روش لوجیت چندگانه برای تبادل طبقه بندی سیاست نرخ ارز با نرم افزار رشد

We model a country's de jure exchange rate policy as the choice from a multinomial logit response conditioned on the volatility of its bilateral exchange rate, the volatility of its international reserves, and the volatility of its effective exchange rate. The category with the highest predictive probability implied by the logit regressions serves as our de facto exchange rate policy. An empirical investigation into the relationship between the de facto classifications and GDP growth finds that growth is higher under stable currency-value policies. For non-industrialized countries, a more nuanced characterization of exchange rate policy finds that those who exhibit ‘fear of floating’ experience significantly higher growth.

Accurate, rigorous, and scientific classifications of exchange rate policy are an important ingredient for assessing the merits between fixed and floating exchange rates. Until relatively recently, empirical research employed the de jure classification, which largely reflects the self-reported policy submitted by a country's central bank to the International Monetary Fund. However, many observers have noted that the de facto currency management for some countries seemed at odds with their de jure management.1 As a result of such discrepancies, the de jure classification has been viewed as unsatisfactory for assessing the role of exchange rate stability in economic performance and has motivated researchers to propose de facto exchange rate classifications that are based on observed properties of the foreign exchange market data. Influential contributions include the pioneering work of Reinhart and Rogoff (2004) (hereafter RR) and Levy-Yeyati and Sturzenegger (2003) (hereafter LYS). RR argue that a natural classification of exchange rate policies should be based on the behavior of the parallel market exchange rates on the grounds that they better reflect underlying market and monetary conditions than do the country's official exchange rates whereas LYS advocate the use of a k-means cluster algorithm to sort and assign countries to the various exchange rate policies.2 In this paper, we propose using a familiar econometric technique to obtain de facto classifications of a nation's exchange rate policy. The procedure uses tools that are familiar to economists, produces sensible results that are easily replicated, modified and updated. Specifically, we see three attractive features in the approach. First, classifier judgment is required primarily in selecting the variables to be included in the exchange rate regime classification model. Modifying and updating the classifications is therefore straightforward since one only needs to adjust the variables or update the data employed in estimation of the response problem. Second, the optimization criteria of our approach is familiar as it is based on the likelihood principle and has well-known properties. Difficulties associated with ‘inconclusive’ regimes often observed in RR and LYS methods are much less problematic. Third, it is feasible with our method to include a potentially large number of policy determinants.3 The idea that underlies our methodology goes like this. It must be the case that for many countries, the de jure exchange rate policy (regime) reported matches the de facto execution of that policy. We assume that these de jure policies would seem to be thoughtful assessments of the degree of perceived and economically relevant exchange rate stability experienced by that country. This is our motivation for modeling the de jure classifications as the outcome of a multinomial logit response conditioned on measures of the volatility of the country's bilateral exchange rate against an anchor currency, the volatility in the country's international reserves, and the volatility in the country's effective exchange rate. The unsystematic component–the error term in the model–captures unobservable factors that cause some countries to deviate from the announced exchange rate policy. The classification that has the highest predictive probability implied by the model serves as the de facto policy. For ease of reference, we refer to them as the LP (logit policy) classifications. Two explanatory variables that we employ, the volatility of a bilateral nominal exchange rate against an anchor currency and the volatility of international reserves, follow directly from the literature. This paper is the first to also use the volatility of the effective exchange rate. We give four reasons why this strategy would seem to make sense.4 First, the incorporation of multilateral factors to assess exchange rate policy makes sense when central banks (such as the Bank of Korea) have increasingly diversified their reserve holdings away from US dollar denominated assets.5 Second, if one restricts the analysis to movements in a single bilateral exchange rate, there are countries (such as the US) where it is not so obvious which exchange rate should be used. LYS based their US de facto classification on the dollar-deutschemark (DM) exchange rate and marked the US as a floater. Such a designation might make sense given the infrequency of US interventions in the