Policy Uncertainty, Irreversibility, and Cross-Border Flows of Capital

Finance and Economics Discussion Series Divisions of Research & Statistics and Monetary Affairs Federal Reserve Board, Washington, D.C. Policy Uncertainty, Irreversibility, and Cross-Border Flows of Capital Brandon Julio and Youngsuk Yook 2013-64 NOTE: Staff working papers in the Finance and Economics Discussion Series (FEDS) are preliminary materials circulated to stimulate discussion and critical comment. The analysis and conclusions set forth are those of the authors and do not indicate concurrence by other members of the research staff or the Board of Governors. References in publications to the Finance and Economics Discussion Series (other than acknowledgement) should be cleared with the author(s) to protect the tentative character of these papers.

Policy Uncertainty, Irreversibility, and Cross-Border Flows of Capital ∗ BRANDON JULIO London Business School YOUNGSUK YOOK† Federal Reserve Board of Governors August 2013 ABSTRACT We examine the effects of government policy uncertainty on cross-border capital flows. FDI flows from US companies to foreign affiliates drop significantly during the periodjustbeforeanelection. TheelectioneffectforFDIislargerthanelectioncyclesin domestic investment. The electoral patterns in FDI flows are more pronounced in countries with higher propensities for policy reversals and when election outcomes are more uncertain. Ouridentificationstrategycomparesvariationindifferenttypesofcapitalflows into the same country around the timing of national elections. The electoral cycles are present in relatively irreversible FDI flows but not in foreign portfolio investment flows, suggesting alikelycausallinkfrompolitical uncertainty toandcapitalflows. ∗ Department of Finance, London Business School, Regent’s Park, London NW1 4SA, United Kingdom; e-mail:bjulio@london.edu;phone:+44(0)2070008254. †Federal Reserve Board of Governors, 20th Street and Constitution Avenue NW, Washington, DC 20551; e-mail: youngsuk.yook@frb.gov; phone: (202)475-6324. The views expressedin this article are those of the authorsandnotnecessarilyof theFederalReserveSystem. We thankLingCen, PasqualeDella Corte, Joshua Gallin, Pedro Magalha˜es, Anamaria Piloiu, Vincenzo Quadrini, and seminar participantsat American University,FederalReserveBoardofGovernors,KoreaUniversityEconomicsDepartment,SeoulNationalUniversity, SungkyunkwanUniversity, University of Maryland, University of Sydney, University of Technology, Sydney, 2012ChinaInternationalConferenceinFinance,theChineseFinanceAssociationBestPaperSymposium2012, GlobalFinanceConference,Lisbon Meetingon Institutionsand PoliticalEconomy,Western Finance AssociationAnnualMeeting,EuropeanFinanceAssociationMeeting,andtheConferenceonPolicyUncertaintyandits EconomicImplicationsattheUniversityofChicagoforhelpfulcomments.

1. Introduction Cross-borderflowsofcapitalhavegrownrapidlyinsizeandimportanceinrecentdecades.1 They are an important source of capital in emerging markets and make up a significant proportion of GDP in many countries around the world. The international investment literature provides an extensive list of determinants of bilateral capital flows, ranging from geographic proximity and macroeconomic conditions to institutional factors. In this paper, we examine whether fluctuations in policy uncertainty leads to variation in foreign investment. Rodrik (1990), for example, argues that even well-meaning government effort such as liberalization or market-oriented reforms may need to take a back seat when it places the sustainability of policies into question as policy uncertainty creates incentives for foreigners to withhold investment. Todate,however,thereislittleempiricalevidencesupportingtheviewthatpolitical uncertaintyaffects foreign investment. While all investments are exposed to the risk that government policies may shift and adverselyaltertheexpectedpayoffs toinvestors,foreign investmentis burdenedwithadditional layers of rules and regulations associated with national boundaries such as capital controls and differential tax treatments. Dixit (2011) highlights the fact that foreign direct investment (FDI) is more sensitive to the political environment than domestic investment as the foreign investorhaslimitedprotectionfromthehostcountry’slegalandpoliticalinstitutions. Foreign investmentsmay beriskierashost governmentslikelyviewtheexpropriationofforeigners as more politically palatable than the expropriation of citizens. Courts in destination countries may have a bias towards domestic firms and investors in the case of disputes (Bhattacharya, Galpin and Haslem (2007)). Among the various types of international capital flows, FDI is thoughtto bemostsensitivetopolicyuncertaintyand institutions. 1AccordingtoUNCTAD(2009),foreigndirectinvestmentinflowsworldwidegrewbyafactorofnearly10 from $208 billion in 1990 to a historic high of $1,979 billion in 2007. A Coordinated Portfolio Investment SurveyconductedbytheInternationalMonetaryFund(IMF)revealsthatforeignportfolioinvestmentholdings worldwidegrewmorethansix-foldbetween1997and2007. 1

The recent global financial crisis and subsequent recession has spawned a fast growing literature investigating the effects of policy uncertainty on economic activity. Cross-border flows of capital also experienced a large contraction and slow recovery.2 A recent debate has focused on why growth in the wake of the financial crisis has been slow to recover. One of the explanations for the sluggish recovery offered by some commentators is that uncertainty about future government policy is abnormally high.3 However, the literature has highlighted thatthetwoempiricalchallengestoestablishingaclearlinkbetweenpoliticaluncertaintyand realoutcomesarefirstmeasuringpoliticaluncertaintyandsecondidentifyingthecausaleffect ofuncertaintyon investment(Baker, Bloomand Davis(2012)). To measure policy uncertainty, we employ the approach of Durnev (2010), Gao and Qi (2013),Jens(2012),JulioandYook(2012)andColak,DurnevandQian(2013)andutilizethe timing of elections as a measure of variation in political uncertainty. Specifically, we examinedirectinvestmentandportfolioinvestmentflowsaroundthetimingofnationalelectionsin destinationcountriesaroundtheworld. TheoutcomesofnationalelectionsarerelevanttoFDI decisions as they have implications for foreign capital controls, trade policy, and taxation as well as other policies that are applicable to both domestic and foreign firms such as industry regulationandfiscal policy. Changesinthesepoliciescanaffect theriskandreturn properties ofrealinvestment. Whenopposingcandidatesinanelectionpromotedifferentpolicies,uncertaintyabouttheelectionoutcomeimpliesuncertaintyaboutwhatpolicieswillbeenactedafter the election. There is empirical evidence supportingour assumptionthat political uncertainty is significantly higher around elections. Bialkowski, Gottschalk, and Wisniewski (2008) and Boutchkova et al. (2012) find that return volatility is significantly higher than normal during 2Annual global foreign direct investmentinflows fell 16% in 2008, and a further 37% to $1,114billion in 2009beforeshowingmodestrecoveryinthefirsthalfof2010(UNCTAD(2010)). BertautandPounder(2009) examine bilateral portfolio investment between the U.S. and the rest of the world and report a considerable pullbackfromcross-borderpositionsduringthefinancialcrisis. Asofmid-2009,theportfolioflowshaveyetto recovertothepre-crisislevel. 3For example, see the Distinguished Speaker presentation by Chester Spatt at the 2009 Western Finance AssociationconferenceandcommentsbyBenBernankeintheJuly22,2010editionoftheWallStreetJournal. 2

election periods around the world. Bernhard and Leblang (2006) document changes in bond yields, exchange rates, and equity volatility around elections, and show that these changes are larger when elections outcomes are close. Additionally, Baker, Bloom and Davis (2012) construct an index of policy-related economic uncertainty in the United States, and note that thisindexspikesupwardduringelections. Thesecond challengein testingwhetherpolicyuncertaintydepresses internationalinvestmentactivityisthelikelyendogeneitybetweenmeasuresofpoliticaluncertaintyandeconomic fundamentals. As Rodrik (1991) notes, it is very difficult to find strong empirical support for uncertainty-driven predictions because political instability and uncertainty are likely endogenous to other factors that affect private investment decisions. Estimating the direction of causality between economic outcomes and policy uncertainty requires employing a variable or event that is correlated with policy uncertainty but uncorrelated with the economic conditions that drive foreign investment. Election timing is admittedly a very broad measure of political uncertainty, capturing not only possible changes in government policy but also changes in the composition of government. The timing of an election in one country is out of the control of any individual firm in another country and indeed fixed in time by constitutional rules for a large number of countries in our sample. Thus, elections around the world providea natural experimentframework for studyingtheeffects ofpolicyuncertainty on FDI flows, allowing us to disentangle some of the endogeneity between economic conditions and politicaluncertainty. Ifpoliticaluncertaintyishigherwhenchangesinnationalleadershipare moreprobable,electionsprovidesomeexogenousvariationinpolicyriskovertimethathelps isolate the impact of policy uncertainty on FDI choices from other confounding factors. In addition, elections around the world take place at different points in time, allowing us to net outglobaltimetrends inFDIflows. Using 184 national elections in 45 countries between January 1994 and June 2010, we examine changes in quarterly FDI flows as political uncertainty fluctuates by comparing the 3

investmentflowsinthequartersleadinguptothenationalelectionoutcomeswiththoseinnonelection quarters. The large body of literature documenting determinants of FDI flows gives us a good benchmark empirical model to gauge abnormal changes in capital flows around the election cycle. We find clear evidence that U.S. FDI flows are significantly lower in the quarter just prior to an election outcome in the destination country. Our empirical results are consistent with the view that policy uncertainty depresses flows of private investment. The election effect remains strong when controlling for various macroeconomic and institutional factors such as GDP growth, exchange rate changes, trade openness, government stability, government expenditure, and stock market returns as well as country and time fixed effects. ThebaselineresultssuggestthattheFDIflowratefallsbyapproximately12%relativetononelection years, all else being equal. The magnitude of decline in the FDI rate compares to an average reduction in domestic corporate investment around election cycles of 4.8% documented by Julio and Yook (2012) and 4.5% by Jens (2012), suggesting that FDI is more sensitive to policy uncertainty than is domestic investment. To address the concern that incumbents may opportunistically time elections to maximize their chance of re-election and thereby induce a correlation between election timing and economic activity, we repeat the tests with the subsample of countries for which elections are fixed in time by electoral law. Theresultsaresimilarinthesubsampleofelectionswithexogenoustiming. We also find that the election effects are stronger when the election race is close, suggesting that a higher degree of uncertainty regarding election outcomes is associated with larger drops in FDI flows in election quarters. The investment cycles are more pronounced in countries with less stable political systems and fewer checks and balances on executive authority. Election effects are smaller when the host country is more open to international trade. Election cycles in FDI flows are present, though less severe, in high income countries as well, suggesting that the depressing effects of policy uncertainty on FDI flows are not just an emerging markets phenomenon. In additionto policyrisk in destinationcountries,we find 4

