of this mechanism isamplified:st =wtltyt=(1? ?)(1 + ?t?t?t)1 + ?(Rt ? 1) +?t?t?Rt(imperfect credit) (3.13)Since the increase in Rt is accompanied with credit rationing implying that?t?tand Rt are positively correlated to each other, a shock to interest ratewill further increase the effective interest rate and influence the labour sharemore adversely. Intuitively, the demand for labour is lowered not only be-cause of the higher cost of borrowing but also the higher credit restrictionsimposed by lenders on firms seeking loans for working capital needs. Notethat since labour decisions affect output which tightens or loosens the creditconstraint, ?t?t?t appears in the nominator. However, because?t?tand ?t aremoving in different directions over the cycle, the impact is mostly driven bythe denominator.If a share of income changes, then some other shares have to, as well.We write down incomes of output and explain the changes in the share:yt = wtlt + rtkt + ?(Rt ? 1)wtlt? ?? ?interest payments on working capital loansInterest cost on wage bill increases when interest rates rise. In a per-fect credit world, capital share remains constant since there is no distortionbetween capital return and marginal product of capital. Therefore, in thebooks of national account, labour income share, wtltyt , falls, and the shareof payments to the rest of the world increases. In the presence of leverageconstraints, rate of return on capital remains higher than it would be underno binding case; therefore, capital share increases as an immediate responsesince capital stock cannot change at the time of shock. Thus, the decline inlabour share would be higher since both the share of interest payments andof capital income rises.573.3. Calibration3.3 CalibrationThe equations (3.2)-(3.13) along with the shock processes constitute thesystem of equations for endogenous variables. These equations are log-linearized, and then solved for the policy functions in terms of endogenousstate, {kt, bt?1}, and exogenous state variables, {At, Rt}. The model is thencalibrated to Mexico quarterly. The sample period is 1987Q1-2008Q4 forwhich we have the Mexican data. Table 3.1 summarizes the parameter val-ues and Figure 3.2 shows the cyclical pattern of the Mexican labour sharewith output, which we attempt to explain.3.3.1 ShocksWe assume that shocks to productivity (in logs) and interest rates (log ofgross interest rate) are correlated simultaneously such that t = [At , Rt ] isdrawn from an i.i.d normal bivariate distribution, N(0,?), with zero meanand covariance, ?. Each shock follows an independent AR(1) process45:A?t = ?AA?t?1 + A,tR?t = ?RR?t?1 + R,t,with?t?t =(?A ?A,R?A?R?A,R?A?R ?R).Solow residuals are used as the measure of productivity. We calcu-late Solow residuals using Mexican GDP from the OECD: lnAt = ln(yt) ??ln(kt)? (1??)ln(lt). Capital exponent is set to match the average labourshare (see below) in Mexico. Employment series come from Neumeyer andPerri (2005) and we extend it to 2008Q4 using series at ILO. In order to findtotal labour input used in production, we calculate total hours by the given45We verified that a VAR estimation of these shocks results in insignificant coefficientsfor the lags, consistent with results from the previous literature (see, for instance, Mendoza(2010)583.3. CalibrationFigure 3.2: Labour Share in MexicoNote: Labour share is at quarterly frequency and seasonally adjusted. It represents totallabour cost as a share of manufacturing value-added. Labour cost is adjusted for self-employment. Cyclical components are calculated as percentage deviations from the HPfiltered trend. Source: OECD.593.3.CalibrationTable 3.1: ParametersName Symbol Value ExplanationDiscount factor ? 0.98 calibratedUtility curvature ? 5 Neumeyer and Perri (2006)Labour curvature ? 2.75 calibratedLabour weight ? varies calibratedCapital exponent ? 0.43 calibratedDepreciation rate ? 0.02 assumedWage bill paid in advance ? 1 or 0.66 assumedBond holding cost ? 0.001 assumedCapital adjustment cost ? varies calibratedInduced leverage ? 1.82 calibratedNet Foreign Debt / GDP ? -0.42 dataPersistence, At ?A 0.75 estimatedPersistence, Rt ?R 0.63 estimatedCorrelation coef. ?A,R -0.45 estimatedStd. deviation, At ?A 0.0134 estimatedStd. deviation, Rt ?R 0.0134 estimated603.3. Calibrationemployment series and hours worked in manufacturing from the OECD forMexico. Capital stock series are constructed using investment perpetualmethod. Particularly, we set depreciation rate, ? = 0.02 and use the bal-anced growth path equation, (?+?)ky =iy . Assuming, there is a constantgrowth rate on the path for the first ten quarters, we find the approximateinitial capital stock. Then, we extend the capital stock data using invest-ment series from the OECD. Detrended Solow residuals suggest an AR(1)coefficient of 0.75 and standard deviation of 1.34% of shocks to TFP.For interest rates, we have two representative series: JP-Morgan EMBIrates and T-bill rates for Mexico. EMBI rates cover only half of the sample.On the other hand, we have T-bill rates for the whole sample period fromthe IFS.46 EMBI rates are constructed using JP-Morgan EMBI+ spreaddata. Since these bonds are denominated by the US dollar, real yields arecalculated using the US inflation as in Chapter 2. For Mexican T-bill rates,we use Mexican GDP deflator to obtain real returns. Since the model doesnot take into account inflation, we subtract the next-period inflation fromnominal interest rates to find the (ex-post) real interest rate in Mexico. Wenote that using different types of expected inflation such as an average ofpast inflation rates still suggest similar volatility in interest rates. However,we think that using current and/or past inflation in an environment withhighly volatile inflation might be fallacious.47 The behaviour of the cyclicalcomponent of domestic interest rates is consistent with Kaminsky et al.(2004). They show that domestic interest rates are volatile and counter-cyclical in most of the developing countries. We find that both interest rateseries have a negative correlation with output above -0.50 in Mexico. Inaddition, both series share a similar persistence of around 0.60.Despite the similar comovement between foreign and domestic currencydenominated bond rates, the volatility differ dramatically. The quarterlyyields from EMBI-constructed detrended interest rates has a standard de-46We take the first two-years observations out of the sample since it represents abnormalchanges from -20% to 100% of real return. This is done in order for the results not to bedriven by these variations.47We also use the trend portion of the current and/or past inflation and observe thatthe results on shock parameters do not change much.613.3. Calibrationviation of 0.44%. On the other hand, domestic interest rates are more thanfour times volatile: 2.01%. This difference is mainly due to the exchangerate risk. Indeed, previous literature on emerging market crises emphasizethe burden of the foreign-currency denominated debt as these countries ex-perience sharp depreciation during their economic slowdowns. Therefore,lending to emerging economies in their own currency is risky. And simi-larly, borrowing in foreign currency is risky in the eyes of firms in emergingmarkets as this increases debt service during the periods of falling outputbecause of depreciation (Calvo (1988), Gumus (2005), Jeanne (2003), Krug-man (1999)).48 Since our model does not include monetary terms in it,assuming a low volatility of rates will overpredict the effect of imperfectcredit. Given large volatility difference, we assume that shocks to interestrates have the same volatility as productivity shocks, that is equal to 1.34%.This corresponds to a number within the range of deviations observed inthose two series.49 The correlation between shocks to TFP and interestrates is estimated to be equal to -0.45 using foreign rates.50For the stochastic leverage ratio, we set ? in equation (8) to be 1.82 sothat the standard deviation of ?t matches the standard deviation of credit-to-gdp ratio for the private sector over the sample period (See Figure 3.3 forthe cyclical properties of the credit-to-GDP ratio in Mexico.) This numbersuggests that debt-to-income rule increases by 1.82% when quarterly interestrates rise by 1%.3.3.2 Other Model ParametersUsing the average interest rate level and depreciation rate, we extract thevalues for discount factor and steady-state shadow price of credit constraintsimultaneously from optimal bond holding and capital equations at steadystate. Calculations result in ? = 0.98 and ?? = 0.01. Capital exponent48Moreover, Kaminsky et al. (2004) show that monetary policies are pro-cyclical withoutput in emerging markets. These policies (such as increasing nominal interest rates toprevent capital outflows) might have an impact on real interest rate.49It is close to other parameter values used in the literature for the deviation of interestrates. See Mendoza (2010) and Li (2011).50The estimated correlation coefficient is -0.48 when T-bill rates are used.623.3. CalibrationFigure 3.3: The Cyclical Component of the Credit-to-GDP Ratio in Mexicocorr(cr,y) = 0.48Mexico?.1?.050.05.1deviations1970 1980 1990 2000 2010tCredit to Nonfinancial firms / Output OutputNote: Credit represents total domestic credit to nonfinancial firms (i.e. private credit).Deviations are in percentage points from the HP filtered trend.633.3. Calibrationis calibrated to match the average labor share in Mexico over the sampleperiod, i.e.,1? ?1 + ?(R? 1)= 0.57.The equation shown above implies ? = 0.43 when ? = 0.6651. The labourshare using the National Account data is very small (around 0.33) since itsuffers from measurement problem such as informal employment and self-employment labour income. Labour compensation data from National Ac-count usually come from the formal sector in developing countries. The dataon the contribution of informal sector to Mexican GDP from 1993-2004 areavailable. Informal sector contributes 12.4% of value added GDP on average.Adjusted share is obtained first by correcting official labour compensationdata using this contribution of the informal sector. The implicit assumptionhere is that labour content is the same across formal and informal sectors.52Then, we further adjust it to account for the labour income of the self-employed people using the self-employment ratio in Mexico, which is closeto 33% of the total employment (see Figure 3.4 for unadjusted and adjustedlabour share as well as the labour share in different sectors of the economy).This implies the average labour share over this period to be equal to 0.57,a number closer to that in developed economies.The functional form for the utility function is the following:u(ct, lt) =11? ?[ct ? ?l?t ]1?? (3.14)Intertemporal elasticity of substitution is set to 0.2, which implies ? = 5following Neumeyer and Perri (2005), who use the same preferences as inthis chapter. Using the optimal labour supply equation at the steady state,labour weight parameter ? in the utility function is set to match l = 0.32which is the fraction of hours worked in the total non-sleeping hours. The51Capital exponent is equal to 0.42 when ? = 152Note that one can imagine a higher labour share for informal sector. A higher share,say 20% higher labour share than in the formal sector, increases the level of labour share inthe overall economy by only a couple of percentage points since informal sector?s contribu-tion is smaller in value. In terms of volatility, a higher share in the informal sector wouldactually make the labour share more volatile because the informal sector is pro-cyclical.643.3. CalibrationFigure 3.4: Adjusted Labour Share and Sectoral Shares in MexicoNote: Adjusted share is annual and obtained by first scaling official labour compensa-tion data using the contribution of the informal sector to the value added. The implicitassumption here is that labour content is the same across formal and informal sectors.Then, the share is further adjusted for self-employment to account for the labour incomeof the self-employed. The sectoral shares are adjusted only for self-employment. Industryrepresents mining and quarrying, manufacturing, electricity, gas and water supply, andconstruction. Private Sector covers all the sectors excluding the public sectors, such ashealth, education, and administration. Source: OECD, INEGI.653.4. Resultstotal hours worked time-series in manufacturing from OECD-MEI datasetis used in the calculation of the steady state value of hours, l. In the model,? determines the Frisch elasticity of labour supply, 1??1 . The empiricalevidence on this parameter is mostly coming from developed markets andthe values used in the literature are in the range [0.5,1]. Considering theirlower income and wealth, we assume that agents in emerging markets standcloser to lower bound of this range and set the value of ? to 2.75 showing anelasticity of labour, 0.57, which implies a standard deviation of hours closerto data. Although this parameter is not crucial for the results on labourshare fluctuations, it changes how the movements in the wage bill are splitbetween the labour input and hourly wages. Finally, we calculate the netforeign asset held by households at the steady state as the average over thesample period using the dataset on countries? external asset positions fromLane and Milesi-Ferretti (2007).3.4 ResultsFigure 3.5 presents the impulse responses from models with perfect andimperfect credits. Table 3.2 further contains the volatility implications ofdifferent versions of the calibrated model along with the second momentsfrom data. Details on data along with sources can be found in Appendix BThe data moments represent quarterly variations after taking logs (exceptnet export-GDP ratio and net interest rate) and HP-filtering that sets thesmoothing parameter to 1600.53 Quarterly labour share data in manufactur-ing are used as a proxy of overall labour share fluctuations in the economy(see Figure 3.2). We check that, at the annual level, series from both man-ufacturing and total economy are highly correlated to each other (0.86) andhave large standard deviations of 4.5% and 3.5% in the manufacturing andoverall economy, respectively. The second column lists the moments fromthe standard SOE-RBC model for comparison, and the remaining columnsdocument the results of the model described above in both cases of perfect53ARIMA-X12 from the Census Bureau is applied to deseasonalize data when there aresignificant seasonal effects.663.4. Resultsand imperfect credit markets for different values of the working capital pa-rameter.SOE-RBC Model. To begin with, the results from the standard SOE-RBC model cannot generate any movements in labour share because theCobb-Douglass production technology implies a constant labour share in acompetitive environment where wage is equal to the marginal product oflabour. Consequently, it cannot account for the volatility in labour marketvariables. As mentioned earlier, real wages are more volatile than outputin emerging markets, but even with the relatively inelastic labour supplyassumed, SOE-RBC is having a hard time explaining highly volatile wages.One of the most distinguishable characteristics for fluctuations in emerg-ing markets emphasized earlier in the literature is that they have highlyvolatile and cyclical consumption and net exports to GDP ratio. The stan-dard model also fails to adequately explain these features in emerging mar-kets since agents tend to smooth their consumption using credit marketswhen the shocks are temporary.54The Model with Working Capital and Perfect Credit. We nowcontinue with the implications from the model with working capital in anenvironment with perfect credit. These results are reported in the thirdcolumn in Table 3.2. The introduction of working capital without any limitson borrowing can generate variations in labour share. Because interest ratesare countercyclical, working capital requirement tends to produce a largerresponse in labour demand than in the standard SOE-RBC model. As aconsequence, it can be seen that wages and hours become more volatilein these models. Having a more volatile wage bill results in a procyclicallabour share consistent with the data. Although the model could predict themovements of labour share with output, the volatility depends heavily on54Aguiar and Gopinath (2007) show that the standard model can explain these featureswhen non-stationary shocks are introduced to the model. However, recently, Garcia-Ciccoet al. (2010) estimate that these shocks have a negligible role in emerging market businesscycles.673.4. ResultsFigure 3.5: Impulse Responses0 10 20?0.500.5LSTo At Shocks0 10 20?202Y0 10 20?202W0 10 20?202To Rt Shocks0 10 20?0.4?0.200 10 20?2?10  Imperfect Cr. and ?