Hello,

I would like a presentation to be made for 2 sections from the attached article!

DOES THE CHOICE OF EXCHANGE RATE REGIME AFFECT THE ECONOMIC GROWTH OF DEVELOPING COUNTRIES? Glauco De Vita Khine Sandar Kyaw* Oxford Brookes University, UK ABSTRACT This paper revisits, empirically, the question of whether the choice of exchange rate regime (floating currency, fixed rate or intermediate arrangement) has a direct effect on the longterm growth of developing countries. We use fixed effects panel estimation on a relatively large sample of 70 developing countries for the period 1981 to 2004. In addition to controlling for the level of income of these economies, we draw from alternative exchange rate regime classification schemes and account for the presence of an explicit monetary anchor to characterize policy. Our results indicate the absence of any robust relation between the choice of exchange rate regime and economic growth. The policy implication of these findings is clear. The choice of exchange rate policy has no direct impact on the long-term growth of developing countries. JEL Classifications: O10, F31 Keywords: Exchange Rate Regimes, Economic Growth, Developing Countries Corresponding Author's Email: gde-vita@brookes.ac.uk INTRODUCTION Over the past decade, the question of the choice of exchange rate regime and the macroeconomic consequences of this choice, has received considerable attention. In addition to effects on trade flows (Rose, 2000; Rose and van Wincoop, 2001; Frankel and Rose, 2002; Glick and Rose, 2002; Rose and Stanley, 2005; Klein and Shambaugh, 2006; Adam and Cobham, 2007; etc.), recent literature identifies empirical regularities between exchange rate regimes and national price levels (Ghosh et al., 2002; Broda, 2006), the transmission of terms of trade shocks (Broda, 2004; Edwards and Levy-Yeyati, 2005), and foreign direct investment flows (Schiavo, 2007; Abbott and De Vita, 2008). In this paper, interest centers on the relationship between exchange rate regimes and economic growth, with a focus on the impact of the choice of regime on the long-term growth of developing countries. At the theoretical level, the traditional view of the neutrality of money suggests that the exchange rate regime should be unimportant for long-term growth performance. Nevertheless, the literature highlights several mechanisms through which a link can be established. The first of such mechanisms can be traced back to the work of Milton Friedman (1953) who, in his essay on the case for flexible exchange rates, argues that flexible regimes are better suited to insulate the economy against economic shocks. Specifically, a flexible regime may foster growth by enabling an economy characterized by 136 nominal rigidities to better absorb and adjust to real domestic and foreign shocks. This business cycle mitigation mechanism (see Barlevy, 2004) implies that fixed regimes induce greater output volatility which, in turn, is expected to affect negatively an economys long-term growth performance (see Ramey and Ramey, 1995; Kneller and Young, 2001). However, it has also been contended that flexible rates are more prone to exchange rate shocks, exacerbating business cycles and dampening growth compared to a fixed regime, especially in the context of developing economies with a weak financial system (Bailliu et al., 2003). Although credibility arguments in favor of fixed exchange rates point to the reduction of a countrys vulnerability to speculative currency attacks and to greater stability as factors conducive to stronger growth performance (see Calvo and Vegh, 1994; and Mundell, 1995), the real interest rate costs associated with the need to defend a peg in the event of a negative external shock, and the consequent uncertainty as to the sustainability of such a regime, are generally considered to be deleterious to growth. The theoretical ambiguity as to the sign of the relationship between exchange rate regimes and economic growth is compounded by the multiplicity of indirect channels through which exchange rate regimes could affect growth. These include the rate of physical capital accumulation (e.g. Aizenman, 1994), the degree of openness to capital flows (Bailliu, 2000) and the level of financial development (Levine, 1997; Aizenman and Hausmann, 2000; Flandreau and Bordo, 2003; Aghion et al., 2006). Whilst this literature posits mostly positive indirect effects of fixed exchange rates on growth over the long run, conflicting predictions as to the growth effects of other key variables that are likely to be influenced by the choice of regime type accentuate the theoretical controversy. For example, increased trade has not always been found to have a positive influence on economic growth (see Slaughter, 2001). Similarly, while within traditional Solow-type neoclassical growth models the