While it is widely acknowledged that political factors contribute to currency crises there have been few efforts at using political variables to improve crisis forecasts. We discuss ways in which political factors can be incorporated into theoretical models of crises, and develop testable hypotheses relating variations in political variables to variations in the probability of a currency crisis. We show that the incorporation of political variables into diverse crisis models substantially improves their out-of-sample predictive performance.
In recent years emerging economies have been severely affected by currency crises. For instance, the currency turmoil that struck Asia in 1997 and 1998 caused Indonesia's gross domestic product to decline by 15% in a single year. Thailand and Malaysia suffered losses amounting to approximately 10% of GDP. These country experiences are far from unique: recent estimates by the IMF put average GDP contractions for emerging markets due to currency crises at 8%.1
Given the high costs of currency crises it is of exceptional importance that the determinants of these catastrophes be well understood. Recognizing this, several scholars have recently responded by developing Early Warning Systems (EWSs) for currency crises. A wide array of EWS models focus on identifying a set of economic fundamentals that are correlated with crises and evaluating the usefulness of these variables in out-of-sample forecasts. What is surprising about the development of this literature is that political variables do not appear in EWS models, despite the fact that few would disagree that politics plays a major role in causing currency crises.2
In this paper we seek to fill this gap in a theoretically informed manner. We discuss ways in which political hypotheses can be drawn from the Morris and Shin (1998) model of currency crises. Specifically, we argue for the plausibility of a causal link between political variables that generate uncertainty, specifically divided government and recent government turnover, and currency crises. Using three previously published econometric models comprised of different economic control variables, different countries, and different conceptualizations/measures of currency crises, we find that the addition of these political variables robustly increases our ability to predict crises out-of-sample and, in some cases helps to reduce the proportion of false crisis warnings.
Our efforts to produce models that help forecast currency crises relate closely to previous efforts by Frankel and Rose, 1996, Sachs et al., 1996, Kaminski et al., 1998, Goldstein et al., 2000, Berg and Pattillo, 1999a, Berg and Pattillo, 1999b and Kamin et al., 2001, and Bussiere and Fratzscher (2002). The distinctive feature of our approach is our focus on political variables.
The structure of the paper is as follows. In Section 2, we discuss the implications that can be drawn from the Morris and Shin crisis model about the relationship between political variables and currency crises. Section 3 contains a description of our data, and we report our main empirical results in Section 4. In Section 5, we address issues related to identification, collinearity, and robustness. Section 6 concludes.
In this paper we derived two hypotheses about the relationship between political variables and the probability of a currency crisis from the Morris and Shin model. We hypothesized that divided government and recent turnover in government would raise the probability of a currency crisis. We found support for these hypotheses across diverse specifications. We also found the marginal effects of political variables, in particular divided government, to be substantial. Furthermore, we found that political variables substantially contributed to our ability to predict currency crises out-of-sample. However, the contribution of political variables to alleviating the problem of false crisis signals was ambiguous.
While adding political variables increases the ability of Early Warning System models to predict crises, a large proportion of crises still remain unpredicted. This paper thus only constitutes the starting point for a research agenda focused on using political variables to improve currency crisis forecasts. One aspect of our agenda includes the application of non-parametric switching models to define currency crises. This effort is called for given the variety of definitions of currency crises that are prevalent in the literature, and the fact that these definitions share the feature that they are fairly arbitrarily chosen by scholars of crises, rather than generated directly from the data. Another aspect of this agenda consists of developing more precise measures of political variables associated with currency crises. The fact that political variables emerge with substantial coefficients despite the likely prevalence of measurement error suggests that such efforts will result in a greater appreciation of the impact of politics on the likelihood of currency crises.