dollar-deutschemark market, but when one considers that US trade shares with China have exceeded those with Germany since 1995 coupled with the Chinese policy of pegging to the dollar, looking exclusively at the dollar-DM exchange rate appears less defensible.6 Third, in cases where a country maintains a bilateral peg, unless that country trades exclusively with the anchor currency country (or within a bloc that pegs to the same anchor), the effective exchange rate will exhibit more instability than the bilateral exchange rate. The non-anchor exchange rates may even be economically more important than the anchor rate, especially if countries engage in relatively little trade with the anchor country. Thus movements of the effective exchange rate may have some weight in how one reports the fixity of ‘the exchange rate.’7 Fourth, while one of the central economic implications behind a hard bilateral peg are the constraints it imposes on monetary policy, instability in non-anchor exchange rates could compel the monetary authorities to pursue policies that are ultimately inconsistent with the peg so that the anchored bilateral rate may appear stable at the present time, but this does not necessarily imply that the country is pursuing a fixed exchange rate policy. Having obtained the LP classifications, we use them to study the impact of exchange rate policies on GDP growth. This is an issue for which economic theory does not have clear-cut predictions. While the trade-offs between fixed and flexible exchange rates have been studied in terms of the exchange rate's effect on stabilization and trade, the effect on growth is imperfectly understood. In the empirical literature, Ghosh et al. (2002) (who use the de jure classifications) and RR report that higher growth outcomes are associated with the more stable exchange rate policies whereas LYS finds that higher growth outcomes are associated with greater exchange rate flexibility. In our analysis using the LP policies to measure de facto exchange rate policy, we find that higher growth outcomes are associated with more exchange rate stability. We find that industrial country growth is not significantly related to the exchange rate regime so this result is driven mainly by the experience of non-industrialized countries. We also consider a more nuanced characterization of exchange rate policy by examining whether differences between what a country says (de jure) and what it does (de facto) matters for growth. Here, we examine growth asymmetries hypothesized by Genberg and Swoboda (2004) between countries that say they fix but float and for countries that say they float but fix. They argue “breach of commitment” when a de jure fixer that floats de facto and that this “has negative consequences for the economy.” On the other hand, a de jure floater that fixes de facto (fear of floating) delivers better than expected exchange rate performance and might is rewarded with superior growth outcomes. Our empirical work provides evidence to support to the Genberg-Swoboda hypothesis. For non-industrialized countries, growth is significantly higher for de jure floaters who are LP de facto fixers. The remainder of the paper is organized as follows. Section 2 presents our regime-response model and discusses features of the LP classifications. Section 3 contains our analysis of the relationship between the exchange rate regime and growth and Section 4 concludes. A description of the data and how variables were constructed is contained in the Appendix.

This paper has proposed an econometric approach to classifying exchange rate policy. The procedure has three attractive features: First, it is based on tools that are familiar to economists. Second, it can be replicated, modified, and updated in a straightforward manner. Third, it produces sensible results. In producing the classifications, we employed information contained in the country's effective exchange rate. The use of the effective exchange rate in our analysis leads to an improvement in classifying policies and underscores the value in taking a multilateral approach in forming a generalized assessment of national policy towards exchange rate management. Our investigation of the impact of exchange rate policies and growth found that the highest growth to be associated with de facto fixers. This is in line with much of the extant literature and is consistent with research that has found trade benefits from currency blocs. Whether the growth advantages that we find are the result of maintaining a stable currency per se or from selection of countries that are members of trade and currency blocs is unanswered but is a problem for future research. While the exchange rate regime adopted de facto appears to matter for growth, we also find evidence that a more nuanced representation of policy based on the fear of floating concept matters. Point estimates imply a rank-ordering of GDP growth from highest to lowest for categories: i) de jure floaters–de facto fixers, ii) de jure fixers–de facto fixers, iii) de jure floaters–de facto floaters, and iv) de jure fixers–de facto floaters. Countries may have a good reason to display fear of floating since those that do experience significantly higher per capita growth.