=0.66Imperfect Cr. and ?=1Perfect Cr. and ?=1Perfect Cr. and ?=0683.4. Results...continued0 10 20?101HTo At Shocks0 10 20012C0 10 20?202NXY0 10 20?1?0.50To Rt Shocks0 10 20?2?100 10 20?505  Imperfect Cr. and ?=0.66Imperfect Cr. and ?=1Perfect Cr. and ?=1Perfect Cr. and ?=0693.4. ResultsTable 3.2: Model Implications for Mexico?=1 ?=0.66Data RBC Perfect Imperfect Perfect ImperfectCredit Credit Credit CreditStandard DeviationOutput 2.19 1.95 2.14 2.30 2.07 2.25Labour share 3.58 0.0 1.68 2.62 1.12 2.13Net exports 2.25 1.80 1.89 1.67 1.85 2.01Standard Deviation(Relative)Wage 1.82 0.64 0.95 1.25 0.83 1.14Hours 0.64 0.36 0.54 0.71 0.47 0.65Consumption 1.35 0.56 0.69 1.08 0.65 1.28Investment 3.45 3.45 3.45 3.45 3.45 3.45Correlationwith OutputLabour Share 0.44 0.0 0.52 0.68 0.48 0.68Interest Rate -0.53 -0.39 -0.52 -0.59 -0.48 -0.55Wage 0.41 1.0 0.93 0.90 0.95 0.92Hours 0.64 1.0 0.93 0.90 0.95 0.92Consumption 0.89 0.87 0.86 0.92 0.86 0.91Investment 0.94 0.30 0.46 0.91 0.41 0.88Net Exports -0.65 0.43 0.24 -0.56 0.30 -0.58Lab. Share and R -0.48 0.0 -1.0 -0.89 -1.0 -0.86Note: Data period is 1987Q1-2008Q1. All variables are in logs (except net interest rate and netexport) and HP-filtered. Quarterly labour share and wages are coming from manufacturing butlabour shares in manufacturing has a very strong correlation with labour share in overall activityat annual level. Investment adjustment cost parameter is set to match investment volatility. Netexport is defined as exports minus imports over output. The last line represents the correlationbetween labour share and interest rates.703.4. Resultsthe working capital parameter, ?. The results from the model with a lowerworking capital requirement can be seen in the fifth column. A smaller valuefor this parameter significantly lowers the volatility of labour share.Although they predict some of the volatility in labor share, the modelswith perfect credit can explain neither the strong countercyclicality in netexports-GDP ratio nor the strong procyclicality in investment. As a result,consumption tends to be less volatile than output. This is because underrelatively less persistent shocks, investment is less cyclical with current out-put and the smoothing behaviour still appears. Consequently, net exportsbecome procyclical (or acyclical) in the perfect credit market. These resultsare different from the ones in Li (2011) where she generates these featuresin a model with working capital and perfect credit. One reason might bethe lower elasticity of the intertemporal substitution set here. However,consumption is still not more volatile than output when using the sameparameter value as in her paper. Another reason is that the interest rateshocks in her paper are twice as volatile as TFP shocks, whereas they arejust as volatile as TFP shocks here.The Model with Working Capital and Imperfect Credit. Finally, theresults from the model with leverage constraint can be seen in columns 4 and6. An imperfect credit market makes labour share more volatile comparedto the case with perfect credit thanks to more volatile wages and labourinput. Now, the model can explain 73% of the variations in the Mexicanlabour share assuming ? = 1. Moreover, even with a lower working capitalparameter (as shown in column 6), the model can explain a significant partof the volatility in labour share (60% of the variations in labour share).Therefore, the presence of borrowing limits allows us to set a lower workingcapital requirement than that used in the literature. This is quite reasonableespecially when one considers that, in some industries, working capital mightbe of less importance.Imperfect credit not only increases the volatility in the labour market,but also significantly improves the implications on the fluctuations of con-sumption, investment, and net exports over the cycle. When the leverage713.5. Implications for Developed Economiesconstraint is introduced, which makes financing even more difficult, smooth-ing behaviour disappears. Therefore, consumption becomes more volatilethan output, and investment becomes strongly procyclical with output. Asa result, the net export-GDP ratio moves inversely with output over thecycle consistent with data.We also note that all models presented here imply very procyclical wages.However, wages are somewhat lower cyclical (0.45) in data. This smallercyclicality is a well-known fact in developed markets as well.55 Wage rigidi-ties through contracting models (see Gomme and Greenwood (1995)) andthe change in skill composition of labour (see Bils (1985)) over the cyclemay make aggregate wages less cyclical.56 In addition, search-and-matchingfrictions in labour market can also contribute to explain wage movementsless cyclical than models with frictionless labour market predict, which isfurther discussed in Chapter 4.3.5 Implications for Developed EconomiesIn this section, we show the performance of the model in a developed market.We calibrate the model to Canada, and then compare and contrast theresults with the implications for Mexico. The model with imperfect creditmarket and a working capital requirement of 0.66 is taken as the baseline inthis section.As mentioned earlier, the literature suggests mechanisms for the coun-tercyclicality of labour share in developed markets through less responsivewage bill to output changes. High unionization (especially in Europe), firingand search costs, and contractual labour market imply sluggishness eitheron wages or on the quantity of labour.57 Based on these explanations, for55See R??os-Rull and Choi (2009) and Li (2011) for the recent wage-output correlationsin the US and other developed markets.56Introducing these features into the model will, in fact, increase the importance ofworking capital and the countercyclical cost of borrowing. This is because those featurestend to lower the volatility of wages in models, and thus work in the opposite direction ofworking capital.57There are also other explanations for the countercyclicality of labour share in de-veloped markets. Hansen and Prescott (2005) introduce occasionally binding capacity723.5. Implications for Developed Economiesa representation of labour market rigidities, we include an adjustment coston labour to understand how they interact with working capital and con-tribute to the variability of labour share. These rigidities or regulations inthe labour market can be expected in emerging markets, as well. Heckmanet al. (2000), for example, show that employment protection is high in LatinAmerican countries.58 Therefore, by having adjustment costs on labour,we would also like to see how they alter the results of the baseline modelin emerging economies explained in the previous section. A convex labouradjustment cost is introduced to the model as follows:?(lt, lt?1) = ?llt?1(lt ? lt?1lt?1)2This cost has a significant effect on the autocorrelation of hours worked.Therefore, ?l parameter is set to match the autocorrelation of hours indata which is 0.69 and 0.66 in Mexico and Canada, respectively. For otherparameters in the calibration of the Canadian economy, we follow similarapproaches done for Mexico above except that we assume a higher labourelasticity (unit elasticity as in the literature) and a lower ? representinga higher level of financial development in Canada. Other parameters forCanada along with Mexican counterparts can be found in Appendix BThe results can be seen in Table 3.3. The first columns for each countryrepresent data moments.59 The baseline model, when calibrated to Canada,can generate a countercyclical labour share since interest rates are slightlyprocyclical in Canada. In response to a positive productivity shock, higherinterest rates mitigate the response of labour demand and wage bill to outputproducing a countercyclical labour share. However, since interest rates donot vary much, the model can only explain a small part of the volatility inlabour share in Canada.On the other hand, the modified version of the model with the adjust-constraints implying procyclical capital share, and consequently countercyclical labourshare with output.58Moreover, the OECD protection index also indicates that Mexico and Turkey havemuch more protection on labour than the average of that among OECD countries.59Refer Appendix B for data details on Canada733.5. Implications for Developed EconomiesTable 3.3: Model Implications for Mexico and CanadaMexico CanadaData Baseline Sluggish Data Baseline SluggishLabour LabourStandard DeviationOutput 2.19 2.25 2.28 1.30 1.02 0.92Labour share 3.58 2.13 1.67 1.05 0.19 0.31Net Exports 2.25 2.01 2.25 0.88 0.21 0.27Standard Deviation(Relative)Wage 1.82 1.10 1.14 0.64 0.49 0.40Hours 0.64 0.65 0.65 0.65 0.49 0.40Consumption 1.35 1.28 1.22 0.72 0.44 0.37Investment 3.45 3.45 3.45 2.55 2.55 2.55Correlationwith OutputLabour Share 0.42 0.68 0.53 -0.62 -0.19 -0.73Interest Rate -0.45 -0.55 -0.49 0.33 0.25 0.25Wage 0.41 0.92 0.92 -0.20 0.99 0.95Hours 0.64 0.92 0.92 0.81 0.99 0.95Consumption 0.89 0.91 0.83 0.69 0.99 0.95Investment 0.94 0.88 0.72 0.72 0.96 0.95Net Exports -0.65 -0.58 -0.30 0.07 0.55 0.68Labour Share and R -0.48 -0.86 -0.89 -0.15 -0.94 -0.12Note: Baseline model is the one with imperfect credit and ? = 0.66. Third and sixth columns addadjustment cost on labour to baseline model. For other details see table 3.2.743.6. Conclusionment cost on labour can explain more variation in the labour share. Sluggishlabour makes the wage bill less responsive to output, and contributes to thecyclicality and volatility of labour share as the previous works in the lit-erature suggest. Therefore, working capital and labour market rigiditieswork in the same direction in explaining the movements of labour share indeveloped markets. Since the interest rates in developed markets are notvolatile, the effect of working capital is minimal. However, emerging mar-kets serve as a good natural experiment with their different behaviour ofinterest rates. In the model economy for the emerging market, introducinglabour market rigidities cannot offset the effect of working capital on labourshare because interest rates are much more volatile than in developed mar-kets. Without working capital, labour share becomes negatively correlatedwith output, which is counterfactual. This suggests that working capitalis the predominant factor in determining the movements of labour share inemerging markets.3.6 ConclusionIn this chapter, we attempt to explain labor share fluctuations in emergingmarkets, namely highly volatile and procyclical labor share. We explorethe effect of financing labour, and show that working capital can be a goodmechanism generating fluctuations in labour share consistent with the data.The liquidity need for labour payments imposes a burden on the cost oflabour, and leads to a more (less) responsive wage bill when interest ratesare countercyclical (procyclical) with output. Since interest rates in differ-ent country groups move in opposite directions over the cycle, the effect ofcost of borrowing varies across these groups implying a procyclical labourshare in emerging markets, and a countercyclical one in developed markets.Introducing other financial problems that emerging economies encounter,such as credit frictions, amplifies the results by making the effective interestrate more volatile than the observed one. Binding leverage constraints notonly contribute to the variability in labour share but also improve the modelperformance in terms of other business cycle regularities in these economies,753.6. Conclusionsuch as highly volatile consumption, strongly countercyclical net exports,and procyclical investment.Following the literature on labour share in developed economies, we alsoinclude an adjustment cost on labour as a representation of the slow ad-justment in the labour market. Without working capital, these models tendto produce counterfactual labour share fluctuations in emerging markets bymaking wage bill less responsive to output. Nonetheless, they can contributeto labour share fluctuations in developed markets. In short, financing labourincome plays an important role on labour share movements in an environ-ment with unstable financial markets associated with highly volatile costsof borrowing. On the other hand, labour market rigidities generating lessresponsive wage bill are more likely to be pronounced than working capitalchannels in countries where financial markets are stable. We leave for futureresearch the task of analyzing the interaction between endogenous rigiditiesspecific to labour market and financing wage payments.76Chapter 4Search Frictions, FinancialFrictions, and LabourMarket Fluctuations inEmerging Markets604.1 IntroductionA number of recent papers have drawn attention to the stylized facts of busi-ness cycles in emerging market economies. In such economies, consumptionis highly volatile and fluctuates more than output while net exports arestrongly countercyclical. Among other papers, Aguiar and Gopinath (2007)attempt to reconcile these findings in the context of a small open econ-omy real business cycle model (SOE-RBC). They argue that for emergingmarkets, the cycle is the trend in that fluctuations in trend productivitycan account for many of the business cycle features. Neumeyer and Perri(2005) introduce countercyclical financial or interest rate shocks in emergingeconomies through a working capital requirement as a propagation mecha-nism. Another difference between developed and emerging economy businesscycles has to do with the behaviour of labour market variables. In contrastto developed economies, real wages are highly volatile and procyclical in sucheconomies, as is the labour share. Boz et al. (2009) examine a model withsearch frictions to account for the labour market findings while Li (2011) and60This chapter benefited from discussions with Sumru Altug, and has become a jointwork with her.774.1. IntroductionChapter 3 of this dissertation seek to rationalize the movements in wagesand labour share with financial frictions in a Walrasian labour market setup.This chapter contributes to this literature by examining variations indifferent margins of labour input: employment (the number of people em-ployed) and hours (worked) per worker. Our motivation for different marginsof labour input stems from the significant variation observed not only in em-ployment but also hours per worker in emerging markets. Figure 4.1 showsthat employment in countries such as Mexico and Turkey fell after the crisesthey experienced, and remained below pre-crisis levels for more than a year.Moreover, hours per worker in these countries displays a U-shaped pattern,too, first declining and then rising over the course of these crises. Not sur-prisingly, these countries experienced increased spreads on their borrowingrates relative to other developed economies and faced difficulties in accessingfunds on the international capital markets, as discussed in Chapter 3. Inour explanation, this behaviour of the cost of borrowing will play a role toexplain the joint comovement between different margins of work.We begin by showing that the variations in the intensive margin of labour(or hours per worker) relative to those in the extensive margin (employment)are more significant in emerging markets than in developed markets. Fur-thermore, both hours per worker and employment are positively correlatedwith output and with each other in emerging economies whereas hours perworker is less cyclical with output and not significantly correlated with em-ployment in developed markets. While changes in employment have beenexamined in the context of search models for emerging economies, previouswork has essentially ignored changes in the intensive margin. We proposea SOE-RBC model that allows for both extensive and intensive margins inthe total hours choice and analyze its implications on cyclical propertiesof labour