assumption of diminishing returns to physical capital means that foreign direct investment (FDI) could at best affect the level of income, with no impact upon the long run growth rate, the advent of endogenous growth theory (Barro and Sala-iMartin, 1995) has spurred a new wave of growth-theoretic models in which FDI induces permanent increases in the rate of output growth in so far as it generates increasing returns in production via externalities and productivity spillovers (see De Mello, 1997). Against this theoretical backdrop, it should come as no surprise that the onus of resolving the question of whether or not there exists a link between exchange rate regimes and growth has fallen upon empirical research. However, to date, empirical work has produced mixed results from which to discern a conventional wisdom is equally difficult. Our contribution adds to this literature in several respects. First, using the best available data for a relatively large panel, we control for most of the variables identified in the empirical growth literature and, drawing on recent advances in the classification of exchange rate regimes, we employ two de facto regime classification schemes to construct the exchange rate regime dummies in addition to the official scheme reported by the IMF. Second, we augment the level of regime aggregation (traditionally based on the three-way classification of fixed, intermediate and flexible rates) by additionally controlling for the monetary policy characterizing the choice of exchange rate regime through further disaggregation between intermediate and flexible rates with and without a nominal policy anchor. Finally, although previous findings suggest that the growth performance of alternative regimes depends on the maturity of member countries (Levy-Yeyati and Sturzenegger, 2003), of their institutions (Bailliu et al., 2003; Husain et al., 2005), and on 137 their degree of involvement in international financial markets (Jadresic et al., 2001), no attempt has been made in earlier literature to estimate the relationship between exchange rate regimes and growth within a framework disaggregated according to the level of income of developing countries. In this paper, we discriminate empirically across different levels of economic development by stratifying our sample of 70 developing countries according to three distinct income level categories, namely low-, lower middle- and upper middle-income countries. The remainder of this paper is organized as follows. The next section offers a concise yet comprehensive literature review of empirical work focusing on the impact of exchange rate regimes on growth. Then we provide a discussion of the exchange rate regime classifications from which we draw to construct our regime dummies. In the following section, the model and the data used are described. This is followed by the presentation and discussion of the estimation results. Finally, the last section draws conclusions. A REVIEW OF PREVIOUS EMPIRICAL WORK Ambiguity at the theoretical level over the relationship between exchange rate regimes and economic growth is reflected in the empirical work. Ghosh et al. (1997) run growth regressions for 136 countries over the period 1960-1989. They controlled for pegged, intermediate and flexible exchange rate regimes using data drawn from the de jure classification published by the International Monetary Fund (IMF) and found no significant differences in growth rates across the exchange rate regimes examined. This result is broadly confirmed in Ghosh et al. (2002) where the sample period extends from 1970 to 1999 and the endogeneity of the regime is controlled for through a treatment effects model. Since failure to capture the extent to which actual exchange rate policy conforms to countries declared commitment to the announced regime can lead to measurement error (Calvo and Reinhart, 2002), Levy-Yeyati and Sturzenegger (2003) (henceforth LYS) compile a de facto exchange rate regime classification to examine the link between exchange rate regimes and growth for a large sample of countries over the period 19742000. They find that while for industrial countries the regime type does not have any significant impact on growth, for developing countries, less flexible (fixed and intermediate) regimes are significantly associated with slower growth. These results prove robust to several alternative specifications (pooled annual data, pooled five-year data and cross-sectional regressions) and pass a number of sensitivity checks. Working with a panel of 60 industrialized and developing countries over the 197398 period, Bailliu et al. (2003) examine the impact of exchange rate regime on growth by means of dynamic GMM estimation. In addition to the traditional tripartite scheme of regime aggregation (flexible, fixes and intermediates) they use an expanded one which further distinguishes between intermediate and floating regimes based on the presence of a nominal policy anchor. In each case, the exchange rate regime is classified according to both the de jure classification, and a de facto classification of their own construction that corrects for observed volatility in the actual behavior of the exchange rate. They find that a pegged regime is positively linked to growth, an intermediate regime without an anchor is negatively associated with growth, and all other regime types have no discernible impact on growth. 