that the source country’s political uncertainty affects FDI flows. Specifically, U.S. investors’ FDI flows are similarly sensitive to elections abroad and to U.S. elections. FDI flows drop significantly in the quarter leading up to the U.S. election and then return to normal levels aftertheelectionresolution. Thissuggeststhatpolicyconsiderationsinmultiplecountriesare relevantformultinationalfirms. A remaining challenge in the political uncertainty literature is the identification of causal effects. While the election timing alleviates many econometric concerns and it is clear that various economic activities vary over the election cycle, an unresolved issue is whether the observed effects are the result of heightened political uncertainty or whether the effects are drivenby someotherpoliticalmechanism,suchas politicalbusinesscycles(Nordhaus 1975). The political business cycle literature has highlighted the incentives of incumbent politicians toattempttomanipulatetheeconomytoimprovetheirre-electionchances. Whilethepolitical business cycle models typically predict a positivejump in economic activity prior to an election,itispossiblethatsuchattemptsmaycrowdoutprivateinvestmentandleadtheresearcher to incorrectly conclude that uncertainty drives the result. Our identification strategy involves comparing two sets of flows into the same country in the same time period. The two sets of flows,FDIandforeignportfolioflows(FPI),havesimilarreturnpropertiesandsensitivitiesto fundamentals. They differ significantly, however, with respect to the ease with which investments can be reversed, allowing us to distinguish uncertainty effects from political business cycleeffects. Irreversibility is an important feature in models of investment under uncertainty. Because investment is costly to reverse, irreversibility increases the information value of waiting to invest (Caballero (1991)), causing investment to vary negatively with fluctuations in policy uncertainty over time. The resulting prediction that the investment-uncertainty relation will be more negativefor more irreversible assets has been examined in various contexts. For example, Bulan (2005) uses asset specificity at the industry level as a measure of the capital 5

irreversibility and documents the negative investment-uncertainty relation for irreversible industries. GuisoandParigi(1999)findmorenegativeuncertainty-investmentrelationforfirms withhighirreversibilitymeasuredbytheiraccesstosecondarymarketsfortheircapitalequipmentandbytheircomovementwithotherfirmswithinanindustry. KimandKung(2013)find a strong relationship between asset redeployability and investment sensitivity to uncertainty. In an international setting, Rajan and Marwah (1998) examine the difference in the degree of irreversibility between exports and FDI, and present a model in which the policies that are perceived as weakly credible lead firms to favor servicing foreign markets through exports rather than by undertaking FDI. FDI flows are, by definition, long-term, relationship-specific investments that are costly to reverse.4 Caballero and Hammour (1998) point out that FDI is like investing in specific assets that ex post cannot be retrieved according to ex-ante terms of trade. While FDI flows are typically considered relatively irreversibledue to specificity, foreign portfolioinvestment(FPI)flowsareconsideredtobeeasiertoreverse(Razin,SadkaandYuen (1998)). In our last set of tests, we incorporatethis intuitionand compare different flows into the same country in the same time period that have similar return properties with respect to fundamentals, but differ in their sensitivityto uncertainty. We compare relatively irreversible FDIflows toFPI flowsaround theelectioncycle. If theresultis drivenby fundamentals,then both flows shouldbe affected by theelection cycle. If political uncertainty is themechanism, thenweexpectFDIflowstodeclinemorethanFPIflows. WefindthatFDIflowsaresensitive to election cycles whereas FPI flows, which can be reversed at a relativelylower cost, are not sensitivetotheelectioncyclesaroundtheworld. Thislendssupporttoourprimaryhypothesis thatpolicyuncertaintyisdrivingourempiricalfindingsandhighlightsthemechanismthrough whichpolicyuncertaintygenerates thetimeseries variationin cross-borderinvestmentflows. 4BalanceofPaymentsManual,fifthedition(IMF,1993). 6

Our empirical predictions are drawn from established theoretical literature related to the effects of political uncertainty. Rodrik (1991) models private foreign investment choices in a setting with policy uncertainty. In his model, foreign investors hold back on investing until a large amount of uncertainty regarding the success of political reform is resolved. Chen and Funke (2003) also model FDI decisions in the face of policy uncertainty and generate similar predictions. In this context, policy uncertainty has a negative effect on private investment when the investment is at least partially irreversible. The impact on investment in this setting is significant. Rodrik demonstrates that under reasonable assumptions even a 10 percent probability of policy reversal requires an investment subsidy of 7.5 percentage points to offset its adverse effects on investment. Thus, policy uncertainty acts like a tax on investment. Theintuitionissimilaringeneralmodelsofinvestmentunderuncertainty,includingBernanke (1983) and Bloom, Bond, and Van Reenen (2007), that the value of waiting increases when uncertainty related to changes in governmentpolicy is high. Pindyck and Solimano (1993) is another example of this literature in which the uncertainty brought about by political factors leads firmsto chooselowerlevelsofinvestmentexpenditures. Our papercontributesto two importantsets ofliterature. First, theFDIliteratureprovides an extensivelistof the determinantsof FDI includingmacroeconomic variablessuch as GDP and exchange rate fluctuations, institutional quality, and firm-level cost considerations5. A number of studies examine the implication of political institutions for FDI. Wei (2000) documents that corruption in the recipient country substantially reduces FDI inflows. Singh and Jun (1995) document that FDI flows are especially sensitiveto political risk in countries that have historically attracted high FDI flows. Daude and Fratzscher (2008) document that FPI is more sensitiveto institutional factors than FDI. Desai et al. (2008) and Desai et al. (2004) document that political risk affects the variability of foreign affiliates’ returns as well as the capital structure decisions of both the parent and affiliates of multinational firms. We depart 5Seesection3.4formorediscussionofFDIdeterminants. 7

from this strand of literature in that we focus on uncertainty surrounding policy rather than policy per se, and investigate whether perceived policy shifts affect the expected payoff to investment. In related work, Wei (1997) documents that uncertainty regarding corruption has importantnegativeeffectsonFDIdecisions. HermesandLensink(2001)documentthatpolicy uncertaintyhasapositiveimpactontheoutflowofdomesticcapital. Ourstudyhasabroadapplication in that, while institutional variables such as corruption and investor protection have applicationsmainlyinlessdevelopedcountries,policyuncertaintyisanimportantconcernfor developedcountries aswell. Our paper also contributes to therecent literaturefocusing on the interaction between politicalchangeand finance. Kim,Pantzalis,and Park(2012)investigatetheimpactofvariation in political geography brought on the outcomes of mid-term elections and find a significant effect on returns. Gao and Qi (2013)showthattheuncertaintyaround gubernatorialelections in theU.S. is reflected in higheroffering yieldsof municipalbonds. Durnev (2010)examines firm investmentaround national elections in an international settingand finds that investment is less sensitive to stock prices in election years. Julio and Yook (2012) find that corporate investment rates drop by an average of around 5% in the pre-election period for a sample of 48 countries. Our paper complements these papers and shows how policy uncertainty affects cross-border capital flows. We also contribute to the literature focused on the causes of the sluggish recovery following the financial crisis. Baker, Bloom, and Davis (2012) construct anindexofeconomicpolicyuncertaintyand findevidenceconsistentwiththehypothesisthat abnormally high levels of policy uncertainty is responsible for a significant amount of unemploymentandslowgrowth. Whilewedonotaddresstherecessiondirectly,ourresultssuggest thatthelinkbetweenpolicyuncertaintyandinvestmentislikelyacausalone. Ourresultsalso support the use of election timing as a proxy variable for fluctuations in political uncertainty overtime. 8

2. Data Description 2.1. Cross-Border Investment Data ThisstudyconsidersinvestmentsabroadbyU.S.investorsintheformofforeigndirectinvestmentandforeignportfolioinvestmentbetween1994andJune2010. Thesampleincludes informationondirectinvestmentto43countriesandportfolioinvestmentto44countries. The FDI data set is drawn from the Survey of U.S. Direct Investment Abroad conducted by the U.S. Bureau ofEconomicAnalysis. U.S. direct investmentabroad isdefined as ownershipby a U.S. investor of at least 10 percent of a foreign business. The direct investor is known as a U.S. parent and the U.S.-owned foreign business is known as a foreign affiliate. FDI flows capture the funds that U.S. parents provide to their foreign affiliates including equity investment, intra-company loans and reinvested earnings. FDI flows (reported in U.S. dollars) are measured on a quarterly frequency, which allows us to track the changes in the flows around the election cycles that cannot be captured in lower-frequency data such as annual data provided by UNCTAD. FDI positions, which are stocks and cumulative, are reported annually and measurethetotaloutstandinglevelofU.S. direct investmentabroad atyear-end. Theforeignportfolioinvestment(FPI)datacontaininformationonnetpurchasesoflong-termforeign securities, both debt and equities, by U.S. residents. We use Bertaut and Tryon’s (2007) estimates of monthly bilateral FPI flows and positions data maintained by the Federal Reserve. Bertaut and Tryon adjust the FPI data collected by the Treasury International Capital (TIC) reporting system to alleviate the biases pointed out by previous studies.6 The resulting estimates are consistent with various officially reported data (Curcuru, Thomas, Warnock and Wongswan(2011)). 6Previous studies suggest that the so-called TIC data need adjustments regarding acquisitions of equity through stock swaps, principal repayment flows on asset-backed corporate securities, and financial center biases, among others. (Chuhan, Claessens, and Mamingi (1998), Griever, Lee, and Warnock (2001), Thomas, Warnock,Wongswan(2004),WarnockandCleaver(2003),andWarnockandWarnock(2005)) 9

Panel A of Table 1 summarizes annual FDI and FPI flows by country (in $US millions). Note that quarterly FDI flows and monthly FPI flows are annualized to generate comparable summary statistics. The average annual FDI flows range from a low of $65 millionin Greece toahighof$24billiontotheUnitedKingdomandtheNetherlands. FPIismadeupofforeign portfolio equity investment(FPEI) and foreign portfolio debt investment(FPDI). The highest average annual FPI flow is $43.6 billion to the United Kingdom with $24.8 billion in equity investmentand$18.8billionindebt investmentwhilethelowestis–$2.1billionto Singapore with $311 million in equity investment and –$2.4 billion in debt investment. The negative figureindicatesthatU.S.investorssoldmoreSingaporeandebtsecuritiesthantheypurchased thesecuritiesduringthesampleperiod. 2.2. Election Data Our measure for variation in policy uncertainty is the timing of national elections held between January 1994 and June 2010. The detailed election information is obtained from a variety of sources. The primary source for election and regime change data is the Polity IV database maintained by the Center for International Development and Conflict Management at the University of Maryland. This database contains annual information on the regime and authority characteristics of all independent states with total populationsgreater than 500,000. The second major source of information is the World Bank Database of Political Institutions. This source provides information about electoral rules and the classification of political platforms for the elected leaders and candidates. We supplement the election data with various internet sources7 for cases in which the election information is missing from the Polity IV databaseortheDatabaseofPoliticalInstitutions. 7Otherinternetsourcesincludehttp://www.cidcm.umd.edu/polity/data/, http://www.binghamton.edu/cdp/era/searchera.html,andhttp://www.electionresources.org/. 10