market variables such as hours per worker, employment and realwages, and goods market variables such as consumption and net exports.The model incorporates a financial friction in the form of a working capitalconstraint and search frictions in the labour market. The financial frictionis motivated by the fact that most firms in emerging economies finance cur-784.1. IntroductionFigure 4.1: Employment and Hours Per Worker during Crises2008?crisis1995?crisisEmployment98100102?2 0 2 4 6quarters1995?crisis2008?crisisHours worked per worker98100102?2 0 2 4 6quartersMexico2008?crisis2001?crisisEmployment95100105?2 0 2 4 6quarters2001?crisis2008?crisisHours worked per worker95100105?2 0 2 4 6quartersTurkeyNote: Data is quarterly and seasonally adjusted. Period zero is the quarter whenthe crisis starts. The value in one quarter before crisis started is scaled to operating costs through short-term bank credit.61 In our framework, asearch-and-matching friction accounts for fluctuations in the extensive mar-gin of work, while Nash bargaining between workers and firms accounts forvariation in the intensive margin along with wages.Our findings suggest once we allow for an endogenous intensive mar-gin, a search-theoretic framework cannot reconcile business cycle puzzlesfor emerging economies. This is due to a strong income effect on hoursworked, which tends to move supply of labour in the same direction asinterest rates, thereby generating counterfactual results for hours and con-sumption. These results are only partially corrected when we utilize alterna-tive forms of preferences which imply a smaller income effect (see Jaimovichand Rebelo (2009)). While search models yield a large response of wages61See Fan et al. (2012) for the importance of short-term debt in emerging economies.794.1. Introductionto the exogenous shocks in the absence of an endogenous hours choice (seeBoz et al. (2009), the smaller income effect tends to lessen this response inthe presence of the endogenous hours choice. It is by allowing for financialfrictions in the form of a working capital requirement that we can explainthe labour market fluctuations. Working capital also generates slightly bet-ter predictions on consumption and net exports.Specifically, the workingcapital requirement tends to amplify the impact of interest rates that workthrough labour demand, in that a higher interest rate depresses the demandfor labour and reduces the hours of work and vacancy postings. Our findingssuggest that there are interactions arising from financial and search frictionsin SOE-RBC models that jointly rationalize the observed responses of thelabour market outcomes together with key macroeconomic time series.This chapter is related to the literature that previously studied the roleof search frictions on aggregate fluctuations in developed markets as well.Andolfatto (1996) and Merz (1995) show that these types of frictions have anamplification effect on labour market variables in closed-economy businesscycle models. Hairault (2002) considers the case of two large open economieswith search and matching frictions. There are also studies that stress therole of differentiating between the intensive and extensive margins in thecontext of search models for developed economies. Yedid-Levi (2009) andMerkl and Wesselbaum (2011) show that hours per worker is a secondarysource of variation in developed economies. Yedid-Levi (2009) further ar-gues that differentiating the different margins of labour is an essential stepfor understanding the comovements across sectors in a model with searchand matching frictions. In a different vein, using a time-varying vector au-toregression framework, Seymen (2011) examines the role of the extensiveversus intensive margins of work in the US and Germany so as to explain theadjustment to cyclical shocks. In a recent paper, Petrosky-Nadeau (2011)combines search frictions with credit imperfections and show that, in a de-veloped market, incentive to hire during credit tightening is lower given thesame benefit to the worker.6262We do not introduce imperfect credit in this chapter so as to concentrate on theinteraction between search and working capital loans.804.2. Fluctuations in the Intensive and Extensive Margins of Labour Input in Emerging MarketsLiterature on intensive margin of labour across developed markets con-centrates on the the role of labour market regulations. Studies such asBurdett and Wright (1989) show that different unemployment compensa-tion schemes may lead to differential responses in employment versus hoursper worker. Specifically, a scheme that allows for partial compensation whenhours are reduced ? as practiced by many European economies ? may yieldlarger variations in the intensive margin relative to a US-type scheme inwhich workers are compensated conditional on being employed or not. InSection 2, we also discuss the labour regulation in emerging markets usingdataset constructed by Botero et al. (2004), and show that despite theirregulatory resemblance with European countries, the movements of labourmargins differ those in Europe. This is why, rather than different insti-tutional frameworks, we study the effect of different financial environmentemerging markets face, such as the higher cost of borrowing during reces-sions, on margins of labour input.The rest of the chapter is organized as follows. Section 2 documents thevolatility and correlation statistics for employment and hours per workerin emerging markets. For comparison, the stylized facts on labour marketfluctuations in developed markets are also included here. Section 3 presentsthe model with both search and financial frictions. Section 4 describes thecalibration strategy, discusses the main findings of the model, and performsrobustness analyses to parameter changes. Section 5 concludes.4.2 Fluctuations in the Intensive and ExtensiveMargins of Labour Input in EmergingMarketsIn this section, we present some evidence regarding variation in the labourinput in emerging economies due to the intensive and extensive margin ofwork. We have data, mostly over the period of 1981-2008, on both marginsof labour input in manufacturing for a set of emerging and developed coun-tries. The data on hours come from industrial surveys in emerging market814.2. Fluctuations in the Intensive and Extensive Margins of Labour Input in Emerging Marketseconomies.63 Hours represent actual hours worked in these surveys, whichis consistent with both national accounts in these economies and the OECDdata for developed economies. Data description and sources are explainedin the Appendix C.We detrend the data using HP-filtering. Since the data is annual, thesmoothing parameter is set to be equal to 6.25. The cyclical properties ofboth margins of labour input are presented in Table 4.1. Specifically, Table4.1 shows the standard deviation of employment and hours per worker as afraction of the standard deviation of real output, the correlation of outputwith employment and hours per worker as well as the correlation of employ-ment and hours per worker in both emerging and developed economies.One of the interesting findings from Table 4.1 is that the relative stan-dard deviations of employment and hours are reversed for emerging versusdeveloped economies. In the developed economies, the variability of em-ployment relative to that of GDP is greater than the variability of hoursper worker relative to that of GDP for almost all countries except Austria,Finland, and Germany. The average value of the relative standard devia-tion of employment is 0.92 versus 0.59 for hours per worker in developedeconomies. By contrast, the difference is not significant for the emergingeconomies. The average value of the relative standard deviation in em-ployment is 0.59 versus 0.64 for hours per worker in emerging economies.64This suggests that labour market behaviour in emerging economies featuressignificant variation in the hours per worker in addition to changes in em-63Some countries report household/labour survey results for the overall economy (seeILO website for this); however, they are often not comparable across countries and overtime. To give a few examples: (i) many countries report hours paid or normal (usual)hours rather than those worked, (ii) some conduct these types of surveys in only one monthduring the year, (iii) labour surveys include many break points which makes it difficultto calculate cyclical components around these points, and (iv) workers surveyed tend toreport overwork (see Mellow and Sider (1983)) which is potentially more problematicduring recessions when hours might be cut. This is why we choose to work on industrial(establishment) surveys in manufacturing. For comparison, we use manufacturing datafor developed markets as well.64Note that the volatility of the intensive margin might be even higher than presentedhere for both groups when we allow for labour utilization, since the effort that each workerputs in during recessions may be lower than in boom times.824.2.FluctuationsintheIntensiveandExtensiveMarginsofLabourInputinEmergingMarketsTable 4.1: Movements in Employment and Hours Worked Per Worker?(e)?(y)?(h)?(y) ?(e, y) ?(h, y) ?(e, h)?(e)?(y)?(h)?(y) ?(e, y) ?(h, y) ?(e, h)EMs DMsArgentina 0.54 0.42 0.83 0.60 0.86 Austria 0.51 0.92 0.68 -0.32 -0.66Brazil 0.72 1.09 0.35 0.57 0.27 Canada 0.71 0.40 0.88 0.43 0.35Columbia 0.51 0.18 0.39 0.36 0.56 Denmark 0.74 0.73 0.73 0.22 -0.01Czech Rep. 0.46 0.70 0.44 0.17 -0.61 Finland 0.40 0.81 0.83 0.56 0.12Hungary 1.01 0.74 -0.03 0.07 0.33 France 1.15 0.92 0.62 -0.08 -0.36Korea 0.67 0.47 0.93 0.77 0.60 Germany 0.36 0.53 0.68 0.06 -0.06Mexico 0.41 0.16 0.73 0.63 0.68 Italy 1.01 0.59 0.45 0.46 0.15Turkey 0.41 1.32 0.55 0.59 0.03 Norway 0.99 0.63 0.54 0.31 -0.16Average 0.59 0.64 0.52 0.47 0.34 Spain 1.34 0.56 0.98 0.12 -0.03Sweden 0.86 0.86 0.47 0.39 -0.20U.S. 0.98 0.64 0.87 0.66 0.38Average 0.82 0.69 0.68 0.26 -0.04Note: The data is HP-filtered using annual smooth parameter, 6.25. The variables are GDP (y), employment (e), and hoursworked per worker (h). See the Appendix C for data sources.834.2. Fluctuations in the Intensive and Extensive Margins of Labour Input in Emerging Marketsployment.65 A second finding from Table 4.1 is the significantly highercorrelation of detrended real GDP with hours per worker for the emergingeconomies. We observe that this quantity is nearly twice as large as that indeveloped economies. Furthermore, the variation is due to countries suchas Argentina, Korea, Mexico and Turkey, where we have typically observedlarge financial shocks and financial crises in the period since the 1980s or1990s.66A third finding from Table 4.1 is that the correlation between employ-ment and hours per worker is also positive in emerging economies and muchlarger than that for the developed economies. On the contrary, there is anegative correlation between employment and hours per worker for manyof the European economies such as Austria, France, Germany, Norway, andSweden.67 These results suggest that the dynamics of labour markets inemerging economies may differ in significant ways from those of developedones. The difference is even more striking when we take into account insti-tutional comparisons. Botero et al. (2004) present a dataset on employmentprotection laws across countries (see Appendix C for a plot of employmentprotection indices across countries examined here). In this dataset, the av-erage employment protection index is 0.45 in emerging markets, a value be-tween the average index for the US and Canada (0.24) and that for Europeancountries (0.67). Yet, emerging economies show even stronger cyclicality ofhours per worker with output and stronger comovement between intensiveand extensive margins than the Anglo-Saxon countries such as Canada and65Since manufacturing industry tends to be more volatile than the aggregate economy,standard deviations of hours per worker for the latter might be lower than presented here.Nevertheless, these results still suggest that the variation in the intensive margin relativeto the extensive margin is more important in emerging markets than in developed ones.66For the case of Argentina and Mexico, the sample period includes the Tequila crisis of1995 as well as the contagionary effects of the 1998 Russian crisis. Moreover, Argentinaexperienced the sovereign debt default of 2002. For Turkey, there are two major financialcrises, the 1994 exchange rate crisis and the 2000-2001 banking and financial crisis. Fora further discussion of the timing of the recessions associated with such crises, see Altug?and Bildirici (2012).67As discussed in Introduction, one reason for the negative correlation between employ-ment and hours in European economies may be the existence of employment protectionlaws.844.2. Fluctuations in the Intensive and Extensive Margins of Labour Input in Emerging MarketsFigure 4.2: Spread during Crises1995?crisis2008?crisis050010001500Spread?2 0 2 4 6quartersMexico2008?crisis2001?crisis2004006008001000Spread?2 0 2 4 6quartersTurkeyNote: CDS data are used for 2008 spreads. This dataset start only after 2002.For earlier crises, we used JP-Morgan EMBI+ spreads from Uribe and Yue (2006)dataset. Period zero is the quarter when the crisis starts.the US in our sample. That is why, rather than institutional differences,we focus on financial differences these countries face over their business cy-cles. As we will show subsequently, financial shocks (countercyclical interestrates) can turn out to be the reason behind such significant comovement ofhours per worker with both output and employment when we allow for afriction in the form of a working capital requirement for firms.In addition to annual data, we present evidence from quarterly dataas well. Figure 4.1 plots the movements of the hours per worker and em-ployment in Mexico and Turkey during both the recent global crisis andtheir own domestic financial crises. The starting dates of domestic crisesare 1994Q4 and 2001Q1 for Mexico and Turkey, respectively. In addition,2008Q4 is chosen to be the starting period for the recent global crisis forboth countries. These are the quarters when GDP declined for the first timeafter a series of quarters with positive growth.68 The evidence from quar-terly data supports the findings using annual data. In particular, Figure4.1 shows that both margins of labour input tend to move similarly, andfollow a U-shape during both crises. Thus, both drop when output drops,68Krugman (1999) also reports 1994Q4 to be the starting period for the balance-of-payment crisis in Mexico.854.2. Fluctuations in the Intensive and Extensive Margins of Labour Input in Emerging MarketsTable 4.2: Correlation between Spread and Labour in Emerging MarketsARG BRA COL HUN KOR MEX TUR?(s, e) -0.60 0.36 -0.27 -0.23 -0.84 -0.76 -0.54?(s, h) -0.48 -0.15 -0.28 0.23 -0.77 -0.69 -0.67?