138 Husain et al. (2005) estimate (with and without fixed country effects) exchange rate regime durability and performance across a large panel of advanced, emerging and developing economies over the period 1970-1999. They find that in developing countries more flexible regimes are associated with high inflation but do not lead to gains in economic growth while fixed or near fixed regimes deliver lower inflation without sacrificing growth. Miles (2006) replicates the LYS growth regressions with a panel of annual data (1976-2000) across a developing countries subset of the LYS original sample. The innovation here consists in the inclusion of a measure of the black market premium on foreign exchange, a variable meant to capture macroeconomic imbalances. His results indicate that once such a measure of domestic distortions is added to the model, exchange rate regimes exert no independent impact on the economic growth of developing countries. More recently, Bleaney and Francisco (2007) use the official (IMF) and four alternative de facto exchange rate regime classifications to examine the relationship with inflation and growth in 91 developing countries over the period 1984-2001. With the exception of the results obtained from the Reinhart and Rogoff (2004) regime classification, which produce quite unfavorable outcomes for flexible regimes (higher inflation and lower growth), their estimates across the other classification schemes suggest that floats have very similar growth rates to soft (easily adjustable) pegs while hard pegs (currency unions and currency boards) have slower growth than other regimes (though the latter result may - by their own admission - be due to fixed country effects). However, it should be noted that they only include a few growth determinants in their regressions, and do not attempt to control for endogeneity. Evidently, the conflicting evidence warrants further empirical research. CLASSIFYING EXCHANGE RATE REGIMES: ISSUES AND SCHEMES As originally noted by Ghosh et al. (1996:2): \"Beyond the traditional fixed-floating dichotomy lies a spectrum of exchange rate regimes. The de facto behavior of an exchange rate, moreover, may diverge from its de jure classification\". The above quote points to two critical issues that give rise to considerable measurement difficulties. The first issue is that in a continuum which spans across fixed rates, hard and soft pegs, crawling or monitoring bands, and floats with heavy, light or no intervention, it is not clear where one should draw the line (Backus, 2005). To the applied economist the issue of a fuzzy continuum poses two additional dilemmas: (i) Which degree of compression should one adopt to aggregate the many fine codes of the original regime classification schemes into a smaller, empirically tractable, yet informative number of regime policy options to be examined?; and (ii) Where do the exact boundaries of intermediate regimes (floating rates but within a predetermined range) lie? These dilemmas assume even greater significance when one considers that, especially in developing and emerging markets, countries regimes are often of the intermediate type (Williamson, 2000). Although it is often contended that intermediate regimes are not sustainable over the long run, as they lack credibility and make economies more susceptible to speculative currency attacks and economic crises (Eichengreen, 1994 and 2000; Fischer, 2001; Glick, 2001; Summers, 2000), Williamson (2000) suggests that intermediate regimes will continue to be seen as a viable option to try to reap the benefits 139 of fixed and flexible rates without having to incur some of their costs. The second issue is that even between de facto regime classification schemes that purport to measure what the regime is (rather than what is declared to be), there is a surprisingly low correlation (Husain et al., 2005; Bleaney and Francisco, 2007). This divergence stems primarily from differences in the way in which admittedly arbitrary correction rules attempt to adjust for the bias between countries publicly stated commitment to a given regime and the observed volatility of the exchange rate. For example, while Levy-Yeyati and Sturzenegger (2003 and 2005) use cluster analysis techniques to group countries regimes