WecollecttheelectiondatafortheU.S.and44destinationcountriesforwhichthedataon bilateral investmentwith the U.S. are available. We focus on elections in which the choice of nationalleaderor executiveauthorityis made. Weincludepresidentialelections forcountries with presidential systems, and legislativeelections for countries with parliamentary systems. Somecountrieshaveahybridsystemcombiningelementsofbothparliamentaryandpresidentialdemocracy; apresidentand aprimeministercoexistwithbothpresidentialand legislative elections held nationally. In such cases, the constitutional framework and practice are examined in greater detail to understand how executive power is divided between the two leaders, andtheelectionassociatedwiththeleaderwho exertsmorepoweroverexecutivedecisionsis selectedforthestudy. Thedatainclude31countrieswithlegislativeelectionsand14countries withpresidentialelections,resultingin totalof184nationalelectionsforoursampleperiod. An important characteristic of national elections is whether the timing of elections is exogenously specified by electoral law. Our identification assumption is that the timing of national elections is correlated with changes in policy uncertainty but uncorrelated with other determinantsofFDI flows. There maybe someconcern that thetimingofelections isa function of economic conditions in a recipient country. In some electoral systems a government can be dissolved before the expiry of its full term for various reasons and an election is then normallycalled toform anewgovernment. Thepotentialcorrelationbetween electiontiming and economic conditions may confound the effect of policy uncertainty on FDI flows. For example,Ito(1990)showsthatthetimingofJapanesegeneral electionsisconsistentwithopportunistictimingofincumbentscallingelectionswheneconomicconditionsaregood. While opportunisticelectiontimingislikelytobias againstthefindingdampenedFDI flowsinelection periods, we classify countries as having either exogenous timing or endogenous timing to address the potential endogeneity. All countries with a record of early elections are classified as having endogenous timing. All presidentialelections in the sampleare held on a fixed basis and as such are classified as having exogenous timing. For the remaining countries, we 11

examine electoral laws and practices as well as the timing classification by Alesina, Cohen, and Roubini (1992).8 Our classification procedure results in 19 countries with fixed election timingand 26countries withflexibletiming. Panel A of Table 2 summarizes the election data. For the countries in our sample, we observe an election in each country every 16.4 quarters on average. The average length of term forelected nationalleaders inoursampleis 4.4years. Ofall theelectionsinthesample, 73.6% take place in parliamentary systems. 45.3% of the elections in our sample are fixed in time by electoral law and hence outside the control of incumbent politicians. The remaining electionsareinsystemsinwhichthereisamechanismforcallingelectionspriortotheexpiry of the term of the government. We observe frequent turnover in leadership, with 56.4% of theelectionsresultinginachangein thegovernmenthead and 48.9%resultinginachangein the ruling party. There is a large amount of dispersionin the magnitudesin margin of victory across elections. On average, the winner received 41.7% of the vote, compared to 28.6% for therunner-up. 2.3. Country Characteristics Thecountry-levelcontrolvariablesaremotivatedbypriorresearchthatexaminethedeterminants of cross-border capital flows.9 Following the literature, we employ several variables that capture macroeconomic and institutional characteristics of the destination country. The WorldBankdatabaseisourprimarysourceforthemacroeconomicvariablesincludingrealper capita GDP, government spending, exports, and imports. We also obtain monthly exchange rate data from IMF International Financial Statistics. Monthly returns on countries’ stock marketindicesaredrawnfromDatastreamandBloomberg. Dataonmonthlygovernmentsta- 8Alesina, Cohen, and Roubini (1992) classify the timing of elections as exogenous or endogenous for 18 developedcountries. 9See,forexample,Albuquerque,Loayza,andServen(2005),FrootandStein(1991),andBlonigen(1997) 12

bilityratingsarefrom PoliticalRiskService’s InternationalCountryRisk Guide(ICRG). The government stability index assigns numbers between 0 and 12, where higher values indicate more stable governments. The government stability index is time-varying and assesses the government’sabilitytocarry outitsdeclared programs anditsabilitytostay inoffice. We obtain an annual measure of the degree of checks and balances for each political system in our sample from the Database of Political Institutions. The metric is intended to capture the number of decision makers whose agreement is necessary for the approval of policy changes. It is calculated as the numberof veto players in the politicalsystem at a givenpoint in time based on the prevailing electoral rules and laws. It also takes into account whether the executiveand legislativebranches of governmentare controlled by the same party, which effectivelyreducesthechecksandbalancesrelativetohavingdifferentpartiescontrollingseparate branches of government. In presidential systems, the count is increased by one for the president and increased by one for each additional legislative body. For parliamentary systems, the count is increased by one for the prime minister and increased by the number of parties included in the governing coalition. The number is reduced if the party of the executiveisthesameasthelargestpartyinanyparticularchamberofgovernment. Inourempirical analysis, the checks and balances measure provides us with variation, both within and across countries,inthedegreetowhich policychanges are likelyfollowinga givenelection. Panel B of Table 2 summarizes the characteristics of 44 destination countries. For our sample, the average government stability rating is 7.82. GDP per capita has a mean value of $9,183 per year and the median of $3,273. The average government consumption is 16% of GDPandtheaveragemonthlyreturnonstockmarketindicesis1%. Tradeopenness,measured as thesumofexportsand importsscaled by GDP, averages 79%ofGDPacross countries. 13

3. Empirical Results This section tests the hypothesis that increased uncertainty around national elections reduces FDI flows. The empirical analysis exploits the timing of national elections around the world to identify variation in policy uncertainty. We begin by discussing how we measure FDI flows. We then turn to a regression analysis to measure the impact of policy uncertainty on FDI flows. We then explore variation across countries and elections and finish with an examinationofforeignportfolioinvestmentflowsaround electioncycles. 3.1. Measuring FDI Flows Followingtheliterature,wescaletheflowsoruseavariancestabilizingtransformation,as described below. The first measure is the growth in the stock of FDI, similar to the measure employedbyBaker,FoleyandWurgler(2009). ThismeasureistheratiooftheU.S.FDIflows to country j in quartert tothecumulativeU.S. FDIpositionin country j at theend ofquarter t−1, as follows: US→j FDI t FDI/Position = , jt US→j Position t−1 where J is the number of countries in our sample. The second measure captures the U.S. FDI flows to country j during quarter t as a proportion of total U.S. FDI flows around the worldin quartert−1. That is, US→j FDI t FDI/Total = . jt (cid:229) J FDI US→j t−1 j=1 14

This measure is intended to capture the share of total FDI flows going to each destination country in our sample. The measure is similar in spirit to that employed by Dewenter (1995) who scales cross-border M&A flows into the U.S. by U.S. domestic acquisition activity. The thirdmeasureweemploy,similartothatusedbyFrootandStein(1991),istheU.S.FDIflows toarecipientcountryinagivenquarterscaledbythelaggedGDPoftherecipientcountry. The finalmeasureisavariationofthelogtransformationusedbyBusseandHefeker(2007). Since the FDI flows are measured on a net basis, some country-quarter observations have negative values. Topreservetheobservationswithnegativevalues,BusseandHefeker(2007)usedthe transformation ln FDI + (FDI2 +1) . (cid:16) jt q jt (cid:17) Forrobustness,werun all regressionswitheach ofthefourmeasures ofFDIflows. Weconstructthecorrespondingfourmeasuresforforeignportfolioinvestment(FPI)flows. Panel B of Table 1 presents the summary statistics for raw FDI and FPI flows, flows to GDP, flowstoposition,andthelogtransformofBusseandHefeker(2007). Foreignportfolioequity investment(FPEI) flows and foreign portfoliodebt investment(FPDI) flows are reported separately. FDI flows are, on average, somewhat larger than FPI flows, while FPI flows display moretime-seriesvariation. 3.2. Measuring Policy Uncertainty As noted by Rodrik (1991), a major obstacle to identifying a link between policy uncertainty and changes in capital flows is the availability of an adequate proxy for variation in uncertainty due to difficulties in measurement and possible endogeneity. Major events that create policy uncertainty are likely correlated with economic conditions, making it difficult to establish the direction of causality between changes in uncertainty and changes in capital flows. To deal with this challenge, we employ the identification strategy of Julio and Yook 15

(2012)andDurnev(2010)andusethetimingofnationalelectionsaroundtheworldasameasure of variation in political uncertainty. Specifically, we create an election timing dummy variable equal to one in quarter t if the election occurs in the second half of quarter t or in the first half of quarter t+1.10 We use election timing as a proxy for variation in political uncertainty of the destination country as the timing of elections is out of the control of the U.S. firms and investors and indeed for a large part of our sample the timing of elections is fixed byelectoral lawand henceindependentofgeneral economicconditions. The identification strategy requires that political uncertainty is indeed higher during electionperiods. A growingliteraturehasdocumentedthattheprobabilityofpolicychangesdoes appear to increase around elections. Several papers have found such evidence in financial markets. Bialkowski, Gottschalk, and Wisniewski (2008) and Boutchkova et al. (2012) find that volatility is significantly higher than normal during election periods around the world. Boutchkova et al. (2012) document that equity return volatility is higher around elections in politically sensitive industries. Bernhard and Leblang (2006) document changes in bond yields, exchange rates, and equity volatility around elections, and show that these changes are larger during close elections. More recently, Baker, Bloom and Davis (2012) construct an index of economic policy uncertainty in the United States composed of news media references to policy uncertainty, future expiration of federal tax code provisions, and forecaster disagreement over inflation and government purchases. Their index spikes upward around U.S. presidential elections, consistent with the premise that political uncertainty is higher in electionperiods. 10Theresultsarerobusttodifferentcutoffpointsfortheelectiontimingdummy. 16

3.3. FDI Flows around Elections As a first step, we examine variation in FDI flows from the U.S. to destination countries around electiondates byestimatingthespecification 2 FDI =g +d + (cid:229) b Election +e , (1) jt j t k j,t+k jt k=−2 where g captures country fixed effects and d time fixed effects of a quarterly frequency. We j t construct an election dummy variable to capture the quarter leading up to the election. The election variable is set equal to one if a national election is held in the second half of a given quarter or in the first half of the next quarter, and zero otherwise. Four additional dummy variablesareincludedtoexaminepossiblechangesinFDIflowsinthetwoquarterspreceding theelectionquarterandthetwoquartersjustfollowingtheelection. Thespecificationin(1)is intended to capture within-country variation in FDI flows around the election cycle, with no additional control variables aside from the fixed effects. Standard errors are clustered at the countrylevel. Table 3 reports the estimationresults for specification (1). Consistent with the hypothesis that policy uncertainty has a depressing effect on FDI, the U.S. FDI flows to a destination countryarelowerinquartersinwhichanationalelectionisheldinthedestinationcountry. The effect is economically and statistically significant. The coefficient for the specification using FDI/Position as the dependent variable suggests that FDI flows from the U.S. to a recipient country are 11.9% lower in election quarters relative to the country mean annual rate of FDI flows. The signs and magnitudes for the other transformations of FDI flows yield similar results. 17

3.4. Including Determinants of FDI Flows We next introduce to our specification various time-varying country characteristics that can potentially affect FDI flows11. GDP per capita, for example, is expected to control for the effect of a host country’s wealth on FDI decisions. Higher volatilityin GDP growth or in realexchangeratesisassociatedwithmacroeconomicinstability,whichisconsideredtodrive awayFDI.Changesintherealexchangerateaffecttherelativewealthlevelsofforeignanddomesticinvestorsandmayleadtochangesininvestors’actualrelativepurchasingpower(Froot andStein(1991)andKleinandRosengren(1994)). Taxisalsoanimportantconsiderationfor thechoiceofFDIlocation(HinesandRice(1994))becausehighertaxesgenerallydiscourage private investment. Further, tax influences the capital structure decision and the choice betweeninternalandexternalfinancingformultinationalfirms(Desaietal. (2004)). Stockmarket valuation may drive FDI, especially cross-border mergers and acquisitions (Baker, Foley, and Wurgler (2009)). Undervaluationin the host-countrystock market may present an attractiveinvestmentopportunityfor international investors (Shleifer and Vishny(1992), Krugman (1998), and Aguiar and Gopinath (2005)). Also, the overvalued market of the source country may provide multinational firms with relatively low-cost funding for overseas investment (Shleifer and Vishny (2003)). Trade openness may influence FDI decisions in two opposite manners. Larger openness may further facilitate FDI if foreign production requires parent firms to supply production parts to their affiliates in host countries. Also, if a firm expects its production presence in a foreign market with oneproduct to generate demand for otherproducts of the firm, larger trade openness may promote FDI (Lipsey and Weiss (1984)). On the otherhand,ifmultinationalshavetochoosebetweenforeignproductionandexportsbasedon considerationson tariffs,transport costs,and locationadvantages (Markusen (1995)), smaller tradeopennessmay lead tohigherFDI. 11See Albuquerque, Loayza, and Serven (2005) and Daude and Fratzscher (2008) for an extensive list of potentialdeterminantsofcross-borderinvestment. 18