(s, y) -0.75 -0.21 -0.31 -0.38 -0.96 -0.62 -0.93Note: The data is HP-filtered using annual smooth parameter, 6.25. The variablesare GDP (y), employment (e), and hours worked per worker (h). See the AppendixC for data sources.and begin to improve after a couple of quarters.One of the interesting features during periods of fallen output is that bothcountries have tended to experience difficulties in financing their externalbalances. During earlier domestic crises, increases in sovereign risk loweredtheir ability to borrow. In the current crisis, we also observe an increase insovereign risk as measured by their CDS spreads (see Figure 4.2), althoughthe increases are typically less than in previous crises.69 Additionally, Table4.2 documents correlations between these spreads, on the one hand, andemployment and hours per worker, on the other, for emerging economies. Itshows that cyclical properties of these margins with spreads instead of withoutput still suggests strong responses in these margins. Therefore, a modelthat emphasizes the importance of financial shocks on margins of labourinput should be able to explain these movements, as well. We show belowthat the model can generate comovement of such labour market outcomes,not only with output but also with interest rates.69In the 2008 global financial crisis, what may have also affected emerging economiesis the illiquidity in international financial markets, suggesting a decreased capability toborrow at longer horizons. While we do not model features such as borrowing constraintsor borrowing at different horizons, these features would just amplify our results by makingthe effective cost of borrowing even more responsive during crisis.864.3. Model4.3 ModelThis section describes a standard small open economy real business cycle(SOE-RBC) model with shocks to total factor productivity (TFP) and in-terest rates modified to incorporate both search and financial frictions. Weuse a Mortensen-Pissarides type of search and matching framework whichmodels employment, unfilled job vacancies and wage determination explic-itly. In the light of the discussion above, we also incorporate the model witha financial friction, namely, a working capital requirement, which requiresthe firm to pay a fraction of wage bill in advance. Moreover, the only assettraded in international financial markets is a non-state contingent real bond.Households trade in this asset for saving purposes while firms make use ofit for their financing needs.4.3.1 The Firm?s ProblemA continuum of a large number of competitive firms produce a single trad-able good at a world-determined price, which is normalized to one. Out-put is produced by a constant returns to scale production function: yt =Atk?t (ntlt)1?? . For the inputs of production, firms hire labour in the form ofthe number of workers, nt, as well as hours per worker, lt, and rent capital,kt from households. As opposed to applications of the search framework foremerging economies, we allow the intensive margin of the labour input, lt, tovary over time and to be chosen as an outcome of Nash-bargaining problem,the details of which will be given in a later section.There are search frictions in the labour market: firms post a job vacancy,vt and pay a recruiting cost, ?, for each vacancy every period. New matchesare formed according to the matching technology, which is a function ofposted vacancies and nonworking population at the beginning of the period:M(vt, ut) = ?v?t u1??t where ut = 1? (1? ?)nt?1.The individual firm faces a market-driven job filling rate given duringthe recruiting process. We denote the job filling rate by ?(?t) =M(vt,ut)vt,where ?t = vtut is a measure of market tightness, and assume that there is anexogenous separation rate between workers and jobs, ?. Then, employment874.3. Modelevolves according to the following law of motion: nt = (1??)nt?1 +?(?t)vt.According to this law of motion, a vacancy can become productive at thesame period of posting.Along with the labour market friction, firms are also subject to a workingcapital requirement. They need to acquire working capital loans to pay afraction of their wage bill before output is available70 and in order to doso they borrow from abroad at the beginning of the period. This type offriction is widely used in macroeconomic papers since wage payments arean important item in the firms? operating expenses as opposed to capitalpayments (business profits) to household.71 The fraction of the wage billthat has to be paid in advance is denoted by ?. Finally, firms use thesame stochastic discount factor as households for the present value of futureprofits, which is ?t,t+1 = ?uc,t+1/uc,t where uc is the marginal utility ofconsumption, the details of which are given in the next section.Given the wage rate, wt, employment, nt?1, labour supplied per worker,lt, an individual firm chooses how much vacancy to post, vt and how muchcapital to rent, kt, and solves the following dynamic problem:V Ft (nt?1, t) = maxvt,ktyt ? (1 + ?(Rt?1 ? 1))wtntlt ? rtkt ? ?vt+Et?t,t+1VFt+1(nt, t+1)s.t. nt = (1? ?)nt?1 + ?(?t)vt. (4.1)where rt is the rental payment to households and t = [At, Rt] is the ex-ogenous state space, namely, the current values of TFP and the interestrate, Rt, on the internationally-traded bond. Following Neumeyer and Perri(2005), firms pay Rt?1 as interest for working capital loans that are bor-rowed at the beginning of the period before consumption and investmentdecisions are made.7270This could be considered as the equivalent version of having to pay labour before thesales are cashed out in an economy where there is a lag between production and cashingout the sales.71See Christiano and Eichenbaum (1992) and Neumeyer and Perri (2005) for the macroimplications of this type of friction.72Using the current interest rate, Rt, does not change the results since the interest rate884.3. ModelThe optimality condition for vacancies is:??(?t)=?V Ft?nt. (4.2)Firms choose the number of vacancies such that the cost of posting an ad-ditional vacancy equals to the value of filling an additional vacancy condi-tional on the vacancy being filled, where the latter phenomenon occurs witha probability of ?(?).Based on the envelope condition with respect to nt, the marginal valueof an additional worker can be written as follows:V Fn ??V Ft?nt=?yt?nt? (1 + ?(Rt?1 ? 1))wtlt + (1? ?)Et?t,t+1?V Ft+1?nt+1. (4.3)This condition tells us that marginal value of an additional worker is themarginal product of one more worker, ?yt?nt , minus the wage cost includingthe interest payments on working capital plus the asset value of not postinga new vacancy and enjoying the pre-existing relationship with the worker inthe next period. We substitute equation (3) into (2) and obtain:??(?t)=?yt?nt? (1 + ?(Rt?1 ? 1))wtlt + (1? ?)Et?t,t+1??(?t+1)(4.4)Additionally, the first-order condition with respect to capital is:rt = At?k??1t (ntlt)1?? = ?ytkt(4.5)This condition is standard and states that firms borrow capital from house-holds to the extent that marginal product of capital is equal to the rentalrate on capital.4.3.2 The Household?s ProblemThe economy is populated with identical and infinitely-lived households onthe interval [0, 1]. Each household is considered as an extended family whichis persistent.894.3. Modelcontains a continuum of family members endowed with one unit of time.Each member derives utility from consumption ct and leisure 1 ? lt wherethe total time that is devoted to labour and leisure is normalized to one.Members in this family either work and supply lt amount of labour or stayunemployed. Employed members earn wt per hour which is determined byNash bargaining along with the amount of working-hours.The utility function for each member is assumed to be twice-continuously-differentiable and concave in consumption and leisure, and exhibits a con-stant relative risk aversion (CRRA). Here, we explore the effects of both sep-arable and non-separable preferences in terms of consumption and leisure.The aggregate utilities for this family-household are:Separable Preferences (SP) : u(.) = U(ct) + nt?eH(1? lt) + (1? nt)?uH(1)Jaimovich-Rebelo Preferences (JR) : u(.) = ntU(ct ?G(xt, lt)) + (1? nt)U(ct + ?u).where U(ct) = c1??1?? and H(l) =(1?l)1??1?? are the utility derived from con-sumption and leisure respectively, and ? > 0 is the inverse of the elasticityof intertemporal substitution. Parameters, ?e and ?u, govern the utilityderived from leisure relative to that from consumption when the memberis employed and unemployed, respectively. We assume perfect risk-sharingagainst unemployment meaning that all family members pool their incomeand face the same prices for contingent consumption, which implies that themarginal utility in consumption is equated across employed and unemployedfamily members. This implies equal consumption levels for both employedand unemployed members in the case of separable preferences.In addition to the standard separable preferences mostly used in searchpapers, we wish to evaluate results from Jaimovich-Rebelo (JR) preferencesas well. These preferences are non-separable across consumption and leisure:U(ct ? G(xt, lt)) =(ct??el?t xt)1???11?? , ? > 1, ?e > 0 if the family memberis employed. The second term in the utility function G(xt, lt) expressesthe disutility of labour which is twice-continuously-differentiable and convexfunction in hours per worker. In addition, xt = c?t x1??t?1 determines thestrength of income effect on labour decisions depending on the parameter ?.904.3. ModelNote that when ? = 0 these preferences show the same characteristics as inpreferences discussed in Greenwood, Hercowitz and Huffman (1988) (GHHhenceforth). These types of preferences eliminate income effect on laboursupply. They are very common in the literature since they tend to generatea more realistic labour movements in open economies.73 The issue is morecrucial for emerging economies where we observe high volatility in interestrates, having a potential to produce larger wealth effects on labour supply.74This is why we analyze the model here under two different preferences.Lastly, the utility function for the unemployed member is assumed to beU(ct+?u) =(ct+?u)1???11?? , where ?u > 0 denotes the minimum consumptionlevel for an unemployed worker.In the model, household also supply capital to firms, kt, at a rental raterkt . In addition to labour and capital income, they earn interest from previ-ous period?s savings, Rt?1bt?1, and get dividends from firms, pit. Given thewage rate, wt, the rental rate of capital, rt, hours per worker, lt, previous pe-riod employment, nt?1, the interest rates on bond, Rt?1, and the probabilityof finding a job, ?t, the household chooses consumption, ct, investment, it,and bond holdings, bt, to solve the following dynamic problem if preferencesare separable:V Ht (kt, bt?1, nt?1, t) = maxct,it,btu(.) + ?EtVHt (kt+1, bt, nt, t+1)s.t. ct + it + bt + ?(bt) = ntwtlt + rkt kt +Rt?1bt?1 + pit (4.6)it = kt+1 ? (1? ?)kt + ?(kt+1, kt) (4.7)nt = (1? ?)nt?1 + ?(?t)utwhere ?(?t) ?M(.)utdenotes the probability of finding a job. Quadraticconvex cost functions, ?(.) and ?(.) make bond holdings and adjustmentsin investment costly. These are standard in SOE-RBC studies to makesure that the model exhibits stationary properties, particularly to solve the73See, for instance, Devereux et al. (1992) and Hairault (2002).74See Neumeyer and Perri (2005), Mendoza (2010), and Li (2011) for further discussionson the wealth effect.914.3. Modelunit-root problem for bond holdings and to prevent excessive investment.75Lastly, the capital stock depreciates at the rate of ? every period.In the case of JR preferences, the household solves a similar problemexcept that the budget constraint must be modified to separately accountfor the consumption of the employed and unemployed family members. Themodified budget constraint can be written as:ntcet + (1? nt)cut + it + bt + ?(bt) = ntwtlt + rkt kt +Rt?1bt?1 + pit (4.8)Furthermore, the agent now chooses the optimal xt subject to an additionalequality constraint as: xt = c?t x1??t?1 .Based on the envelope condition with respect to nt, we obtain the value ofan additional worker to the household, depending on the type of preferences,as:V Hn ??V Ht?nt=???(ue ? uu) + ?twtlt + ?Et?V Ht+1?nt+1(1? ?)(1? ?(?t)) (SP )(ue ? uu) + ?t[wtlt ? cet + cut ] + ?Et?V Ht+1?nt+1(1? ?)(1? ?(?t)) (JR)(4.9)where ue and uu denotes the utilities for employed and unemployed familymembers respectively, and ?t represents marginal utility of consumption ofa family member. For both the separable and non-separable cases, the firstterm illustrates the net utility loss from being unemployed relative to beingemployed, conditional on the level of the consumption. The second termdiffers depending on the nature of preferences: in the separable case, thesecond term shows the marginal value of being employed in terms of thevalue of wage payments. By contrast, in the non-separable case, the addi-tional term reflects the idea that when a family member becomes employed,s/he is entitled to a different amount of consumption and hence, there is anexpenditure difference on consumption (See, for instance, Hall and Milgrom(2008)). Finally, the last term is the discounted expected future marginalvalue of future employment, where the weighting factor reflects the probabil-ity of still being employed in the next period, 1??, as well as the probability75See Schmitt-Grohe and Uribe (2003) for more details.924.3. Modelof finding a job when separation occurs, ??(?t), plus the opportunity costof taking up employment, ?(?t). In the next section, we derive the solu-tion to the Nash bargaining problem between the household and the firm,which is based on their relative valuations of additional employment shownin equations (4.3) and (4.9).The solution to the household?s problem also yields the Euler equationsfor the optimal bond holdings and capital accumulation as in the standardSOE-RBC models:1 +??(bt)?bt= ?Et[?t+1?tRt](4.10)1 +??(kt, kt+1)?kt+1= ?Et[?t+1?t(1 + rt+1 ? ? ???(kt+1,kt+2)?kt+1)]. (4.11)These conditions tell us that bond holdings and capital are at their optimallevel when the marginal cost of additional bonds (or capital) is equal to theirexpected discounted value to households.Before ending this section, we note that the different preference specifica-tions have implications for the behaviour of the marginal utility of consump-tion. When preferences are separable, the marginal utility of consumptionfor each family member is given by:?t = c??t , (4.12)which implies equal consumption across employed and unemployed house-hold members since we assume that there is perfect risk sharing within thefamily. However, in the case of JR preferences, equal marginal utilities donot necessarily imply equal consumption levels across the employed and un-employed since labour enters the marginal utility of consumption as?t = (cet ? ?el?t xt)?? + ?t?xtct(4.13)?t = (cut + ?u)??. (4.14)If we denote the marginal value of the consumption habit by ?t, then theoptimality condition for the consumption stock, xt, in JR preferences states934.3. Modelthat:(cet ? ?el?t xt)???el?t + ?t = ?Et[(1? ?)?t+1xt+1xt]. (4.15)4.3.3 Nash BargainingAfter the vacancy is filled, wages and working hours per worker are setthrough a Nash Bargaining game between the firm and the worker as in thefollowing manner:(wt, lt) = arg maxw,l(V Hn?t)?(V Fn)1??where ? is the bargaining power for the worker. The problem above is sub-ject to the value of an additional worker to the firm and household derivedearlier as equations (3) and (9), respectively. Here, V Hn is divided by ?t inorder to express everything in terms of consumption units.Taking the derivative with respect to wages, we have the sharing rulebetween the firm and the household, which states that the total matchingsurplus is shared between parties according to their bargaining power:V Hn /?tV Fn=?1? ?. (4.16)We can also obtain the condition for the optimal level of working hoursby taking the derivative of the problem above with respect to hours perworker:?ul?t=ynh(1 + ?(Rt?1 ? 1))(4.17)where ul ? ?u?l and ynl ??y?nl where nl is the total labour input. Thisequation implies that at the optimal point the marginal loss of increasingone more labour hour in units of the consumption good has to be equalto the value of additional product the firm earns. Note that an additionallabour-hour increases the value of production less than it would be if therewere no working capital requirement because of the interest burden on thefirms.944.4. Quantitative Analysis4.3.4 Equilibrium Prices and AllocationGiven the initial conditions and a sequence of exogenous interest rates, Rt,and At, a search equilibrium consists of a sequence of a state-contingent se-quence of allocations {ct[cet , cut , xt], lt, kt+1, it, bt, nt, vt} and of prices {wt, rt}such that(i) the allocations solve the firm and household problems at the equilibriumprices,(ii) the Nash Bargaining solutions are satisfied.(iii) The market for capital clears, i.e, firms? capital demand is equal to thecapital supplied from households: kdemandt = ksupplyt = kt.(iv) Goods markets clear:ct + it + nxt + ?vt + ?(bt) = yt (4.18)which implies that the goods that are not spent on consumption, investment,the cost of recruiting activities or of bond holdings represent the net exportfor the economy, nxt.4.4 Quantitative AnalysisThe model is solved by log-linearizing the equilibrium conditions aroundthe steady state,76 which is then parameterized so that the deterministicsteady state of the model matches several average ratios of macroeconomicaggregates of the Mexican economy. The period in the calibration is 1993Q1-2008Q4 for which we have quarterly data from OECD and Mexican NationalStatistics, (INEGI). The summary of parameter values from calibration canbe seen in Table 4.3.76We use the Dynare routine to solve the log-linearized equilibrium conditions. SeeAdjemian et al. (2011).954.4. Quantitative Analysis4.4.1 Calibration of Parameter ValuesParameter values. The values of the quarterly depreciation rate ? andthe investment-output ratio, i/y, determine the value of capital-output ra-tio, k?