on the basis of the volatility of the exchange rate, of exchange rate changes and of reserves, Bailliu et al. (2003) develop a two-step mechanical rule whereby countries regimes are grouped according to a flexibility index for each country based on its degree of exchange rate volatility relative to the group average for each year. Accordingly, for the sake of comprehensiveness and for comparative purposes, our analysis draws from three different regime classification schemes that vary considerably in the way in which they deal with the above mentioned issues. The first regime classification that we employ is that published by the IMF in its Annual Report on Exchange Rate Agreements and Exchange Restrictions (various issues). Since 1999 the IMF moved from a purely de jure classification based on what countries report they do, to a hybrid one which combines information based on the IMF officials informed judgment about the actual behavior of the exchange rate. Despite this, concerns about the regime classification published by the IMF have prompted researchers to develop alternative schemes that attempt to characterize more accurately countries de facto regimes. The second regime classification from which we draw is the one developed by Levy-Yeyati and Sturzenegger (2003). As noted earlier, LYS use a de facto classification that groups countries annual observations according to the behavior of three dimensions of exchange rate policy: the volatility of the exchange rate relative to the relevant anchor currency (or basket of currencies); the volatility of exchange rate changes (measured by the standard deviation of monthly percentage changes); and the volatility of reserves (to quantify the degree of active intervention). Using cluster analysis, LYS then group observations into flexible, intermediates, fixed and inconclusive (unclassifiable) regimes according to the degree of significant variability recorded in each dimension. In spite of its complexity, this classification too has not escaped criticism on the grounds of a non-trivial proportion of unclassifiable observations, a large number of recorded regime switches as well as its inherent inability to distinguish between genuine floats and devaluations due to extreme inflation, where exchange rates are flexible by necessity. Finally, we use the Reinhart and Rogoff (2004) (henceforth RR) regime classification scheme, which by virtue of its characteristics seems particularly suited to deal with the challenges of the analysis ahead. The RR classification seeks to address potential misclassification by isolating flexible rates characterized by episodes of very high inflation (which are often followed by devaluations and tend to be associated with very low growth) through a separate \"freely falling\" category (resulting in a much smaller number of observations recorded as floats). Another distinguishing feature of the RR classification lies in the identification of the market-determined exchange rate. This market-determined exchange rate is defined as either the official rate (where no black market premium exists), the parallel rate (if market-determined), or the black market exchange rate, where/when it 140 exists. This definitional aspect is particularly important in the context of the present analysis since the black market premium on foreign exchange was already used by Miles (2006) as a measure of macroeconomic imbalances and, as noted earlier, he found that its inclusion in the growth equations left no significant independent effect of exchange rate regimes. Finally, the RR classification employs a more stable rolling five-year horizon that better captures the longer-term regimes (rather than just picking up transient regime shifts or short-term spells within a regime) and, in so doing, is likely to be more suitable for the analysis of long term economic growth. The use of the above exchange rate regime classification schemes to define fixed, intermediate and floating rates, is necessary but not sufficient to properly represent policy. While a fixed exchange rate regime represents a coherent monetary order (see Laidler, 1999), without further information, knowledge that the regime is flexible or intermediate is not enough to characterize policy. This is because, as Bailliu et al. (2003) point out, flexible and intermediate regimes simply define the exchange rate arrangement and are not informative about the framework in which monetary policy is conducted. For this reason, we later supplement the information on the regime type by further disaggregating the sample according to the monetary policy pursued by countries adopting flexible and intermediate regimes. MODEL AND DATA The basic econometric model used in our regression is as follows: yi ,t i ,t i ,t t i i ,t where the dependent variable, at time t, i,t (1) y i ,t , is the growth