توزیع های آماری و شناسایی بحران ارز

We use extreme value theory to identify periods of currency crisis for a broad cross-section of Asian, European, and Latin American countries. We argue that our methodology improves upon the more conventional methodology for defining currency crises because fewer parametric assumptions need to be satisfied. We compare the incidence of currency crises using the conventional method with the incidence obtained using our method. We conclude that identifying currency crises using extreme value theory is a good alternative to the conventional method.

A voluminous literature on currency crises has developed in response to the severe dislocations that economies experience when their currencies come under attack. This literature has focused on many issues, seeking answers to a variety of questions including: What are the determinants of currency crises? Are currency crises contagious? Can crises be predicted? Are certain exchange rate systems more prone to currency crisis? This paper addresses none of these interesting issues. Instead, our objective is more basic and fundamental—to evaluate the existing methodology used for identifying currency crises and to present an alternative methodology that makes fewer parametric assumptions. We apply this alternative methodology to the economies of a broad sample of European, Latin American, and Asian countries. In what follows we first discuss how currency crises have more commonly been defined. In general, researchers declare a currency crisis to take place when large values for a currency index are observed, with ‘large’ defined as observations that are 1, 2 or 3 standard deviations above the mean value for the index. We question the appropriateness of these methods to define extreme values given the non-conformity of speculative price series to the parametric assumptions needed to employ such methods. We suggest instead using extreme value theory for identifying large values of the currency index. In place of making an a priori assumption about the shape of the distribution and using that assumption to define extreme values, we estimate the shape of the distribution of the currency index and use this information directly to infer the frequency of extreme observations. Our contribution, therefore, lies in avoiding parametric assumptions in identifying the extreme or tail values of the distribution of the currency index. We demonstrate the application to data spanning 1965 to 1997. In conclusion, we compare the performance of the methodology based on extreme value theory to that of the conventional methodology using data from 16 nations located in Asia, Europe and Latin America.

In this paper, we compare the performance of the extremal value method with the ER&W method to identify currency crises. The number of crisis episodes computed using these two methods for the 16 countries in our sample are summarized in Table 7. It is evident from the results in this table that the extremal value method is more sensitive than the ER&W method for identifying crisis, signaling more periods of speculative pressure in all cases. According to the extremal method, 160 quarters of crisis are identified, while using the ER&W method we identify only 64 quarters of crisis. Furthermore, the country-specific crisis incidence appears more sensible. While the ER&W methodology suggests the incidence of crisis in the UK and in Argentina to be about equal at 4% (four out of 101 quarters of crisis in the UK and two out of 48 quarters of crisis in Argentina), the incidence of crisis in the two countries differs substantially using our methodology. In particular, we find more realistic crisis odds using extremal analysis, with the UK equaling about 5% and Argentina equaling about 15% (five out of 101 quarters of crisis in the UK and seven out of 48 quarters in Argentina).At the regional level, the incidence of crises computed using the two alternative methods also differ. According to Table 7, under the extremal value method, Latin America is by far the most crisis-prone with an incidence of 21%; Asia is least crisis-prone with an incidence of 7.8%; while Europe falls in between (though more similar to Asia) with a probability of crisis of 10.6%. As is the case for all other approaches used to identify currency crisis periods, our approach may not provide an unambiguous standard that can be used to verify that what we identify as a currency crisis is indeed a currency crisis. There is no formal definition of currency crisis derived from theory, and multilateral organizations do not systematically categorize crisis countries or crisis periods. Hence, there is no way to ‘grade’ the accuracy of these multiple approaches. Nonetheless, our approach appears to dominate the ER&W-type approaches on statistical grounds by avoiding a priori assumptions regarding the underlying distribution of the EMP series. This is particularly important considering the existing uncertainty regarding the true distribution of speculative price series. In addition to the statistical rationale for employing extreme value theory, the results conform better with our expectations about the propensity of currency crises in the examined economies. In sum, we have identified a promising method to distinguish currency crises using extreme value theory, which may help us to better measure speculative pressures and understand the determinants, development, and spreading of currency crises.