سیستم های هشدار دهنده زود هنگام برای بحران ارز: مورد ترکیه ای

Different severe financial crises episodes occurred in the Turkish economy in the last two decades. These crises led to severe economic and social consequences for Turkey in terms of increasing interest rates, large reserves losses, considerable currency depreciations, high output losses and high unemployment rates. This paper aims to illustrate the essential determinants of these crises by developing a multivariate logit model which estimates the predictive ability of sixteen economic and financial indicators in a sample that covers the period from January 1990 to December 2008. The empirical findings show that the Turkish crises are mainly due to excessive fiscal deficits, high money supply growths, sharp rises in short-term external debt, growing riskiness of the banking system (in particular currency and liquidity mismatches), and external adverse shocks.

After relative stability in the post-World War II period, the world economy again became familiar with recurrent crisis episodes following the collapse of the Bretton Woods system in 1971. After the Latin American economies in the early 1980s, financial crises hit European countries in 1992–1993, Mexico in 1994, South-East Asia in 1997–1998, Russia in 1998, Brazil in 1999, Argentina in 2001, not omitting the last global financial crisis that has been affecting the world economy since late 2008. The global economic and financial instability of the 1990s and 2000s also touched the Turkish economy, which underwent three severe currency crises in April 1994, February 2001 and October 2008, and a high depreciation episode in May 2006. These repeated crisis episodes stimulated a large discussion on the theoretical specification of crisis models and the empirical analyses that aim at identifying crises determinants.1 Following the first wave of crises in the early 1980s, Krugman (1979) and Flood and Garber (1984) developed the so-called first generation models in which a currency crisis is related to persistent and growing macroeconomic problems. When these problems become unsustainable, investors attack the domestic currency, which leads to its devaluation. The outbreak of the European Rate Mechanism crisis in 1992–1993 led to the development of new crisis models, in particular by Obstfeld, 1994 and Obstfeld, 1996. In these so-called second generation models, a crisis can be triggered without significant deterioration of macroeconomic fundamentals ex ante. Therefore, even if economic policies are consistent with the fixed exchange regime, a speculative attack may occur if investors change their expectations towards the sustainability of the exchange rate. Unlike the first generation models, where policymakers are supposed to show mechanical behavior against a speculative attack (selling foreign reserves and then floating the currency when the reserves stock is exhausted), policymakers in the second generation models are supposed to show optimizing behavior by adjusting their policy to the shift of investors’ expectations. This interaction between the government and investors creates multiple equilibria that may lead to the occurrence of self-fulfilling crises. The outbreak of the 1997 Asian crisis caused a reorientation of crisis models, since the dominant crisis theories failed to understand the consecutive crisis episodes which started with the Thai baht devaluation in July 1997. Several theoretical studies were then conducted in order to explain the nature of these violent and contagious crises which resulted largely from the banking sector weaknesses in a financially liberalized economy (Krugman, 1998, Krugman, 1999, Radelet and Sachs, 1998, Corsetti et al., 1999, Aghion et al., 2000, Chang and Velasco, 1998 and Chang and Velasco, 2001, inter alia). The high costs of crises for the public sector as well as for private investors also led to a proliferation of empirical studies that aim to illustrate the key determinants of crises in order to predict future crisis episodes. In this regard, these studies are frequently called “early warning systems”, as they are likely to inform policymakers as well as investors about the occurrence of a crisis in the near future. These empirical studies (Kaminsky and Reinhart, 1999, Berg and Pattillo, 1999b, Aziz et al., 2000, Caramazza et al., 2000, Bussière and Fratzscher, 2002 and Bussière, 2007, among others) focused on very large country samples (industrialized and/or developing countries), utilized very different sample periods (from 1900 to our days) and used a very large set of indicators (representing macroeconomic, financial, political sectors), generally on a basis of panel data or cross-section analysis. The studies that adopt a multi-country approach give a general view of currency crises; however, there is not enough match among studies concerning what the key determinants of crises are. This result is due to the different sample countries and/or sample periods used across studies. Since the developing and industrialized countries have different structural characteristics, the origin of crises may differ from one group of countries to the other. Moreover, even if the same variables affect crisis probabilities for some “homogenous” countries, the degree to which they affect the likelihood of a crisis occurring may differ from one country to the next (Mariano et al., 2004), as shown empirically in Cartapanis et al. (1998), Abiad (2003) and Edison (2003). The heterogeneity across countries also leads to a poor in-sample and out-of-sample forecast performance of the empirical papers that adopt a multi-country-approach (Berg et al., 2005). This is why we need a more country-specific analysis which allows policymakers or international organizations to take or suggest country-specific measures for avoiding crises. In this paper, we consider the Turkish economy, which suffered from different currency crisis episodes in a relatively short fifteen-year period. Therefore, exploring the cause of these crises becomes important in order to avoid future crises in the Turkish economy. There are few empirical studies that examine the currency crises that occurred in the Turkish economy.2 These studies generally focus on the 1994 and/or 2001 crises, paying no attention to the post-2001 crisis period. Hence, this study aims to fulfill this shortage by covering the entire post-liberalization era (1990–2008) rather than focusing on certain crisis episodes. Moreover, the study tests a larger set of sixteen potential leading indicators representing different sectors of the economy. Besides, the approach used in this paper enables us to illustrate the relative importance of first-, second-, and third-generation theoretical models in the occurrence of the Turkish crises. Lastly, to the best of our knowledge, this is the first Turkey-specific currency crisis study in which the out-of-sample forecast performance of the model is assessed. The paper is organized as follows. Section 2 presents a brief history of the Turkish crises. Section 3 details the development of a performing EWS for the Turkish economy. Section 4 presents the estimation results and assesses the forecast performance of the model both in-sample and out-of-sample. Section 5 discusses some policy implications to avoid future crisis episodes in the Turkish economy.

This paper aimed at illustrating the recent “history of crises” of the Turkish economy. After summarizing the stylized facts of the Turkish economy in the last three decades, we constructed an EWS model for the Turkish economy which identified the determinants of the Turkish crisis episodes very well. Besides, the model performed quite well in predicting these Turkish crises both in-sample and out-of-sample. The very good forecast performance of the models should drive modelers to adopt a single-country approach for constructing EWS models. According to the estimation results, the Turkish crises are due to a combination of different macroeconomic imbalances (excessive budget deficits, high money supply growths, real exchange rate overvaluations, sharp rises in short-term external debt), banking sector weaknesses and external shocks (terms of trade). However, the model cannot indicate clearly the relative importance of the first-, second-, and third-generation theoretical models in the occurrence of the Turkish crises. This stems from the lack of a complete theoretical model that confirms the statements of Cartapanis (2004) in the literature. Moreover, our model shows that, even if the Turkish economy seems more stable in the post-2001 period thanks to the large reform process implemented since the 2001 crisis, it still remains vulnerable to external shocks, as the May 2006 and October 2008 events confirmed. The economic growth strategy, which is heavily dependent on export earnings and short-term capital inflows, is the main reason behind these vulnerability problems. In that regard, reducing the dependence on foreign capital requires an increase in public and private savings that may finance a dynamic and sustainable economic growth. This transformation would also break the parallelism between economic growth and widening current account deficits which are a source of vulnerability for the domestic economy. This study could be extended for further research by including other explanatory variables like political factors, governance indicators, contagion indicator and more variables that take into account private sector weaknesses. Besides, other methods like the signaling approach and/or the Markov-switching approach might be utilized in order to compare the estimation results.