In addition to the time-varying country characteristics, we also include country fixed effects to control for time-invariant country characteristics associated with FDI decisions. Examples of such country characteristics include geographic and language proximity and legal origin (Daude and Fratzscher (2008)). Capital controls may also have a direct or indirect implication on FDI by influencing the foreign affiliates’ borrowingenvironments or repatriation decisions (Desai et al. (2006)), or by influencing the volatility of macroeconomic conditions (Aizenman (2003), Bekaert, Harvey, and Lundblad (2006)). The liberalization dummy variable, a common measure of capital control, is largely time-invariant for our sample period because all countries in our sample with one exception have already been liberalized prior to thebeginningofoursampleperiod.12 We estimatetheregression 2 FDI =g +d + (cid:229) b Election +X ′q +e , (2) jt j t k j,t+k jt k=−2 whereX isavectorofcontrolvariables,whichincludeGDPpercapita,GDPgrowth,volatility ofGDPgrowth,theICRGgovernmentstabilityratings,governmentconsumptionasaproportion of GDP, lagged stock market returns and volatility, changes in exchange rates, exchange rate volatility, and trade openness of the recipient country. Country and time (quarterly frequency)fixedeffectsareincludedineachspecification. Thecoefficientontheelectiondummy variable can be interpreted as the difference in the within-countryconditional mean FDI rate, controllingfortheotherdeterminantsofFDIflows. Table 4 reports the results of the FDI regressions controlling for known determinants of FDI. The inclusion of country control variables does not change the economic magnitude of the election effects. Controlling for country characteristics, we find that FDI flows from the 12Wecross-checkoursamplecountriesagainsttheliberalizationdateprovidedbyBekaertandHarvey(2000) andCampbellR.Harvey’swebsite. Among38samplecountriesforwhichtheliberalizationdateinformationis available,onlyonecountryliberalizeditsfinancialmarketafter1994(SouthAfricain1996). 19

U.S. to other countries decline by 11.2% in an election quarter relative to non-election quarters. The reduction in FDI flows in the quarter leading up to the election is both statistically and economically significant. The control variables exhibit signs consistent with those documentedbyextantstudies. GDPgrowth,whichmeasuresimprovementsinoverallproductivity as reflected in economic growth, has a positive sign across the board as predicted. The positive sign on trade openness suggests that larger trade openness serves to attract FDI into the recipient country. This is consistent with the previous country-level studies documenting net complementarityeffectsbetweenexportsandFDI13.GDPgrowthvolatilityshowsnegativeassociationwithFDIinflowsaspredicted,buttheresultisnotstatisticallystrong. Exchangerate volatility is insignificant in all four regressions. This may be because our sample consists of bothdevelopedandless-developedcountrieswhilemacroeconomicinstabilityisanimportant concern primarily in less-developed economies. Previous-quarter market return is significant inonlyonecase,suggestingthattheassociationbetweenFDIflowsandlocalmarketvaluation is ratherweak for oursample. An increase in governmentexpendituresis expected toact as a deterrentforFDIinflowsasincreasedgovernmentspendingfundedbyhighertaxationislikely to discourage private investment. However, the results show that the ratio of government expenditure to GDP is insignificant except for one case. Finally, changes in real exchange rates is insignificant. This is consistent with the empirical literature documenting that the effect of changes intheexchangerate onFDI isunclear(seeBlonigen (1997)forliteraturereview). 3.5. Domestic and Foreign Sources of Uncertainty Firms and investors are exposed to two sources of policy uncertainty in a cross-border setting, that of the home country and that of the destination country. In this section, we 13Whilelittleevidenceofsubstitutioneffectsisfoundontheaggregatelevel,someevidenceisdocumentedin lessaggregatedata(Blonigen(2001)). 20

augmenttheempiricalspecificationandincludetheU.S.electionstoestimatepossibleeffects ofsourcecountryelectionson FDIflows. We estimatetheregression 2 2 FDI =g +d + (cid:229) b Election + (cid:229) d Election +X ′q +e , (3) jt j t k j,t+k l US,t+l jt k=−2 l=−2 where j denotes country and t indexes time. We construct additional dummy variables designed to capture the U.S. election effects in the quarter leading up to the election as well as the two quarters before and after the election quarter. As before, we include the election dummiesforthedestinationcountriesandcontrolvariables. Yearfixedeffectsareincludedto control for global trends in FDI. Note that time fixed effects of a quarterly frequency are replaced by year fixed effects whenever U.S. election dummies are included in the regressions. Because all flows in our sample originate from the US, The US election effects are harder to identify than the destination country effects in which different countries have elections in differentquarters. Table5reportstheestimationresultsforthespecificationincludingtheU.S.electiondummies. Tosavespace,wedonotreportthecoefficientsforthecontrolvariablesandtheelection dummiesfort±2. Asbefore,wefindthatU.S.FDIflowstoadestinationcountryarelowerin country-quartersinwhichthecountryholdsanationalelection,controllingforchangesinthe economic environment. The economic magnitude is considerable. The results suggest that, controlling for country-level characteristics and the timing of U.S. elections, FDI flows as a percentage ofcumulativeFDI stock in thegivenrecipient country drop by 12% in thequarter leadinguptoanationalelectionintherecipientcountry. WealsoobservethatFDIflowsoriginatingfromtheU.S. tendto beloweringeneral duringU.S. elections,suggestingthatpolicy uncertainty in the source country depresses flows to host countries until the source country’s electoral uncertainty is resolved. Taken together, the results suggest that firms and investors respondto bothforeignand domesticsources ofpoliticaluncertainty. 21

3.6. Exogenous vs. Endogenous Election Timing One concern with the above analysis is that, for some countries in our sample, national electionsmaybecalledearlybythenationalleaderorlegislativebody. Earlyelectionsraisea possibility that election timing may be correlated with economic conditions and cause a bias inourestimatesoftheelectioneffects. Whilesuchcorrelationdoesnotappeartobegenerally observed14 in the literature, there is some evidence for such correlation in Japan. Ito (1990) findsthatelectionsinJapanareheldinperiodsofeconomicexpansion,suggestingopportunistic behavior of the incumbent politicians. In our sample, we find that within-country GDP growthis,onaverage,1.96%higherintheperiodjustbeforeanelectionincountriesthathave flexibility over election timing, while we find no statistical difference in growth rates around the election cycle for countries with fixed election timing15. Higher GDP growth around the election cycle is consistent with either opportunistic timing or a reluctance to call elections when growth is relatively low. We note, however, that the results in Table 4 show that FDI flows are strongly pro-cyclical and hence the possible opportunistic behavior of incumbents is likely to act as a bias against finding a negative effect attributable to electoral uncertainty. To address the concern that FDI flows may be confounded with strategic election timing, we estimate the FDI regressions for the subsampleof countries for which the timing of elections isfixed intimeby electoral lawand henceorthogonaltothebusinesscycle. Table 6 reports the results for the subsamplewith exogenous election timing. For brevity, wehaveonlyreportedthecoefficients fortheelectiondummyvariables,althoughthecountry 14Alesina,Cohen,andRoubini(1992)examine14OECDcountrieswithflexibleelectiontimingandfindthat such an association between election timing and economic conditionsis not present in any of those countries excludingJapan. 15Specifically,weestimatetheregression GDPgrowth =a +g +b Election +e , jt j t jt jt wherea isacountryfixedeffectandg isayear/quarterfixedeffect. Weestimatetheregressionseparatelyfor j t countrieswithanoptiontocallanearlyelectionandforthosecountrieswithfixedelectiontime. Thecoefficient b capturesthewithin-countrydifferenceinGDPgrowthbetweenelectionandnon-electionquarters.Resultsare availableuponrequest. 22

controls were included in the regression. The main results are present in the countries for which election timing is fixed. The magnitude of the coefficients for elections in both destination countries and the U.S. are similar to that for the full sample. FDI flows to countries holdinganelectioninaparticularquarterdropby9.5%comparedtonon-electionyearsinthe exogenouselection sample. This confirms that ourresultsare notdrivenby factors correlated withtheopportunistictimingofelections. 3.7. Variation in Electoral Uncertainty: Close Elections The Rodrik (1991) model suggests that the reluctance to invest in a recipient country will be higher when the country has a higher degree of uncertainty over future policy. To the extent that different candidates have different policy preferences, election uncertainty translates into policy uncertainty when the outcome is uncertain. In some cases, election outcomes are predicted with a great deal of confidence prior to the election day. Singapore, for example, has not experienced a change in the ruling party for many decades. However, some elections arecharacterizedbyverycloseracesinwhichtheoutcomeishighlyuncertainuntilthedayof theelection. In thissection,weinvestigatevariationinelectoral uncertaintyby usingelection voteturnoutsasaproxyforelectionuncertaintybeforetherevelationoftheelectionoutcome. We construct a dummy variable equal to one if the margin of victory for a given election is in the lowest quartile of the sample distribution of victory margins. In our sample, the 25th percentileforthemarginofvictoryis7.1%. Wetheninteractthecloseelectionindicatorwith the election dummy. We also construct an indicator variable to capture elections with wide victory margins and therefore likely to be associated with less uncertainty. We set a dummy variable equal to one if the margin of victory for a given election is in the highest quartile of the distribution. We include this dummy in some specifications to capture whether elections withmorecertain outcomesalso create cyclesinFDI flows. 23

Table 7 reports the results of the specification with the close election interaction. For the sake of brevity, we report only the regressions with the FDI/position variable as the left hand sidevariable. Thefirst two columnsreport the estimatesfor thefull sampleof countries and the last two columns report the results for only the countries with exogenous election timing. The interaction term for close margins of victory is statistically significant in all four regressions. This finding suggests that cycles in FDI flows around national elections have a larger magnitude when the uncertainty regarding the election outcome is higher. This result is in line with Rodrik (1991), which predicts the effect of policy uncertainty is increasing in thelikelihoodofpolicychange. Thisresultalsostrengthenstheinterpretationthatthepatterns we are finding in the data are related to policy uncertainty and not likely related to any other underlying mechanism. The coefficient on the wide margin of victory interaction term is positive and statistically insignificant for both the full and exogenous timing samples. The combined results suggest that declines in FDI are largest when the margin of victory is very tightandnegligiblewhen marginsofvictoryare wide. 3.8. Variation Across Countries In this section, we investigatethe interaction between election effects and the factors that capture the potential likelihoodand magnitudeof policy shifts after elections. The prediction is that countries that are more susceptible to policy reversals will experience larger election cyclesinFDIflows. Weexaminevariationalongfourdimensions. First,welookatdifferences inICRGgovernmentstabilityratings. Second,weexaminedifferencesinthedegreeofchecks and balances on executive authority, based on counts of the number of veto players within a politicalsystemat anypointintimeas measuredbytheWorldBank. Third,wesortcountries accordingtoWorldBank’sdevelopmentindex,basedontheideathatlessdevelopedcountries may be more exposed to policy uncertainty compared to wealthier countries. Finally, we 24

examinewhetherthedegreeoftradeopennessaffectsthesensitivityofFDItoahostcountry’s politicalcycles. We estimatetheregression 2 FDI =g +d +a ·Z +a ·Z ·Election + (cid:229) b Election +X ′q +e , jt j t 1 jt 2 jt jt k j,t+k jt k=−2 whereZ isatime-varyingcountrycharacteristicintendedtocapturedifferencesinthepropenjt sity for large policychanges after elections. Table 8 reports theresults of theFDI regressions including the interaction terms with the four factors listed above. To save space, we only report election dummy variables and the interaction terms. We run regressions using all four FDImeasuresbutonlyreportresultsusingtheFDI/positionvariable. Thefirstcolumnreports the results including an interaction between the election dummy variable and ICRG government stability ratings. The point estimateof the interaction term is positiveand significant.16 Whilethe statisticalevidenceis not overwhelming,it is consistentwith the viewthat election effects in countries withmore stablegovernments(higher ICRG rating)are mitigatedrelative to countries with less stable political systems. This suggests that policy uncertainty is more materialin countriesinwhich governmentsare lessstable. Column2ofTable8presentstheresultswithaninteractiontermbetweenchecksandbalancesandtheelectiondummy. Thecoefficientontheinteractiontermhasthesignaspredicted thoughnotsignificant. TheregressionsusingtheotherFDImeasures(unreported)producethe samesignwithstatisticalsignificance. Theresultssuggestthattheeffectofpolicyuncertainty is less severe in countries where the power of the national leader is relatively restricted in terms of making policy changes after taking office. Electoral uncertainty, therefore, appears to havelarger effects on FDI flows when election outcomes may lead to relativelyunchecked policychanges by thenationalleader. 16RegressionsusingotherFDImeasuresproducesimilarresults. 25

