انتخاب رژیم نرخ ارز در کشورهای در حال توسعه: تجزیه و تحلیل پانل های چندگانه

This paper analyses the choices of exchange rate regimes in developing countries since 1980. Static and dynamic random-effects multinomial panel models are estimated using simulation-based techniques. Explanatory variables include OCA fundamentals, stabilization considerations, currency crises factors, and political and institutional features. The results reveal strong state dependence in regime choices. We also find evidence for non-monotonic relationship between regime determinants and the degree of regime flexibility.

The choice of exchange rate regimes has long been a controversial topic among academic economists and policy makers. For most developing countries, it is commonly regarded as their single most important macroeconomic policy decision, which strongly influences the making and efficacy of other macroeconomic policies. The collapse of the Bretton Woods System in 1973 provided countries with a far wider range of choices than before, but many countries continued to apply some kind of exchange rate pegs. Since the mid-1980s a trend toward more flexible regimes has emerged. However, independently floating exchange rates comparable to those of the major international currencies remain rare in the developing world. Instead, various types of intermediate arrangements have been adopted in attempts to combine exchange rate stability with policy flexibility. The variation in exchange rate regimes has generated research interest concerning the determination of these choices. The Optimum Currency Area (OCA) theory of the 1960s views the exchange rate primarily as an expenditure-switching device and develops a list of criteria for favouring fixed-rate against flexible-rate regimes, such as the absence of asymmetric demand shocks, high factor mobility (Mundell, 1961), small economic size and high economic openness (McKinnon, 1963), and high production diversity (Kenen, 1969). The literature of the 1970s focuses on the automatic-stabilizer property of exchange rates and concludes that fixed-rate (flexible-rate) regimes perform better in terms of output stability if nominal (real) shocks are the main source of disturbances (Boyer, 1978 and McKinnon, 1981). Following the analysis of Barro and Gordon (1983) on the credibility of monetary policy, the literature in the 1980s discusses the possibility of using exchange rates as nominal anchor to improve the credibility of the domestic monetary authority's efforts to contain inflation (Goldstein, 1980, Melitz, 1988 and Fratianni and von Hagen, 1992). Empirical research on exchange-rate regime choices started in the late 1970s, when more diverse regime choices began to be observed. The early studies selected potential regime determinants based mainly on the OCA criteria (Heller, 1978 and Dreyer, 1978), while those in the 1980s added variables reflecting different types of shocks and stabilization strategies (Melvin, 1985 and Savvides, 1990). Some authors also included institutional and political variables as potential regime determinants (Edwards, 1996, Bernhard and Leblang, 1999 and Méon and Rizzo, 2002). A comprehensive approach covering a wide range of regime determinants is adopted by many recent studies (Rizzo, 1998, Poirson, 2001, Juhn and Mauro, 2002 and Von Hagen and Zhou, 2005). As summarized in Table 1, the empirical results of these studies seem to be sensitive to the sample composition, data construction, and model specification.1Table 1 indicates that most studies (10 out of 14 under our review) use a simple binary structure (denoted by “B” for regimes) to classify exchange rate regimes into either fixed or flexible ones. Seven studies include intermediate regimes as a separate option and use an ordered-choice classification (denoted by “O”), assuming that the degree of regime flexibility is monotone in the regime determinants. Only two studies use a multinomial-choice structure (denoted by “M”), which is a more general and flexible framework able to capture both the diversity in regime choices and the complexity in the response of regime choices to the changes in the determinants. One commonly used estimation method is cross section analysis (denoted by “CS” for methods). Exchange rate regime choices of a given year are typically explained by the average values of the independent variables over several previous years. While this can dampen the effects of temporary disturbances in the regime determinants and attenuate endogeneity problems, it is less appropriate when substantial volatility is observed in the economic environment or the exchange rate regime itself undergoes frequent changes. Using past averages to explain current choices in such a constellation may result in misleading inferences on the role of some factors. Another commonly used estimation method is pooled panel analysis (denoted by “PP”). Observations from different countries and different years are assumed to be drawn from an identically and independently distributed data-generating process. Assuming an identical distribution ignores cross-country heterogeneity in unobserved factors, while assuming independence over time ignores the correlation of exchange rate regime choices over time. This simplification may result in biased estimates of the covariance matrix making statistical inference inaccurate. In reality, and especially when the country coverage is wide, it should be more sensible to allow for country heterogeneity than simply to assume it away. Moreover, an empirical regularity in exchange rate regime choices is that choices made by the same country tend to be correlated over time. This suggests that countries that have selected a particular exchange rate regime in the past are more likely to select the same regime in the future than are countries that have not selected the regime. There are two explanations for this empirical regularity (Heckman, 1981a). One is that macro economic conditions or constraints relevant to current regime choices are affected by past exchange rate regime choices. For example, having had an exchange-rate peg in the past may have promoted trade with the anchor country and encouraged unhedged foreign-currency borrowing, both of which re-enforce the desirability of a fixed-rate regime. State dependence may also work through reputation mechanisms, as the credibility of a central bank's commitment to defending a fixed exchange rate may deter speculative attacks and reduce the need to maintain large foreign currency reserves. Under such circumstances, past regime choices have a genuine structural effect in the sense that an otherwise identical country which has not chosen a given exchange rate regime in the past is more likely to make different choices in the current period than a country which has chosen this regime. This structural dynamic linkage between past and current regime choices is labelled “true” state dependence. Alternatively, regime choices may be correlated over time simply because the economic and political factors impacting them are correlated over time. More specifically, if cross-country differences in these factors are correlated over time, and if these differences are not properly controlled for in the empirical analysis, previous regime choices may appear to be important determinants of current choices simply because they are good proxies for persistent unobservable factors that affect regime choices. This type of dynamic linkage between past and current regime choices is labelled “spurious” state dependence. Distinguishing between true and spurious state dependence has important implications when assessing the effect of policy measures or macro economic development on the choice of exchange rate regimes. If regime choices are truly state dependent, one-time policy actions or other macro economic shocks that have led to the adoption of an exchange rate regime have persistent effects on the probability of choosing that regime in later periods. Furthermore, policy efforts to enhance the sustainability of an exchange rate regime or to improve a country's ability to live with it, which may be politically costly in the short run, may pay-off in the longer run by making the macro economic fundamentals more conducive to choosing this regime. With spurious state dependence, in contrast, the determinants of exchange rate regime choices are not affected by short-term macro economic developments and policy efforts in the long run. These two types of state dependence call for different econometric techniques. Accounting for true state dependence in dynamic regime choices is possible by including lagged regime choices as explanatory variables of current regime choices. Accounting for spurious state dependence is possible by adding a country-specific error component as a proxy for unobserved country heterogeneity, which generates serial correlation in overall regression errors. However, due to technical difficulties in the estimation of panel discrete-choice models, especially due to the heavy computational burden of numerical integrations, panel discrete-choice models are rarely implemented in the literature. This paper studies exchange rate regime choices in more than 100 developing countries, emerging market economies, and transition economies during the 1980s and the 1990s. The model allows three choices – fixed, intermediate, and flexible regimes – in a non-ordered way and can be easily extended to choice structures with more alternatives. The dynamic linkage among regime choices is modelled by including country-specific random effects and lagged regime choices to account for both types of state dependence. The technical difficulties involved in the numerical integrations are solved by a simulation-based estimation approach. We follow a comprehensive approach when selecting potential regime determinants. Conventional OCA fundamentals, stabilization strategies, currency crises risks, and political and institutional features are added sequentially to the model. They explain most of the variation in observed regime choices, leaving little room for unobservable factors to play a role. As a result, spurious state dependence is found of little importance in regime choices. We do find, in contrast, strong evidence for true state dependence in regime choices, as past regime choices strongly influence current choices. Our results suggest that the empirical importance of some regime determinants be overestimated by the existing literature employing static models. We also find that some regime determinants have non-monotonic influence on regime choices. This suggests that our non-ordered multinomial choice structure is more appropriate than an ordered or binary structure commonly adopted in the existing literature. The rest of the paper is organized as follows. Section 2 discusses the classification of exchange rate regimes as well as the potential regime determinants. Section 3 presents our multinomial panel model for exchange rate regime choices and sketches the estimation procedures. Section 4 discusses the estimation results and Section 5 concludes.