/y?. The optimality condition for capital demand and the arbitragecondition for bond and capital holdings at the steady state, R? = 1 + r?k ? ?,then yield the value of the real interest rate. We set ? = 0.025 and i/y = 0.19using Mexican data. Setting the capital share as ? = 0.36 together with theremaining parameters yields a value for the capital-output ratio as k?/y? = 7.6.This ratio is equal to 1.9 at the annual frequency and is close to the annualfinding for Mexico (2.09) in Nehru and Dhareshwar (1993) using historicalover the period of 1950-1990. Given the capital-output ratio, the steadystate interest rate is calculated as 2.24%.For the parameters of preferences, the coefficient of relative risk aversion,?, is set equal to 2. We then calibrate the parameters for the elasticity oflabour supply in separable and JR preferences, ? and ?, respectively, so thatthe model matches a Frisch elasticity of labour supply of 0.5, which is inthe range given by Blundell and Macurdy (1999)).77 This implies parametervalues for ? and ? equal to 1.77 and 2.66. Hours per worker at the steadystate, l?, is set to 0.53 which is the ratio of total hours worked and non-sleeping hours per employee in Mexico. The leisure weight coefficients, ?eand ?u, for separable preferences are determined using the optimality condi-tion for hours worked and optimal wage equation at the steady state. Theseparameters affect the consumption ratios across employed and unemployedagents in JR preferences. The previous literature based on evidence for theUS has suggested that the unemployed have a 15% lower consumption thanthe employed.78 We set ?e in JR preferences to 1.29 to match this ratio be-tween the employed and the unemployed. The parameter, ?u, is then foundto be -0.01 using the equality of marginal utility of consumption across theemployed and the unemployed. The parameter that governs the strength of77Blundell and Macurdy (1999) estimate the elasticity of labour supply to be in therange of [0.5,1] for the US. We used the low end for Mexico as incomes are much lowerthan in the US.78See Hall and Milgrom (2008), Shimer (2009) and Hall (2009).964.4. Quantitative Analysisthe income effect, ?, is assumed to be 0.5, which is the mid-point of feasiblevalues.For the matching parameters, the natural breakup rate, ?, is set equal to0.06 following the estimates in Bosch and Maloney (2008) for Mexico. Thesteady state employment is chosen equal to one minus unemployment rate(3.6%) in Mexico over the sample period. The steady state value of matches,M? , follows from M? = ?n?. We then calculate the vacancy rate at the steadystate, v?, assuming the job filling rate to be 0.7 following Boz et al (2009),which implies an average vacancy duration of 45 days. Following Andolfatto(1996), the matching exponent, ?, is assumed to be 0.5. Using the steadystate values for matches, employment and vacancies together with the valueof ?, the matching efficiency parameter, ?, is obtained as 0.66. We assumean equal bargaining power (? = 0.5) between the firm and the worker dueto lack of empirical evidence for Mexico. The recruiting cost parameter,?, is then calculated as 0.78 so that the wage equation implies an equalbargaining power. This suggests recruiting costs, ?v?, to be around 4% ofoutput.The working capital parameter, ? is assumed to be 1 as in Neumeyer andPerri (2005), which implies that workers have to be paid three months beforethe sales are cashed out.79 The net foreign asset ratio held by households atthe steady state, b?y? , is calculated using the debt ratios from Lane and Milesi-Ferretti (2007) that are estimated from the external wealth of countries. Theratio of net foreign assets to GDP for Mexico in this dataset is equal to -0.40.Note that, in our model, net foreign assets is the household?s foreign bondholdings net of working capital loans of the country: bt??wtlt. Accordingly,the ratio b?y? is calculated as 0.17. The ratio of economic profit and outputis found to be 0.0028 in the model, which is different than zero due to the79As we discussed in Chapter 3, working capital requirement might be less importantfor some sectors of the economy, such as the self-employed. That is why we set a value lessthan one in that chapter. However, note that the model in 3 had other financial friction,such as a collateral constraint; it is shown that the introduction of this constraint hasimplications similar to those of a higher working capital requirement. Therefore, giventhat we do not have any type of collateral constraints here, we believe that the assumedvalue for ? could be reasonable.974.4. Quantitative Analysisfrictions in the labour market. Using the steady state values of profits,household earnings from output, investment, bond holdings and interestrate, we calculate the consumption-output ratio at the steady state as 0.76,which is close to the Mexican consumption-output ratio of 0.75.Finally, the functional form for the quadratic convex cost functions forbond holdings and capital adjustments can be written as ?(bt) =?b2 yt(btyt?b?y? )2 and ?(kt, kt+1) =?k2 kt(kt+1?ktkt)2 following the literature in small openeconomies. The cost parameter for bond holdings, ?b, is set to be as smallas 0.01 so that it does not change the business cycle volatilities but ensuresthat the model is stationary. For the parameter, ?k, we follow the estimatesin the literature using similar functional forms and set it to 25, which impliesan investment volatility nearly three times output volatility as observed inthe data for Mexico.Shock Processes. The recent literature has shown the importance ofshocks to interest rate in the fluctuations in emerging markets.80 Emergingmarkets differ from developed markets in terms of the behaviour of the in-terest rate they face. Interest rates are countercyclical mainly because of de-fault risk that is negatively correlated with the output. Following this liter-ature, we assume that shocks to productivity (in logs) and interest rates (logof gross interest rate) are correlated simultaneously such that t = [At , Rt ] isdrawn from an i.i.d normal bivariate distribution, N(0,?), with zero meanand covariance, ?. Each shock follows an independent AR(1) process:A?t = ?AA?t?1 + A,tR?t = ?RR?t?1 + R,t,with?t?t =(?A ?A,R?A?R?A,R?A?R ?R).80See Neumeyer and Perri (2005), Arellano 2008), Mendoza (2010) on the role of interestrates in the output fluctuations in emerging markets.984.4. Quantitative AnalysisWe construct Solow residuals for the TFP shocks over the sample periodin Mexico using the seasonally adjusted real GDP, total hours worked andcapital stock series. We have data on real GDP from the OECD, and totalhours worked and employment in manufacturing from INEGI. In order tocalculate hours worked in the overall economy, we first divide hours workedin manufacturing by the employment in manufacturing and then multiply itby total employment each quarter in Mexico. We take historical employmentseries from Neumeyer and Perri (2005) and extend it to 2008Q4 using thethe growth of employment series from ILO.The capital stock series is constructed using seasonally-adjusted quar-terly investment series from the OECD and the depreciation rate.81 Usingseries on the capital stock and the labour input, we now calculate the Solowresiduals as lnAt = lnYt? ?lnKt? (1? ?)lnLt. The HP-filtered series yieldsthe persistence as ?A = 0.80 and the standard deviation as ?A = 1.1%.The persistence is quite lower than the counterparts for developed coun-tries. This raises a question about measurement problems in constructingthese series. That is why we will also consider different shock processes inthe sensitivity analysis in order to check if the results are robust to morepersistent shocks.For the interest rates, we have two different representative costs of bor-rowing: EMBI+ dataset from Uribe and Yeu (2006) and real domestic inter-est rates on Mexican T-bills from IFS. EMBI+ dataset documents spreadsfor traded debt instruments for various countries including Mexico. In Chap-ter 3, we show that the interest rates constructed using these spreads showa significantly smaller volatility than that of domestic rates. In fact, thestandard deviation of Mexican rates from EMBI is around 0.55% at quar-terly levels whereas the domestic rates are almost four times more volatile,2%.82 Although EMBI+ rates have been widely used in the literature as the81Initial capital-output ratio is assumed to be equal to 10 so that the average ratioover sample period corresponds to 7.6 as in the model. Note that Mexico experienced aneconomic boom in the initial year, 1993 ? the year before the 1994 crisis.82Here, it is crucial to take the quarterly yields on these bonds since gross interest rates,not net ones, are logged. Therefore, taking the annualized value at the quarterly frequencywill mistakenly increase the volatility of interest rates by four times.994.4. Quantitative Analysisrepresentative cost of borrowing, this discrepancy between two rates raisesquestions on the other potential costs of external borrowing which EMBI+ignores.83 That is why we take the volatility of EMBI+ rates as the lowerbound of the volatility of the cost of borrowing and that of domestic ratesas the upper bound. When we assume that shocks to interest rates havethe same volatility as the productivity shocks, ?A = ?R = 1.1%, the interestrates in the model represent a volatility very close to the average of standarddeviations of the two interest rate series.For the persistence of the shock to interest rate, ?R, domestic and EMBI-constructed rates show autocorrelation coefficients of 0.59 and 0.68, respec-tively. In our calibration, we set this parameter equal to 0.64, the average ofthose autocorrelation coefficients. We set the correlation parameter betweenshocks to productivity and interest rates, ?A,R , to the average correlationbetween interest rate and TFP from two series of interest rates, -0.58. WithEMBI+ rates this correlation is -0.56 and with domestic rates, the correla-tion is - Characterization of Equilibrium and ImpulseResponsesWithout Working Capital. We now discuss the role of search and finan-cial frictions and how endogenous decisions at the intensive margin affectthe amplification that such frictions generate. We start by discussing thecase with only search frictions to show their sole effects on the fluctuationsas a response to shocks to productivity and interest rates. The key equationin the models with search frictions is the wage equation. Regardless of theseparability in preferences, we can write down the optimal wage equation by83These might include the limited and varying access to financial markets, exchange-raterisk exposure in the eyes of domestic agents, strategic issuance of bonds and withdrawalsfrom financial markets. For example, recently, Fostel and Geanakoplos (2008) shows thatan emerging market asset can have leverage cycles because of asymmetric informationproblems in the financial markets even when the price of the debt does not change forthat particular assets in international markets.1004.4.QuantitativeAnalysisTable 4.3: Parameter ValuesParameter Value Source Parameter Value SourceI. Preferences III. ProductionDiscount factor, ? 0.98 calibrated Capital exponent, ? 0.36 literatureRelative risk aversion, ? 2 literature Working capital, ? 1 assumedSeparable Pref.Labour curvature, ? 1.77 calibrated IV. ShocksCoef. of leisure (emp.), ?e 0.36 calibrated Persistence, ?A 0.80 estimatedCoef. of leisure (unemp.), ?u -1.36 calibrated Persistence , ?R 0.64 estimatedJR Pref. Correlation coef., ?A,R -0.58 estimatedLabour curvature, ? 2.6 calibrated Std. deviation, ?A 0.011 estimatedCoef. of leisure (emp.), ?e 1.29 calibrated Std. deviation, ?R 0.011 estimatedCoef. of leisure (unemp.), ?u -0.01 calibratedIncome effect, ? 0.50 assumedII. Search V. OtherElas. of job matches, ? 0.5 literature Bond holding cost, ?b 0.01 literatureMatching efficiency, ? 0.66 calibrated Capital adj. cost, ?k 25 literatureCost of posting vacancy, ? 0.75 calibrated Depreciation rate, ? 0.025 literatureJob separation rate, ? 0.06 BM (2008) SS bond holdings, b?y? -0.40 dataBargaining power, ? 0.5 assumed1014.4. Quantitative Analysiscombining the share rule in equation (16) with equations (2), (3) and (9):wtlt =11 + ??(Rt?1 ? 1){?(yn + ?Et?t,t+1?t+1) + (1? ?)uu ? ueuc},(4.19)where we have substituted for ?t in equation (8) using the marginal utilityof consumption as uc.The above condition tells us that labour income of the employed (wagebill per job) is a combination of the worker?s contribution to output atthe margin ? which is the marginal productivity and the average savingsin vacancy costs ? and the worker?s outside option, which is the utility offoregone leisure evaluated in terms of the marginal utility of consumption,depending on the bargaining power of the household. The second term whichmakes the wage equation different from that in the standard RBC modelbecomes much more important in a model with interest rate shocks. As aresponse to a negative productivity shock, for example, interest rates tendto increase, which causes a larger drop in consumption than in the case withjust productivity shocks84. As a result, the wage bill falls not only becausemarginal productivity decreases but also because the worker?s outside optiondrops due to a higher marginal utility of consumption. In other words, theexpected value of staying unemployed and searching for a job in the nextperiod becomes smaller.However, note that the amplification of this mechanism on wages de-pends on the changes in the hours per worker, lt. Movements in the hoursper worker will affect the fluctuations in wages through not only their effecton the wage bill on marginal consumption. In order to illustrate this point,we now explore the optimal decision for the intensive margin. The log-linearized version of equation (16) for each form of utility can be expressed84See Neumeyer and Perri (2005) for details on the effect of interest rates on consumption1024.4. Quantitative AnalysisFigure 4.3: Impulse Responses0 5 10 15012YTo At Shocks0 5 10 15012W0 5 10 15?0.500.5L0 5 10 15?101To Rt Shocks0 5 10 15?2?100 5 10 15?202  Fixed lSeparable prefJR prefBaseline1034.4. Quantitative Analysis...continued0 5 10 1500.05NTo At Shocks0 5 10 1500.51C0 5 10 15?101NXY0 5 10 15?0.1?0.050To Rt Shocks0 5 10 15?2020 5 10 15?505  Fixed lSeparable prefJR prefBaseline1044.4. Quantitative Analysisas below:l?t =A?t + ?k?t ? ?nt ? ?c?t?l?1?l?+ ?(separable preferences) (4.20)l?t =A?t + ?k?t ? ?nt ? x?t? ? 1 + ?(JR preferences) (4.21)The difference in these equations comes from the strength of the incomeeffect they have on the labour supply decisions. This effect is captured by?c?t in separable preferences and is stronger than in JR preferences wherex?t = ?c?t + (1 ? ?)x?t?1. We can see how this income effect changes hoursdecisions across different preferences in Figure 4.3, which plots impulse re-sponses of selected variables to one standard error shock in TFP and interestrates. These impulse responses show us that hours per worker increases as aresponse to a shock to interest rates in the model with separable preferences(Figure 4.3, blue line). This also generates an increase in output when theinterest rate goes up. This result is inconsistent with the data where bothoutput and hours per worker tend to be decreasing during the periods withhigher interest rates in emerging markets as discussed in section 2.Note that this counterfactual result makes wages more responsive to in-terest rates. In order for the wage bill to fall, the wage has to decreasesharply as a response to interest-rate shocks. The wage response is still highwhen we take the intensive margin as fixed (Figure 4.3, dotted line). Thus,the wage responds more in these cases at the expense of movements in theintensive margin. On the other hand, when we take JR preferences (Figure4.3, black line), which have a smaller income effect on the intensive margin,responses of hours per worker improve both to productivity and interest rateshocks. This improvement in hours come with a cost: the wage responsebecomes smaller to productivity shocks. Therefore, intensive margin playsan important role in the contribution of search frictions on wage fluctuations.With Working Capital. Although JR preferences reduce the incomeeffect on labour supply, the impulse responses still show a slight increase in1054.4. Quantitative Analysishours per worker as a response to a positive shock in interest rates. One canfurther decrease the strength of the income effect, however; as we will showin the sensitivity analyses, this will result in a large drop in the responseof wages. Instead, here we show that the presence of a financial friction,namely working capital, dominates the income effect on labour supply andpredicts the right movements for hours per worker. In the presence of aworking capital requirement, equation (20) can be rewritten as:l?t =A?t + ?k?t ? ?nt ? x?t ??