Capital Controls and the International Transmission of U.S. Money Shocks

Board of Governors of the Federal Reserve System International Finance Discussion Papers Number 778 September 2003 CAPITAL CONTROLS AND THE INTERNATIONAL TRANSMISSION OF U.S. MONEY SHOCKS Jacques Miniane and John H. Rogers NOTE: International Finance Discussion Papers are preliminary materials circulated to stimulate discussion and critical comment. References in publications to International Finance Discussion Papers (other than an acknowledgment that the writer has had access to unpublished material) should be cleared with the author or authors. Recent IFDPs are available on the Web at www.federalreserve.gov/pubs/ifdp/.

Capital Controls and the International Transmission of U.S. Money Shocks Jacques Miniane John H. Rogers ∗ July 10, 2003 Abstract In thispaper weassesswhether capitalcontrolseffectivelyinsulate countriesfromU.S.monetaryshocks,lookingsimultaneouslyatalarge range of country experiences in a unified estimation framework. We estimatethe effect of identified U.S. monetary shocks on theexchange rateandforeigncountryinterestrates,andtestwhethercountrieswith lessopencapitalaccountsexhibitsystematicallysmallerresponses. We findessentiallynoevidenceinfavorofthisnotion. Othercountryfactorssuchastheexchangerateregimeordegreeofdollarizationexplain more of the cross-country differences in responses. The significant differencesinresponseswedofindaremorepronouncedatshorthorizons. JEL classification: F32, F34 1 Introduction The question of whether the existing “architecture” of the world financial system can cope with the size and nature of modern-day capital flows has once again come with force to the attention of policymakers and academics. Key to the resurgence of interest is the late 1990’s crises in Asia, Russia and Brazil, aided by the perception that Malaysia avoided a harsher fate Miniane: DepartmentofEconomics,JohnsHopkinsUniversity. Rogers: BoardofGov- ∗ ernorsoftheFederalReserveSystem. Email: jminiane@att.netandjohn.h.rogers@frb.gov. The authors wish to thank Kirran Bari for assistance with the data and Matt Shum, Jonathan Wright, and seminarparticipantsatJohnsHopkins University, the InterAmerican Development Bank and the International Finance Division of the Federal Reserve Board for useful comments and suggestions. The views in this paper are solely the responsibility of the authors and should not be interpreted as reflecting the views of the Board of Governors of the Federal Reserve System, or of other members of its staff. 1

by imposing controls on capital outflows. The recent fallout in Argentina following a decade of capital account liberalization suggests that the debate on capital controls is not likely to die soon, whether or not open capital accounts had much to do with Argentina’s problems. A vast academic literature surveying a wide range of experiences with capital controls has yet to produce consensus on their effectiveness (Dooley (1996) and Eichengreen (2002) provide overviews). One strand of the literature undertakes cross-sectional studies, typically dealing with the relationship between capital controls and real variables such as investment and growth. Rodrik (1998) finds no evidence of a positive correlation between capital account opennessand growthor investment/GDP ratios, and argues against capital account convertibility: given the periodicity and devastating power of financial crises, it is not wise to remove capital controls since, at worst, they don’t seem to affect welfare. Quinn (1997) and Edwards (2001) reach the opposite conclusion, using a different measure of capital controls developed by Quinn (see Grilli and Milesi-Ferretti (1995) also). Some studies have looked for the effects of capital controls in places other than output. Chinn and Ito (2002) find that financial development, measured by private credit creation and stock market capitalization, is negatively correlated with the extent of capital controls, a correlation that is stronger in developed countries with solid institutional frameworks. Desai, Foley, andHines (2003) use the Bureauof Economic Analysis annual survey ofU.S.DirectInvestmentAbroadtostudytheimpactofhostcountries’capitalcontrolsonU.S.multinationalscorporatefinancedecisions. Theauthors find that controls domatter: affiliates incountries withlessopen capital accounts are smaller in size, overinvest in capital assets to the detriment of financial assets, tend to rely more heaviliy on initial parent equity infusions oronretainedearningsratherthanondebtfinancingtoavoidhigherborrowing costs, and use intra-firm pricing to evade controls on profit remittances. AnIMFcollectionoffourteencountrycasestudiesinvolvingcapital account liberalization (Ariyoshi et al. (2000)) concludes that capital controls cannot act as a substitute for sound macroeconomic policy, but also acknowledges that controls cannot be easily dismissed as a buffer to external shocks or to provide breathing space in which to adopt sound policy. Lackingconsensusonhowtobest measure capital account openness and how to define “effectiveness”, a second strand of the literature on capital controls has taken a case study approach. Researchers focus on a well identified sequence of events, comparing the behavior of relevant variables such 2

as exchange rates, interest rates, or capital flows after the imposition or removal of controls. Consider the case of Malaysia, which imposed controls in September 1998, and whose exchange rate, interest rate (T-bill) and foreign reserves are displayed in Figure 1. Reserves, which had fallen substantially beforeSeptember, startedrisingimmediately andpromptlyreturnedtoprecrisis levels. The exchange rate was eventually stabilized and interest rates fell(thoughthisstartedabitbeforetheimpositionofcontrols). Moreformal analysis of Malaysia’s experience by Kaplan and Rodrik (2001) and Edison andReinhart(2001)findthat,assuggestedbyFigure1,capitalcontrolshad indeed been effective. Chile’s unremunerated reserve requirement (URR), designed to discourage short-term flows in favor of long term investments, is examined by Cardoso and Laurens (1998) and de Gregorio et al. (2000). Both studies find that the URR has tilted the composition of flows towards longer maturities, but has had little effect on interest rate differentials and the real exchange rate.1 Thequestionweaskinthispaperiswhethertheapparenteffectivenessof capitalcontrolsisconfinedtoparticularexperienceslikeMalaysiaandChile, as the case-study literature suggests, or is a more general phenomenon. We examine whether more stringent capital controls better insulate countries from foreign monetary shocks. Although this metric of ”effectiveness” is an implication of almost any theoretical model that incorporates restrictions on capital flows, we are not aware that it has been examined in the crosscountry literature. Our notion of effectiveness is not the only relevant one, but it does allow us to look systematically across a large number of country experiences, includingemergingmarketsandindustrializedcountries. Using standardtools from the VAR literature, we estimate the response of the foreign (i.e., non-U.S.) interest rate and the exchange rate toan identified U.S. monetary policy shock.2 We then test whether those responses are smaller 1Otherpapersinthecase-studyveinfocusonhowwellcontrolsareabletosustaindual exchangerateregimes. Gros(1988)findsawedgebetweendualratesin1980sBelgiumand Mexico that lasts for about six months. A similar finding applies to returns on domestic bank deposits with similar deposits offered by off-shore branches (Gros (1987)). Studies concentrating on currency or balance of payments crises such as Fieleke (1994) or Edison andReinhart(2001)findthatcountrieswhichimposecontrolsinthemidstofacrisismay in some cases reduce the volatiliy of interest rates while trying to defend their currency, but typically cannot avoid a realignment in the end. 2This is not the only type of disturbance we could consider. It would be interesting to extend this analysis to, e.g., terms-of-trade shocks. The advantage of our approach is thattheidentificationof(U.S.)monetarypolicyshockshasbeenstudiedmuchmorethan other types of shocks. Thus, there is something of a consensus (or at least a roadmap) in the literature, relative to identification of other types of disturbances. 3

for countries with more stringent capital controls. We consider different response horizons, from six and twelve months up to 48 months. Because a country’s interest rate and exchange rate responses might differ on the basis of factors other than the degree of capital controls, we also control for the type of exchange rate arrangement, degree of liability dollarization, and degree of trade integration with the United States. We find virtually no evidence that capital controls are effective in this sense. We do find some evidence that the exchange rate response to a U.S. monetary policy shock is significantly smaller in countries with more stringent capital controls, but the effect is not large and not robust, as it goes away at horizons greater than one year and/or when we control for the additional country characteristics. We find little evidence that the interestrateresponseissmallerforcountrieswithhighcapitalcontrols. The exchange rate regime and degree of liability dollarization are significant in determining how exchange rates and interest rates respond to the shock, in the way predicted by theory. In the next section we describe the empirical strategy and data used in the paper. We present results in section 3, and discuss the implications of those results in the final section. 2 Empirical Strategy and Data Key to our estimation work is properly identifying U.S. monetary policy shocks. The current predominant approach is to use structural vector autoregression (VAR) techniques. Researchers have traditionally relied on assuming a recursive structure to the system and using a priori restrictions, either on the contemporaneous or long-run relationship between variables.3 We follow in the traditional, ”short-run restrictions” vein. Assume the economy is represented by the following model: A(L)Y =e t t 3Open economy applications taking the traditional approach include Clarida and Gali (1994),EichenbaumandEvans(1995),andRogers(1999),whileKimandRoubini(2000)) useshort-runrestrictionsbutdonotrelyexclusivelyonarecursivestructure. Morerecent contributions include Faust and Rogers (2003) and Faust, Rogers, Swanson and Wright (2002), which are discussed below. 4

where Y is an (n 1) vector of data, e is a (n 1) vector of strucutural t t × × shocks, and A(L) is a (n n) matrix whose typical element is a polynomial × inthe lagoperator. The shocks are assumedtobe serially and mutually uncorrelated. Sincethematrixofcontemporaneousrelationsbetweenvariables A(0) need not be diagonal, one cannot estimate this model. The estimable reduced form version of the model is: B(L)Y = u t t where B(L) = A(0) 1 A(L) − × u = A(0) 1 e t − t × U is the vector of errors in the reduced-form equations. If one knew A(0) t thenthe parameters inthe structural model and the structural shocks could be recovered by estimating this reduced form. Notice that if one normalizes the structural model for the shocks e to have unit variance, then: t Ω=A(0) 1 (A(0) 1) − − 0 × where Ω is the estimable variance-covariance matrix of reduced-form errors. Given that Ω contains n(n+1)/2 free parameters, identifying A(0) requires imposing n(n 1)/2 restrictions. − We concentrate on abenchmark eight-variable model includingan index of commodity prices (CP), U.S. industrial production (Y), U.S. consumer prices (P), foreign industrial production (Y*), foreign 3-month interest rate (R*),theU.S.FederalFundsRate(FFR),theratioofnon-borrowedreserves to total reserves in the U.S. (M), and the nominal exchange rate with the U.S. dollar (S). This model adds commodity prices to the 7-variable model estimatedbyEichenbaumandEvans(1995). Inour”baseline”identification, we impose the following recursive ordering on the variables: [CP, Y, P, Y*, R*, FFR, M, S]. In this setting, an exogenous monetary shock is identified as an orthogonal shock to the Fed Funds Rate, normalized to be positive (contractionary money shock) and equal to 25 basis points on impact. Practitionersandcriticsalike acknowledgethatthe traditional approach to identification of structural monetary policy shocks is contentious. There are an insufficient number of uncontroversial assumptions available to fully identifystandard-sizedopeneconomyVARs. FaustandRogers(2003)apply an approach to identification, originally developed by Faust (1998), that allows one to do inference in partially-identified models. They investigate 5

the sensitivity of various results in the literature to implausible identifying assumptions. Unfortunately, the results of interest are often found to be sensitive to the identification scheme.4 In a related vein, Faust, Rogers, Swanson, and Wright (2002) bring high-frequency data to the identification problem, imposing that the impulse responses of the exchange rate and interestratesintheVARmatchtheresponsesestimatedfromhigh-frequency financial market data. Motivated by this, we examine the sensitivity of our conclusions to different identification schemes. One could argue that applying any particular statistical model to a diverse set of countries is unwarranted. This criticism should be weighed against the desirability of identifying a shock that is common for all countriesunderconsideration. Asdiscussedintheresultsbelow,thisissomething we seem to have achieved. While case studies allow for greater tailoring of the statistical model to the specific event under study, they preclude the kind of statistical inference presented in this paper and in the cross-country studies discussed above. Ultimately, one can get a complete picture about the effectiveness of capital controls only after considering both types of evidence. 2.1 Data We study twenty-six countries covering a wide spectrum in terms of capital account restrictions, economic development, geographic location, exchange rate regime, etc.5 Our VARs are monthly and span the period 1975:1 to 1998:12. As is common practice in monthly VARs we impose six lags in the estimation. All variables are expressed in logs except for interest rates. An importantfactormotivatingourchoiceofcountriesisthedesiretohavebetter measures of capital controls than the crude zero/one dummy published by the International Financial Statistics (IFS) of the IMF until 1996. The 4For example, Eichenbaum and Evans (1995) find thatmonetary policy shocks induce largedeviationsfromuncoveredinterestparityanddollarappreciationsthatpeakaslateas twotothreeyearsaftertheshock-so-calleddelayedovershooting. FaustandRogers(2003) findthatalthoughthefirstresultishighlyrobusttoidentification,delayedovershootingis very sensitive to allowing simultaneity among asset market variables. In addition, papers estimatingtheshareofmonetarypolicyshocksinaccountingforexchangeratevariability reportestimatesrangingfromafewpercentagepointstomorethanhalf(ClaridaandGali (1994), Eichenbaum and Evans (1995), and Rogers (1999)). 5Thecountriesare, in alphabeticalorder: Australia, Austria,Belgium,Canada, Chile, Colombia, Denmark, Finland, France, Germany, Greece, India, Italy, Japan, Korea, Malaysia, Mexico, The Netherlands, Norway, The Philippines, Portugal, South Africa, Spain, Sweden, Turkey, and the United Kingdom. 6