Column 3 of Table 8 reports the results of the regression that includes a high-income interaction term with the election dummy. A high-income dummy is set equal to one if a country is classified as a high-income country in a given year by World Bank. World Bank’s development classification is based on the country’s gross national income (GNI) each year. Somewhatsurprisingly,wefindthatelectioneffectsonFDIflowsarenotsignificantlydifferent between high and low-income countries, suggesting that policy uncertainty is not limited to emergingmarketsandlessdevelopedcountriesbuthasimplicationsindevelopedcountriesas well. Evenrelativelywell-developedcountriesexperiencecyclesinFDIflowsaroundelection time. The results remain similar when the other FDI measures are employed (unreported). Alternatively,werepeat thetestusingGDPpercapitaand find similarresults.17 Column4reportstheresultsincludinganinteractionbetweentheelectiondummyvariable and trade openness dummy variable. The coefficient on the interaction term is positive and significant,suggestingthatwhenaneconomyisopentointernationaltrade,FDIdecisionsare less sensitive to local political environment. This is consistent with the extant studies documentingthatwhenaneconomyismoreopen,capitalflowsarelesscorrelatedwithacountry’s institutional quality. Albuquerque, Loayza, and Serven (2005) show that FDI is increasingly more dependent on global factors and less dependent on country-specific factors, suggesting that increased market integration leads to a greater role of global risk factors. Examining industry-leveltradeopenness,GiovanniandLevchenko(2009)showthatasectorthatismore opentointernationaltradeislesscorrelatedwithdomesticeconomiccycles. Inarelatedvein, Fratzscher and Imbs (2009) and Ju and Wei (2010) examinehow financial openness interacts withacountry’sinstitutionalqualityandshowthatthesensitivityofFDItoinstitutionalfactors dependson thedegreeofthecountry’sfinancial openness. 17Wesetacountry-quartertooneiftheGDPpercapitainagivencountry-quarterisabovethemedianofthe distribution and to zero otherwise. We also simply interact GDP per capita with the election dummy and find similarresults. 26

3.9. Additional Robustness Weperformedseveralrobustnessteststochecktheconsistencyofourresults. Inresultsnot reported here18, we perform the followingrobustnesstests: (1) we use raw FDI flows and the naturallogofFDIflowsasthedependentvariable;(2)weclusterstandarderrors attheyearly andquarterlylevelratherthanatthecountrylevel;(3)weincludeadditionalcontrolvariables, suchasvolatilityoftermsoftradeandleadandcontemporaneousvaluesofstockreturns;and (4) we estimate the FDI regressions on a country-by-country basis and take the average of the coefficients across the country regressions to make inferences; and (5) we consider only elections in which the incumbent national leader is not running for re-election. In every case listedabove,theresultsare similarto thosereported inthetables. 4. Policy Uncertainty and Irreversibility In this section we provide evidence that the negative relationship between FDI flows and election timing reflects increased political uncertainty rather than some other election related mechanism. The political uncertainty literature has documented election effects on corporate investment (Durnev (2010), Julio and Yook (2012), Jens (2012)), borrowing costs (Gao and Qi (2013), and the IPO decision (Colak, Durnev and Qian (2013)). An unresolved question is whetherelectiontimingcaptures uncertaintyor whethersomeothertypeofpoliticalmechanism is causing the observed relationships. For example, opportunistic models of political business cycles (PBC), beginning with Nordhaus (1975), incumbents attempt to manipulate fiscalandmonetarypolicytoincreasetheprobabilityofre-election. WhileopportunisticPBC models typically predict an increase in economic activity prior to an election, it is possible thatFDIflowsmaydeclinebecauseactionstostimulatetheeconomypriortotheelectionmay crowd out privateinvestment. Julio and Yook (2012)show that governmentspending, money 18Availablefromtheauthorsuponrequest 27

supply, interest rates and inflation do not vary across the election cycle in a similar sample of countries. However, it is possible that other, unobservable political activities near election timecouldhavesomeeffect on FDIflows. Our identification strategy allows us to disentangle uncertainty effects from other mechanism by comparing two sets of flows into the same country and same time period that have different sensitivities to uncertainty but otherwise share the same return properties. Specifically, we compare FDI and FPI flows (both equity and debt) around the election cycle. The investment under uncertainty literature such as Bernanke (1983) and Rodrik (1991), among others, shows that irreversibility of investment generates an incentive to wait when uncertainty is high. When capital investment is costly to undo, investment decisions become very sensitive to the information environment, and firms and individuals have a strong incentive to wait for some degree of uncertainty to unravel before committing to investment projects. To the extent that government policy choices are relevant to expected payoffs for investment, irreversible investment will be sensitive to the policy uncertainty, as in Rodrik (1991). Empirically, Guiso and Parigi (1999) show that uncertainty has a stronger effect on firms that cannot easily resell capital equipment in secondary markets and Kim and Kung (2013) show thatfirmswithrelativelylessassetredeployabilitydecreaseinvestmentmorewhenuncertainty ishigh. FDI flows are, by definition, long-term, relationship-based investments. The IMF defines FDIas follows: The BPM519 defines FDI as a category of international investment that reflects the objective of a resident in one economy (the direct investor) obtaining a lasting interest in an enterprise resident in another economy (the direct investment enterprise). The lasting interest implies the existence of a long-term relationship between the direct investor and the direct investment enterprise, and a significant 19BalanceofPaymentsManual,fifthedition(IMF,1993). 28

degree of influence by the investor on the management of the enterprise. A direct investment relationship is established when the direct investor has acquired 10 percent or moreoftheordinarysharesorvotingpower of anenterpriseabroad. Caballero and Hammour (1998) classify FDI as relationship-specific and argue that the specificity reduces the flexibility of decisions. While FDI flows are typically considered relatively irreversible due to specificity, FPI flows are considered to be easier to reverse (Razin, SadkaandYuen(1998)). Comparingdifferenttypesofinternationalequityinvestments,GoldsteinandRazin(2006)arguethatFDIismorecostlytoreversethanFPIifinvestorsfacedwith liquidity shocks need to sell their investments before maturity. Because direct investors who act effectively as managers of firms are more informed than portfolio investors, they would be forced to sell at a lower price that reflects the discount for information asymmetry. Thus, our identification strategy compares two sets of equity flows into the same country that have similar sensitivityto the macroeconomic environment but differ with respect to the degree of reversibilityand hencesensitivityto politicaluncertainty. To comparetheeffectsofelection timingonFDI andFPI flows, weestimatethesystem 2 2 FDI =g +d + (cid:229) b Election + (cid:229) d Election +X ′q +e (4) jt j t k j,t+k l US,t+l jt k=−2 l=−2 2 2 FPI =g ′ +d ′ + (cid:229) b ′ Election + (cid:229) d ′ Election +X ′h +n , jt j t k j,t+k l US,t+l jt k=−2 l=−2 where FPI represents foreign portfolio equity investments (FPEI), foreign portfolio debt investments (FPDI), or the sum of debt and equity investment flows. The right-hand-side variables include election dummy variables for both the destination and source countries and a collection of control variables as defined previously. Since both FDI and FPI share similar determinants, we estimate the system using seemingly unrelated regression estimation. The estimationprocedurealso allowsus totestdifferences in coefficientsacross equations. 29

Table 9 reports the estimation results for the seemingly unrelated regressions. The first column reports the coefficients from the FDI regression. As with the previous results, FDI flows are significantly lower in the pre-election period. The following three columns report theestimatesfortheequityFPI,debtFPI,andcombinedFPIflows. Thetableshowsthatthere arenosignificantchangesinFPIflowsacrosstheelectioncycleforeitherequityordebtflows. The difference in coefficients on the election indicator variable between FDI and equity FPI flows is significantly different at the 1% level. The coefficients on the election indicator are indistinguishablefrom zero also forthedebtflows and totalFPI flows. Thesensitivityof FDI to the election cycle and the absence of an effect for FPI flows suggests that the underlying mechanism driving the pre-election declines in FDI flows is heightened political uncertainty priortotheelectionoutcome. 5. Conclusion In this paper, we examine the relationship between cross-border flows of capital and uncertainty over future government policy. Using the timing of national elections as a proxy for exogenous variation in policy uncertainty, we find that policy uncertainty has a negative impactonFDIflowsfromtheU.S. parentfirms totheiraffiliatesin43countries. Specifically, we document cycles in FDI flows around the timing of elections in both destination countries and the source country. The average FDI rate drops by approximately 12% compared to non-election years, all else equal. The results suggest that the uncertainty related to electionoutcomesleadseconomicagentstopostponeprivateinvestmentabroaduntilsomedegree of the uncertainty is resolved. The magnitudes of the declines in FDI flows are significantly larger than the effects of policy uncertainty on domestic investment, suggesting that foreign flowsofcapitalaremoresensitivetothepolicyenvironment,ashypothesizedbyDixit(2011). We find that the effect is stronger around elections with close outcomes and in countries with 30

less stable political systems and fewer checks and balances on executive power. Election effectsaremitigatedwhentradeopennessislarge. Wealsofind thattheresultsarerobusttothe possible endogeneity of election timing as the results are similar for the sample of countries forwhichthetimingofelectionsis fixedby electoral law. Two additional findings emerge from the empirical analysis. First, we find that policy uncertainty is not only an emerging market phenomenon. In fact, developed countries in our sample display mild cycles in FDI flows around elections, suggesting that policy uncertainty is important in the developed world as well, although the cycles are more amplified in developingeconomies. Second, we find thatFDI flows, whichare considered to berelativelymore irreversible than FPI flows, are more sensitive to policy uncertainty than are FPI flows. The difference in sensitivity between relatively irreversible FDI and FPI flows suggests a likely causal linkbetween heightenedpoliticaluncertaintyaroundelectiontimeanddeclinesin FDI flows. Any alternative theory would have to explain not only the reductions in FDI, but also the differential sensitivities of FDI and FPI flows to the election cycle. Among the existing theories,thepoliticaluncertaintymechanismbest fitsthetotalofourempiricalresults. Weviewourresultsashavingseveralcontributions. First,ourresultsarelargelyconsistent withtheimplicationsofvariousmodelsofdirectinvestmentunderuncertainty,inparticularthe Rodrik(1991)modelofpolicyuncertaintyandprivateforeigninvestment. Asfarasweknow, we are the first to provide evidence of political cycles in FDI flows. Second, we contribute to the literature on the determinants of FDI flows by identifying a political factor that leads to variation in FDI flows over time. Third, our results support the increasing use of election timing as a proxy for variation in political uncertainty. FDI flows are sensitive to the election cycle but FPI flows are not, suggesting that uncertainty is the mechanism underlying the results. Finally, we view our results as contributing to the recent debate over whether policy uncertainty depresses economicactivity in general and especially in the wake of the financial 31

crisis. Whilewedonotaddressthepost-crisisrecoverydirectly,ourresultsaresuggestivethat periodsofhighuncertaintyregardingpoliticaloutcomesdo havereal effects. 32