(R??1)1+?(R??1)R?t? ? 1 + ?Since ? > 0, the above equation implies that a positive shock to interestrates will have a negative effect on hours per worker. This suggests thatdemand for labour in the model economy as a response to a positive interestrate shock is lower in the case with working capital than before. In fact,impulse responses in the presence of working capital, which will be denotedas ?the baseline model? henceforth, show that hours per worker tends todecrease as a response to an interest rate shock in the second period (Figure4.3, red line). This generates an output drop after an increase in the interestrate consistent with the data. The model with working capital also increasesthe responsiveness of wages to interest rates significantly. This result canbe seen through the wage equation in the model with working capital. If wedenote w1t l1t as the wage bill in equation (18), the wage bill in the presenceof working capital, w2t l2t will be:w2t l2t =w1t l1t1 + ??(Rt?1 ? 1)Therefore, introducing working capital not only affects the composition ofthe wage bill between wages and hours per worker but also makes labourincome more responsive to shocks. As a result, wages become more volatilewithout sacrificing movements in hours per worker.For the other impulse responses, it is worth to remark the responsesof employment and goods market variables, such as consumption and net1064.4. Quantitative Analysisexports, as well. The baseline scenario produces the most amplified responseof employment among others due to the effect of cost of borrowing on labourdemand, and consequently on vacancy decisions. On the other hand, thepreferences and frictions discussed here play a smaller role on goods market(consumption and net exports) than labour market variables, except thatconsumption responses less to a TFP shock under separable preferences andfixed intensive margin because less responsive employment and hours perworker generate a smaller output.4.4.3 Quantitative ResultsWe summarize the results from simulations of various versions of the modelin Tables 2 and 3. We first simulate the model with only search frictionswith a fixed intensive margin as in Boz et al (2009). We, then, continue withresults from endogenizing intensive margin for both types of preferences. Indoing this, our aim is to analyze how much search frictions contribute tofluctuations in emerging markets discussed in earlier sections, particularlythe positive correlation between intensive margin and extensive margin oflabour, highly volatile real wages and labour share, procyclical wages, labourinput and labour share, highly volatile consumption and countercyclical netexports. We, lastly, simulate the model with working capital to investigatethe effect of a financial friction on these fluctuations over search frictions.Fixed and Endogenous Intensive Margin. The first column in Table4.4 shows the business cycle moments in data for Mexico. The second andthird columns, then, report the implications of the standard frictionlessSOE-RBC model and the search model with a fixed intensive margin andseparable preferences. When hours per worker is taken constant as in Boz etal (2009), search model can explain some distinguishable characteristics ofthe fluctuations in emerging markets such as countercyclical interest rates,very procyclical and highly volatile consumption and countercyclical netexports. However, SOE-RBC alone, with fixed labour input does a similarjob in terms of these fluctuations. The contribution of search frictions on1074.4. Quantitative Analysisthese fluctuations is then minimal.Table 4.4 shows that the search model has a stronger effect on the re-sponse of labour market variables compared to the standard SOE-RBC.These findings are similar with those of Andalfotto (1996) and Merz (1995),finding significant contribution of search frictions to the fluctuations in thelabour market rather than goods market. For instance, the third columnshows a significant increase in wage volatility compared to the one in SOE-RBC. However, as we have shown above, wages here are the only variableto respond in the wage bill per job since the intensive margin is fixed. Nev-ertheless, the search model gives a momentum to the labour share whereasthis is constant in the frictionless SOE-RBC model.Although search frictions seem to amplify the responses of labour mar-ket variables, particularly wages and labour share under separable utility,the results change considerably when we let the intensive margin respondendogenously. The last column shows that the model implies a large incomeeffect and consequently a very strong positive correlation between interestrates and hours per worker, which is the opposite of what we observe in data.This makes interest rates no longer countercyclical even though productiv-ity and interest rate shocks are negatively correlated because an increasein hours per worker as a response to an increase in interest rates tends tohave a positive impact on output. As a result, a large number of cyclicalproperties of the search model fail to predict the ones in the data. Sincetotal hours tend to be weakly correlated with output, labour share becomesacyclical in this model whereas it is significantly procyclical in the data. Inaddition, wages become less cyclical and slightly more volatile at the expenseof counterfactual movements in the intensive margin. These failures withseparable preferences shows that it is important to understand the intensivemargin decisions in order to assess the contribution of search frictions to thefluctuations in emerging economies.Baseline Model. The results with JR preferences can be seen in Table4.5. Our baseline model (the second column) that has JR preferences andworking capital improves the implications of the model significantly. Witha reduced income effect due to JR preferences and a more effective labour1084.4. Quantitative AnalysisTable 4.4: Results with Separable PreferencesData RBC Search SearchFixed l Fixed l Endo. lStandard DeviationOutput 2.39 1.34 1.36 1.31Net Exports 1.10 1.65 1.67 2.24Labour Share 3.56 0.0 1.11 0.76Standard Deviation(Relative)Wage 1.79 1.00 1.57 1.58Hours per worker 0.24 0.0 0.0 0.81Employment 0.43 0.0 0.07 0.08Total hours 0.61 0.0 0.07 0.76Consumption 1.03 1.25 1.23 1.21Correlation with YInt. Rate -0.54 -0.56 -0.56 -0.10Wage 0.40 1.0 0.91 0.55Hours per worker 0.58 0.0 0.0 0.13Employment 0.57 0.0 0.63 0.57Total hours 0.68 0.0 0.63 0.20Labour Share 0.47 0.0 0.55 0.04Consumption 0.92 0.69 0.70 0.26Net Exports -0.81 -0.12 -0.13 0.33Correlation with RHours per worker ,l -0.28 0.0 0.0 0.96Employment, n -0.48 0.0 -0.88 -0.83Corr(n, l) 0.48 0.0 0.0 -0.68Note: This table shows the results of endogenizing intensive margin under separablepreferences with only search frictions. For data sources, see the Appendix C.1094.4. Quantitative Analysisdemand due to working capital, the model can mimic the cyclical propertiesof the variables that are related to the labour market. As well, the modeldoes a better job on the cyclical properties of consumption and net exportsrelative to the alternatives.In the baseline model, the comovement between hours per worker andinterest rates has the right sign. Consequently, interest rates are counter-cyclical with output. In addition, since both margins of the work similarlyresponds to both shocks in the model, hours per worker and employmentare positively correlated with each other, as in the data that we displayedin Table 4.1. This is in contrast to the implications from the model withseparable preferences discussed earlier. For more comparisons, we includethe results from the frictionless SOE-RBC model (third column) and the re-sults from the model with only search frictions; i.e.; without working capital(fourth column). The baseline model performs better in terms of gener-ating countercyclical interest rates. In addition, without working capital,the model cannot explain the strong comovement between hours per workerand the number of people employed, mainly because it fails to explain theresponse of the hours per worker to interest rates even under JR preferencethat implies a smaller income effect. Thus, although a smaller income ef-fect can partially improve the correlation between the margins of the work;without the financial requirement ? working capital ? it is not enough toexplain the comovement we observe in the data.85The high volatility of wages in emerging markets is noted in Chapter 2 ofthis dissertation. The baseline model here can explain such highly volatilewages thanks to the presence of both search and financial frictions. In termsof the comovement with output, the model overpredicts the procyclicalityof wages with output. However, we should mention that they become moreprocyclical with current output in the data when the leads of wages areconsidered. For instance, the correlation between current output and wages85The reader might question the contribution of working capital as opposed to justlowering the strength of income effect further. Hence, to address this question, a discussionon the sensitivity of the parameter that governs the strength of income effect,?, will followthis section.1104.4. Quantitative AnalysisTable 4.5: Results with JR PreferencesData Baseline RBC Search UncorrelatedModel Only Shocks? = 0 ?A,R = 0Standard DeviationOutput 2.39 1.64 1.58 1.52 1.63Net Exports 1.10 1.55 1.56 1.55 2.10Labour Share 3.56 1.73 0.0 0.84 1.80Standard Deviation(Relative)Wage 1.79 1.54 0.78 1.19 1.27Hours per worker 0.24 0.27 0.40 0.24 0.32Employment 0.43 0.10 0.0 0.07 0.07Total hours 0.61 0.34 0.40 0.26 0.36Consumption 1.03 0.96 0.89 0.98 0.82Correlation with YInt. Rate -0.54 -0.52 -0.41 -0.47 0.03Wage 0.40 0.81 0.93 0.84 0.50Hours per worker 0.58 0.81 0.69 0.58 0.83Employment 0.57 0.83 0.0 0.84 0.58Total hours 0.68 0.88 0.69 0.77 0.86Labour Share 0.47 0.52 0.0 0.37 -0.05Consumption 0.92 0.77 0.74 0.73 0.42Net Exports -0.81 0.01 0.15 0.03 0.48Correlation with RHours per worker, l -0.28 -0.21 0.20 0.26 0.27Employment, n -0.48 -0.76 0.0 -0.78 -0.66Corr(n, l) 0.48 0.64 0.0 0.07 0.40JR preferences are used in all models here. The baseline model has both searchfrictions and working capital with the calibrated parameters described in the text.?Search Only? stands for the baseline model without working capital (? = 0). Thelast column lists the results from the baseline model with zero correlation betweenproductivity and interest rate shocks.1114.4. Quantitative AnalysisTable 4.6: More Sensitivity AnalysesData Baseline Full Income Zero Income MoreModel Effect Effect Volatile? = 1 ? = 0 ShocksStandard DeviationOutput 2.39 1.64 1.56 2.03 2.39Net Exports 1.10 1.55 1.77 1.41 2.28Labour Share 3.56 1.73 1.71 1.73 2.40Standard Deviation(Relative)Wage 1.79 1.54 1.65 1.13 1.67Hours per worker 0.24 0.27 0.29 0.52 0.26Employment 0.43 0.10 0.10 0.08 0.10Total hours 0.61 0.34 0.35 0.60 0.29Consumption 1.03 0.96 1.07 0.96 1.12Correlation with YInt. Rate -0.54 -0.52 -0.47 -0.60 -0.54Wage 0.40 0.81 0.79 0.86 0.87Hours per worker 0.58 0.81 0.57 0.94 0.81Employment 0.57 0.83 0.0 0.89 0.82Total hours 0.68 0.88 0.71 0.94 0.84Labour Share 0.47 0.52 0.50 0.62 0.60Consumption 0.92 0.77 0.69 0.89 0.87Net Exports -0.81 0.01 0.03 -0.07 -0.18Correlation with RHours per worker, l -0.28 -0.21 0.23 -0.62 -0.35Employment, n -0.48 -0.76 -0.77 -0.76 -0.88Corr(n, l) 0.48 0.64 0.41 0.95 0.71JR preferences are used in all models here. The baseline model has both searchfrictions and working capital with the calibrated parameters described in the text.The third and fourth column show the results from different values for the param-eter, ?, that governs the strength of income effect. ?Search Only? stands for thebaseline model without working capital (? = 0), and the last column presents theimplications from the baseline model with alternative shock process described inthe text.1124.4. Quantitative Analysisafter two quarters is 0.56 compared to 0.40, contemporaneous correlation.We leave this lagging property of wages for future research and focus onthe volatility in wages in this work since they present a strikingly highervolatility in emerging markets than in developed markets.The baseline model can also address a large fraction of the volatility inconsumption relative to output volatility. Consumption is more volatile thanoutput and very procyclical in emerging markets, which tend to generatecountercyclical net exports in these economies. The model predicts a morevolatile and a more procyclical consumption than the frictionless model,consistent with data. As a result, the cyclical properties of net exportsare closer to data than the frictionless model predicts, albeit still acyclical.Compared to alternatives, the cyclical properties of these series improve inthe baseline model. We show in 3 that borrowing constraints can help toexplain the strong countercyclicality of net exports along with the strongprocyclicality of consumption with output.Since the model can explain highly volatile wages and produces the rightsign of comovement of labour market variables with output, the labour sharebecomes more volatile than in the ?search only? model, closer to the data.The model can address more than half of the variation in the labour shareand procyclicality with output for this variable. The part that is not ex-plained in this model is related to employment fluctuations. The model canonly explain about one-fourth of the employment fluctuations due to searchfrictions. Boz et al (2009) introduce shocks to matching efficiency, ?, whichincreases the volatility for this variable.86One failure of the model is the underestimation of output volatility. Aswe discussed earlier, this could stem from the measurement problems inproducing shocks using data. In the sensitivity analysis, we will use highervolatility for the shocks assuming that output can be explained fully bythese shocks and analyze the results compared to baseline model.The differences in the results between the third and fourth column givesus a better understanding of the sole contribution of search frictions to86In addition, credit imperfections similar to those in Chapter 3 tend to generate morevariability in employment (see Petrosky-Nadeau (2011).1134.4. Quantitative Analysisfluctuations in emerging markets. As in the case with the fixed intensivemargin, the main contribution of search frictions is its impact on variablesrelated to labour markets as opposed to a frictionless model. Search frictionscan produce movements in employment and the labour share. In addition,wages become more volatile than in the frictionless model. However, withoutworking capital, the model underestimates the volatility in wages, labourshare and output, the negative correlation between interest rate and hoursper worker, and the positive correlation between intensive and extensivemargins of labour input.Lastly, we discuss the results when we assume a zero correlation betweeninterest rates and TFP shocks, a situation closer to that in developed mar-kets.87 The last column in Table 4.5 show that uncorrelated shocks imply a%40 smaller correlation between intensive and extensive margins of the workthan does the baseline scenario. This result is consistent with our findingson different fluctuations of hours per worker across country groups in thedata.4.4.4 Sensitivity AnalysisThe Strength of the Income Effect: The results in Table 4.5 