choice of countries, and in particular the desire to preserve a reasonablysized sample of countries (which includes nine of the original eleven euro area members), in turn dictates our ending the sample period in 1998:12. With the launch of the euro, the exchange rates and interest rates of nine out of our 26 countries became one, making it impossible to consider them independent cross-sectional observations after 1998:12. We exploit a new set of measures of capital controls restrictions developed by Miniane (2000). The indexes use publicly available information contained in the Exchange Arrangements and Exchange Restrictions publication of the IMF to extend back to 1983 the IMF’s post-1996 practice of reporting dummies in twelve different categories of capital transactions.6 Before 1996, the IMF only reported a single zero/one dummy for capital account restrictions. Miniane (2000) shows in detail how the more disaggregated measure captures several episodes of capital account liberalization that are not picked up by pre-1996 IFS data. Miniane (2000) also adds information on dual exchange rates. In the current paper, we use as our capital controls measure the average of the dummies over all categories and all years for each country.7 To reduce the chance of making spurious inferences about the effectivenessofcapitalcontrols,wealsoaccountforothervariablesthatcouldexplain differencesinresponsestoU.S.shocks. Weconcentrateonthreesuch”country factors”: the exchange rate regime, degree of dollarization, and degree of trade integration with the United States. The exchange rate regime is classified using information from IFS. We assign countries to one of four categories, where a higher index denotes a more flexible regime. Pure peg regimes are category one. Limited flexibility regimes, category two, include pegs to baskets including the U.S. dollar or bindingbandsofnomorethanplusorminus7.5%aroundthecentralparity. Managed floats comprise category three and pure floating regimes is the 6The twelve categories are: Capital Market Securities, Money Market Instruments, Collective Investment Securities, Derivatives and Other Instruments, Commercial Credits, Financial Credits, Guarantees and Financial Back-Up Securities, Direct Investment, Repatriation of Profits or Liquidation of Direct Investment, Real Estate, Specific Provisions to Commercial Banks and Other Credit Institutions, and Specific Provisions to Institutional Investors. 7We also construct a sub-index using only the first four and the last three categories which are more directly concerned with asset markets; the correlation between the two is 0.93. We do not report any results with the sub-index. 7

final category.8 The country index is obtained by averaging this data over the period 1975-1998. Some studies such as Levy Yeyati and Sturzenegger (1999) and Reinhart and Rogoff (2002) have criticized the IFS exchange regime data for being a de jure and not de facto classification. We also utilize an index based on Reinhart and Rogoff. The sample correlation with our IFS-based index is above 0.6, as seen in Table 1, and estimation results were not much changed when we used these instead. The degree of dollarization in the country is measured by the share of the country’s banking sector claims that are denominated in U.S. dollars. Several studies, including Aghion et al. (2000) and Hausmann et al. (2001), have stressed that foreign currency liabilities are an important explanation why countries with de jure floats may work hard to smooth exchange rate fluctuations. Finally, trade openness is measured as the sum of exports to the U.S. and imports from the U.S., as a share of the country’s GDP. All data sources are detailed in the appendix. Table1showsthevalueofeachcountryfactortogetherwiththeirsample correlations. Asexpected,developednationstendtohavemoreopencapital accounts on average, with the U.K. the least restrictive. Chile stands at the otherextreme. EventhoughthecorrelationbetweentheMiniane(2000)and thetraditionalIFSmeasuresishigh,thereareimportantdifferencesbetween thetwo. Whereasseveralcountriesarereportedashavingnocontrolsinany year according to IFS measure, this never happens in the new measure. At the other extreme, only Chile is classified by the Miniane (2000) measure as having a value of one for all years, whereas nine countries do so according to the traditional measure. As a result, there is much more clustering of countriesattheextremesintheIFSmeasure. Inaddition, note that there is alargenegativecorrelationbetweentheexchangerateregimeandthecapital controlsindex,perhapsbecausecountrieswithpegsresorttocapitalcontrols inanattempttogainsomemonetarypolicyindependence. Thedollarization and trade integration measures are as expected: countries in the Americas displayhighertradeintegrationwiththeUSandhigherliabilitydollarization than European countries, with Asia being a middle ground. 8We slightly modify IFS information to reflect the fact that we are interested in responsesvis-a-visU.S.shocks: forinstance,wecategorizeFranceorGermanyaspurefloats even though the IFS assigns them to the “cooperative agreements” category. 8

2.2 Panel VARs As a first approach to estimation, we separate countries in two groups depending on whether they have high or low levels of capital controls, fixed or floating exchange rate regimes, high or low levels of dollarization, and high or low levels of trade integration with the United States. We separate countries in a straightforward manner. For capital controls, we group countries accordingtowhether their index is below orabove 0.5. This leads to groups of12(lowcontrols)and14(highcontrols)countriesinitially(andeventually 12 and 12 when we ultimately exclude Mexico and Turkey). For the other variables,wesimplysortcountriesbyascendingleveloftherelevantvariable and divide the sample in two. For each sub-group, we estimate our benchmark eight-variable model in a panel setting, imposing a common constant and common coefficients amongcountriesinthesub-group. Wethencomputeimpulseresponsestoan identified U.S. monetary shock and check whether we can detect significant differencesbetweenthetwogroupsintheirforeigninterestrateandexchange rate responses. 2.3 Bilateral VARs Thepanelapproachhasanadvantageinitseaseofexposition, butislimited in that it allows us to make inferences about the importance of only one country factor at a time. Furthermore, separating the sample in two is a rathercrudewaytoassesstheeffectofthecountryfactoronthetransmission of U.S. shocks. Thus, we also estimate VARs for each country separately, and compute the cumulative exchange rate and interest rate responses. We then regress the country-specific responses at horizons six, twelve, twentyfour, and forty-eight months on the (country-specific) values of the capital controls index, exchange rate regime dummy, degree of dollarization, and degree of trade integration with the United States. We refer to the latter as our ”second stage” regressions. One can deduce the expected sign of the relationships as follows.9 If capital controls have an effect on short-term capital flows, then countries with such controls could keep interest rates low following a contractionary U.S.shockwithoutsufferingalargedepreciationoftheircurrency. Countries 9TextbookMundell-Flemingtheoryisenoughtodeducethesignsofthecapitalcontrols and exchange rate regime dummies. See Aghion et. al. for theoretical explanation of the signs of the relationships with our dollarization measure. 9

without controls would have to choose between low interest rates or little depreciation, but in principle cannot achieve both. For the exchange rate regime, ceteris paribus, the group of “fixers” should raise interest rates in response to a contractionary U.S. shock, since an interest rate differential with the U.S. would be inconsistent with a fixed exchange rate (and open capital markets, implicitly a part of our ceteris paribus assumption). The group of “floaters”, on the other hand, could absorb part of the effect on the domestic interest rate by allowing their currency to depreciate against the U.S. dollar. Finally, countrieswithhighdegreesofdollarizationortrade integration with the U.S. may behave as de facto fixers even if their official exchange rate regime is a float. Thus, one would expect the following signs in our regressions: Exchange rate responses Interest rate responses Capital Controls - - Exchange Regime + - Dollarization - + Trade integration - + We need to account for uncertainty in the VAR and its effect on our second-stage regression estimates. To do so, we begin with the standard Runkle(1987)bootstrappingmethodtoestimateconfidenceintervalsaround impulseresponses. Wegenerateonethousandartificalseriesfromtheactual data and reduced-form residuals, re-estimate the VAR, use the cumulative impulse responses toestimate our second stage regression a thousand times, and obtain the empirical distribution of the second stage coefficients. We report the second-stage coefficients from the original regression along with the 16.5th and 83.5th percentile of the empirical distribution.10 3 Empirical Results 3.1 Panel Results Figure 2 displays the impulse responses of each variable to an identified U.S. monetary policy shock when the sample is divided between high and 10Thus,thereportedbandincludestwo-thirdsoftheestimates,oraroundplusorminus one standard deviation in a standard normal distribution, as in Leeper, Sims, and Zha (1996), Rogers (1999), and Faust and Rogers (2003). Note that this confidence band is ”conservative”,inthesensethatitwillleadustofindmore evidenceofasignificanteffect fromcapitalcontrols,comparedtousingawiderplus-minustwostandarddeviationsband. Thismakesourfindingofessentiallyno significanteffectofcapitalcontrolsevenstronger. 10

low capital controls countries, along with their confidence bands.11 The responses ofU.S. variables are familiar: adropin NBRXand rise inthe Fed Funds rate is accompanied a negative and hump-shaped response of U.S. industrial production (see Strongin (1995) and Faust and Rogers (2003) among many others). The price puzzle is present despite the inclusion of commodity prices. Foreign industrial production falls by less than in the United States. Foreign interest rates rise and the currency depreciates against the U.S. dollar, but with some differences in magnitudes across groups. As noted above, if capital controls were binding one would expect both a lower interestrate response andsmallercurrencydepreciationamongthe highcontrols group. Infact, Figure 2does showonimpact asignificantly smallerdepreciation for the group of high capital control countries, although the difference with the low controls group is not large. We also see that interest rates rise more in high capital control countries, a difference that is statistically significant at the peak response (around 6 months). Examination of the individual country responses indicates that Mexico and Turkey (both in the high controls group) have unusually large, positive interest rate responses. To investigate whether the results are driven to any great extentbythese potential outliers, weexclude MexicoandTurkey from the sample. The results, depicted in Figure 2-b, are very similar to those for the full sample. InFigures3and4wecompareinterestrateandexchangerateresponses, respectively, when the panel VARs are divided based on the other country factors. The top panel repeats the Figure 2 comparison based on capital controls. We exclude the responses of variables other than S and R* for space consideration. However, these are virtually identical across groups and to those of Figure 2. The results go in the expected directions. The group of “fixers” raises interest rates by more than the countries with more flexible exchange rates (Fig. 3, row 2). The difference is significant at the peak. That difference is not trivial: peak interest rate responses are ten basis points higher for the fixers following a shock of twenty-five basis points. Also, the currencies of fixers depreciate by less on impact vis-a-vis 11Bootstrapping for the panel was done by generating initial conditions separately for each country as in Runkle (1997) but sampling from the entire panel vector of residuals. This was done to account for possible cross-country correlations. Drawing bands by sampling from each country separately leads to virtually identical results. 11

the U.S. dollar (Fig. 4, row 2). This is in line with Broda (2002) who shows in a similar panel VAR setting that floating exchange rate regimes insulate countries from terms of trade shocks. In highly dollarized countries and countries that have relatively high trade integration with the U.S., interest rates rise by more and currencies depreciate by less compared to their “low” counterparts, according to Figures 3 and 4. In two of the four comparisons the differences are statistically significant. This is consistent with Haussman et al., who find - using very different techniques - that liability dollarization pushes countries to smooth exchange rate fluctuations. To understand the apparently contradictory result that countries with high capital controls have both higher interest rate responses and lower exchange rate responses, note that subdividing countries by capital controls intensity leads to groupings similar to those obtained by dividing according to the other country factors (Table 1). The capital controls groupings might in fact be picking up exchange arrangement or dollarization effects. ThuswenextestimateVARscountry-by-countryandperform“second-stage regressions” controlling for all country factors simultaneously. 3.2 Bilateral results In Appendix 1 we display the impulse responses of all countries andall variables. To aid in ocular inference making, countries are depicted in ascending order of the capital controls index, that is, from the most open country (U.K.) to the most closed (Chile).12 Inspection of the individual country impulse responses suggests that our baseline VAR generally produces a reasonable representation of an identified U.S. monetary policy shock, by the standards of the literature.13 Tables 2A and 2B contain the results of our second-stage regressions. We present regressions with each country factor included separately, as well asone regressioncontainingall fourcountry factorstogether. We report the coefficients from the original impulse responses and the 16.5th and 83.5th 12The last five countries require a different scale and so are mixed. 13The responses of the U.S. variables are uniformly sensible, closely mirroring those of the panel VARs described above. Exceptions to the general pattern of “reasonable” responses include the occasional positive response of foreign output (e.g., the U.K. and Germany) and somewhat more pervasive price puzzle. 12

