Nor Azuana  Ramli, Hooy Chee  Wooi, Mohd Tahir  Ismail, 
    Do the Choices of Indicators Affect the Predictability in Modeling an Early Warning System for Currency Crisis? An Analysis,
    International Business Management,
    Volume 8,Issue 2,
    2014,
    Pages 64-75,
    ISSN 1993-5250,
    ibm.2014.64.75,
    (https://makhillpublications.co/view-article.php?doi=ibm.2014.64.75)
    Abstract: Currency crisis is like a never ending episode in the economic story. It occurs anytime and sometimes without any warning. Previous studies showed that researchers can actually see the sign of the crisis before it occurs. But the questions left are how accurate the sign given and what variables should researchers look up to see this &#145signal’? This study’s aim to see the signal through International Country Risk Guide indicators by using the famous method by Kaminsky and Reinhart in modeling an early warning system. Researchers compared the performance of International Country Risk Guide with selected 12 macroeconomic indicators for 10 countries. Besides studying the signal of crisis from different types of data, researchers also have conducted an experiment in predicting future currency crisis by using logit method and another method from the previous study on macroeconomic indicators which is k-nearest neighbour method. The findings showed that the ICRG indicators give better prediction than macroeconomic indicators. The signal for crisis shown in every country by using the ICRG indicators is almost accurate except for South Africa. For the prediction results, k-nearest neighbour method is quite comparable to logit method.
    Keywords: signaling approach;International country risk guide;k-nearest neighbour method;logit;episode;currency crisis