suggestthat with a smaller income effect, the models tend to explain the data mo-ments better compared to results with separable preferences. With JR pref-erences, this income effect is governed by the parameter, ?, which is set tobe equal to 0.50 in the baseline scenario. Now we discuss the sensitivity ofthe results to this parameter.Table 4.6 presents the results from sensitivity analyses. The first twocolumns show the data moments and the results from the baseline model.The third and fourth columns display the implications of the model withfull income effect (? = 1) and zero income effect (? = 0), respectively.88Although higher income effect produces cyclical properties closer to data87Note that we are changing only the value of the correlation coefficient; we are notcalibrating the whole model to developed markets. Thus, by doing this, our objective isnot to explain labour market fluctuations in developed markets. Rather, we aim to discussthe direction of the results with no correlation between shocks.88When ? = 1, preferences converge to those in King et al. (1988).1144.4. Quantitative Analysisfor the labour input, it fails to explain the response of hours per workerto interest rates. As a result, the comovement between margins of work issmaller than in the data. In addition, the model with higher income effectgenerates less cyclical interest rates and consumption than we observe in thedata.On the other hand, a model with zero income effect; i.e., GHH pref-erences, makes hours per worker more cyclical over the cycle compared tobaseline model. Hours per worker are, now, highly correlated with output,employment, and interest rates, which is not consistent with the data. Highvolatility and cyclicality of labour input comes at the expense of smallerfluctuations in wages. In contrast, the baseline model produces less cyclicalhours and wages and more volatile wages, closer to the data.Robustness to Shocks: In this section, we analyze how robust the resultsare to different shock processes. We calibrate shock parameters so that theymatch particular data moments in Mexico as an alternative to estimatesused in the previous models. We do this alternative calibration since thereare some potential measurement problems in constructing Solow residuals.Specifically, TFP estimates from the data face measurement problems af-fecting factor shares, utilization rates on factors, and adjustment costs oncapital.The calibration technique used to find alternative parameters are as fol-lows. We assume that the persistence of productivity shocks, ?A, is the sameas in the US and set this parameter to 0.95, which implies a persistence ofoutput not higher than in the data.89 The standard deviation of the shockto productivity is set to match the output volatility in Mexico.90 Therefore,?A = ?R = 1.52. Lastly, the correlation between shocks is calibrated tomatch the correlation between output and interest rates, which results in89In the data, the autocorrelation of output is 0.84 and the model presented here withthe assumed persistence of productivity predicts this number to be 0.76, which is betterthan the prediction with the estimated persistence, 0.67.90This type of identification analysis for productivity parameters has been often used inRBC analysis. See, for example, Greenwood et al (1988), Mendoza (1991) and Neumeyerand Perri (2005).1154.5. Conclusion?A,R = ?0.72.The last column in Table 4.6 shows the implications of the baselinemodel with these alternative parameters in the shock processes. The resultschange only insignificantly for variables related to the labour market, whichimplies the findings analyzed above do not depend on the shock process weestimated in the quantitative analyses. The more significant effect of theseshocks appears on the fluctuations of goods market variables such that, withmore persistent TFP shocks and more volatile shocks, the model producesmore volatile and cyclical consumption and net exports, consistent withdata.4.5 ConclusionThe implications of small open economy real business cycle models for ex-plaining emerging market business cycles have been studied extensively inrecent years. Many of these studies have concentrated on the implicationsof such models for generating the observed responses of key macroeconomicvariables. Yet increasingly researchers and policy-makers are interested inunderstanding the cyclical response of labour market variables in emergingeconomies. During the recent global financial crisis, it is well known thatemerging market economies have seen increases in their unemployment ratesdespite the absence of a negative domestic shock. We here document thathours per employee dropped, too, during the crisis. We also show that thiscomovement between hours per worker and employment has actually been acharacteristics over the cycle in emerging economies, in contrast to advanceeconomies.In our framework, changes in the cost of borrowing coupled with a fric-tion such as a working capital requirement can lead to changes to bothemployment and hours of work in a search-theoretic framework. Hence, ourframework has the potential to account for the observed changes in emergingeconomies during the recent global financial crisis when the economy facesan external shock, as much as accounting for emerging economy businesscycles more generally.116Chapter 5ConclusionLabour market fluctuations in EMs are characterized by highly volatile realwages. I provide four additional facts on these fluctuations and comparethem with those in DMs: (1) Labour share is more volatile in EMs. (2)Labour share is also more procyclical in these economies. (3) Volatility ofhours per worker relative to the number of people employed is higher in EMs.(4) Hours per worker is procyclical with output and positively correlatedwith employment in EMs, while they are less cyclical with output and notcorrelated with employment in DMs.The findings in Chapter 2 suggest that labour share drops in recessionsthat are associated by an increase in interest rates, such as debt crises.Labour share then recovers after the crisis, implying a procyclicality withoutput. Since interest rates are countercyclical in most of the EMs, we tendto observe a procyclical labour share, on average, in these economies. Chap-ter 3 shows that working capital mechanisms together with countercyclicalcost of borrowing can explain some of the volatility in the labour share.When borrowing limits are introduced, the model implications on labourshare are amplified. A varying borrowing criterion also accounts for someof the business cycle puzzles in emerging markets, such as highly volatileconsumption, strongly procyclical investment and consumption, and coun-tercyclical net exports.Chapter 4 attempts to answer two more questions: (1) How do we rec-oncile the stylized facts of the cyclical properties of hours per worker inEMs, and specifically why is hours per worker more correlated with outputand employment in EMs than in DMs? (2) How much do search frictionscontribute to wage volatility under endogenous hours choice? This chap-ter shows that a model which incorporates working capital requirement and117Chapter 5. Conclusionsearch-and-matching frictions together with countercyclical interest ratescan account for those observations on the movements in the intensive mar-gin of labour input. Previous work provide explanations on high volatilityin real wages using search frictions. However, it is shown in this chapterthat these frictions account for a smaller part of the volatility in wages inthe presence of endogenous hours choice. Financial frictions still contributesignificantly to explain movements in real wages and labour share.This dissertation also offers some possible avenues for future research.First of all, I do not explore any policy decisions here. Policy makers wouldtake into account the effects of their actions on real interest rates as theymight lead to changes in labour share, i.e., large responses of labour marketsto economic shocks. Fiscal policy, for instance, might choose to be morecautious than in standard models to achieve stability in real interest rates. Inaddition, monetary policy have potential to alter aggregate demand not onlythrough traditional Keynesian channels but also through lowering/increasingthe income share of labour, which presumably have higher propensity toconsume in an economy. Understanding labour share dynamics are alsocrucial for the estimation of New Keynesian Philips Curve. Finally, it alsoremains a future work to explore the effects of working capital requirementacross firms? external and internal financing under the presence of financialfrictions and its implications on goods and labour market fluctuations.118BibliographyAdjemian, S., Bastani, H., Juillard, M., Mihoubi, F., Perendia, G., Ratto,M., and Villemot, S. (2011). Dynare: Reference manual, version 4. DynareWorking Papers 1, CEPREMAP.Aghion, P., Bacchetta, P., and Banerjee, A. (2001). Currency crises andmonetary policy in an economy with credit constraints. European Eco-nomic Review, 45(7):1121?1150.Aguiar, M. and Gopinath, G. (2007). Emerging market business cycles: Thecycle is the trend. Journal of Political Economy, 115(1):69?102.Aiyagari, S. and Gertler, M. (1999). Overreaction of Asset Prices in GeneralEquilibrium. Review of Economic Dynamics, 2(1):3?35.Altug?, S. and Bildirici, M. (2012). Business cycles in developed and emerg-ing economies: Evidence from a univariate markov switching approach.Emerging Markets Finance and Trade, 48(6):4?38.Altug, S. and Kabaca, S. (2013). Search Frictions, Financial Frictions, andLabor Market Fluctuations in Emerging Markets. Working Paper.Ambler, S. and Cardia, E. (1998). The cyclical behaviour of wages and prof-its under imperfect competition. Canadian journal of economics, pages148?164.Andolfatto, D. (1996). Business cycles and labor-market search. The Amer-ican Economic Review, 86(1):112?132.Arellano, C. (2008). Default risk and income fluctuations in emergingeconomies. American Economic Review, 98(3):690?712.119BibliographyArteta, C. and Hale, G. (2008). Sovereign debt crises and credit to theprivate sector. Journal of International Economics, 74(1):53?69.Ball, L. and Romer, D. (1990). Real rigidities and the non-neutrality ofmoney. The Review of Economic Studies, 57(2):183?203.Barth III, M. J. and Ramey, V. A. (2002). The cost channel of monetarytransmission. pages 199?256.Bentolila, S. and Bertola, G. (1990). Firing costs and labour demand: howbad is eurosclerosis? The Review of Economic Studies, 57(3):381?402.Bentolila, S. and Saint-Paul, G. (1998). Explaining Movements in theLabour Share. CEPR Discussion Papers.Bernanke, B., Gertler, M., and Gilchrist, S. (1999). The financial acceleratorin a quantitative business cycle framework. Handbook of Macroeconomics,1:1341?1393.Bils, M. (1985). Real wages over the business cycle: evidence from paneldata. The Journal of Political Economy, 93(4):666?689.Blundell, R. and Macurdy, T. (1999). Labor supply: A review of alternativeapproaches. In Ashenfelter, O. and Card, D., editors, Handbook of LaborEconomics, volume 3 of Handbook of Labor Economics, chapter 27, pages1559?1695. Elsevier.Boldrin, M. and Horvath, M. (1995). Labor contracts and business cycles.Journal of Political Economy, 103(5):972?1004.Botero, J. C., Djankov, S., La Porta, R., Lopez-de Silanes, F., and Shleifer,A. (2004). The regulation of labor. The Quarterly Journal of Economics,119(4):1339?1382.Boz, E., Durdu, C. B., and Li, N. (2009). Labor market search in emergingeconomies.Burdett, K. and Wright, R. (1989). Optimal firm size, taxes, and unemploy-ment. Journal of Public Economics, 39(3):275?287.120BibliographyCaballero, R. and Krishnamurthy, A. (2001). International and domesticcollateral constraints in a model of emerging market crises. Journal ofMonetary Economics, 48(3):513?548.Calvo, G. (1998). Capital flows and capital-market crises: the simple eco-nomics of sudden stops. Journal of Applied Economics, 1(1):35?54.Calvo, G. A. (1988). Servicing the public debt: The role of expectations.The American Economic Review, pages 647?661.Danthine, J.-P., Donaldson, J. B., and Siconolfi, P. (2005). Distribution riskand equity returns. CEPR Discussion Papers 5425, C.E.P.R. DiscussionPapers.De Soto, H. (1990). The other path: The invisible revolution in the thirdworld.Devereux, M. and Yetman, J. (2009). Leverage Constraints and the Inter-national Transmission of Shocks. NBER Working Papers.Devereux, M. B., Gregory, A. W., and Smith, G. W. (1992). Realistic cross-country consumption correlations in a two-country, equilibrium, businesscycle model. Journal of International Money and Finance, 11(1):3?16.Diwan, I. (2002). The labor share during financial crises: new results. WorldBank. Washington, DC. Processed.Fan, J. P., Titman, S., and Twite, G. (2012). An international comparisonof capital structure and debt maturity choices. Journal of Financial andQuantitative Analysis, 47(1):23.Fostel, A. and Geanakoplos, J. (2008). Leverage cycles and the anxiouseconomy. The American Economic Review, 98(4):1211?1244.Garcia-Cicco, J., Pancrazi, R., and Uribe, M. (2010). Real business cyclesin emerging countries? American Economic Review, 100(5):2510?31.121BibliographyGiammarioli, N., Messina, J., Steinberger, T., and Strozzi, C. (2002). Eu-ropean labor share dynamics: An institutional perspective. Technicalreport, European University Institute.Gollin, D. (2002). Getting income shares right. Journal of Political Economy,110(2):458?474.Gomme, P. and Greenwood, J. (1995). On the cyclical allocation of risk.Journal of Economic Dynamics and Control, 19(1-2):91?124.Gourinchas, P., Valdes, R., Landerretche, O., Talvi, E., and Banerjee, A.(2001). Lending Booms: Latin America and the World [with Comments].Economia, 1(2):47?99.Greenwood, J., Hercowitz, Z., and Huffman, G. (1988). Investment, capacityutilization, and the real business cycle. The American Economic Review,78(3):402?417.Guajardo, J. (2004). Financial Frictions and Business Cycles in DevelopingCountries. Econometric Society 2004 Latin American Meetings.Gumus, I. (2005). Debt denomination and default risk in emerging markets.Macroeconomic Dynamics, pages 1?26.Hairault, J.