percentiles of the bootstrap distributions. We present results at horizons of 6, 12, 24, and 48 months. The results are very consistent with the evidence from the panel regressions. Consider first the regressions explaining exchange rate responses (Table2A).Inthesingle-regressorestimatesallvariablesareoftheexpected sign and statistically significant at the six-month horizon (upper left panel, first four columns). For example, the point estimate for the capital controls index is -1.49, with a confidence band of -2.23 to -0.45. The point estimate means that a twenty-five basis points contractionary shock in the U.S. induces a one hundred basis points larger depreciation in a relatively open country like Canada than in a more restrictive country like Korea, whose capital controls index is 0.65 higher than Canada’s. The capital controls index is also negative and statistically significant at twelve months (upper right panel, first column), but insignificant at longer horizons (lower panels). AccordingtotheregressionR-squaredvalues,thecapitalcontrolsindex explains between six and twelve percent of the cross-sectional variation in exchange rate responses at the short horizons, but virtually none at longer horizons. However, the (short-horizon) significance of the capital controls index vanishes when the other country factors are included. This can be seen from the final column of each panel, where the confidence band around the estimate for KC includes zero at all horizons considered. Due to the collinearity of capital controls and the other country factors, effects that might be attributed to capital restrictions are as likely to be a result of other country characteristics. In our case, the exchange rate regime dummy EAgenerallyhaslargereffectsthanthecapitalcontrolsindex. Theexchange rate regime is statistically significant at all horizons in the single-regressor estimates, has a reasonably high R-squared value, and retains significance uptothe24-monthhorizonwhenallregressorsare included(finalcolumnof each panel). Finally, note that for all regressions, the R-squared decreases with the length of the horizon. This makes sense, as one would expect that most of the reaction in asset markets occurs soon after the shock. Turning to interest rate responses (Table 2B), we see that the capital controls index is never statistically significant, consistent with the panel regressions. In contrast, our dollarization measure is significant and of the hypothesized sign in both individual and general regressions and at both short and long horizons. Countries with higher liability dollarization raise 13

interestratesbymorefollowingacontractionaryshockintheUnitedStates. The corresponding R-squared of fifteen to twenty percent at short-horizons is again relatively high.14 3.3 Robustness The results so far suggest that the degree to which U.S. money shocks are transmitted to foreign interest rates and exchange rates is essentially unaffected by the barriers to capital flows imposed by the foreign economy. The impulseresponsesofthegroupofhighcapitalcontrolscountriesarenotvery different from those of the low capital controls countries. Capital controls are found to be less important than the exchange rate regime or degree of liabilitydollarizationasafactordeterminingthetransmissionofU.S.money shocks across countries. In this section we examine the robustness of these results to the choice of sample period, the use of an alternative measure of capital controls, the identification scheme adopted in the VAR, and the possible endogeneity of capital controls. 3.3.1 Sample period To check whether the baseline results are robust to the choice of sample period, we re-estimate our panel VARs for the nine sub-periods containing at least two-thirds of the full sample, i.e., 1975-90, 1976-91, ..., 1983-1998. We ask, is there any sub-period for which the baseline VAR produces reasonable impulse responses and capital controls are found to be “effective”? Appendix 2 depicts the impulse responses for the high and low capital controlscountriesforeachsub-period.15 Inthe firstsixsub-periodsthe impulse responses suggest that the monetary shock is reasonably identified (again withapervasive price-puzzle). The interest rate responses R* are uniformly larger in high capital control countries, although the differences are not always statistically significant. The more restricted countries experience a smaller depreciation of their currencies, as seen from the plots labeled S, but the difference is typically small and significant only at short horizons. This mirrors the results from the full sample period. 14Note the Tables 2A and 2B results exclude Mexico and Turkey. With Mexico and Turkey, capital controls are found to have the opposite effect as the one hypothesized: more stringent controls lead to larger interest rate responses. 15Wekeepthesamecountrysub-divisionsasinthefullsampletomaintainconsistency. 14

For the last three sub-periods - and in particular the last two (1982-97 and 1983-98) - the impulse responses do not suggest that a reasonable U.S. monetary policy shock has been identified: domestic and foreign output rise following a contractionary shock, non-borrowed reserves rise and then fall, and the price puzzle becomes very pronounced. However, even in these last three sub-periods our primary conclusions hold: if anything, interest rates responses in high controls countries become much larger. Further experimentation reveals that the ”reasonableness” of the output, money, and price responses (though to repeat, not our conclusions concerning the effect of capital controls onthe R*andSresponses) issensitive toexcluding the years 1979-81. In order to obtain more reasonable impulse responses for the sub-periods that include the last few years of the sample, therefore, we re-estimate the panel VARs for the periods 1980-96, 1980-97, and 1980- 98. As can be seen from the Appendix, the impulse responses now suggest a more reasonable monetary policy shock, and once again our conclusions about the (in)effectiveness of capital controls is found to be robust. 3.3.2 An alternative measure of capital controls We wish to make sure that our main fndings are not driven by the particular measure of capital account restrictions. The Miniane (2000) measure, despite being demonstrably preferable to the pre-1996 IFS index, is still imperfect both for being de jure and for not taking explicitly into account the severity of each individual restriction. In this section, we examine the robustness of our results to using an alternative measure of capital controls derived from the capital account openness measures of Lane and Milesi- Ferretti (2001). The Lane-Milesi-Ferretti (LMF) openness index is the ratio of the stock of portfolio and direct investment assets and liabilities to GDP. The measure is different from our baseline measure in that it is a de facto metric. It is also well-known. The disadvantage of LMF as a measure of capital account openness is that changes in asset prices can lead to changes in the measure even with no changes in underlying restrictions or in the underlying financial position. Values of the LMF index are displayed in Table 1. Note that the correlation between LMF and our baseline capital controls measure is negative (-0.556), as it should be since the former measures capital account openness. First, we use the LMF measure to separate our sample of countries between high and low capital controls groups and examine in our panel VAR settingwhetherinterestrate andexchange rateresponsesare smallerforthe 15

former.16 As can be seen in Figure 5, the foreign interest rate responses R* areinsignificantlydifferentinthetwogroups. Theexchangeratedepreciates by less in the high capital controls group, as seen in the panels labeled S, but once again the difference is small and not persistent. These results are thus robust to using the alternative measure of capital account openness. Second,inTables3Aand3Bwepresentthecountry-by-country,”secondstage” regressions by analogy to Tables 2A and 2B. For the exchange rate responses, the alternative measure is always insignificantly different from zero (Table 3A). For the interest rate responses, the KC-LMF measure of openness is positive and significant at the six-month horizon in both the singleregressorandmultipleregressorspecifications. Itssignificancedeclines as the horizon increases, but KC-LMF is still positive and significant in the multiple regressor specifications up to 24 months. These results, which indicate that the interest rates of countries with de facto less open capital accounts (even if not more severe de jure more stringent capital account restrictions themselves) are less sensitive to U.S. monetary policy shocks, constitute the strongest evidence we have found of the ”effectiveness” of restricted capital accounts. Note, however, that we have again excluded Mexico and Turkey to be consistent with the baseline estimates of Tables 2A and 2B. Including Mexico and Turkey renders the KC-LMF coefficient in all R* regressions insignificant, thus making even this evdence of capital controls effectiveness disappear.17 3.3.3 Identification Recalling the controversies in the monetary VAR literature, it is important totest whether our results depend on the particular scheme used to identify the shocks. We re-estimate the “baseline” model using one hundred twenty different (recursive) identificationschemes. In all cases, we fix CP, Y, and P at the topof the ordering, arguingthat these variables do not react within a 16Ofcourse,thegroupingissomewhatdifferentundertheKC-LMFmeasure;otherwise thisexercisewouldnotbeinformative. The”lowcontrols”groupforourbaselinemeasure KCis: Australia,Austria,Canada,Denmark,France,Germany,Italy,Japan,Netherlands, Norway, Portugal, and the UK; for KC-LMF it is: Austria, Belgium, Canada, Chile, France,Germany,Malaysia,Netherlands,Norway,Spain,Sweden,andtheUK.Thus,7out of12countriesoverlap. TheinclusionofChileandMalaysiaamongthe”lowcontrols/high openness”countriesintheLMFmeasureisarguablycounter-intuitive,butisaby-product of the de facto nature of their measure. 17Recall that these countries have very closed capital accounts yet huge interest rate sensitivities to U.S. monetary shocks. 16

monthtoforeignoutput,pricesandassetmarketvariables. Wethenallowall possibleorderingsamongtheremainingfivevariablesinthesystem, yielding one hundred twenty specifications. Although this strategy falls short of the exhaustive search conducted in Faust and Rogers (2003), which is in principle over an infinite number of alternative schemes, it does represent a large set of plausible alternatives. Further, Faust, Rogers, Swanson, and Wright (2002) find that the baseline recursive structure we estimate in this paper cannot be rejected in favor of a large number of alternatives for one of the two currencies they examine. Table 4 summarizes our results by showing the percentage of rotations in which the coefficient of interest is of the hypothesized sign and statistically significant at the 67% confidence level. The top row presents results for the exchange rate responses at various horizons, while the bottom row is forthe interest rate responses. Thisrobustness exercise largely confirms our original results. In the exchange rate equations, the capital controlsindex is only significant in one third of the rotations at the shortest horizon (top left panel), but this significance fully disappears when one includes the other regressors (top right panel). In contrast, the significance of the exchange rate regime appears very robust. This country factor is significant in more than half of the rotations at all horizons in single-regressor specifications, and in almost all of the rotations at all horizons in the joint estimation. For interest rate responses, capital controls are never significant at any horizon. The dollarization index is almost always significant, either alone or controlling for the other regressors. The new result emerging from this exercise is that the exchange rate regime also appears to be important for explaining interest rate responses, at least in the single-regressor estimates. In short, the exchange rate regime and dollarization seem to be the key variables explaining cross-sectional differences in responses, while the effect of capital controls is very weak at best. 3.3.4 Endogeneity of capital controls It is possible that countries choose to restrict capital flows when vulnerable to foreign monetary disturbances. We would then observe in the data that high capital controls are associated with large interest rate and exchange responses. Even if capital controls ”worked” in the sense that they subsequently insulated the inherently-vulnerable country from foreign disturbances, we might findinourregressionsthatcapitalcontrolsare insignificant. 17

To control for the possible endogeneity of our capital controls index, we instrument for it using Transparency International’s well-known Corruption Perception Index. The index is an average of several survey-based measures of corruption from business executives, in-house experts at international consulting firms, and/or other people in the country in question. We use theaveragevalueofthecorruptionindexfrom1995to1999(theindexexists since 1995, measuring the year 1994). Values of the index are reported in Table 1. Note that a higher value corresponds to a less corrupt country. Time variation in the index is minimal: for our countries the correlation between the 1995 and 2002 values is 0.94. This gives us some comfort that using the average value of the index from 1995-99 is not far different from whatwewouldhavecomputedifwehadthedataoverourfullsampleperiod. Two facts make the corruption index a good instrument for capital controls. First, the correlation between the corruption and capital controls measures is high, at -0.75 (Table 1). Countries with higher levels of corruption are likely to have more red tape and in particular heavier regulation of capital flows. This is consistent with Wei (2000a,b), who uses the index to examine the effect of corruption on international investment. Second, there is good reason to believe on a priori grounds that corruption in a foreign country is not likely to be driven by U.S. monetary disturbances. Tables 5A and 5B present the results of two-stage least squares estimation for the exchange rate and foreign interest rate responses, respectively, by analogy to the baseline results in Tables 2A and 2B. Since the only country factor we instrument for is capital controls, we display only those specifications that include capital controls. The instrumented capital controls factor is, with one exception, insignificant at all horizons and in both single- and multiple-regressor specifications, while the exchange rate regime and the dollarization measures are significant and of the correct sign. The capital controls factor is negative and (barely) significant only for the R* response at horizon six (Table 5b, upper left panel). This investigation indicates that the possible endogeneity of capital controls does not appear to be spuriously driving our main findings. 3.3.5 Exogeneity of U.S. variables In our bilateral VARs, U.S. variables can respond (with a lag) to foreign variables. Because of this, the response of U.S. output, prices, and money toatwentyfivebasispointsFedFundsrateshockcanpotentiallyvaryacross 18

our country-specific VARs. Inspection of the impulse responses by country indicatesthatinfacttheydovary, albeitnotbymuch. Toensureuniformity of the U.S. responses across country VARs, we make CP, Y, P, FFR, and M block exogenous, and re-estimate the interest rate and exchange rate responses.18 It turns out that the responses of interest under the block exogeneity assumption are virtually no different than in the baseline case. The correlations of the new impulse responses with those obtained under the baseline specification are above 0.8 at all horizons, indicating that the U.S. block was not responding much to foreign variables in the first place. Tables 6A and 6B show the results from the new second-stage regressions. In this case, the finding that capital controls are ineffective is even stronger than in our baseline estimates, as the index is not significantly negative in any specification at any horizon. The exchange rate regime and the degree of dollarization factors are significant and of the hypothesized sign once again, while the trade integrationmeasure becomes more significant in the interest rate regression. 4 Conclusion In posing the provocative question ”Who needs capital account convertibility?”, Rodrik (1998) argues that if countries’ investment and output growth rates are not inhibited by capital controls, as he finds, the potential costs associated with capital account openness far exceed the benefits. Eichengreen’s (2002) survey concludes that capital controls may or may not affect investment and output growth across countries, and points to the difficulty of teasing such effects out of the data. In this paper, we pursue an implication of capital controls that is more direct and arguably more easily tested: controlsoughttoinsulateassetmarketsfromforeigndisturbances. Toadopt Rodrik’s line of reasoning from the other end, since there are costs, at least administrative ones, to imposing capital controls, it is also good to know whether controls are insulating home asset markets as they should. Our particular strategy is to test whether or not high capital controls countriesexhibit smallerresponsesoftheirinterest ratesandexchangerates 18We do not assume U.S. block exogeneity in our benchmark estimations to allow for easier replication of our results. Note also that inspection of the impulse responses under blockexogeneity(notdisplayed)indicatesthattheresponsesofU.S.variableswereindeed identical across countries once the VAR was restricted. 19