rate of real per capita GDP in country i is a vector of explanatory variables, i,t is a vector of exchange rate t are time specific effects, i are country specific effects, i,t are error terms and the s and s are parameters to be estimated. The estimators are also regime dummies, designed to incorporate individual effects and time effects (Greene, 2008) to handle the systematic tendency of i,t to be higher for some individual countries than for others and possibly higher for some time periods than for others. Although our selection of countries was constrained by data availability, particularly with respect to exchange rate regimes data, we were able to use annual data for 70 developing countries from 1981 to 2004. The full list of countries in our sample is reported in Table 1. Further details on the definition of the variables and data sources are provided in Table 2. For each classification scheme, we create dummies for flexible (FLEX), intermediate (INTER) and fixed regimes (the latter used as the default benchmark). We then augment the standard tripartite scheme of regime aggregation by further distinguishing between flexible and intermediate regimes with and without a (nominal) policy anchor. The International Financial Statistics (IFS) provides information on the monetary policy framework for many countries, including data on the presence of a publicly announced nominal anchor in the form of an exchange rate anchor, a monetary target, or an inflation target. 141 TABLE 1. COUNTRIES GROUPED BY INCOME LEVEL Low Income Lower Middle Income Upper Middle Income Cote d'Ivoire Gambia Ghana Haiti Albania Algeria Armenia Bolivia Argentina Botswana Brazil Bulgaria Kenya Cameroon Chile Liberia China Costa Rica Madagascar Colombia Gabon Malawi Congo, Rep. Jamaica Mozambique Dominican Republic Latvia Nigeria Ecuador Kazakhstan Niger Egypt, Arab Rep. Lithuania Pakistan El Salvador Malaysia Senegal India Mexico Tanzania Guatemala Panama Togo Honduras Poland Uganda Indonesia Romania Zambia Iran, Islamic Rep. Russian Federation Zimbabwe Jordan Slovak Republic Moldova South Africa Mongolia Turkey Morocco Uruguay Nicaragua Venezuela, RB Paraguay Peru Philippines Sri Lanka Syrian Arab Republic Thailand Tunisia Ukraine Note: For the income level categorization we used the 2002 per capita GNI income categories calculated by means of the most recent World Bank Atlas method. 142 TABLE 2. DEFINITION OF VARIABLES AND SOURCES Variable Description Source RPCG Rate of growth of real per capita GDP World Bank, World Development INVT Investment as percentage of GDP World Bank, WDI. IGDP Initial per capita GDP World Bank, WDI. GC(-1) Lagged government consumption as a percentage of GDP World Bank, WDI. POPG Population growth in annual percent World Bank, WDI. EDN Proportion of the population aged 25 and over that has a tertiary Barro and Lee (2001) Indicators (WDI). education qualification TRD Total trade (sum of exports and imports) as a percentage of GDP World Bank, WDI. INFL(-1) Lagged annual percentage change in the GDP deflator World Bank, WDI. PCR Private sector credit as a percentage of GDP World Bank, WDI. BCS Dummy variable for banking crisis Bank of England. PST Index of government stability that measures the ability of a Political Risk Services. government to carry out its declared program(s) and its ability to stay in office DHIC Dummy variable for highly indebted countries World Bank, WDI. ADAPE Dummy variable for South and East Asia and the Pacific World Bank, WDI. ADLAC Dummy variable for Latin America and the Caribbean World Bank, WDI. ADMA Dummy variable for Middle East, and North and Sub-Saharan World Bank, WDI Africa Exchange Equals one for each of the classifications defined IMF Annual Report (1981-2005); rate Levy-Yeyati & Sturzenegger regime (2005); Reinhart & Rogoff dummies (2004). 143 In addition to the exchange rate regime dummies, many factors identified in the empirical growth literature (e.g., Levine and Renelt, 1992) are accounted for. Furthermore, to control for the possibility that countries with considerable macroeconomic problems may self-select into choosing fixed exchange rates, we add measures of poor macroeconomic performance as well as dummies for banking crises and for highly indebted countries. Regional dummies are also included in the model to control for a potential skew in the regional distribution of floating observations. As recently noted by Bleaney and Francisco (2007), de facto exchange rate regime classifications, particularly the RR classification scheme, may have a bias due to a considerably low proportion of floats in East and South Asia (the fastest-growing region) and a very high proportion in the slowest-growing one (sub-Saharan Africa). RESULTS The annual rate of GDP growth averaged 1.2 percent over our sample, with noticeable differences in various exchange