The political economy of exchange rate regimes in transition economies

See the essays in Bernhard et al. (2003) as well as contributions from Bernhard and Leblang (1999) and Broz (2002).

This argument is developed in greater detail and is tested on 120 years of data from advanced industrial countries in Eichengreen and Leblang (2002).

We could not obtain data on key variables such as agricultural and manufacturing employment for Armenia, Tajikistan, Turkmenistan, and Uzbekistan and therefore we do not include these countries in the analysis.

One could group the exchange rate regimes otherwise, e.g., by distinguishing between soft pegs and hard pegs (currency boards) or by redefining floats to include managed exchange rate regimes with no preannounced boundaries as in Markiewicz (2006) but the general tendencies are similar.

As can be expected, there are differences between the announced policies and the actual policies. For example, after 2000, only 10 percent of the de jure floats were actually floats and 15% of the de jure pegs were actually managed floats.

Export oriented industries benefit from devaluation only if their operations do not depend heavily on imported inputs. Similarly, import competing interests stand to gain from devaluation unless they are in an industry with low pass through where exchange rates have little influence on prices. Testing these hypotheses requires detailed data on the structure of the economy that are not available for a large number of countries and years. In general, while we believe that outcomes are driven by preferences operating through a political process, the lack of data on the size and influence of groups (particularly cross-nationally) forces us to look at aggregate economic features as a proxy for the size of underlying interest groups.

Table 2 shows that the correlation of trade openness and the dummy variables for de facto and de jure pegs is positive and statistically significant. Greater FDI is also associated with a greater likelihood of de facto pegs. Conversely, the correlation of the share of agricultural employment and the peg dummies is negative and statistically significant. The correlations of the share of manufacturing employment and the exchange rate dummies are not statistically significant.

The empirical evidence on these hypotheses in the transition economies is mixed. Greater trade openness is associated with a greater likelihood of pegging in Von Hagen and Zhou (2005b) and in Aisen (2004). The effect is not statistically significant or is opposite to expected in Klyuev (2002), Markiewicz (2006), and Von Hagen and Zhou (2005a). Trade concentration has a positive effect on the likelihood of adopting a peg in Von Hagen and Zhou (2005b) and a negative effect in Markiewicz (2006). Similarly, Von Hagen and Zhou (2005b) and Markiewicz (2006) find that financial development decreases the likelihood of a peg whereas Von Hagen and Zhou (2005a) report the opposite effect. Klyuev (2002) and Markiewicz (2006) find that high inflation is associated with a lower likelihood of a peg and Von Hagen and Zhou (2005b) find that the effect of inflation is not statistically significant. There is some evidence that small budget deficits are associated with a greater likelihood to peg whereas the effect of high unemployment is not significant (Klyuev 2002).

The index of central bank independence obtained from Cukierman et al. (2002) provides one value for all years. Therefore we cannot trace the changes to central bank independence over time.

We also estimated a conditional logistic model grouping the data on the country level. The model yields similar results to what we report in terms of the signs on the estimated coefficients. However, due to the short time span of our data, introducing fixed effects reduces substantially the statistical significance on our key explanatory variables.

The CEEC are Albania, Bulgaria, Croatia, the Czech Republic, Estonia, Hungary, Latvia, Lithuania, Macedonia, Poland, Romania, the Slovak Republic, and Slovenia.

The same effect was obtained using other measures of trade concentration: trade with Germany as percent of overall trade, trade with the EMU members as percent of GDP, and trade with the EMU members as percent of overall trade.

Markiewicz (2006) does not investigate in further detail the negative impact of trade concentration on the likelihood of pegging but conjectures that economies with significant geographic concentration of trade are more prone to shocks and would, therefore, prefer the flexibility of a float or a managed float. Here we explore an alternative explanation.

The Czech Republic, Hungary, Slovakia, and Slovenia, which are the transition countries with the greatest volume of trade with Germany and the EMU as percent of their GDP never had a de facto peg against the DM or the euro during our sample period.

The correlation of trade with Germany and trade with the EMU countries is high: 0.78.

We also included variables for checks and balances with veto power as well as interaction terms of democracy with other economic variables such as the trade balance, the current account balance, and inflation. The interactions were not statistically significant. We checked the magnitude and the statistical significance of all interaction variables using the procedure developed by Norton et al. (2004) as the standard coefficient estimates and significance tests for interaction terms in nonlinear models can be misleading. As in Ai and Norton (2003), the interaction terms varied, with an S-shaped form, at different levels of the predicted variable. However, the interaction effects were not statistically significant at any of those levels, confirming the results reported in the paper. In addition to the political variables, we also included the standard deviation of economic growth rates as a proxy for the likelihood of economic shocks and the correlation coefficient of GDP growth rates with the growth rates of Germany as a proxy for similar business cycles but did not obtain statistically significant results.

We also estimated a hazard rate model explaining the duration of floating and intermediate exchange rate regimes before a switch to a peg. All transition countries start off without a peg and then some adopt pegs over time. We lose about 50 observations because the model drops all observations after the switch has occurred. Also, we don’t have many switches due to the short length of the time period—only 11 years. As a result, the coefficients have the same signs as in our probit model but are not as significant. Still, agricultural employment and trade concentration are significant, and the remaining variables are significant at about the 15% level.

The estimation was performed using two-stage least squares with robust standard errors, clustered on the country level. We performed two tests to validate the choice of instruments. The null hypothesis of the Kleibergen-Paap rk LM Test is that the endogenous regressors are not identified by the instruments. The null was rejected at the 1.7% level providing evidence that the excluded instruments explain the endogenous variable. This was confirmed by the F-statistics that were significant at the 1% level. The null hypothesis of the Hansen J Test is that the excluded instruments are uncorrelated with the error term in the stage two equation. We fail to reject the null of the Hansen J Test indicating that the instruments are uncorrelated with the residuals in the stage two equation.

EMU entry cannot be reversed easily, which may serve to boost expectations of financial stability; but it also removes the option for currency policy in the event of a large shock. EMU entry also gives a country voice in the setting of monetary policy, although the transition countries would be relatively minor actors among the numerous other current and future members of EMU.