In this paper we apply simulation-based estimation techniques to the analysis of the choices of exchange rate regimes in developing countries since the fall of the Bretton Woods System. We expand the conventional fixed-vs.-flexible dichotomy into a trichotomous choice structure, with fixed, intermediate, and flexible regimes as three options. We use a non-ordered multinomial framework to allow the possibility that the influence of some variables on regime choices are not monotonically increasing or decreasing in the underlying regime flexibility. Moreover, we model the persistence in the regime choices of the same country by including country-specific time-invariant heterogeneity and past regime choices in the decision on the current ones. We consider a wide range of potential regime determinants, including the OCA fundamentals, stabilization strategies, currency crises risks, and political and institutional features. Together, they explain well the observed choices of exchange rate regimes in our sample. We find that the use of a multinomial choice model is more appropriate than an ordered-choice model, as the impact of the explanatory variables on the choice of exchange rate regimes is often non-monotonous as the latter model assumes. Furthermore, we find that there is considerable regime persistence, which is best explained in terms of true state dependence in the choice of exchange rate regimes. In contrast, spurious state dependence caused by persistent country heterogeneity is of much less importance for regime choices. Ignoring the dynamics of regime choice tends to result in an overestimation of the effects of changes in the explanatory variables and to miss the importance of the status-quo regime on the direction of changes in regime choices. The existence of true state dependence in exchange rate regime choices opens up a set of interesting questions as to the channels through which regime persistence works and its policy implications. We hope to address these questions in future research.