Interdependence between markets and autoregressive modelling

Abstract

In February 1990, on the day following the elections in the Lower House of the Japanese Diet, the Nikkei index at the Tokyo Stock Exchange lost 3%. At this time, Tokyo was the world largest Stock Market in terms of capitalization. In the following months, there was a steady decline of the Nikkei index; a key issue then was whether or not this drop would be transmitted to European and American markets, in which case a downward spiralling of stock prices could result. This did not happen, however, and the relative disconnection between Tokyo and other major stock markets came as a surprise to most analysts. This example concerning stock markets could be rephrased for commodity markets as well. It highlights the importance of being able to assess the strength of interactions between markets. Price differentials which have been discussed in chapters 3 and 4 do in fact provide a measure of market interdependence, but this measure has but a limited range of applicability since it is only appropriate for transactions in one commodity and in the same currency; it is for instance useless when it comes to analysing the coupling between the fluctuations of the Nikkei and the Dow-Jones indexes. A number of general methods have been developed by statisticians and econometricians in order to analyse the inter-relationship between two time-series:

• 1. Correlation analysis

• 2. Cross-spectral analysis

• 3. Autoregressive modelling and causality tests

• 4. Cointegration tests