References Aguiar, M., and G. Gopinath. 2005. Fire-Sale FDI and Liquidity Crises. Review of EconomicsandStatistics87, 439–52. Albuquerque, R., N. Loayza, and L. Serven. 2005. World Market Integration Through the Lens ofForeign Direct Investors. JournalofInternationalEconomics66,267–295. Alesina,A.,G.Cohen,andN.Roubini. 1992. MacroeconomicPolicyandElectionsinOECD Democracies. Economicsand Politics5, 1–30. Aizenman, J. 2003. Volatility, Employment and the Patterns of FDI in Emerging Markets. JournalofDevelopmentEconomics72, 585–601. Baker, M., C. Foley, and J. Wurgler. 2009. Multinationals as Arbitrageurs: The Effect of StockMarketValuationsonForeignDirectInvestment. ReviewofFinancialStudies22,337– 369. Baker, S., N. Bloom, and S. Davis. 2012. Measuring Economic Policy Uncertainty. Working paper, Stanford University. Bekaert,G.,andC.Harvey. 2000. ForeignSpeculatorsandEmergingEquityMarkets. Journal ofFinance55, 565–613. Bekaert, G., C. Harvey, and C. Lundblad. 2006. Growth Volatility and Financial Liberalization. JournalofInternationalMoneyandFinance 25,370–403. Bernanke, B.1983. Irreversibility,Uncertainty,andCyclicalInvestment. QuarterlyJournalof Economics,85–106. Bernhard, W., and D. Leblang. 2006. Democratic Processes and Financial Markets: PricingPolitics,CambridgeUniversityPress. Bernhard, W., S. Riley, and P. Vaaler. 2012. How do International Bankers Vote in DevelopingCountryElections? WorkingPaper. Bertaut, C., and L. Pounder. 2009. The Financial Crisis and U.S. Cross-Border Financial Flows. Federal ReserveBulletin95,A147–A167. Bertaut, C., and R. Tryon. 2007. Monthly Estimates of U.S. Cross-Border Securities Positions. International Finance Discussion Papers 910, Board of Governors of the Federal ReserveSystem(U.S.). 33

Bhattacharya, U., N. Galpin, and B. Haslem. 2007. The Home Court Advantage in InternationalCorporateLitigation,JournalofLaw andEconomics,50, 625–659 Bialkowski, J., K. Gottschalk, and T. Wisniewski. 2008. Stock Market Volatility around NationalElections. JournalofBankingand Finance32, 1941–1953. Blonigen, B. 1997. Firm-Specific Assets and the Link Between Exchange Rates and ForeignDirect Investment. American EconomicReview87, 447–465. Blonigen,B. 2001. Insearch ofSubstitutionbetweenForeignProductionandExport. Journal ofInternationalEconomics53,81–104. Bloom, N., S. Bond, and J. Van Reenen. 2007. Uncertainty and Investment Dynamics. ReviewofEconomicStudies74, 391–415. Boutchkova, M., H. Doshi, A. Durnev, and A. Molchanov. 2012. Precarious Politics and Return Volatility. ReviewofFinancialStudies 25,1111–1154. Bulan, L. 2005. Real Options, Irreversible Investment and Firm Uncertainty: New Evidence from U.S. Firms. ReviewofFinancial Economics14, 255–279. Busse, M., and C. Hefeker. 2007. Political Risk, Institutions and Foreign Direct Investment. European JournalofPoliticalEconomy23,397–415. Caballero, R. 1991. On the Sign of the Investment-UncertaintyRelationship. American EconomicReview81, 279-288. Caballero, R., and M. Hammour. 1998. The Macroeconomics of Specificity. Journal of PoliticalEconomy106,724–767. Chen, Y., and M. Funke. 2003. Option Value, Policy Uncertainty, and the Foreign Direct InvestmentDecision. Hamburg InstituteofInternationalEconomicsdiscussionpaper. Chuhan, P., S. Claessens, and N. Mamingi. 1998. Equity and Bond Flows to Latin America and Asia: the Role of Global and Country Factors. Journal of Development Economics 55,439–463. Colak, G., A. Durnev, and Y. Qian. 2013. Derailed by Political Uncertainty: U.S. GubernatorialElectionsand IPO Activity,Workingpaper. Curcuru, S., C. Thomas, F. Warnock, and J. Wongswan. 2011. U.S. International Equity Investmentand Pastand ProspectiveReturns. AmericanEconomicReview 101,3440–3455. 34

Daude, C., and M. Fratzscher. 2008. ThePecking OrderofCross-Border Investment. Journal ofInternationalEconomics74,94–119. Desai, M., C. Foley, and J. Hines. 2004. A Multinational Perspective on Capital Structure Choiceand Internal CapitalMarkets. JournalofFinance59,2451-2488. Desai,M.,C.Foley,andJ.Hines. 2006. CapitalControls,Liberalizations,andForeignDirect Investment. ReviewofFinancial Studies19, 1433–1464. Desai, M., C. Foley, and J. Hines. 2008. Capital Structure with Risky Foreign Investment. JournalofFinancial Economics88, 534–553. Dewenter, K. 1995. Do Exchange Rate Changes Drive Foreign Direct Investment? Journal ofBusiness68,405–433. Dixit, A. 2011. International Trade, Foreign Direct Investment, and Security. Annual ReviewsofEconomics3, 191–213. Durnev, A. 2010. The Real Effects of Political Uncertainty: Elections and Investment SensitivitytoStock Prices. Workingpaper. Fratzscher, M., and J. Imbs. 2009. Risk Sharing, Finance, and Institutions in International Portfolios. JournalofFinancial Economics94, 428–447. Froot, K., and J. Stein. 1991. Exchange Rates and Foreign Direct Investment: An Imperfect Capital MarketsApproach. Quarterly JournalofEconomics106, 1191–1217. Gao, P., and Y. Qi. 2013, Political uncertainty and public financing costs: Evidence from U.S. municipalbond markets,workingpaper. Giovanni,J., and A. Levchenko. 2009. Trade Openness and Volatility. Review of Economics and Statistics91,558-585. Goldstein, I., and A. Razin. 2006. An Information-Based Trade-Off between Foreign Direct InvestmentandForeignPortfolioInvestment. JournalofInternationalEconomics70,271–95. Griever,W.,G.Lee, andF. Warnock. 2001. TheU.S.SystemforMeasuringCross-BorderInvestmentinSecurities: APrimerwithaDiscussionofRecentDevelopments. FederalReserve Bulletin,633–650. Guiso, L., and G. Parigi. 1999. Investment and Demand Uncertainty. Quarterly Journal ofEconomics,185–227. 35

Hermes, N., and R. Lensink. 2001. Capital Flight and the Uncertainty of Government Policies. EconomicsLetters 71,377-381. Hines, J., and E. Rice. 1994. Fiscal Paradise: Foreign Tax Havens and American Business. Quarterly JournalofEconomics109, 149-182. Ito, T. 1990. The Timing of Elections and Political Business Cycles in Japan. Journal of AsianEconomics1, 135–156. Jens,C.2012. InvestmentaroundU.S.GubernatorialElections. Workingpaper,Universityof Rochester. Ju, J., and S. Wei. 2010. Domestic Institutions and the Bypass Effect of Financial Globalization. American EconomicJournal: EconomicPolicy2010,2(4), 173-204. Julio,B., and Y. Yook. 2012. Political Uncertaintyand Corporate InvestmentCycles. Journal ofFinance67, 45–84. Kim, C., C. Pantzalis, and J.C. Park. 2012, Political Geography and Stock Returns: The ValueandRiskImplicationsofProximitytoPoliticalPower,JournalofFinancialEconomics, 106,196–228.. Kim, H., and H. Kung. 2013. How Uncertainty Affects Corporate Investment: The Asset RedeployabilityChannel. Workingpaper. Klein, M., and E. Rosengren. 1994. The Real Exchange Rate and Foreign Direct Investmentin theUnitedStates. JournalofInternationalEconomics36,373–389. Krugman,P.1998. Fire-SaleFDI.PreparedforNBERConferenceonCapitalFlowstoEmergingMarkets. mimeo,MIT. Lipsey,R.,andM.Weiss. 1984. ForeignProductionandExportsofIndividualFirms. Review ofEconomicsandStatistics66,304-307. Markusen, J. 1995. The Boundaries of Multinational Enterprises and the Theory of InternationalTrade. JournalofEconomicPerspectives9, 169-189. Nordhaus, W. 1975. The Political Business Cycle. Review of Economic Studies 42, 169– 190. Panousi, V., and D. Papanikolaou. 2012. Investment, Idiosyncratic Risk, and Ownership, Journalof Finance67,1113–1148. 36

Pastor, L., and P. Veronesi. 2013. Political Uncertainty and Risk Premia. Forthcoming em JournalofFinancial Economics. Pindyck,R., andA.Solimano. 1993. EconomicInstabilityandAggregateInvestment. NBER MacroeconomicsAnnual8, 259–303. Rajan,R.,andS.Marwah. 1998. TheEffectsofPolicyUncertaintyontheChoiceandTiming of Foreign Direct Investment: An Exploratory Firm-Level Assessment. Journal of Economic Development23,37–58. Razin, A., E. Sadka, and C. Yuen. 1998. A Pecking Order of Capital Flows and InternationalTax Principles. JournalofInternationalEconomics44,45–68. Rodrik, D. 1990. How Should Structural Adjustment Programs be Designed? World Development18,933–947. Rodrik,D.1991. PolicyUncertaintyandPrivateInvestmentinDevelopingCountries. Journal ofDevelopmentEconomics36, 229–242. Shleifer, A., and R. Vishny. 1992. LiquidationValues and Debt Capacity: A Market EquilibriumApproach. JournalofFinance47, 1343–66. Shleifer, A., and R. Vishny. 2003. Stock Market Driven Acquisitions. Journal of Financial Economics70, 295–311. Singh, H., and K. Jun. 1995. Some New Evidence on Determinants of Foreign Direct Investmentin DevelopingCountries. TheWorld Bank PolicyResearch WorkingPaper 1531. Thomas, C., F. Warnock, and J. Wongswan. 2004. The Performance of International Portfolios. FRB InternationalFinanceDiscussionPaper No. 817. United Nations Conference on Trade And Development (UNCTAD), 2009. World InvestmentReport 2009: TransnationalCorporations,AgriculturalProductionand Development. United Nations Conference on Trade And Development (UNCTAD), 2010. World InvestmentReport 2010Overview: Investingina Low-CarbonEconomy. Warnock, F., and C. Cleaver. 2003. Financial Centers and the Geography of Capital Flows. InternationalFinance6, 27–59. Warnock, F., and V. Warnock. 2005. International Capital Flows and U.S. Interest Rates. Board of Governors of the Federal Reserve System International Finance Discussion Papers 37