toU.S.monetaryshocks. Lookingsystematicallyatalargerange ofcountry experiences in a uniform estimation framework, we find essentially no evidence that capital controls are effective in this sense. Countries with more stringent controls apparently do experience smaller currency depreciations, but this result holds only at short horizons (as one would expect) and only if we do not condition on other relevant country factors like the degree of dollarization or the type of exchange rate arrangement. We also find that countries with more stringent capital controls do not experience smaller interest rate increases in response to contractionary U.S. monetary shocks. These findings are highly robust to the choice of sample period, the identification scheme adopted in the VAR, and the possible endogeneity of capital account restrictions. The results are also largely robust to an alternative, de facto proxy for capital controls. Why might capital controls have little or no effect? One reason is that controls are hard to enforce and can be evaded at small cost. If the opportunity cost of not evading controls - missing higher returns abroad - is high, agents will take their chances. Our measure of capital controls does not give any indication whether the government is making an effort to enforce them. Unfortunately, there is no systematic source of information regarding enforcement. Note that our results come from regressing average responses on the average level of capital controls, where averages are taken over periods relatively long periods of sixteen years or more. It is possible that controls are effective for a brief time after being established and are then left in place by bureaucratic inertia. In this case, the short-lived effect may not be picked up by our averaging. Nonetheless, note that important effects coming from the exchange rate regime and the degree of dollarization are detected despite our averaging. 5 Bibliography Aghion,P.,Bachetta,P.,andA.Banerjee,“CurrencyCrisesandMon- • etary Policy in an Economy with Credit Constraints,” mimeo, UCL, 2000. Ariyoshi,A.etal.,“CapitalControls: CountryExperienceswithTheir • Use and Liberalization”, IMF Occasional Paper 190, May 2000. Broda, C., “Terms of Trade and Exchange Rate Regimes in Develop- • ing Countries”, Federal Reserve Bank of New York Staff Report 148, March 2002, Journal of International Economics, forthcoming. 20

Cardoso, J., andB. Laurens, “Managing Capital Flows: Lessons From • the Experience of Chile,” IMF Working Paper 98/168, 1998. Clarida, R., and J. Gali, “Sources of Real Exchange Rate Fluctu- • ations: How Important Are Nominal Shocks?,” Carnegie-Rochester Conference Series on Public Policy, vol. 41, 1994, pp. 1-56. Chinn, M. D., and H. Ito, 2002, ”Capital Account Liberalization, In- • stitutions, and Financial Development, ” NBER Working Paper 8967. De Gregorio, J., Edwards, S., and R. Valdés, “Controls on Capital • Inflows: DoTheyWork?,”JournalofDevelopment Economics, vol.63, October 2000, pp. 59-83. Desai, M. A., Foley, C. F., and J. R. Hines Jr., 2003, ”Capital Con- • trols, Liberalizations, and Foreign Direct Investment,” available at http://wpweb2k.gsia.cmu.edu/wfa/wfapdf/dfhcapcontapr03.pdf. Dooley, M. P., “ASurveyofLiterature onControlsOverInternational • Capital Transactions,” IMF Staff Papers, vol. 43, December 1996, pp. 639-87. Edison, H., and C. M. Reinhart, “Stopping Hot Money,” Journal of • Development Economics, vol. 66, December 2001, pp. 533-553. Edwards, S., 2001, ”CapitalMobilityandEconomic Performance: Are • Emerging Economies Different?” NBER Working Paper 8076. Eichenbaum, M., and C. L. Evans, “Some Empirical Evidence on the • Effects of Shocks to Monetary Policy on Exchange Rates,” Quarterly Journal of Economics, vol. 110, November 1995, pp. 975-1009. Eichengreen, B., 2002, ”Capital Account Liberalization: What do the • Cross-Country Studies Tell Us?” World Bank Economic Review. Faust,J.,“TheRobustnessofVARConclusionsAboutMoney,”Carnegie— • Rochester Conference on Public Policy, vol. 49, December 1998, pp. 207-244. Faust, J., and J. H. Rogers, “Monetary Policy’s Role in Exchange • RateBehavior,”JournalofMonetaryEconomics,forthcomingOctober 2003. 21

Faust, J., Rogers, J.H., Swanson, E., Wright, J.H., “Identifying the • EffectsofMonetaryPolicyShocksonExchangeRatesUsingHighFrequency Data,” Journal of the European Economic Association, forthcoming. Fieleke, N. S., “International Capital Transactions: Should They Be • Restricted?”, New England Economic Review, Federal Reserve Bank of Boston, March-April 1994, pp. 27-39. Grilli, V., and G. M. Milesi-Ferretti, “Economic Effects and Struc- • tural Determinants of Capital Controls,” IMF Staff Papers, vol. 42, September 1995, pp. 517-51. Gros, D., “The Effectiveness of Capital Controls: Implications for • Monetary Autonomy in the Presence of Incomplete Market Separation,” IMF Staff Papers, vol. 34, December 1987, pp. 621-642. Gros, D., “DualExchangeRatesinthePresenceofIncompleteMarket • Separation: Long-Run Effectiveness and Policy Implications,” IMF Staff Papers, vol. 35, September 1988, pp. 437-460. Hausmann, R., Panizza, U., and E. Stein, “Why Do Countries Float • the Way They Float?”, Journal of Development Economics, vol. 66, December 2001, pp. 387-414. Kaplan, E., and D. Rodrik, “Did the Malaysian Capital Controls • Work?”, NBER Working Paper 8142, February 2001. Kim, S., and N. Roubini (2000): Exchange Rate Anomalies in the • Industrial Countries: a Solution with a Structural VAR Approach, Journal of Monetary Economics, 45, 561—586. Lane, P., and G.M. Milesi-Ferretti, “The External Wealth of Nations: • Measures of Foreign Assets and Liabilities for Industrial and Developing Countries,” Journal of International Economics, vol. 55, 2001, pp. 263-294. Leeper, E., Sims, C., and T. Zha, “What Does Monetary Policy Do?,” • Brookings Papers on Economic Activity, 2:1996, pp. 341-369. Miniane, J., “A New Set of Measures on Capital Account Restric- • tions,” unpublished draft, Johns Hopkins University, October 2000. 22

Reinhart, C., and K. S. Rogoff, “The Modern History of Exchange • RateArrangements: AReinterpretation,”NBERWorkingPaper8963, May 2002. Rodrik,D,1998,”WhoNeedsCapitalAccountConvertibility?”Prince- • ton Essays in International Finance, 207, pp. 55-65. Rogers, J., “Monetary Shocks and Real Exchange Rates,” Journal of • International Economics, vol. 48, 1999, pp. 269-288. Runkle, D. E., “Vector Autoregressions and Reality,” Journal of Busi- • ness and Economic Statistics, vol. 5, 1987, pp. 437-442. Quinn, D. P., 1997. ”The Correlates of Changes in International Fi- • nancial Regualtion,” American Political Science Review, 91, pp. 531- 551. Strongin,S.,“TheIdentificationofMonetaryPolicyDisturbances: Ex- • plaining the Liquidity Puzzle,” Journal of Monetary Economics, vol. 35, 1995, pp. 463-498. Wei, S. J., ”How Taxing is Corruption on Local Investors?”, Review • of Economics and Statistics, vol. 82, February 2000a, pp. 1-11. Wei, S. J., ”Local Corruption and Global Capital Flows,” Brookings • Papers on Economic Activity, 2000b, pp. 303-354. 6 Data appendix All VAR data are monthly, spanning the period January 1975 - December 1998unlessindicated. Non-seasonallyadjusteddatawereadjustedusingthe Fed’s X-11 routine. For foreign interest rates, we tried to obtain the closest available equivalent to the U.S. 3-month Tbill. Note that FRBI stands for Federal Reserve Board International Database, FRBD for its counterpart domestic database, and IFS for International Financial Statistics. Nominal Exchange Rates: IFS, all countries. • Industrial Production Index: (1) IFS for Austria, Canada, Colom- • bia, France, Germany, Greece, India, Italy, Japan, Korea, Malaysia, Mexico, Netherlands, Norway, Philipines, South Africa, Spain, Sweden, UK; (2) Turkish central bank for Turkey; (3) FRBI for Australia, 23

Belgium, Chile, Denmark, Finland, Portugal; and (4) FRBD for the U.S. Consumer Price Index: IFS for all countries except Australia (FRBI) • and U.S. (FRBD). Commodity Price: Producer Price Index - All Commodities, FRED- • Saint Louis. Federal Funds Rate: FRED-Saint Louis. • U.S. Non-borrowed Reserves and Total Reserves: FRBD. • Interest Rates: Australia: 13-week T-bill Rate, IFS; Austria: Money • Market Rate, IFS; Belgium: Call Money Rate, IFS; Canada: T-bill Rate, IFS; Chile: Lending Rate, IFS; Colombia: Discount Rate, IFS; Denmark: Call-MoneyRate,IFS;Finland: LendingRate,IFS;France: T-bill Rate, IFS; Germany: Call Money Rate, IFS; Greece: Central Bank Rate, IFS; India: Call Money Rate, IFS; Italy: T-bill Rate, IFS; Japan: Call Money Rate, IFS; Korea: Money Market Rate, IFS; Malaysia: T-bill Rate, IFS; Mexico: T-bill Rate, IFS; Netherlands: Money Market Rate, IFS; Norway: Call Money Rate, IFS; Philippines: T-bill Rate, IFS; Portugal: T-billRate, Eurostat; SouthAfrica: T-bill Rate, IFS; Spain: Money Market Rate, IFS; Sweden: Money Market Rate, IFS; Turkey: 3-month Deposit Rate, Central Bank of the Republic of Turkey; UK: T-bill Rate, IFS. United States: Industrial Production Index from FRBD, Consumer • Price Index from FRBD, Producer Price Index - All Commodities from FRED-Saint Louis, Federal Funds Rate from FRED-Saint Louis, Non-borrowed Reserves and Total Reserves from FRBD. DataonexchangerateregimeswasobtainedfromtheExchangeArrangements and Exchange Restrictions publication of the International Monetary Fund. Trade integration data starts in 1980 and was taken from the Direction of Trade Statistics of the IMF. International banking claims data starts in 1990 and comes from the Bank of International Settlements. Capital Controls data starts in 1983 and was taken from Miniane (2000). The Lane and Milesi-Ferretti measure was kindly provided by the authors. The Transparency International Corruption Perceptions Index can be found at www.transparency.org 24

Table1: Country Factors Country KC KC-IFS KC-LMF EA-IFS EA-RR USTR DOL CORRUP Australia 0.492 0.077 0.6 3.594 2.979 0.041 0.454 8.732 Austria 0.481 0.615 1.779 4 4 0.016 0.185 7.486 Belgium 0.553 0 3.506 4 4 0.127 0.295 5.928 Canada 0.192 0 2.306 4 3 0.388 0.550 9.066 Chile 1 1 0.872 1.976 2.969 0.092 0.838 6.898 Colombia 0.970 1 0.618 2.313 2.635 0.112 0.861 2.7 Denmark 0.269 0.385 0.773 4 4 0.024 0.225 9.718 Finland 0.558 0.615 0.732 2.861 4.024 0.027 0.275 9.41 France 0.462 0.538 1.065 4 4 0.023 0.271 6.784 Germany 0.231 0 0.888 4 4 0.030 0.121 8.108 Greece 0.548 1 0.461 2.882 3.326 0.015 0.438 4.84 India 0.917 1 0.082 2.549 2.191 0.021 0.556 2.792 Italy 0.428 0.538 0.471 4 3.660 0.022 0.185 4.148 Japan 0.471 0.077 0.423 4 3.878 0.044 0.518 6.428 Korea 0.841 1 0.247 1.875 1.899 0.122 0.795 4.32 Malaysia 0.846 0 1.614 2.191 3.028 0.222 0.535 5.202 Mexico 0.882 1 0.705 2.441 2.892 0.262 0.842 3.168 The Netherlands 0.149 0 4.806 4 4 0.063 0.325 8.886 Norway 0.409 0.923 0.848 3.014 3.538 0.044 0.533 8.86 The Philippines 0.894 1 0.395 2.250 2.667 0.149 0.757 3.082 Portugal 0.495 0.769 0.625 2.861 4 0.025 0.263 6.452 South Africa 0.861 1 0.555 3.531 3.160 0.051 0.669 5.29 Spain 0.603 0.846 0.885 4 3.500 0.018 0.269 5.452 Sweden 0.519 0.769 1.619 2.861 3.747 0.039 0.328 9.24 Turkey 0.773 1 0.135 3.375 4.566 0.020 0.541 3.57 U.K. 0.077 0 3.585 4 3.080 0.050 0.344 8.506 Mean 0.574 0.583 1.177 3.253 3.413 0.079 0.460 6.349 Standard deviation 0.269 0.420 1.167 0.767 0.652 0.090 0.226 2.311 Country Factors: Sample Correlations KC KC-IFS KC-LMF EA-IFS EA-RR USTR DOL CORRUP KC 1 KC-IFS 0.692 1 KC-LMF -0.556 -0.632 1 EA-IFS -0.776 -0.606 0.419 1 EA-RR -0.513 -0.307 0.227 0.625 1 USTR 0.139 -0.187 0.205 -0.244 -0.417 1 DOL 0.735 0.498 -0.306 -0.726 -0.710 0.492 1 CORRUP -0.749 -0.556 0.452 0.470 0.442 -0.117 -0.540 1 Note: KC stands for Miniane's (2000) capital controls index, KC-IFS for the IMF's pre-1996 zero/one capital controls dummy, KC-LMF for Lane and Milesi-Ferretti's (2001) capital controls measure, EA-IFS for the IMF's exchange rate classification, EA-RR for Reinhart and Rogoff's (2002) exchange rate classification, USTR for the ratio of exports to the U.S. plus imports from the U.S. as a share of the country's GDP, DOL for the share of the country's foreign currency banking sector claims that are in U.S. dollars, and CORRUP for Transparency International's corruption perception index. All variables are expressed as the country's sample period average, given data availability constraints (see text). A higher value for the KC-LMF measure corresponds to a more open capital account, thus explaining the negative correlation between KC and KC-LMF.