-O. (2002). Labor-market search and international business cy-cles. Review of Economic Dynamics, 5(3):535?558.Hall, R. E. (2009). Reconciling cyclical movements in the marginal value oftime and the marginal product of labor. Journal of Political Economy,117(2):281?323.Hall, R. E. and Milgrom, P. R. (2008). The limited influence of unemploy-ment on the wage bargain. American Economic Review, 98(4):1653?74.Hansen, G. and Prescott, E. (2005). Capacity constraints, asymmetries, andthe business cycle. Review of Economic Dynamics, 8(4):850?865.122BibliographyHeckman, J., Page?s-Serra, C., Edwards, A., and Guidotti, P. (2000). Thecost of job security regulation: Evidence from latin american labor mar-kets [with comments]. Economia, 1(1):109?154.Holmstrom, B. and Tirole, J. (1997). Financial Intermediation, Loan-able Funds, and The Real Sector*. Quarterly Journal of Economics,112(3):663?691.Hornstein, A. (1993). Monopolistic competition, increasing returns to scale,and the importance of productivity shocks. Journal of Monetary Eco-nomics, 31(3):299?316.IMF (2009). World Economic Outlook October-2009.IMF (2012). World Economic Outlook October-2012.Jaimovich, N. and Rebelo, S. (2009). Can news about the future drive thebusiness cycle? American Economic Review, 99(4):1097?1118.Jeanne, O. (2003). Why do emerging economies borrow in foreign currency?International Monetary Fund.Jermann, U. and Quadrini, V. (2006). Financial innovations and macroeco-nomic volatility. NBER Working Paper.Kaminsky, G., Reinhart, C., and Ve?gh, C. (2004). When it rains, it pours:procyclical capital flows and macroeconomic policies. NBER Macroeco-nomics Annual, pages 11?53.King, R. G., Plosser, C. I., and Rebelo, S. T. (1988). Production, growthand business cycles: I. the basic neoclassical model. Journal of MonetaryEconomics, 21(2):195?232.King, R. G. and Watson, M. W. (1996). Money, prices, interest rates andthe business cycle. The Review of Economics and Statistics, 78(1):35?53.Kiyotaki, N. and Moore, J. (1997). Credit cycles. Journal of Political Econ-omy, 105(2):211?248.123BibliographyKocherlakota, N. (2000). Creating business cycles through credit constraints.Federal Reserve Bank of Minneapolis Quarterly Review, 24(3):2?10.Krugman, P. (1999). Balance sheets, the transfer problem, and financialcrises. pages 31?55.Kydland, F. E. and Prescott, E. C. (1982). Time to build and aggregatefluctuations. Econometrica, 50(6):1345?1370.Lane, P. and Milesi-Ferretti, G. (2007). The external wealth of nationsmark II: revised and extended estimates of foreign assets and liabilities,1970?2004. Journal of International Economics, 73(2):223?250.Li, N. (2011). Cyclical wage movements in emerging markets comparedto developed economies: the role of interest rates. Review of EconomicDynamics, 14(4):686?704.Lorenzoni, G. (2008). Inefficient credit booms. Review of Economic Studies,75(3):809?833.Ludvigson, S. (1999). Consumption and credit: a model of time-varyingliquidity constraints. Review of Economics and Statistics, 81(3):434?447.Maarek, P. and Orgiazzi, E. (2010). Which factor bears the cost of currencycrises?Mandelman, F. S. (2011). Business cycles and the role of imperfect compe-tition in the banking system*. International Finance, 14(1):103?133.Mankiw, N. (1985). Small menu costs and large business cycles: a macroe-conomic model of monopoly. The Quarterly Journal of Economics,100(2):529?537.Mellow, W. and Sider, H. (1983). Accuracy of response in labor marketsurveys: Evidence and implications. Journal of Labor Economics, pages331?344.124BibliographyMendoza, E. (2010). Endogenous Sudden Stops in a Business Cycle Modelwith Collateral Constraints: A Fisherian Deflation of Tobin?s Q. forth-coming in American Economic Review.Mendoza, E. and Smith, K. (2006). Quantitative implications of a debt-deflation theory of Sudden Stops and asset prices. Journal of InternationalEconomics, 70(1):82?114.Mendoza, E. and Terrones, M. (2008). An anatomy of credit booms: evi-dence from macro aggregates and micro data. NBER Working Paper.Mendoza, E. G. (1991). Real business cycles in a small open economy.American Economic Review, 81(4):797?818.Merkl, C. and Wesselbaum, D. (2011). Extensive versus intensive marginin germany and the united states: any differences? Applied EconomicsLetters, 18(9):805?808.Merz, M. (1995). Search in the labor market and the real business cycle.Journal of Monetary Economics, 36(2):269?300.Nehru, V. and Dhareshwar, A. (1993). A new database on physical capitalstock: sources, methodology and results. Revista de Analisti Econo?mico.Vol, 8(1):37?59.Neumeyer, P. A. and Perri, F. (2005). Business cycles in emerging economies:the role of interest rates. Journal of monetary Economics, 52(2):345?380.Petrosky-Nadeau, N. (2011). Credit, vacancies and unemployment fluctua-tions.Ranciere, R., Tornell, A., and Westermann, F. (2008). Systemic Crises andGrowth*. Quarterly Journal of Economics, 123(1):359?406.R??os-Rull, J. and Choi, S. (2009). Understanding the Dynamics of LaborShare: The Role of Non-competitive Factor Prices . Working paper.R??os-Rull, J. and Santaeulalia-Llopis, R. (2009). Redistributive Shocks andProductivity Shocks. forthcoming in Journal of Monetary Economics.125BibliographyRotemberg, J. J. and Woodford, M. (1999). The cyclical behavior of pricesand costs. Handbook of macroeconomics, 1:1051?1135.Sarquis, S. (2008). Interest Rate and Business Cycles in a Credit ConstrainedSmall Open Economy. Cycles, Contagion and Crises, page 238.Schmitt-Grohe?, S. and Uribe, M. (2003). Closing small open economy mod-els. Journal of International Economics, 61(1):163?185.Schneider, M. and Tornell, A. (2004). Balance sheet effects, bailout guaran-tees and financial crises. Review of Economic Studies, 71(3):883?913.Seymen, A. (2011). Labor market institutions and the business cycle ingermany and the us. Working paper.Shimer, R. (2009). Convergence in macroeconomics: The labor wedge.American Economic Journal: Macroeconomics, 1(1):280?97.Stiglitz, J. and Weiss, A. (1981). Credit rationing in markets with imperfectinformation. American Economic Review, 71(3):393?410.Thomas, J. J. (1992). Informal economic activity.Uribe, M. and Yue, V. Z. (2006). Country spreads and emerging countries:Who drives whom? Journal of international Economics, 69(1):6?36.Vermeulen, P. (2007). Can adjustment costs explain the variability andcounter-cyclicality of the labour share at the firm and aggregate level?Working Paper Series 772, European Central Bank.Yedid-Levi, Y. (2009). Why does employment in all major sectors movetogether over the business cycle? Stanford University working paper.126Appendix AAppendices to Chapter 2A.1 DataCompensation of Employees: We obtain labor cost data mainly fromUN. They are compatible with 1993 System of National Accounts. For adetailed source description for each country, see A.1Interest Rates: The cost of borrowing is obtained mostly from IFS. Theyare either lending rates, or average cost of borrowing, or T-bill rates depend-ing on data availability for different countries. See A.1 for country specificinterest rate definition and the source.GDP deflator: The data are of the same source as the interest rates foreach country.EMBI rates: The data on EMBI spreads for emerging economies comefrom Uribe and Yue (2006) dataset. For the recent global financial crisis,we have the data from JP-Morgan EMBI database.Self-employment ratios: The sources are the OECD for Mexico, Korea,and Turkey; and ILO for the rest of the countries.Wages: For countries such as Argentina, Mexico, and Turkey, we have wagedata for only manufacturing. Wages are total gross earnings per worker ex-cept for Argentina, Chile, and Turkey in which they are hourly earnings.The source for wages is ECLAC (Economic Commission for Latin Amer-ica and the Caribbean) for Brazil, Chile, Costa Rica, and Mexico; ILO for127A.1. DataPhilippines; INEC and TURKSTAT (National sources) for Argentina andTurkey, respectively.Employment: The source is ILO for Chile, Costa Rica, Peru and Philip-pines. OECD is the source for Korea, Mexico, and Turkey. For Brazil, weobtain the data from IPEA (national source). For Argentina, employmentdata comes from the World Bank Development Indicators and the time se-ries of employees from ILO.Sectoral value added: The data on the contribution of the major sectors(agriculture, industry, and services) are from the World Development Indi-cators (World Bank).Quarterly data: The data sources are IPEADATA (national source) forBrazil, the OECD for Korea and Mexico, and TURKSTAT (national source)for Turkey for labor share and output. See the above table for the source ofthe interest rates.Informal employment: Informal employment refers to the self-employedin their own informal sector enterprises, contributing family workers (un-paid), members of informal producers cooperatives (not established as legalentities), employees holding informal jobs (i.e., jobs not subject to nationallabor legislation, income taxation, social protection or entitlement to certainemployment benefits, such as sick leave); own-account workers engaged inproduction of goods exclusively for own final use by their household. There-fore, informal employment covers the employment in the informal sector aswell as the informal employment in the formal sector (informal jobs in theformal enterprises). The source ILO KILM 8 Indicators.128A.1. DataTable A.1: Data Sources for Labor Compensation and Interest RatesPeriod Compensation Interest& GDP (VA) RatesEmerging MarketsArgentina 1993-2007 UN lending rate (IFS)Brazil 1992-2007 UN T-bill rate (IFS)Chile 1981-2007 UN lending rate (IFS)Columbia 1992-2007 UN lending rate (IFS)Costa Rica 1982-2007 UN lending rate (IFS)Czech Rep. 1992-2008 OECD lending rate (IFS)Egypt 1996-2006 UN lending rate (IFS)Hungary 1995-2008 OECD lending rate (IFS)India 1981-2002 UN lending rate (IFS)Israel 1995-2007 UN lending rate (IFS)Korea 1981-2008 OECD corporate bond rate (IFS)Mexico 1981-2008 OECD ave cost of borr. (IFS)Peru 1986-2006 UN lending rate (IFS)Philippines 1992-2007 UN T-bill rate (IFS)Poland 1991-2008 OECD T-bill rate (OECD)Russia 1995-2008 UN lending rate (IFS)South Afr. 1981-2008 UN T-bill rate (IFS)Turkey 1987-2006 UN money market (IFS)Developed MarketsAustralia 1981-2008 OECD T-bill rate (OECD)Austria 1990-2008 OECD T-bill rate (OECD)Canada 1981-2008 OECD T-bill rate (OECD)Denmark 1987-2008 OECD T-bill rate (OECD)Finland 1987-2008 OECD T-bill rate (OECD)France 1981-2008 OECD T-bill rate (OECD)Germany 1981-2008 OECD T-bill rate (OECD)Greece 1981-2008 OECD T-bill rate (IFS)Iceland 1988-2008 OECD T-bill rate (OECD)Ireland 1984-2008 OECD T-bill rate (OECD)Italy 1981-2008 OECD T-bill rate (OECD)Netherlands 1986-2008 OECD T-bill rate (OECD)New Zealand 1981-2008 OECD T-bill rate (OECD)Norway 1981-2008 OECD T-bill rate (OECD)Spain 1981-2008 OECD T-bill rate (OECD)Sweden 1982-2008 OECD T-bill rate (OECD)United Kingdom 1981-2008 OECD T-bill rate (OECD)United States 1981-2008 OECD T-bill rate (OECD)Note: Compensation is the compensation of employees in national accounts from incomeapproach. 129A.1. DataFigure A.1: Correlation of Labor Share and EMBI Interest Rates with Out-putARGBRACHLKORMEXPERPHLZAFTUR CZHISRRUSAUSAUTCANNLDNZLESPSWEUKUSDNKFIN FRAGERGRCISLIRLITANOR?.50.5corr(s,y)?.5 0 .5corr(r,y)Note: corr(s,y) and corr(r,y) denote the correlation of labor share with output and ofinterest rate with output, respectively. Interest rate data cover 1994Q1-2005Q1 for manycountries and constructed using EMBI spread data from Uribe and Yue (2006) exceptArgentina. Interest rate data for Argentina (1983Q1-2005Q1) come from Neumeyer andPerri (2005) until 2001Q2, we then extend the data using EMBI rates from Uribe and Yue(2006).130A.1. DataFigure A.2: Labor Share (Self-Employment-Adjusted) vs. Interest RateFluctuationsARGBRACHL CRIKORMEXTUR CZHEGYHUN?.4?.,y)?.6 ?.4 ?.2 0 .2 .4corr(r,y)ARGBRACHLCRIKORMEXTURCZHEGYHUN123456sigma(s)2 4 6 8 10sigma(r)This figure updates Figure 1 with adjusted labor share using methods described in thetext for emerging markets. See Table 4 for further information on which method is usedfor each country. See also Figure 1 for information on notations.131Appendix BAppendices to Chapter 3B.1 DataMexican Labor Variables:The longest quarterly labor share for Mexico comes from the manufactur-ing sector, covering the period between 1987Q1-2008Q1. The source is theOECD. The correlation between cyclical component of labor share in man-ufacturing and that in the total economy is 0.86 at the annual frequency.Although it has a similar comovement with output, manufacturing laborshare is 28% more volatile than that in the total economy, 4.5% and 3.5% re-spectively. Wages respresent hourly earnings in manufacturing from OECDmonthly economic indicators (MEI). Hours are the proxy for the total hoursworked in the total economy. They are constructed as hours per workerin manufacturing multiplied by the total civilian employment. Total hoursworked and employment come from INEGI (industrial survey in manufac-turing). Total employment is from Neumeyer and Perri (2005) for the periodbetween 1987Q1 and 2001Q1. We then extend the data using the ILO laborstatistics database until 2008Q1.Other Variables for Mexico:The source for interest rates, GDP, and GDP deflator is IFS. Interest ratesare the average cost of borrowing in the total economy. We also use JP-Morgan EMBI+ data from Uribe and Yue (2006). Other variables for Mex-ico, such as consumption, net export, and investment come from OECDquarterly database.132B.2. Parameters for Mexico and CanadaCanadian Labor Variables:Data on labor income share and employment are from OECD, and cover thewhole economy. For wages, we have quarterly data for the overall economyfrom ILO. We take the same data period as in the case for Mexico (1987Q1-2008Q1).Other Variables for Canada:Interest rates are the short-term treasury bills from OECD. All other vari-ables including GDP, GDP deflator, consumption, investment, and net ex-ports are taken from OECD quarterly database.B.2 Parameters for Mexico and CanadaIn Chapter 3, we extend the baseline model by including labor adjustmentcosts to compare implications between Canada and Mexico in Section 3.3.The calibration for Mexico is explained in the text. Here we discuss thecalibration for Canada.We set discount factor to be equal to 0.99, which suggests a 4% annu-alized interest rate at the steady state. Labor curvature parameter is setto be 2 to match a Frisch elasticity of labor of 1, implying an elasticity oflabor higher than in Mexico. This value is within the range of estimatesused in the literature for developed markets. The capital exponent is set tobe equal to 0.40 following the literature in business cycles. The elasticityof credit-income criterion to interest rates is assumed to be equal to 0.75,less than half of the value calibrated in Mexico. By doing this, we assumethat Canada has better financial institutions; i.e, a less imperfect financialindustry. Net foreign asset ratio at the steady state is calculated using thedata on external asset positions from Lane and Milesi-Ferretti (2007). Therest of the parameters are the same as those for Mexico.133B.2. Parameters for Mexico and CanadaTable B.1: Parameter Values for Mexico and CanadaName Symbol ValueMexico CanadaDiscount factor ? 0.98 0.99Utility curvature ? 5 5Labor curvature ? 2.75 2.0Labor weight ? varies variesCapital exponent ? 0.43 0.40Depreciation rate ? 0.02 0.02Wage bill paid in advance ? 1 or 0.66Bond holding cost ? 0.001 0.001Capital adjustment cost ? varies variesInduced leverage ? 1.82 0.75Net Foreign Debt / GDP ? -0.42 -0.25Table B.2: Shock ProcessesMexicoA?t = 0.75A?t?1 + at ?t?t =(0.01342 ?0.000093?0.000093 0.01342)R?t = 0.60R?t?1 + rtCanadaA?t = 0.58A?t?1 + at ?t?t =(0.0072 0.00000380.0000038 0.00252)R?t = 0.70R?t?1 + rt134Appendix CAppendices to Chapter 4C.1 DataTable C.1 lists data sources for each economic variable and country. Sincedata from most of the emerging market start after 1980s, we take observa-tions after 1981 for developed markets, as well. Hours represent total hoursworked in manufacturing from industrial surveys conducted by nationalsources in emerging economies. We divide total hours by total employmentin manufacturing to find hours worked per worker for these economies. Wedid the same for developed economies using total hours worked and employ-ment in manufacturing from OECD or national sources. For both groups,hours are consistent with National Accounts in the sense that they bothrepresent hours worked rather than normal or paid hours.Total employment is the civilian employment. For Turkey, employmentdata represent the number of employees in the overall economy. The nationalsource releases employment data including unpaid family workers. However,the strong cultural practices might hinder the real labor market outcomesas a response to output variations. This is why we exclude family workers.In the other countries, this is not an issue because this type of employmentconstitute a very small part of employment.135C.1. DataTable C.1: Data SourcesPeriod GDP Employment Hours Employment(Manufacturing)Emerging MarketsArgentina 1995-2008 INDEC INDEC INDEC INDECBrazil 1994-2008 WDI WDI IPEA IPEAColumbia 1990-2008 WDI DANE DANE DANECzech Rep. 1991-2008 OECD OECD OECD OECDHungary 1995-2008 OECD OECD OECD OECDKorea 1981-2008 OECD OECD OECD OECDMexico 1993-2008 OECD OECD INEGI INEGITurkey 1988-2008 OECD TURKSTAT TURKSTAT TURKSTATDeveloped MarketsAustria 1995-2008 OECD OECD OECD OECDCanada 1981-2008 OECD OECD OECD OECDDenmark 1981-2008 OECD OECD OECD OECDFinland 1981-2008 OECD OECD OECD OECDFrance 1990-2005 OECD OECD OECD OECDGermany 1991-2008 OECD OECD OECD OECDItaly 1981-2008 OECD OECD OECD OECDNorway 1981-2008 OECD OECD OECD OECDSpain 1981-2008 OECD OECD OECD OECDSweden 1993-2008 OECD OECD OECD OECDUnited States 1987-2008 OECD BLS BLS BLS136C.1. DataFigure C.1: EPL Index across EMs ad DMsARGBRAKORMEXTURCOLCZHHUNAUTCANESP SWEUSDNKFINFRAGERITANOR. index0 10000 20000 30000 40000GDP per capitaNote: Employment protection index comes from (see Botero et al., 2004). Highernumbers indicate more regulation of labor markets through employment laws, col-lective bargaining laws, and social security laws. GDP per capita (PPP adjustedin US dollars) in 2000 is taken for the income level.137


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



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"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items