Quasi-experimental analysis of the impact of exchange rate regime selection on crisis recovery: evidence from the Asian Financial Crisis

Abstract

Research typically treats exchange rate regime selection as exogenous. Using the Asian Financial Crisis as a case study, we show that countries that peg in 1996 and countries that float in 1996 are, on average, different from each other on variables that affect the outcomes of interest. After accounting for endogenous exchange rate regime selection using propensity score matching, we find that a country’s exchange rate regime choice in 1996 had no significant impact on the size of the shock to real income levels, but reduced subsequent income growth and weakly increased inflation.

Keywords:

• exchange rate regime
• propensity score matching
• Asian Financial Crisis

JEL Classification:

• F33
• C18

Notes

¹ See Ghosh et al. (1997), Levy-Yeyati and Sturzenegger (2003) and Andrew Rose (2011).

² As early as 2002, Calvo and Reinhart (2002) forwarded the hypothesis that countries endogenously selected their exchange rate regimes, often defending a de facto exchange rate that diverged sharply from the declared exchange rate. Subsequently, authors have attempted to estimate the regime selection decision (see Vonhagen and Zhou (2007) and Levy-Yeyati et al. (2010)).

³ Under kernel-weighted matching, a treatment group member’s outcome is compared to a weighted average of nearby treatment group members’ outcomes. The weight is proportional to a control member’s ‘distance’ from the treatment group member, in terms of the difference of their propensity scores. This matching method is preferred to other matching methods in cases where there are many individuals in the control group (Hirano et al., 2003).

⁴ As a check, we re-estimated our PROBIT model using only 1996 data and found estimates of the treatment effect that are not qualitatively different from those described in this section.

⁵ Readers may worry about movement between the treatment and control groups after the onset of the Asian Financial Crisis. Because this article investigates the effects of a country’s choice of exchange rate regime in 1996, only, and matches control and treatment countries on the 1996 propensity scores, it is unnecessary for us to account for treatment assignment slippage during the crisis. If, for example, a country fixed in 1996 and due to the crisis had to break the peg, which resulted in a higher interest rates or capital flight and thus lower post-crisis period economic growth, vis-à-vis a 1996-floating exchange rate country, then our article would attribute this result to the country’s exchange rate regime choice in 1996. From the standpoint of our framework, any action a floating exchange rate regime country is forced to take after 1996 that affects real income, income growth or inflation is part of the ATE of selecting a floating exchange rate in 1996. If we have properly matched, then we have controlled for relevant pre-crisis differences between countries so any post-crisis differences represent the treatment effect.

⁶ As a robustness check, we perform our analysis using the Levy-Yeyati and Sturzenegger (2005) de facto exchange rate classification method and find qualitatively similar results.