Table 2a: Regressions for Exchange Rate Responses at Various Horizons Six-Month Cumulative ER Response Twelve-Month Cumulative ER Response Constant 1.5707 -1.5455 1.0179 1.7739 -0.9696 Constant 1.8636 -3.6519 0.7868 1.6739 -5.6991 (0.81 2.04) (-2.40 -0.50) (0.61 1.22) (1.05 2.23) (-3.06 0.78) (0.43 3.24) (-5.90 -1.20) (-0.03 1.41) (0.29 3.00) (-9.78 -0.53) KC -1.4863 0.429 KC -2.4476 0.3099 (-2.23 -0.45) (-0.78 1.81) (-4.56 -0.27) (-2.72 3.26) EA 0.6993 0.6074 EA 1.2685 1.6642 (0.35 0.96) (0.20 1.01) (0.51 1.95) (0.59 2.48) USTR -3.6491 -1.8379 USTR -3.7488 -3.2168 (-5.14 -1.49) (-3.67 0.51) (-8.24 0.67) (-8.13 2.42) DOL -2.3226 -0.853 DOL -2.6365 1.8454 (-3.23 -1.05) (-2.24 0.58) (-5.37 0.01) (-2.18 4.89) R-squared 0.1176 0.2201 0.0707 0.1932 0.2638 R-squared 0.0537 0.122 0.0126 0.0419 0.1347 Twenty Four-Month Cumulative ER Response Forty Eight-Month Cumulative ER Response Constant 1.5805 -8.0206 0.0295 1.0758 -15.9727 Constant 1.6668 -11.2581 -0.0707 1.9617 -18.8136 (-0.64 4.80) (-12.07 -1.55) (-0.91 1.88) (-0.98 4.56) (-20.40 -1.45) (-0.61 7.83) (-18.50 0.60) (-0.67 3.44) (-0.34 8.66) (-21.70 9.01) KC -3.6746 1.5456 KC -4.9211 3.8105 (-8.37 0.68) (-5.20 6.87) (-13.72 1.74) (-8.15 9.46) EA 2.3068 3.8164 EA 3.1098 4.5773 (0.58 3.60) (0.78 4.81) (0.16 5.64) (-0.90 5.64) USTR -6.5169 -7.569 USTR -13.3382 -10.4263 (-17.12 1.89) (-17.28 4.19) (-30.34 0.50) (-25.68 8.31) DOL -3.4595 6.123 DOL -6.8333 3.1765 (-9.98 1.34) (-4.18 10.45) (-18.86 1.73) (-14.34 10.19) R-squared 0.0248 0.0827 0.0078 0.0148 0.1093 R-squared 0.015 0.0507 0.011 0.0195 0.0626 Note 1: Results exclude Mexico and Turkey Note 2: In parenthesis are the 16.5th and 83.5th percentiles of the coefficient's distribution Note 3: The R-squared are computed from the regression using the original impulse responses

Table 2b: Regressions for Foreign Interest Rate Responses at Various Horizons Six-Month Cumulative R* Response Twelve-Month Cumulative R* Response Constant -0.0044 0.6854 0.1611 -0.1427 -0.2753 Constant 0.0934 1.5419 0.4442 -0.2377 -0.3514 (-0.20 0.18) (0.05 1.25) (0.08 0.23) (-0.35 0.06) (-0.81 0.19) (-0.29 0.46) (0.21 2.49) (0.24 0.55) (-0.59 0.19) (-1.44 0.68) KC 0.41 -0.1383 KC 0.8229 -0.5648 (-0.08 0.86) (-0.49 0.21) (-0.14 1.63) (-1.17 0.29) EA -0.1412 0.0355 EA -0.3028 0.0479 (-0.30 0.03) (-0.07 0.15) (-0.57 0.04) (-0.18 0.28) USTR 0.829 -0.2813 USTR 1.4129 -1.2604 (0.24 1.39) (-0.88 0.32) (-0.04 2.67) (-2.53 0.31) DOL 0.827 1.0836 DOL 1.7813 2.601 (0.20 1.42) (0.53 1.65) (0.41 2.75) (1.24 3.33) R-squared 0.055 0.0551 0.0224 0.1505 0.1572 R-squared 0.065 0.0744 0.0191 0.2049 0.2258 Twenty Four-Month Cumulative R* Response Forty Eight-Month Cumulative R* Response Constant 0.1251 2.7521 0.9556 -0.1145 0.3954 Constant 0.1703 3.7576 1.3687 0.2416 2.9258 (-0.42 0.76) (0.06 3.62) (0.47 1.00) (-0.56 0.66) (-1.71 2.30) (-0.65 0.96) (-0.13 4.55) (0.45 1.23) (-0.60 1.04) (-1.28 4.72) KC 1.519 -0.4979 KC 1.917 -0.4342 (-0.35 2.34) (-1.47 1.12) (-0.67 2.83) (-2.04 1.69) EA -0.5447 -0.0957 EA -0.7703 -0.5916 (-0.83 0.13) (-0.48 0.37) (-1.04 0.18) (-0.97 0.28) USTR 0.1233 -3.4567 USTR -1.884 -4.8335 (-3.05 2.50) (-5.26 0.23) (-4.98 2.58) (-6.57 1.09) DOL 2.4458 3.2026 DOL 2.2399 1.9074 (-0.33 3.31) (0.18 3.68) (-1.01 3.49) (-1.29 3.57) R-squared 0.0887 0.0964 0.0001 0.1547 0.1888 R-squared 0.124 0.1693 0.0119 0.1139 0.2296 Notes: see notes in Table 2a.

Table 3a: Robustness to Alternative Measures of Controls (Exchange Rate Responses) Six-Month Cumulative ER Response Twelve-Month Cumulative ER Response Constant 0.594 -0.5876 Constant 0.2662 -5.4623 (0.23 0.89) (-2.16 0.84) (-0.59 1.09) (-8.75 -1.56) KC-LMF 0.1251 -0.0273 KC-LMF 0.1971 -0.1204 (-0.05 0.28) (-0.25 0.2) (-0.17 0.59) (-0.62 0.39) EA 0.549 EA 1.6748 (0.13 0.93) (0.65 2.51) USTR -1.9677 USTR -2.7538 (-3.57 -0.14) (-7.03 1.75) DOL -0.6487 DOL 1.8798 (-1.91 0.8) (-1.7 4.83) R-squared 0.016 0.262 R-squared 0.0068 0.1362 Twenty Four-Month Cumulative ER Response Forty Eight-Month Cumulative ER Response Constant -1.0207 -14.563 Constant -1.9274 -15.2606 (-2.23 1.1) (-17.62 -2.8) (-3.27 2.04) (-18.31 6.31) KC-LMF 0.4598 -0.0127 KC-LMF 0.705 0.1679 (-0.31 1.23) (-0.94 1.05) (-0.45 1.85) (-1.13 1.69) EA 3.5612 EA 3.8438 (0.87 4.58) (-0.79 5.13) USTR -8.5095 USTR -13.8471 (-16.75 2.13) (-28.01 5.88) DOL 6.9553 DOL 5.4524 (-3.14 10.82) (-12.79 10.9) R-squared 0.0077 0.108 R-squared 0.006 0.06 Notes: see notes in Table 2a.

Table 3b: Robustness to Alternative Measures of Controls (Foreign Interest Rate Responses) Six-Month Cumulative R* Response Twelve-Month Cumulative R* Response Constant 0.1414 -0.343 Constant 0.4094 -0.7518 (0.02 0.25) (-0.75 -0.01) (0.13 0.6) (-1.56 0.07) KC-LMF 0.0651 0.1508 KC-LMF 0.112 0.299 (0.02 0.11) (0.08 0.2) (-0.02 0.19) (0.13 0.37) EA -0.0201 EA -0.013 (-0.12 0.1) (-0.22 0.24) USTR -1.0246 USTR -2.5436 (-1.86 -0.16) (-4.02 -0.39) DOL 1.1774 DOL 2.6277 (0.57 1.76) (1.11 3.54) R-squared 0.0273 0.2576 R-squared 0.0237 0.3344 Twenty Four-Month Cumulative R* Response Forty Eight-Month Cumulative R* Response Constant 0.8565 0.107 Constant 1.3495 2.6485 (0.22 1.05) (-1.23 1.68) (0.26 1.49) (-0.66 3.63) KC-LMF 0.0873 0.4295 KC-LMF -0.0965 0.3081 (-0.12 0.24) (0.08 0.46) (-0.36 0.18) (-0.16 0.39) EA -0.2362 EA -0.6794 (-0.53 0.22) (-0.88 0.18) USTR -5.5049 USTR -6.2521 (-6.99 -0.68) (-7.52 0.95) DOL 3.4126 DOL 2.0158 (0.28 4.09) (-1.46 3.51) R-squared 0.0058 0.2834 R-squared 0.0062 0.2717 Notes: see notes in Table 2a.

Table 4: Robustness to Identification Single Regressor Estimation Joint Regressors Estimation S6 S12 S24 S48 S6 S12 S24 S48 KC 33.33 9.17 0 0 0 0 0 0 EA 50 74.17 65 52.5 95.83 100 100 73.33 USTR 35.83 0 0 7.5 0 0 0 0.83 DOL 35 0.83 0 0 0 0 0 0 Single Regressor Estimation Joint Regressors Estimation RF6 RF12 RF24 RF48 RF6 RF12 RF24 RF48 KC 0 0 0 0 0 0 0 0 EA 86.67 65.83 49.17 23.33 0 0 0 0 USTR 100 91.67 50 50 31.67 31.67 29.17 49.17 DOL 100 100 89.17 59.17 100 100 94.17 55 Notes: Figures correspond to the percentage of rotations in which the coefficient is significant (and of the right sign) at 67% confidence in the boostrapping distribution. S6 stands for the six-month exchange rate response.

Table 5a: Instrumental Variables Regressions for Exchange Rate Responses Six-Month Cumulative ER Response Twelve-Month Cumulative ER Response Constant 0.9825 -1.3194 Constant 1.1235 -6.0578 (0 1.87) (-2.99 0.61) (-0.68 3.42) (-9.51 -0.76) KCI -0.4223 1.3212 KCI -1.1088 1.143 (-1.92 1.03) (-0.3 2.69) (-5.12 1.77) (-3.03 3.91) EA 0.5943 EA 1.6632 (0.20 0.91) (0.61 2.29) USTR -1.5755 USTR -2.9496 (-3.62 0.63) (-7.77 2.15) DOL -1.1241 DOL 1.5773 (-2.4 0.28) (-1.87 4.62) R-squared 0.005 0.2963 R-squared 0.006 0.1388 Twenty Four-Month Cumulative ER Response Forty Eight-Month Cumulative ER Response Constant 1.9829 -13.3439 Constant 5.0127 -10.2714 (-1.11 6.59) (-17.14 -0.76) (-0.75 11.61) (-17.03 11.86) KCI -4.4027 -2.1605 KCI -10.9747 -8.9943 (-11.7 1.87) (-9.41 4.31) (-20.56 1.88) (-16.43 5.6) EA 3.4571 EA 3.5261 (0.69 4.31) (-0.95 4.91) USTR -9.468 USTR -16.6162 (-19.21 2.26) (-28.42 5.6) DOL 7.7972 DOL 8.7091 (-2.69 11.83) (-11.88 12.17) R-squared 0.0187 0.1113 R-squared 0.0392 0.0785 Notes: see notes in Table 2a. The estimates of KCI come from a 2SLS regression in which the Miniane capital controls measures are instrumented for with the period average of the Transparency International's Corruption Perception Index, whose values are listed in Table 1.