840. Wei,S. 1997. WhyisCorruptionsoMuchMoreTaxingthanTax? ArbitrarinessKills. NBER WorkingPaper6255. Wei, S. 2000. How Taxing is Corruption on International Investors. Review of Economics and Statistics82,1–11. 38

Appendix A: Variable Descriptions Variable Description FDIFlows QuarterlydirectinvestmentflowsthatU.S.parentsprovidetotheirforeignaffiliates,whereaU.S.parentisdefined asaU.S.investorofatleast10percentofaforeignbusiness. FPIFlows Monthlynetpurchasesoflong-termforeignsecurities,bothdebtandequities,byU.S.residents. FDI/Totaljt TheratioofU.S.FDIflowtotherecipientcountry jinagivenquarterttototalU.S.FDIflowsinquartert−1. FDI/Positionjt TheratioofU.S.FDIflowtotherecipientcountry jinagivenquarterttothecumulativeU.S.FDIpositioninthe country jasoftheendofquartert−1. FDI/GDPjt TheratioofFDIflowstotherecipientcountry jinagivenquarterttotheGDPofthecountry j. Electiont Electiont takesavalueofoneifthedestinationcountryholdtheelectioninthesecondhalfofquartertorinthe firsthalfofquartert+1,and0otherwise. Checks Thenumberofvetoplayersinapoliticalsystem,updatedannuallyandtakenfromtheWorldBankDatabaseof PoliticalInstitutions. Close Anindicatorvariablesetequaltooneifthevotedifferenceislessthanthefirstquartilevalue,andzerootherwise, wherevotedifferenceisdefinedasthedifferencebetweentheproportionofthevotesgarneredbythewinnerandthat receivedbytherunner-up. Gov.Stability Thegovernmentstabilityindexassessesthegovernment’sabilitytocarryoutitsdeclaredprograms,anditsabilityto stayinoffice.Theindexassignsnumbersbetween1and12,wherehighervaluesindicatemorestablegovernments. ThedataareupdatedonamonthlybasisandobtainedfromInternationalCountryRiskGuide(ICRG)producedby PoliticalRiskServices. Gov.Expenditure CentralgovernmentexpensesasapercentageofGDP,takenfromWorldDevelopmentIndicatorsprovidedbythe WorldBank. GDPPerCapita RealGrossDomesticProductpercapita,obtainedfromtheWorldBank. GDPGrowth AnnualpercentagechangesinrealpercapitaGDP GDPGrowthVolatility StandarddeviationofgrowthrateofrealpercapitaGDPover3years(yearst,t-1,andt-2). DomesticMarketReturn Quarterlyreturnsonacountry’smarketindex,calculatedusingdatafromDatastreamandBloomberg. ExchangeRate Realeffectiveexchangeratebetweentherecipientcountry’slocalcurrencyandtheU.S.dollar,takenfromIMF InternationalFinancialStatistics. TradeOpenness SumofexportsandimportsscaledbyGDP,whereexportsandimportsdataaredrawnfromtheWorldBank. 39

Table1 CapitalFlowsSummary Statistics Thistablereporttheannualizedcross-borderflowsofcapital(unit:$millions)averagedbycountry.Thefirstcolumnshowsaverageforeigndirectinvestment (FDI)flowsperyearfromU.S.toeachofthe43recipientcountries. Thenextcolumnpresentsforeignportfolioequityinvestment(FPEI)flowsfromthe U.S.toeachofthe44countries.Thefinalcolumnreportforeignportfoliodebtinvestment(FPDI)fromtheU.S.toeachofthe44countries.PanelBprovides summarystatisticsforvariousmeasuresofcapitalflows. ThefirsttworowssummarizerawFDIflowsaswellasofthethreemeasuresofFDIflows. The nexttworowsreporttheresultsforFPEIflowsandthefinaltworowsreporttheresultsforFPDIflows. SeetheAppendixforvariabledescriptionsaswell asthevariablesources. PanelA:AnnualizedFlowsAveragedbyCountry Country FDI FPEI FPDI Country FDI FPEI FPDI Argentina 1,152.6 113.1 32.7 Malaysia 825.5 403.7 254.3 Australia 4,295.6 504.5 6,029.6 Mexico 6,513.9 -428.7 862.8 Austria 235.2 33.8 -683.4 Netherlands 24,481.8 -652.2 -250.6 Belgium 3,333.8 -297.0 18,468.1 NewZealand 170.7 8.3 2,668.8 Brazil 3,314.8 4,788.4 988.7 Norway 537.8 -365.9 51.8 Canada 13,892.8 5,874.7 6,145.8 Pakistan . 54.2 54.1 Chile 1,470.1 65.5 24.7 Peru 378.9 74.5 79.1 Colombia 495.1 31.2 694.0 Philippines 71.8 122.1 719.9 CzechRepublic 236.8 -43.1 -25.2 Poland 438.8 74.8 117.7 Denmark 447.7 197.4 -491.1 Portugal 196.5 155.5 317.5 Finland 177.8 47.8 313.4 Russia 1,041.2 30.8 -179.3 France 3,368.4 3,146.6 -1,036.6 Singapore 4,935.2 310.5 -2,428.9 Germany 4,828.8 3,538.4 -4,377.8 SouthAfrica 376.5 460.4 410.5 Greece 65.0 82.4 -346.6 SouthKorea 1,684.7 1,623.7 666.9 Hungary 570.5 8.8 -107.3 Spain 2,327.5 585.5 -787.1 India 1,011.3 703.1 344.8 Sweden 640.7 -160.8 211.4 Indonesia 811.1 247.6 387.9 Switzerland 6,679.7 614.8 948.2 Ireland 7,818.2 -38.7 1,212.5 Taiwan 905.4 2,811.2 -227.8 Israel 854.4 753.7 373.7 Thailand 715.7 178.6 87.9 Italy 2,141.1 139.4 -1,957.4 Turkey 450.9 307.7 313.5 Japan 4,287.9 12,402.9 -1,448.6 UnitedKingdom 24,192.1 24,834.4 18,764.1 Luxembourg 6,867.4 -532.1 19,241.2 Venezuela 885.8 60.4 802.2 (continued) 40

Table1–Continued PanelB:SummaryStatistics(AnnualizedFlows) Flow($millions) Flow/GDP Flow/Position ln(Flow+ (Flow2+1)) p FDI Mean 3,355.3 1.13% 13.00% 6.00 StandardDeviation 7,795.6 4.18% 2.02% 5.10 FPEI Mean 1,485.3 0.15% 3.08% 2.25 StandardDeviation 6,502.9 0.72% 17.41% 6.67 FPDI Mean 1,132.7 0.61% 8.11% 0.06 StandardDeviation 8,590.1 5.85% 92.93% 7.27 41

Table2 ElectionSummary Statistics Panel A reports summary statistics for 184 national elections held between 1994 and 2010 in the 45 sample countries including the U.S. Panel B summarizes various characteristics of 44 destination countries. See the Appendixforvariabledescriptionsaswellasthevariablesources. PanelA:ElectionCharacteristics Mean Median St.Dev. ElectionFrequency(unit:quarters) 16.4 16.0 2.3 LengthofTerm(unit:years) 4.4 4.0 0.7 PercentofVotesWoninanElection Winner(%) 41.7 40.0 14.2 Runner-up(%) 28.6 27.0 10.1 Thirdplace(%) 11.5 11.4 5.5 TypeofElections Legislative(%) 73.6 Presidential(%) 26.4 ProportionofElectionswithExogenousTiming(%) 45.3 ChangeofGovernmentHead(%) 56.4 ChangeofRulingParty(%) 48.9 PanelB:DestinationCountryCharacteristics ChecksandBalances 4.06 4.00 1.92 ICRGGovernmentStabilityRating 7.82 8.00 2.01 GovernmentConsumption/GDP 0.16 0.15 0.06 GDPPerCapita($US) 9,183.1 3,273.1 12,842.1 GDPGrowth 0.078 0.077 0.130 StockMarketReturn(Monthly) 0.010 0.012 0.078 ChangeinExchangeRate(Monthly) 0.003 0.000 0.117 TradeOpenness 0.789 0.626 0.586 42

Table3 FDIFlowsaround Elections Thistablereportsestimatesofthefollowingspecification: 2 FDI =g +d + (cid:229) b Election +e , jt j t k j,t+k jt k=−2 Thedependentvariable,FDI,ismeasuredinfourways.Flow/PositionistheU.S.FDIflowtoarecipientcountry inagivenquarterscaledbytheU.S.FDIpositioninthatcountryattheendofthepreviousquarter. Flow/Total isdefinedastheU.S.FDIflowtorecipientcountry j inagivenquarterasaproportionoftotalU.S.FDIflows aroundtheworldattheendofthepreviousquarter. Flow/GDPistheU.S.FDIflowtoarecipientcountryina givenquarterscaledbythelaggedGDPofthecountry.Thefinalmeasureisasign-preservinglogtransformation used by Busse and Hefeker (2007). Election is set equal to one if the recipient country under consideration holdsa nationalelectioninthesecondhalfofthegivenquarterorinthefirsthalfofthenextquarter,andzero otherwise. Seeappendixfordetailedvariabledescriptions. Countryandtime(quarterlyfrequency)fixedeffects areincluded. Standarderrorsareclusteredatthecountrylevelandthecorrespondingt-statisticsarereportedin brackets. FDI/Position FDI/Total FDI/GDP ln(FDI+ (FDI2+1)) p Election t−2 0.0012 0.0014 0.0013 0.3302 [0.542] [0.609] [0.356] [0.841] Election t−1 -0.0118* -0.0035* -0.0052 -0.1369 [1.701] [-1.760] [1.193] [0.279] Election -0.0155*** -0.0088*** -0.018** -0.4434** t [-3.520] [-3.256] [-2.446] [-2.080] Election 0.0147 0.0060 0.0037 0.1564 t+1 [1.484] [1.195] [0.895] [1.117] Election 0.0023 0.0032 -0.0110 -0.0625 t+2 [0.221] [1.107] [-0.658] [-0.123] Constant 0.0253*** 0.0126*** -0.0020 5.7530*** [2.719] [4.225] [-0.072] [11.307] Observations 2,512 2,802 2,802 2,802 Adj. R2 0.184 0.298 0.304 0.164 43