rate regimes. As shown in Table 3, a preliminary pass at the data suggests that countries with intermediate exchange rates had higher average growth rates than those fixing or pegging their currency under all three classification schemes. Intermediate regimes also seem to outperform floaters under both the IMF and RR classifications (the average growth rates drop from 3.6 and 1.9 percent to 0.7 and 1.1 percent, respectively) though, according to the LYS classification scheme, floating exchange rates have an average growth rate 0.2 percent higher than intermediates. As shown in Table 4, this overall pattern emerges mainly because of the upper middleincome countries, where intermediate regimes have higher average growth rates. However, both the de facto regime classification schemes indicate that for low-income countries, growth was somewhat higher under a fixed regime. The fact that developing countries with intermediate regimes have higher average growth, does not in itself mean that an intermediate regime causes higher growth in these countries. Causality issues aside, it is first necessary to investigate whether a statistically significant association can be established. Table 5 reports the results of estimating our baseline equation for the various classification schemes. Although we report the estimated coefficients for all the variables included in our specification, our interest inevitably centres upon the interpretation of the regime dummies. The INTER and FLEX dummies compare average growth rates for intermediate and flexible regimes with those of the omitted (benchmark) category of fixed rates. Although under the IMF classification scheme neither of the differential regime dummies is statistically significant, according to both of the de facto classification schemes (the RR and LYS classifications), the intermediate regime dummy has a positive and significant coefficient (of 0.05 and 0.08, respectively). But do these results hold to our augmented specification that distinguishes between different monetary policy frameworks for intermediate and floating regimes? Table 6 reveals that once the presence of a nominal policy anchor is controlled for, the coefficients of all regime dummies (INTER and FLEX with and without an anchor), turn insignificant, a result consistent across all exchange rate classifications. 144 TABLE 3. AVERAGE RATES OF REAL PER CAPITA GDP GROWTH (19812004) Exchange Rate Regimes Intermediate Regime Float IMF classification Observations Fixed Exchange Rate Regime -0.16 (0.53) 515 3.57 (2.27) 394 0.75 (1.43) 354 RR classification Observations 1.74 (1.42) 326 1.91 (2.21) 583 1.09 (1.82) 186 LYS classification Observations 0.95 (1.27) 532 1.36 (0.58) 372 1.60 (2.19) 353 Note: Since average growth rates may be overly sensitive to extreme values stemming from extraordinary growth volatility (due, for example, to periods of wars), we also report, in parentheses, the medians. 145 TABLE 4. AVERAGE RATES OF REAL PER CAPITA GDP GROWTH DISAGGREGATED BY INCOME GROUPS IMF Classification Low income Observations Lower middle income Observations Upper middle income Observations Fixed Exchange Rate Regime 0.59 (-1.06) 189 -0.92 (1.05) 216 -0.63 (2.24) 110 Intermediate Regime 1.10 (1.77) 97 3.32 (2.38) 182 4.67 (2.21) 115 Float 1.65 (1.41) 116 0.68 (1.62) 156 -0.34 (1.09) 82 Reinhart and Rogoff Classification Low income Observations Lower middle income Observations Upper middle income Observations Fixed Exchange Rate Regime Intermediate Regime Float 2.35 (-0.22) 155 1.85 (2.83) 89 0.36 (1.58) 82 1.38 (1.24) 143 1.96 (2.28) 293 2.29 (3.35) 147 0.54 (0.90) 73 1.42 (3.14) 61 2.03 (1.73) 52 Levy-Yeyati and Sturzenegger Classification Low income Observations Lower middle income Observations Upper middle income Observations Fixed Exchange Rate Regime Intermediate Regime Float 1.31 (-0.14) 225 1.08 (1.95) 176 0.32 (2.61) 131 0.57 (0.54) 89 0.66 (1.15) 168 3.96 (-1.68) 115 0.89 (1.67) 91 1.07 (2.05) 160 2.68 (3.54) 102 Note: Figures in parentheses are the medians. 146 TABLE 5. BASELINE GROWTH REGRESSION Variable IMF Reinhart & Rogoff Levy-Yeyati & Sturzenegger INVT 0.02 0.02 0.07 (1.75) (0.69) (1.82) IGDP -0.78 -0.55 -0.81* (-1.92) (-1.87) (-2.45) GC(-1) -0.01 -0.00 -0.01** (-1.46) (-1.84) (-2.86) POPG 0.04 0.06* 0.04 (1.83) (1.97) (1.67) EDN 0.01* 0.01** 0.01* (2.17) (2.77) (2.34) TRD 0.01* 0.03 0.08* (1.96) (1.78) (2.26) INFL(-1) -0.91 -0.82 -1.20* (-1.18) (-0.82) (-2.37) PCR -0.11 -0.09 -0.86 (-0.63) (-1.68) (-0.98) BCS -1.30 -0.28 -1.00 (-1.82) (-1.43) (-1.73) PST -0.01 -0.01 -0.01 (-0.53) (-1.75) (-0.32) DHIC 0.00 0.00 0.00 (1.26) (1.18) (0.76) ADAPE -0.01* -0.05 -0.06 (1.98) (-1.41) (-1.92) ADLAC -0.02 -0.02 -0.01 (-1.24) (-1.28) (-1.10) ADMA -0.08 0.07 0.05 (-1.59) (1.32) (0.82) INTER 0.03 0.05* 0.08** (1.38) (2.25) (2.61) FLEX 0.03 -0.02 0.03 (1.46) (-0.47) (1.73) Observations 