Table 5b: Instrumental Variables Regressions for Foreign Interest Rate Responses Six-Month Cumulative R* Response Twelve-Month Cumulative R* Response Constant 0.2795 0.0304 Constant 0.5249 -0.1249 (0.02 0.5) (-0.59 0.5) (-0.07 0.84) (-1.41 0.73) KCI -0.1038 -0.769 KCI 0.0421 -1.3229 (-0.48 0.35) (-1.37 -0.05) (-0.7 1.13) (-2.13 0.35) EA 0.0242 EA 0.0839 (-0.08 0.14) (-0.1 0.32) USTR -0.507 USTR -1.4347 (-1.33 0.28) (-2.8 0.55) DOL 1.3024 DOL 1.5887 (0.6 1.99) (1.08 3.81) R-squared 0.0018 0.2275 R-squared 0 0.2791 Twenty Four-Month Cumulative R* Response Forty Eight-Month Cumulative R* Response Constant 0.7659 0.908 Constant 0.1485 2.4098 (-0.39 1.33) (-1.38 2.26) (-1.28 1.11) (-1.77 3.71) KCI 0.3597 -1.7232 KCI 1.9563 0.2112 (-1.29 2.15) (-2.73 1.28) (-0.99 3.89) (-1.9 3.31) EA -0.089 EA -0.5086 (-0.41 0.36) (-0.82 0.29) USTR -3.8392 USTR -4.4627 (-5.43 0.46) (-5.79 2.04) DOL 3.59 DOL 1.5887 (-0.04 4.21) (-1.87 3.04) R-squared 0.0026 0.2285 R-squared 0.0678 0.2282 Notes: see notes in Table 2a. The estimates of KCI come from a 2SLS regression in which the Miniane capital controls measures are instrumented for with the period average of the Transparency International's Corruption Perception Index, whose values are listed in Table 1.

Table 6a: Regressions for Exchange Rate Responses under US Block Exogeneity Six-Month Cumulative ER Response Twelve-Month Cumulative ER Response Constant 1.4493 -1.4555 0.9976 1.6468 -1.4686 Constant 1.1375 -2.8703 0.4563 0.8446 -6.19 (0.39 1.69) (-2.11 0.01) (0.4 1.06) (0.49 1.8) (-3.6 0.54) (-1.14 1.9) (-4.6 0.6) (-0.83 0.69) (-1.69 1.46) (-10.67 -0.55) KC -1.2961 0.6062 KC -1.5757 0.3486 (-1.85 0.1) (-0.87 1.88) (-3.34 1.47) (-3.36 2.91) EA 0.6669 0.6852 EA 0.9559 1.5859 (0.15 0.83) (0.16 1.02) (-0.24 1.41) (0.2 2.28) USTR -3.594 -2.0644 USTR -2.577 -3.423 (-4.3 -0.45) (-3.76 0.69) (-4.74 4.95) (-7.6 3.41) DOL -2.0715 -0.5217 DOL -1.3103 2.9729 (-2.6 -0.13) (-1.58 1.44) (-3.13 3.09) (-0.39 7.72) R-squared 0.0853 0.1907 0.065 0.1466 0.2303 R-squared 0.0206 0.0639 0.0054 0.0095 0.0872 Twenty Four-Month Cumulative ER Response Forty Eight-Month Cumulative ER Response Constant 0.4266 -6.1311 -0.3134 -0.2062 -14.682 Constant 2.1558 -12.5836 1.653 3.9444 -22.0321 (-2.94 2.81) (-10.09 1.45) (-2.07 0.87) (-3.91 2.22) (-20.8 -0.16) (-2.26 7.37) (-21.52 2.56) (-1.36 3.92) (-1.92 8.82) (-32.61 7.51) KC -2.1677 0.7678 KC -4.697 7.7826 (-6.54 3.31) (-7.23 5.91) (-14.68 4.62) (-8.21 13.5) EA 1.6332 3.2625 EA 3.7004 5.6108 (-0.64 2.76) (0.03 4.34) (-0.59 6.13) (-0.67 7.47) USTR -6.2201 -10.0243 USTR -28.4172 -22.5862 (-12.58 7.73) (-17.49 4.36) (-38.92 1.28) (-35.02 4.83) DOL -1.2815 7.9742 DOL -9.938 1.2278 (-6.8 6.52) (-1.36 15.39) (-20.39 5.48) (-13.74 15.4) R-squared 0.0089 0.0426 0.0073 0.0021 0.0806 R-squared 0.0131 0.0687 0.0478 0.0395 0.1156 Notes: see notes in Table 2a.

Table 6b: Regressions for Interest Rate Responses under US Block Exogeneity Six-Month Cumulative R* Response Twelve-Month Cumulative R* Response Constant -0.035 0.8118 0.2055 -0.159 -0.3192 Constant 0.1291 1.6892 0.6254 -0.1414 -0.3835 (-0.23 0.14) (0.22 1.39) (0.11 0.27) (-0.37 0.03) (-0.8 0.17) (-0.25 0.55) (0.19 2.66) (0.39 0.72) (-0.51 0.34) (-1.36 0.78) KC 0.5205 -0.0823 KC 0.94 -0.4947 (0.08 0.94) (-0.44 0.22) (-0.18 1.74) (-1.3 0.26) EA -0.1704 0.0397 EA -0.3171 0.0748 (-0.33 -0.02) (-0.07 0.15) (-0.59 0.07) (-0.17 0.3) USTR 0.641 0.6094 USTR 0.316 -2.5885 (0.07 1.24) (-1.23 0.06) (-1.19 1.62) (-3.87 -0.73) DOL 0.933 1.2054 DOL 1.7906 2.8351 (0.35 1.52) (0.67 1.75) (0.22 2.77) (1.29 3.7) R-squared 0.0673 0.0607 0.0102 0.1455 0.1535 R-squared 0.0693 0.0667 0.0008 0.169 0.2091 Twenty Four-Month Cumulative R* Response Forty Eight-Month Cumulative R* Response Constant 0.1696 3.2139 1.4293 0.1421 0.3631 Constant -0.0967 3.995 1.9082 0.6459 2.554 (-0.5 0.96) (0.07 4.56) (0.86 1.44) (-0.5 1.02) (-1.69 2.49) (-1.15 1.16) (-1.32 6.17) (0.85 1.8) (-0.76 1.8) (-1.95 5.08) KC 1.9168 -0.1112 KC 2.6179 1.2312 (-0.44 3.1) (-1.6 1.4) (-1.37 4.55) (-1.68 3.01) EA -0.6048 -0.0438 EA -0.8059 -0.4775 (-1.01 0.18) (-0.51 0.46) (-1.45 0.53) (-1.04 0.48) USTR -2.7177 -6.6754 USTR -7.5729 -9.5715 (-6.1 0.64) (-8.69 -2.08) (-12.73 1.11) (-13.1 -1.19) DOL 2.4636 3.5422 DOL 1.5965 0.8792 (-0.67 3.96) (0.32 4.48) (-3.02 4.77) (-2.62 3.86) R-squared 0.1073 0.0903 0.0215 0.1192 0.2246 R-squared 0.1548 0.124 0.1292 0.0387 0.3309 Notes: see notes in Table 2a.

Figure 1: Malaysia Before and After Controls 2 Exchange Rate T-Bill Rate Reserves 1.8 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 1 3 5 7 9 1 1 3 5 7 9 1 1 3 5 7 9 1 M M M M M 1 M M M M M 1 M M M M M 1 7 7 7 7 7 M 8 8 8 8 8 M 9 9 9 9 9 M 9 9 9 9 9 7 9 9 9 9 9 8 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 1 1 1 1 1 9 1 1 1 1 1 9 1 1 1 1 1 9 1 1 1

Low Capital Controls 0 -0.1 -0.2 -0.3 0 5 10 15 20 25 30 35 40 45 Y High Capital Controls 0 -0.1 -0.2 -0.3 0 5 10 15 20 25 30 35 40 45 0.05 0 -0.05 -0.1 0 5 10 15 20 25 30 35 40 45 P 0.05 0 -0.05 -0.1 0 5 10 15 20 25 30 35 40 45 0.1 0 -0.1 0 5 10 15 20 25 30 35 40 45 PC 0.1 0 -0.1 0 5 10 15 20 25 30 35 40 45 30 20 10 0 0 5 10 15 20 25 30 35 40 45 RFF 30 20 10 0 0 5 10 15 20 25 30 35 40 45 0 -0.2 -0.4 0 5 10 15 20 25 30 35 40 45 XRBN 0 -0.2 -0.4 0 5 10 15 20 25 30 35 40 45 20 10 0 0 5 10 15 20 25 30 35 40 45 *R 20 10 0 0 5 10 15 20 25 30 35 40 45 0.6 0.4 0.2 0 0 5 10 15 20 25 30 35 40 45 S 0.6 0.4 0.2 0 0 5 10 15 20 25 30 35 40 45 0.05 0 -0.05 -0.1 -0.15 0 5 10 15 20 25 30 35 40 45 *Y Figure 2: Comparison of Panel Impulse Responses 0.05 0 -0.05 -0.1 -0.15 0 5 10 15 20 25 30 35 40 45

Low Capital Controls 0 -0.1 -0.2 -0.3 0 5 10 15 20 25 30 35 40 45 Y High Capital Controls 0 -0.1 -0.2 -0.3 0 5 10 15 20 25 30 35 40 45 0.05 0 -0.05 -0.1 0 5 10 15 20 25 30 35 40 45 P 0.05 0 -0.05 -0.1 0 5 10 15 20 25 30 35 40 45 0.1 0 -0.1 0 5 10 15 20 25 30 35 40 45 PC 0.1 0 -0.1 0 5 10 15 20 25 30 35 40 45 30 20 10 0 0 5 10 15 20 25 30 35 40 45 RFF 30 20 10 0 0 5 10 15 20 25 30 35 40 45 0 -0.2 -0.4 0 5 10 15 20 25 30 35 40 45 XRBN 0 -0.2 -0.4 0 5 10 15 20 25 30 35 40 45 15 10 5 0 0 5 10 15 20 25 30 35 40 45 *R 15 10 5 0 0 5 10 15 20 25 30 35 40 45 0.4 0.2 0 0 5 10 15 20 25 30 35 40 45 S 0.4 0.2 0 0 5 10 15 20 25 30 35 40 45 0.05 0 -0.05 -0.1 -0.15 0 5 10 15 20 25 30 35 40 45 *Y Figure 2b: Comparison excluding Turkey and Mexico 0.05 0 -0.05 -0.1 -0.15 0 5 10 15 20 25 30 35 40 45

15 10 5 0 0 10 20 30 40 CK woL 15 10 5 0 0 10 20 30 40 CK hgiH 15 10 5 0 -5 0 10 20 30 40 LOD woL 15 10 5 0 -5 0 10 20 30 40 LOD hgiH 20 10 0 0 10 20 30 40 srexiF 20 10 0 0 10 20 30 40 sretaolF 15 10 5 0 -5 0 10 20 30 40 edarT SU woL 15 10 5 0 -5 0 10 20 30 40 edarT SU hgiH Figure 3: Comparison of Interest Rate Responses

0.4 0.2 0 0 10 20 30 40 CK woL 0.4 0.2 0 0 10 20 30 40 CK hgiH 0.4 0.2 0 0 10 20 30 40 LOD woL 0.4 0.2 0 0 10 20 30 40 LOD hgiH 0.6 0.4 0.2 0 0 10 20 30 40 srexiF 0.6 0.4 0.2 0 0 10 20 30 40 sretaolF 0.4 0.2 0 0 10 20 30 40 edarT SU woL 0.4 0.2 0 0 10 20 30 40 edarT SU hgiH Figure 4: Comparison of Exchange Rate Responses

Low Capital Controls 0 -0.1 -0.2 -0.3 0 5 10 15 20 25 30 35 40 45 Y High Capital Controls 0 -0.1 -0.2 -0.3 0 5 10 15 20 25 30 35 40 45 0.05 0 -0.05 -0.1 0 5 10 15 20 25 30 35 40 45 P 0.05 0 -0.05 -0.1 0 5 10 15 20 25 30 35 40 45 0.1 0 -0.1 0 5 10 15 20 25 30 35 40 45 PC 0.1 0 -0.1 0 5 10 15 20 25 30 35 40 45 30 20 10 0 0 5 10 15 20 25 30 35 40 45 RFF 30 20 10 0 0 5 10 15 20 25 30 35 40 45 0 -0.2 -0.4 0 5 10 15 20 25 30 35 40 45 XRBN 0 -0.2 -0.4 0 5 10 15 20 25 30 35 40 45 15 10 5 0 0 5 10 15 20 25 30 35 40 45 *R 15 10 5 0 0 5 10 15 20 25 30 35 40 45 0.4 0.2 0 0 5 10 15 20 25 30 35 40 45 S 0.4 0.2 0 0 5 10 15 20 25 30 35 40 45 0.05 0 -0.05 -0.1 -0.15 0 5 10 15 20 25 30 35 40 45 *Y Figure 5: Comparison using the Lane and Milesi-Ferretti measure 0.05 0 -0.05 -0.1 -0.15 0 5 10 15 20 25 30 35 40 45