Table4 FDIRegressions: Country Controls Thistablereportsestimatesofthefollowingspecification: 2 FDI =g +d + (cid:229) b Election +X ′q +e , jt j t k j,t+k jt k=−2 where j indexescountryandt indexestime. X is a vectorof controlvariablesincludinggovernmentstability, GDPper capita, GDPgrowth, growthvolatility, governmentexpendituresto GDP, laggedstock marketreturn, stock return volatility, exchange rates, exchange rate volatility, and trade openness. Each column reports the estimates from the regression with differenttransformations of FDI flows. Election is set equal to one if the countryunderconsiderationholdsa nationalelectioninthe secondhalfofthe givenquarterorin the firsthalf ofthenextquarter,andzerootherwise. ThecoefficientsforElection t−2 andElection t+2 arenotreportedtosave space. See appendix for detailed variable descriptions. Country and time (quarterly frequency) fixed effects areincluded. Standarderrorsareclusteredatthecountrylevelandthecorrespondingt-statisticsarereportedin brackets. FDI/Position FDI/Total FDI/GDP ln(FDI+ (FDI2+1)) p Electiont−1 -0.0124 -0.0033* -0.0047 -0.0969 [-1.520] [-1.604] [1.126] [0.253] Electiont -0.0146*** -0.0090*** -0.0177** -0.4211* [-2.908] [-2.956] [-2.293] [-1.086] Electiont+1 0.0118 .0062 0.0031 0.1428 [1.195] [1.024] [0.866] [0.979] GovernmentStability 0.0006 -0.0002 0.0034 0.3113** [0.301] [-0.129] [1.405] [2.705] GDPPerCapita -0.0000 0.0000** 0.0000*** -0.0000 [-0.479] [2.084] [2.849] [-0.163] GDPGrowth 0.1158*** 0.0184** 0.1057* 4.9200*** [5.264] [2.448] [1.727] [4.048] GDPGrowthVolatility 0.0057 -0.0616 -0.0740* -4.101** [0.442] [-0.969] [-1.794] [-2.117] GovernmentExpenditures/GDP 0.0026** 0.0021 0.0745 0.0299 [2.451] [1.026] [1.266] [0.203] DomesticMarketReturn 0.0015** -0.0001 -0.0002 0.0364 [2.082] [-0.776] [-0.170] [1.172] ReturnVolatility -0.0008 0.0001 -0.0007* -0.0234 ‘ [-0.857] [0.578] [-1.748] [-0.661] D ExchangeRate -0.0113 -0.0006 0.0037 -0.8851 [-0.437] [-0.171] [0.279] [-1.430] ExchangeRateVolatility 0.0000 0.0000 0.0000 0.0001 [1.555] [0.215] [1.316] [0.535] TradeOpenness 0.0002 0.0002 0.0028** 0.0306** [1.424] [1.633] [2.300] [2.684] Observations 1,800 1,928 1,928 1,928 Adj.R2 0.194 0.244 0.347 0.169 44

Table5 FDI Regressions: Including U.S. Elections Thistablereportsestimatesofthefollowingspecification: 2 2 FDI =g +d + (cid:229) b Election + (cid:229) d Election +X ′q +e jt j t k j,t+k l US,t+l jt k=−2 l=−2 where j indexescountryandt indexestime. X is a vectorof controlvariablesincludinggovernmentstability, GDPper capita, GDPgrowth, growthvolatility, governmentexpendituresto GDP, laggedstock marketreturn, stock return volatility, exchange rates, exchange rate volatility, and trade openness. Each column reports the estimates from the regression with differenttransformations of FDI flows. Election is set equal to one if the countryunderconsiderationholdsanationalelectioninthesecondhalfofthegivenquarterorinthefirsthalfof thenextquarter,andzerootherwise.Thecoefficientsforthecontrolvariablesandtheelectiondummiesfort±2 are notreportedto savespace. See appendixfordetailed variabledescriptions. Countryandyearfixedeffects areincluded. Standarderrorsareclusteredatthecountrylevelandthecorrespondingt-statisticsarereportedin brackets. FDI/Position FDI/Total FDI/GDP ln(FDI+ (FDI2+1)) p Election t−1 -0.0103* -0.0050* -0.0038 -0.1527 [-1.741] [-1.999] [-1.100] [-0.253] Election -0.0156*** -0.0089*** 0.0140** -0.4692** t [-3.541] [-3.228] [2.324] [-2.029] Election 0.0136 0.0054* 0.0037 0.1507 t+1 [0.486] [1.694] [0.080] [0.107] USElection t−1 -0.0168** 0.0000 -0.0124* -1.1540* [-2.393] [0.006] [-1.767] [-1.748] USElection -0.0147*** -0.0054*** -0.0405*** -1.9882*** t [-2.916] [-2.699] [-2.690] [-3.137] USElection 0.0366** -0.0000 0.0249* 1.2502** t+1 [2.218] [-0.005] [1.850] [2.043] Observations 1,800 1,928 1,928 1,928 Adj. R2 0.144 0.238 0.332 0.134 45

Table6 FDI Regressions: Exogenous Timing ofElections Thistablereportsestimatesofthefollowingspecification: 2 FDI =g +d + (cid:229) b Election +X ′q +e jt j t k j,t+k jt k=−2 where j indexescountryandt indexestime. X is a vectorof controlvariablesincludinggovernmentstability, GDPper capita, GDPgrowth, growthvolatility, governmentexpendituresto GDP, laggedstock marketreturn, stock return volatility, exchange rates, exchange rate volatility, and trade openness. Each column reports the estimatesfromtheregressionwithdifferenttransformationsofFDIflows. Theanalysisconsidersthesubsample ofcountrieswithfixedelectiontimingonly.Electionissetequaltooneifthecountryunderconsiderationholds anationalelectioninthesecondhalfofthegivenquarterorinthefirsthalfofthenextquarter,andzerootherwise. See appendixfor detailed variabledescriptions. Country andyear/quarterfixedeffectsare included. Standard errorsareclusteredatthecountrylevelandthecorrespondingt-statisticsarereportedinbrackets. (1) (2) (3) (4) FDI/Position FDI/Total FDI/GDP ln(FDI+ (FDI2+1)) p Election t−1 -0.0041 -0.0054 -0.3362 -0.0634 [-1.109] [-1.459] [-0.826] [-1.142] Election -0.0124** -0.0110** -1.1085* -0.4949* t [-2.202] [-2.103] [-1.815] [-1.892] Election 0.0048* 0.0053 0.2586 0.1068 t+1 [1.825] [0.798] [0.624] [0.283] Observations 764 712 764 764 Adj. R2 0.284 0.189 0.310 0.170 46

Table7 FDI Flowsaround CloseElections Thistablereportsestimatesofthefollowingspecification: FDI jt =g +d +a ·Close +a ·Wide + (cid:229) 2 b Election +X ′q +e , j t 1 jt 2 jt k j,t+k jt Position j,t−1 k=−2 where jindexescountryandt indexestime.Closeisadummyvariableequaltooneifthemarginofvictoryfor agivenelectionisinthelowestquartileofthemarginofvictordistribution.Wideisadummyvariableequalto oneifthemarginofvictoryforagivenelectionisinthehighestquartileofthemarginofvictordistribution. X isavectorofcontrolvariablesincludinggovernmentstability,GDPpercapita, GDPgrowth,growthvolatility, governmentexpenditurestoGDP, laggedstockmarketreturn,stockreturnvolatility,exchangerates, exchange ratevolatility,andtradeopenness.EachcolumnreportstheestimatesfromtheregressionwithdifferenttransformationsofFDIflows.. Electionissetequaltooneifthecountryunderconsiderationholdsanationalelection inthesecondhalfofthegivenquarterorinthefirsthalfofthenextquarter,andzerootherwise. Seeappendix fordetailedvariabledescriptions. Thefirsttwocolumnsreporttheresultsforthefullsampleandthefinaltwo columns report results for the sample of countries with exogenouselection timing. Country and year/quarter fixedeffectsareincluded.Standarderrorsareclusteredatthecountrylevelandthecorrespondingt-statisticsare reportedinbrackets. FullSample ExogenousTimingSample Election t−1 -0.0099* -0.0100* -0.0042 -0.0042 [-1.737] [-1.739] [-1.101] [-1.109] Election -0.0052* -0.0049* -0.0077* -0.0075* t [-1.701] [-1.708] [-1.702] [-1.698] CloseElectionInteraction -0.0096** -0.0099** -0.0108* -0.0112* [-2.552] [-2.547] [-1.922] [-1.926] WideMarginofVictoryInteraction 0.0062 0.0089 [1.191] [0.996] Election 0.0116 0.0118 0.0048* 0.0047* t+1 [0.587] [0.588] [1.827] [1.822] Observations 1,800 1,800 764 764 Adj. R2 0.155 0.162 0.290 0.297 47

Table8 Interactions withMeasures ofGovernmentStability Thistablereportsestimatesofthefollowingspecification: FDI jt =g +d +a ·Z +a ·Z ·Election + (cid:229) 2 b Election +X ′q +e , j t 1 jt 2 jt jt k j,t+k jt Position j,t−1 k=−2 whereZ isatime-varyingcountrycharacteristicmeanttocapturedifferencesinthepropensityforlargepolicy jt changesafterelections.FourmeasuresofZareutilized:ICRGgovernmentstabilityratings,checksandbalances on executive authority, World Bank classification of high income countries, and a dummy variable indicating whether a country’s trade openness is above the median in a given year across countries. X is a vector of control variables including governmentstability, GDP per capita, GDP growth, growth volatility, government expenditurestoGDP,laggedstockmarketreturn,stockreturnvolatility,exchangerates,exchangeratevolatility, andtradeopenness. Electionissetequaltooneifthecountryunderconsiderationholdsanationalelectionin thesecondhalfofthegivenquarterorinthefirsthalfofthenextquarter,andzerootherwise. Seeappendixfor detailedvariabledescriptions.Countryandtime(quarterlyfrequency)fixedeffectsareincluded.Standarderrors areclusteredatthecountrylevelandthecorrespondingt-statisticsarereportedinbrackets. (1) (2) (3) (4) Govt.Stability Checks&Balances HighIncome TradeOpenness Election -0.0103*** -0.0066** -0.0232*** -0.0231*** t [-2.955] [-2.129] [-2.840] [-3.321] Stability×Election 0.0022* t [1.950] Checks×Election 0.0017 t [1.591] HighIncome×Election 0.0119 t [1.022] HighOpenness×Election 0.0206** t [2.183] Observations 1,800 1,800 1,800 1,800 Adj. R2 0.145 0.145 0.146 0.146 48

Table9 Election Cyclesand WorldPortfolioInvestment Flows Thistablereportsseeminglyunrelatedregressionestimatesforthefollowingequations: FDI jt =g +d + (cid:229) 2 b Election +X ′q +e j t k j,t+k jt Position j,t−1 k=−2 FPI jt =g ′ +d ′ + (cid:229) 2 b ′ Election +X ′h +n , Position j,t−1 j t k=−2 k j,t+k jt whereFPI representsforeignportfolioequityinvestment(FPEI)flows, foreignportfoliodebtinvestment(FPDI) flows, orthe sum of the two flows. We employtheflows/positionmeasureasthedependentvariable.X isavectorofcontrolvariables,whichincludesarecipientcountry’sGDPgrowth,GDPper capita,volatilityofGDPgrowth,governmentconsumptionscaledbyGDP,lagged,lead,andcontemporaneousstockmarketreturn,stockmarketvolatility, theU.S.marketreturn,changeinexchangerate,volatilityofexchangerates,andtradeopenness. Seeappendixfordetailedvariabledescriptions. Country andtime(quarterlyfrequency)fixedeffectsareincluded. Standarderrorsareclusteredatthecountrylevelandthecorrespondingt-statisticsarereportedin brackets. FDI EquityFPI DebtFPI TotalFPI Election t−1 -0.0075 -0.0004 0.0250 -0.0022 [-1.170] [-0.056] [0.236] [-0.349] Election -0.0172*** 0.0019 -0.0073 0.0009 t [-2.831] [0.287] [-0.203] [0.149] Election 0.0069 -0.0045 -0.1321 0.0021 t+1 [1.009] [-0.772] [-1.450] [0.388] Observations 1,649 1,649 1,649 1,649 Adj. R2 0.223 0.084 0.066 0.203 Test: (b −b =0) FDI FPI Difference -0.0191*** -0.0099** -0.0181*** [-3.56] [2.45] [-3.19] 49