1263 1095 1257 R2 0.2538 0.3032 0.3729 Note: Figures in parentheses are t ratios obtained from robust standard errors adjusted for heteroscedasticity and serial correlation using the Newey-West technique (Newey and West, 1987). The adjusted R-squared statistic was computed by using the technique suggested in Chamberlain (1982). Year dummies are also included in the regression. ** indicates significance at 1 percent. * indicates significance at 5 percent. The results presented in Table 6 may still mask important differences in the growth performance of alternative regimes and monetary policies across different groups of developing countries. We therefore re-run growth regressions on a disaggregated sample, stratified according to the level of economic development of the countries under examination. As shown in Table 7 (that only reports the regime dummies of interest), with the sole exception of a significant intermediate regime without an anchor (in low- 147 income countries under the LYS classification, and in lower middle-income countries under the IMF classification), all other exchange rate arrangements have no statistically significant differential effect on growth vis--vis the alternative o a fixed exchange rate regime. TABLE 6. EXCHANGE RATE REGIMES AND GROWTH REGRESSIONS CONTROLLING FOR MONETARY POLICY ANCHOR Variable IMF Reinhart & Rogoff Levy-Yeyati & Sturzenegger INVT 0.04 (1.94) -0.80 (-1.87) -0.00 (-1.25) 0.04 (1.73) 0.01* (2.29) 0.01** (2.61) -0.01 (-0.44) 0.00 (0.87) -0.02* (-1.96) -0.02 (-1.18) 0.05 (1.10) -1.20 (-1.23) -0.85 (-0.89) -1.20 (-1.68) -0.03 (-1.32) -0.02 (-0.91) -0.02 (-0.79) 0.00 (0.57) 1263 0.4162 0.01 (0.75) -0.49* (-2.29) -0.00* (-1.96) 0.05** (2.96) 0.05 (1.63) 0.04* (2.17) -0.00 (-0.93) 0.00 (1.27) -0.62 (-1.59) -0.02 (-1.30) 0.06 (0.95) -1.02 (-1.14) -0.11 (-1.73) -0.85 (-1.85) -0.03 (-1.12) -0.02 (-0.96) -0.02 (-0.48) -0.05 (-0.23) 1095 0.4295 0.08 (0.47) -0.79* (-2.18) -0.01** (-2.94) 0.05 (1.59) 0.03* (2.40) 0.07* (1.96) -0.00 (-0.57) 0.01 (0.98) -0.81* (-2.23) -0.02 (-1.09) 0.06 (0.86) -1.01* (-2.10) -0.51 (-0.17) -0.72 (-1.65) -0.01 (-0.39) -0.03 (-1.08) -0.03 (-1.41) 0.00 (0.93) 1257 0.4852 IGDP GC(-1) POPG EDN TRD INFL(-1) PCR BCS PST DHIC ADAPE ADLAC ADMA INTER with anchor INTER without anchor FLEX with anchor FLEX without anchor Observations R2 Note: Figures in parentheses are t ratios obtained from robust standard errors adjusted for heteroscedasticity and serial correlation using the Newey-West technique (Newey and West, 1987). The adjusted R-squared statistic was computed by using the technique suggested in Chamberlain (1982). Year dummies are also included in the regression. ** indicates significance at 1 percent. * indicates significance at 5 percent. 148 The analysis would be incomplete without a formal check on potential regime endogeneity problems. With this aim in mind, we re-run the regressions using a system generalized methods-of-moments (SYS-GMM) estimation technique. SYS-GMM estimation not only exploits the time series variation in the data while accounting for unobserved country specific effects, it also allows to control for both a possible correlation between the regressors and the error term, and endogeneity bias. i The results of this exercise (available from the authors upon request) do not change the conclusions reached heretofore. ii TABLE 7. EXCHANGE RATE REGIMES AND GROWTH ESTIMATIONS BY INCOME GROUPS Low Income IMF Reinhart & Rogoff Levy-Yeyati & Sturzenegger 0.03 (-1.92) 0.02 (-1.91) 0.02 (-0.79) 0.00 (-0.83) 402 0.3785 -0.02 (-1.12) -0.03 (-0.96) -0.02 (-0.48) 0.10 (0.53) 371 0.3870 -0.01 (-0.39) -0.03* (-2.08) -0.03 (-1.41) -0.01 (-1.25) 405 0.4312 INTER with anchor 0.02 (1.14) 0.02 (0.48) 0.01 (0.63) INTER without anchor FLEX with anchor 0.03* (2.23) 0.04 (1.59) 0.06 (0.96) 554 0.4058 0.03 (0.62) 0.02 (0.68) 0.03 (0.67) 0.01 (0.02) 443 0.4102 0.02 (0.25) -0.02 (-1.85) -0.01 (-0.31) 0.00 (0.24) 504 0.4512 -0.01 (-0.21) 0.08 (1.29) 0.06 (0.84) -0.04 (-0.75) 307 0.3551 0.04 (0.48) 0.03 (1.24) 0.01 (0.98) 281 0.3864 0.07 (1.17) 0.08 (1.24) -0.03 (-0.02) 348 0.4258 INTER with anchor INTER without anchor FLEX with anchor FLEX without anchor Observations R2 Lower Middle Income FLEX without anchor Observations R2 Upper Middle Income INTER with anchor INTER without anchor FLEX with anchor FLEX without anchor Observations R2 Note: Figures in parentheses are t ratios obtained from robust standard errors adjusted for heteroscedasticity and serial correlation using the Newey-West technique (Newey and West, 1987). The adjusted R-squared statistics were computed by using the technique suggested in Chamberlain (1982). All other control variables and year dummies are also included in the regression. ** indicates significance at 1 percent. * indicates significance at 5 percent. 