0.2 0 -0.2 -0.4 0 10 20 30 40 Y UK 0.2 0 -0.2 0 10 20 30 40 P 0.4 0.2 0 -0.2 0 10 20 30 40 PC 30 20 10 0 -10 0 10 20 30 40 RFF 0.2 0 -0.2 -0.4 0 10 20 30 40 XRBN 20 0 -20 0 10 20 30 40 *R 1 0 -1 0 10 20 30 40 S 0.4 0.2 0 -0.2 -0.4 0 10 20 30 40 *Y Appendix 1: Impulse responses for all countries Netherlands Canada Germany 0.2 0.2 0.2 0 0 0 -0.2 -0.2 -0.2 -0.4 -0.4 -0.4 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0.2 0.2 0.2 0 0 0 -0.2 -0.2 -0.2 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0.4 0.4 0.4 0.2 0.2 0.2 0 0 0 -0.2 -0.2 -0.2 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 30 30 30 20 20 20 10 10 10 0 0 0 -10 -10 -10 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0.2 0.2 0.2 0 0 0 -0.2 -0.2 -0.2 -0.4 -0.4 -0.4 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 20 20 20 0 0 0 -20 -20 -20 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 1 1 1 0 0 0 -1 -1 -1 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0.4 0.4 0.4 0.2 0.2 0.2 0 0 0 -0.2 -0.2 -0.2 -0.4 -0.4 -0.4 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40

0.2 0 -0.2 -0.4 0 10 20 30 40 Y Norway 0.2 0 -0.2 0 10 20 30 40 P 0.4 0.2 0 -0.2 0 10 20 30 40 PC 30 20 10 0 -10 0 10 20 30 40 RFF 0.2 0 -0.2 -0.4 0 10 20 30 40 XRBN 20 0 -20 0 10 20 30 40 *R 1 0 -1 0 10 20 30 40 S 0.4 0.2 0 -0.2 -0.4 0 10 20 30 40 *Y Italy France Japan 0.2 0.2 0.2 0 0 0 -0.2 -0.2 -0.2 -0.4 -0.4 -0.4 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0.2 0.2 0.2 0 0 0 -0.2 -0.2 -0.2 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0.4 0.4 0.4 0.2 0.2 0.2 0 0 0 -0.2 -0.2 -0.2 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 30 30 30 20 20 20 10 10 10 0 0 0 -10 -10 -10 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0.2 0.2 0.2 0 0 0 -0.2 -0.2 -0.2 -0.4 -0.4 -0.4 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 20 20 20 0 0 0 -20 -20 -20 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 1 1 1 0 0 0 -1 -1 -1 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0.4 0.4 0.4 0.2 0.2 0.2 0 0 0 -0.2 -0.2 -0.2 -0.4 -0.4 -0.4 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40

0.2 0 -0.2 -0.4 0 10 20 30 40 Y Austria 0.2 0 -0.2 0 10 20 30 40 P 0.4 0.2 0 -0.2 0 10 20 30 40 PC 30 20 10 0 -10 0 10 20 30 40 RFF 0.2 0 -0.2 -0.4 0 10 20 30 40 XRBN 20 0 -20 0 10 20 30 40 *R 1 0 -1 0 10 20 30 40 S 0.4 0.2 0 -0.2 -0.4 0 10 20 30 40 *Y Australia Portugal Sweden 0.2 0.2 0.2 0 0 0 -0.2 -0.2 -0.2 -0.4 -0.4 -0.4 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0.2 0.2 0.2 0 0 0 -0.2 -0.2 -0.2 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0.4 0.4 0.4 0.2 0.2 0.2 0 0 0 -0.2 -0.2 -0.2 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 30 30 30 20 20 20 10 10 10 0 0 0 -10 -10 -10 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0.2 0.2 0.2 0 0 0 -0.2 -0.2 -0.2 -0.4 -0.4 -0.4 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 20 20 20 0 0 0 -20 -20 -20 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 1 1 1 0 0 0 -1 -1 -1 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0.4 0.4 0.4 0.2 0.2 0.2 0 0 0 -0.2 -0.2 -0.2 -0.4 -0.4 -0.4 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40

0.2 0 -0.2 -0.4 0 10 20 30 40 Y Greece 0.2 0 -0.2 0 10 20 30 40 P 0.4 0.2 0 -0.2 0 10 20 30 40 PC 30 20 10 0 -10 0 10 20 30 40 RFF 0.2 0 -0.2 -0.4 0 10 20 30 40 XRBN 20 0 -20 0 10 20 30 40 *R 1 0 -1 0 10 20 30 40 S 0.4 0.2 0 -0.2 -0.4 0 10 20 30 40 *Y Belgium Finland Spain 0.2 0.2 0.2 0 0 0 -0.2 -0.2 -0.2 -0.4 -0.4 -0.4 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0.2 0.2 0.2 0 0 0 -0.2 -0.2 -0.2 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0.4 0.4 0.4 0.2 0.2 0.2 0 0 0 -0.2 -0.2 -0.2 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 30 30 30 20 20 20 10 10 10 0 0 0 -10 -10 -10 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0.2 0.2 0.2 0 0 0 -0.2 -0.2 -0.2 -0.4 -0.4 -0.4 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 20 20 20 0 0 0 -20 -20 -20 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 1 1 1 0 0 0 -1 -1 -1 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0.4 0.4 0.4 0.2 0.2 0.2 0 0 0 -0.2 -0.2 -0.2 -0.4 -0.4 -0.4 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40

0.2 0 -0.2 -0.4 0 20 40 Y Korea 0.2 0 -0.2 0 20 40 P 0.4 0.2 0 -0.2 0 20 40 PC 30 20 10 0 -10 0 20 40 RFF 0.2 0 -0.2 -0.4 0 20 40 XRBN 20 0 -20 0 20 40 *R 1 0 -1 0 20 40 S 0.4 0.2 0 -0.2 -0.4 0 20 40 *Y Malaysia S. Africa India Colombia 0.2 0.2 0.2 0.2 0 0 0 0 -0.2 -0.2 -0.2 -0.2 -0.4 -0.4 -0.4 -0.4 0 20 40 0 20 40 0 20 40 0 20 40 0.2 0.2 0.2 0.2 0 0 0 0 -0.2 -0.2 -0.2 -0.2 0 20 40 0 20 40 0 20 40 0 20 40 0.4 0.4 0.4 0.4 0.2 0.2 0.2 0.2 0 0 0 0 -0.2 -0.2 -0.2 -0.2 0 20 40 0 20 40 0 20 40 0 20 40 30 30 30 30 20 20 20 20 10 10 10 10 0 0 0 0 -10 -10 -10 -10 0 20 40 0 20 40 0 20 40 0 20 40 0.2 0.2 0.2 0.2 0 0 0 0 -0.2 -0.2 -0.2 -0.2 -0.4 -0.4 -0.4 -0.4 0 20 40 0 20 40 0 20 40 0 20 40 20 20 20 20 0 0 0 0 -20 -20 -20 -20 0 20 40 0 20 40 0 20 40 0 20 40 1 1 1 1 0 0 0 0 -1 -1 -1 -1 0 20 40 0 20 40 0 20 40 0 20 40 0.4 0.4 0.4 0.4 0.2 0.2 0.2 0.2 0 0 0 0 -0.2 -0.2 -0.2 -0.2 -0.4 -0.4 -0.4 -0.4 0 20 40 0 20 40 0 20 40 0 20 40

0.2 0 -0.2 -0.4 -0.6 0 20 40 Y Denmark 0.2 0.1 0 0 20 40 P 0.4 0.2 0 0 20 40 PC 30 20 10 0 -10 0 20 40 RFF 0.2 0 -0.2 -0.4 0 20 40 XRBN 20 0 -20 0 20 40 *R 2 0 -2 0 20 40 S 1.5 1 0.5 0 -0.5 0 20 40 *Y Turkey Mexico Philippines Chile 0.2 0.2 0.2 0.2 0 0 0 0 -0.2 -0.2 -0.2 -0.2 -0.4 -0.4 -0.4 -0.4 -0.6 -0.6 -0.6 -0.6 0 20 40 0 20 40 0 20 40 0 20 40 0.2 0.2 0.2 0.2 0.1 0.1 0.1 0.1 0 0 0 0 0 20 40 0 20 40 0 20 40 0 20 40 0.4 0.4 0.4 0.4 0.2 0.2 0.2 0.2 0 0 0 0 0 20 40 0 20 40 0 20 40 0 20 40 30 30 30 30 20 20 20 20 10 10 10 10 0 0 0 0 -10 -10 -10 -10 0 20 40 0 20 40 0 20 40 0 20 40 0.2 0.2 0.2 0.2 0 0 0 0 -0.2 -0.2 -0.2 -0.2 -0.4 -0.4 -0.4 -0.4 0 20 40 0 20 40 0 20 40 0 20 40 150 150 150 150 100 100 100 100 50 50 50 50 0 0 0 0 -50 -50 -50 -50 0 20 40 0 20 40 0 20 40 0 20 40 2 2 2 2 0 0 0 0 -2 -2 -2 -2 0 20 40 0 20 40 0 20 40 0 20 40 1.5 1.5 1.5 1.5 1 1 1 1 0.5 0.5 0.5 0.5 0 0 0 0 -0.5 -0.5 -0.5 -0.5 0 20 40 0 20 40 0 20 40 0 20 40

Low Capital Controls 0 -0.1 -0.2 0 5 10 15 20 25 30 35 40 45 Y High Capital Controls 0 -0.1 -0.2 0 5 10 15 20 25 30 35 40 45 0.05 0 -0.05 -0.1 0 5 10 15 20 25 30 35 40 45 P 0.05 0 -0.05 -0.1 0 5 10 15 20 25 30 35 40 45 0 -0.1 -0.2 0 5 10 15 20 25 30 35 40 45 PC 0 -0.1 -0.2 0 5 10 15 20 25 30 35 40 45 30 20 10 0 0 5 10 15 20 25 30 35 40 45 RFF 30 20 10 0 0 5 10 15 20 25 30 35 40 45 0 -0.2 -0.4 0 5 10 15 20 25 30 35 40 45 XRBN 0 -0.2 -0.4 0 5 10 15 20 25 30 35 40 45 15 10 5 0 -5 0 5 10 15 20 25 30 35 40 45 *R 15 10 5 0 -5 0 5 10 15 20 25 30 35 40 45 0.8 0.6 0.4 0.2 0 0 5 10 15 20 25 30 35 40 45 S 0.8 0.6 0.4 0.2 0 0 5 10 15 20 25 30 35 40 45 0.1 0 -0.1 0 5 10 15 20 25 30 35 40 45 *Y Appendix 2: Comparison for the sub-period 1975-90 0.1 0 -0.1 0 5 10 15 20 25 30 35 40 45

Low Capital Controls 0 -0.1 -0.2 0 5 10 15 20 25 30 35 40 45 Y High Capital Controls 0 -0.1 -0.2 0 5 10 15 20 25 30 35 40 45 0.05 0 -0.05 -0.1 0 5 10 15 20 25 30 35 40 45 P 0.05 0 -0.05 -0.1 0 5 10 15 20 25 30 35 40 45 0.05 0 -0.05 -0.1 -0.15 0 5 10 15 20 25 30 35 40 45 PC 0.05 0 -0.05 -0.1 -0.15 0 5 10 15 20 25 30 35 40 45 20 10 0 0 5 10 15 20 25 30 35 40 45 RFF 20 10 0 0 5 10 15 20 25 30 35 40 45 0 -0.2 -0.4 0 5 10 15 20 25 30 35 40 45 XRBN 0 -0.2 -0.4 0 5 10 15 20 25 30 35 40 45 10 5 0 -5 0 5 10 15 20 25 30 35 40 45 *R 10 5 0 -5 0 5 10 15 20 25 30 35 40 45 0.4 0.2 0 0 5 10 15 20 25 30 35 40 45 S 0.4 0.2 0 0 5 10 15 20 25 30 35 40 45 0 -0.1 -0.2 0 5 10 15 20 25 30 35 40 45 *Y Comparison for the sub-period 1976-91 0 -0.1 -0.2 0 5 10 15 20 25 30 35 40 45

Low Capital Controls 0 -0.1 -0.2 0 5 10 15 20 25 30 35 40 45 Y High Capital Controls 0 -0.1 -0.2 0 5 10 15 20 25 30 35 40 45 0.05 0 -0.05 0 5 10 15 20 25 30 35 40 45 P 0.05 0 -0.05 0 5 10 15 20 25 30 35 40 45 0.1 0.05 0 -0.05 0 5 10 15 20 25 30 35 40 45 PC 0.1 0.05 0 -0.05 0 5 10 15 20 25 30 35 40 45 20 10 0 0 5 10 15 20 25 30 35 40 45 RFF 20 10 0 0 5 10 15 20 25 30 35 40 45 0 -0.2 -0.4 0 5 10 15 20 25 30 35 40 45 XRBN 0 -0.2 -0.4 0 5 10 15 20 25 30 35 40 45 10 5 0 -5 0 5 10 15 20 25 30 35 40 45 *R 10 5 0 -5 0 5 10 15 20 25 30 35 40 45 0.4 0.2 0 0 5 10 15 20 25 30 35 40 45 S 0.4 0.2 0 0 5 10 15 20 25 30 35 40 45 0 -0.05 -0.1 -0.15 0 5 10 15 20 25 30 35 40 45 *Y Comparison for the sub-period 1977-92 0 -0.05 -0.1 -0.15 0 5 10 15 20 25 30 35 40 45

