A market based study of foreign exchange exposure - the effect of time horizon and survivorship bias

The overall impact of exchange rates on firms is extremely complex especially as indirect exposures are difficult to identify and measure. Consequently a large number of market based studies have failed to identify any residual impact of exchange rate volatility on the firm’s market value. There are various possible explanations for these findings, including the time horizons selected for the research and the sample selection process, both of which are addressed in this paper. We examine the existence of a contemporaneous relationship between the stock returns of UK non-financial firms and fluctuations in foreign exchange rates using three sub periods. The sample was split into three equal periods covering the years 1985-1989, 1990-1994, and 1995-1999. The findings show the market values of approximately 15% of firms are exposed at the 10% level to exchange rate volatility. There are also a high proportion of positive exposures among the sample firms with significant exchange rate exposures, indicating a higher proportion of firms benefit from an appreciation of the £. The findings of this subset analysis indicate that exposure to exchange rate movements are not stable over time and increase in periods of economic uncertainty. Secondly, the study examines the impact of survivorship bias that can arise as firms drop out of the sample over the three periods. In order to test and correct for survivorship bias a probit sample selection model is used. The results indicated that survivorship bias is a problem in times of economic uncertainty: survivorship bias was significant during the period 1990-1994 where the UK economy exchange rates and interest rates were volatile. We conclude that the extent and significance of foreign exchange rate exposure is low and consistent with the reported findings of earlier studies. However, the full extent of firm exposure to exchange rate shocks may be better understood if researchers consider survivorship bias in their sample data sets