Possibilities of financial crises forecasting with latent semantic indexing

Abstract

Our study contains application of Latent Semantic Indexing on financial crises prediction. Hypothesis to test was that equity markets are able to predict even sharp changes in monetary policy during a quarter ahead of such a change (which was searched during two quarters that followed). This hypothesis, tested on sample of 36 countries between years 1985 and 2007, has been confirmed according to interest rate and foreign exchange expert interpretation. The studied application of LSI even though it timed several crises on their exact start day is not suitable for financial crises prediction but can be recommended for specification and analysis of fragile countries which are or could be prone to a crisis.