Low Capital Controls 0 -0.1 -0.2 0 5 10 15 20 25 30 35 40 45 Y High Capital Controls 0 -0.1 -0.2 0 5 10 15 20 25 30 35 40 45 0.05 0 -0.05 0 5 10 15 20 25 30 35 40 45 P 0.05 0 -0.05 0 5 10 15 20 25 30 35 40 45 0.05 0 -0.05 0 5 10 15 20 25 30 35 40 45 PC 0.05 0 -0.05 0 5 10 15 20 25 30 35 40 45 20 10 0 0 5 10 15 20 25 30 35 40 45 RFF 20 10 0 0 5 10 15 20 25 30 35 40 45 0 -0.2 -0.4 0 5 10 15 20 25 30 35 40 45 XRBN 0 -0.2 -0.4 0 5 10 15 20 25 30 35 40 45 15 10 5 0 -5 0 5 10 15 20 25 30 35 40 45 *R 15 10 5 0 -5 0 5 10 15 20 25 30 35 40 45 0.4 0.2 0 0 5 10 15 20 25 30 35 40 45 S 0.4 0.2 0 0 5 10 15 20 25 30 35 40 45 0.05 0 -0.05 -0.1 -0.15 0 5 10 15 20 25 30 35 40 45 *Y Comparison for the sub-period 1978-93 0.05 0 -0.05 -0.1 -0.15 0 5 10 15 20 25 30 35 40 45

Low Capital Controls 0 -0.1 -0.2 0 5 10 15 20 25 30 35 40 45 Y High Capital Controls 0 -0.1 -0.2 0 5 10 15 20 25 30 35 40 45 0.05 0 -0.05 0 5 10 15 20 25 30 35 40 45 P 0.05 0 -0.05 0 5 10 15 20 25 30 35 40 45 0.05 0 -0.05 0 5 10 15 20 25 30 35 40 45 PC 0.05 0 -0.05 0 5 10 15 20 25 30 35 40 45 20 10 0 0 5 10 15 20 25 30 35 40 45 RFF 20 10 0 0 5 10 15 20 25 30 35 40 45 0 -0.2 -0.4 0 5 10 15 20 25 30 35 40 45 XRBN 0 -0.2 -0.4 0 5 10 15 20 25 30 35 40 45 15 10 5 0 -5 0 5 10 15 20 25 30 35 40 45 *R 15 10 5 0 -5 0 5 10 15 20 25 30 35 40 45 0.4 0.2 0 0 5 10 15 20 25 30 35 40 45 S 0.4 0.2 0 0 5 10 15 20 25 30 35 40 45 0 -0.1 -0.2 0 5 10 15 20 25 30 35 40 45 *Y Comparison for the sub-period 1979-94 0 -0.1 -0.2 0 5 10 15 20 25 30 35 40 45

Low Capital Controls 0.1 0 -0.1 -0.2 -0.3 0 5 10 15 20 25 30 35 40 45 Y High Capital Controls 0.1 0 -0.1 -0.2 -0.3 0 5 10 15 20 25 30 35 40 45 0.05 0 -0.05 0 5 10 15 20 25 30 35 40 45 P 0.05 0 -0.05 0 5 10 15 20 25 30 35 40 45 0.1 0.05 0 -0.05 0 5 10 15 20 25 30 35 40 45 PC 0.1 0.05 0 -0.05 0 5 10 15 20 25 30 35 40 45 20 10 0 -10 0 5 10 15 20 25 30 35 40 45 RFF 20 10 0 -10 0 5 10 15 20 25 30 35 40 45 0 -0.1 -0.2 -0.3 0 5 10 15 20 25 30 35 40 45 XRBN 0 -0.1 -0.2 -0.3 0 5 10 15 20 25 30 35 40 45 20 10 0 -10 0 5 10 15 20 25 30 35 40 45 *R 20 10 0 -10 0 5 10 15 20 25 30 35 40 45 0.4 0.2 0 0 5 10 15 20 25 30 35 40 45 S 0.4 0.2 0 0 5 10 15 20 25 30 35 40 45 0.1 0 -0.1 -0.2 0 5 10 15 20 25 30 35 40 45 *Y Comparison for the sub-period 1980-95 0.1 0 -0.1 -0.2 0 5 10 15 20 25 30 35 40 45

Low Capital Controls 0 -0.2 -0.4 0 5 10 15 20 25 30 35 40 45 Y High Capital Controls 0 -0.2 -0.4 0 5 10 15 20 25 30 35 40 45 0.1 0 0 5 10 15 20 25 30 35 40 45 P 0.1 0 0 5 10 15 20 25 30 35 40 45 0.15 0.1 0.05 0 0 5 10 15 20 25 30 35 40 45 PC 0.15 0.1 0.05 0 0 5 10 15 20 25 30 35 40 45 30 20 10 0 -10 0 5 10 15 20 25 30 35 40 45 RFF 30 20 10 0 -10 0 5 10 15 20 25 30 35 40 45 0.1 0 -0.1 -0.2 0 5 10 15 20 25 30 35 40 45 XRBN 0.1 0 -0.1 -0.2 0 5 10 15 20 25 30 35 40 45 30 20 10 0 -10 0 5 10 15 20 25 30 35 40 45 *R 30 20 10 0 -10 0 5 10 15 20 25 30 35 40 45 0.5 0 -0.5 0 5 10 15 20 25 30 35 40 45 S 0.5 0 -0.5 0 5 10 15 20 25 30 35 40 45 0.2 0 -0.2 0 5 10 15 20 25 30 35 40 45 *Y Comparison for the sub-period 1981-96 0.2 0 -0.2 0 5 10 15 20 25 30 35 40 45

Low Capital Controls 0 -0.2 -0.4 0 5 10 15 20 25 30 35 40 45 Y High Capital Controls 0 -0.2 -0.4 0 5 10 15 20 25 30 35 40 45 0.1 0 0 5 10 15 20 25 30 35 40 45 P 0.1 0 0 5 10 15 20 25 30 35 40 45 0.2 0 0 5 10 15 20 25 30 35 40 45 PC 0.2 0 0 5 10 15 20 25 30 35 40 45 30 20 10 0 0 5 10 15 20 25 30 35 40 45 RFF 30 20 10 0 0 5 10 15 20 25 30 35 40 45 0.1 0 -0.1 -0.2 0 5 10 15 20 25 30 35 40 45 XRBN 0.1 0 -0.1 -0.2 0 5 10 15 20 25 30 35 40 45 40 20 0 0 5 10 15 20 25 30 35 40 45 *R 40 20 0 0 5 10 15 20 25 30 35 40 45 1 0.5 0 -0.5 0 5 10 15 20 25 30 35 40 45 S 1 0.5 0 -0.5 0 5 10 15 20 25 30 35 40 45 0.2 0 -0.2 0 5 10 15 20 25 30 35 40 45 *Y Comparison for the sub-period 1982-97 0.2 0 -0.2 0 5 10 15 20 25 30 35 40 45

Low Capital Controls 0.2 0 -0.2 -0.4 -0.6 0 5 10 15 20 25 30 35 40 45 Y High Capital Controls 0.2 0 -0.2 -0.4 -0.6 0 5 10 15 20 25 30 35 40 45 0.2 0 0 5 10 15 20 25 30 35 40 45 P 0.2 0 0 5 10 15 20 25 30 35 40 45 0.5 0 0 5 10 15 20 25 30 35 40 45 PC 0.5 0 0 5 10 15 20 25 30 35 40 45 30 20 10 0 -10 0 5 10 15 20 25 30 35 40 45 RFF 30 20 10 0 -10 0 5 10 15 20 25 30 35 40 45 0.2 0 -0.2 0 5 10 15 20 25 30 35 40 45 XRBN 0.2 0 -0.2 0 5 10 15 20 25 30 35 40 45 80 60 40 20 0 0 5 10 15 20 25 30 35 40 45 *R 80 60 40 20 0 0 5 10 15 20 25 30 35 40 45 2 1 0 0 5 10 15 20 25 30 35 40 45 S 2 1 0 0 5 10 15 20 25 30 35 40 45 0.4 0.2 0 -0.2 0 5 10 15 20 25 30 35 40 45 *Y Comparison for the sub-period 1983-98 0.4 0.2 0 -0.2 0 5 10 15 20 25 30 35 40 45

Low Capital Controls 0 -0.2 -0.4 0 5 10 15 20 25 30 35 40 45 Y High Capital Controls 0 -0.2 -0.4 0 5 10 15 20 25 30 35 40 45 0.05 0 -0.05 -0.1 0 5 10 15 20 25 30 35 40 45 P 0.05 0 -0.05 -0.1 0 5 10 15 20 25 30 35 40 45 0.1 0 -0.1 0 5 10 15 20 25 30 35 40 45 PC 0.1 0 -0.1 0 5 10 15 20 25 30 35 40 45 20 10 0 -10 0 5 10 15 20 25 30 35 40 45 RFF 20 10 0 -10 0 5 10 15 20 25 30 35 40 45 0 -0.1 -0.2 -0.3 0 5 10 15 20 25 30 35 40 45 XRBN 0 -0.1 -0.2 -0.3 0 5 10 15 20 25 30 35 40 45 20 10 0 -10 0 5 10 15 20 25 30 35 40 45 *R 20 10 0 -10 0 5 10 15 20 25 30 35 40 45 0.5 0 -0.5 0 5 10 15 20 25 30 35 40 45 S 0.5 0 -0.5 0 5 10 15 20 25 30 35 40 45 0.1 0 -0.1 -0.2 -0.3 0 5 10 15 20 25 30 35 40 45 *Y Comparison for the sub-period 1980-96 0.1 0 -0.1 -0.2 -0.3 0 5 10 15 20 25 30 35 40 45

Low Capital Controls 0 -0.2 -0.4 0 5 10 15 20 25 30 35 40 45 Y High Capital Controls 0 -0.2 -0.4 0 5 10 15 20 25 30 35 40 45 0.05 0 -0.05 0 5 10 15 20 25 30 35 40 45 P 0.05 0 -0.05 0 5 10 15 20 25 30 35 40 45 0.1 0.05 0 -0.05 0 5 10 15 20 25 30 35 40 45 PC 0.1 0.05 0 -0.05 0 5 10 15 20 25 30 35 40 45 20 10 0 -10 0 5 10 15 20 25 30 35 40 45 RFF 20 10 0 -10 0 5 10 15 20 25 30 35 40 45 0 -0.1 -0.2 -0.3 0 5 10 15 20 25 30 35 40 45 XRBN 0 -0.1 -0.2 -0.3 0 5 10 15 20 25 30 35 40 45 20 10 0 -10 0 5 10 15 20 25 30 35 40 45 *R 20 10 0 -10 0 5 10 15 20 25 30 35 40 45 0.4 0.2 0 -0.2 -0.4 -0.6 -0.8 0 5 10 15 20 25 30 35 40 45 S 0.4 0.2 0 -0.2 -0.4 -0.6 -0.8 0 5 10 15 20 25 30 35 40 45 0.1 0 -0.1 -0.2 -0.3 0 5 10 15 20 25 30 35 40 45 *Y Comparison for the sub-period 1980-97 0.1 0 -0.1 -0.2 -0.3 0 5 10 15 20 25 30 35 40 45

Low Capital Controls 0 -0.2 -0.4 0 5 10 15 20 25 30 35 40 45 Y High Capital Controls 0 -0.2 -0.4 0 5 10 15 20 25 30 35 40 45 0.08 0.06 0.04 0.02 0 -0.02 -0.04 0 5 10 15 20 25 30 35 40 45 P 0.08 0.06 0.04 0.02 0 -0.02 -0.04 0 5 10 15 20 25 30 35 40 45 0.1 0.05 0 0 5 10 15 20 25 30 35 40 45 PC 0.1 0.05 0 0 5 10 15 20 25 30 35 40 45 20 10 0 -10 0 5 10 15 20 25 30 35 40 45 RFF 20 10 0 -10 0 5 10 15 20 25 30 35 40 45 0 -0.1 -0.2 -0.3 0 5 10 15 20 25 30 35 40 45 XRBN 0 -0.1 -0.2 -0.3 0 5 10 15 20 25 30 35 40 45 20 10 0 -10 0 5 10 15 20 25 30 35 40 45 *R 20 10 0 -10 0 5 10 15 20 25 30 35 40 45 0.4 0.2 0 -0.2 -0.4 0 5 10 15 20 25 30 35 40 45 S 0.4 0.2 0 -0.2 -0.4 0 5 10 15 20 25 30 35 40 45 0.1 0 -0.1 -0.2 0 5 10 15 20 25 30 35 40 45 *Y Comparison for the sub-period 1980-98 0.1 0 -0.1 -0.2 0 5 10 15 20 25 30 35 40 45