149 How do our result compare with those from previous studies? Although different country samples and estimation periods make straightforward comparisons difficult, our results are in stark contrast to those by Levy-Yeyati and Sturzenegger (2003) who found that, for developing countries, less flexible exchange rate regimes are associated with slower growth. However, they did not control for the monetary policy framework that accompanies the regime type. Furthermore, their two-step treatment effects model to correct for endogeneity problems may in itself introduce a non-linearity bias in the second stage regression that is likely to result in inconsistent estimates (Miles, 2006; Angrist and Krueger, 2001). Like Bailliu et al. (2003), we find that controlling for a monetary anchor in the characterization of the policy choice makes a significant difference in unveiling the independent effect of the regime type. Their results, however, suggest that more flexible exchange rate regimes with an anchor (jointly considered flexible and intermediates with anchor) are positively linked to growth, those without an anchor are negatively linked to growth, and fixed rates have a positive effect on growth. Nevertheless, their findings are based on a smaller panel (60 countries, of which only 41 class as developing countries, over a time period ending in 1998) and their model specification accounts for the influence of fewer regressors of interest. Our results are more in line with those by Ghosh et al. (1997), who found no systematic link between regime type and growth, Bleaney and Francisco (2007), who found that growth rates in developing countries are similar under soft pegs and floats (for hard pegs they find it difficult to distinguish between a regime effect and fixed country effects), and the conclusion reached by Miles (2006), who found no independent effect of exchange rate regimes on growth once hard-to-observe policy distortions and macroeconomic imbalances are accounted for. CONCLUSION The paper investigated empirically the question of the impact of exchange rate regime choice on economic growth using a relatively large panel of developing countries over the period 1981-2004. The analysis, which further disaggregates the sample of countries according to their level of economic development, draws on recent advances in exchange rate regime classification schemes and supplements de jure and de facto regime type data with information on the monetary policy framework. After controlling for the monetary framework that accompanies the choice of exchange rate regime, intermediate and flexible exchange rate arrangements are found not to be more or less pro-growth than the default regime of a fixed exchange rate. These results, which prove robust to different estimation techniques and sensitivity checks, hold across exchange rate classification schemes and apply irrespective of the level of economic development of the developing countries included in our sample. Notwithstanding the value of the findings uncovered by the present study, two caveats ought to be borne in mind. First, IFS data are, to date, the only available source to gauge information on the policy anchor but the reliability of such data is yet to be confirmed since countries can just as easily misreport this aspect of policy to the IMF as they could misreport exchange rate policy. Second, although this paper was exclusively concerned with testing, empirically, whether the choice of exchange rate regime affects growth, the resulting findings highlight the necessity to explain, theoretically, why the role of the accompanying monetary policy anchor appears 150 to be particularly important in influencing the relationship of interest. This evidence, therefore, provides a profitable avenue for future work aimed at the development of relevant theory. ENDNOTES * We are indebted to Andrew Abbott, for his early work in collecting data, and to an anonymous referee of this journal for helpful comments. 1 Consistency of the SYS-GMM estimates requires evidence of the validity of the chosen instruments and of significant first order correlation but no higher order serial correlation. The Hansens J-test (Hansen, 1982) for instrument validity as well as second order serial correlation tests did not reject the chosen econometric specification which appeared to fit the data well. 2 We also checked whether the results were driven by contamination across regimes due to regime switches. Specifically, if a regime is not sustained, then it is questionable whether the growth performance under that regime can be attributed to it. Accordingly, we re-estimated the model while dropping the first two years following a regime switch and found that this modification did not alter the results. REFERENCES Abbott, A. and G. De Vita, \"Evidence on the Impact of Exchange Rate Regimes on Foreign Direct Investment Flows\