Explaining forecasting bias: The case of real exchange rate variance

Abstract

Recent tests of the rational expections hypothesis (REH), based on the assumption that agents are risk-neutral, have yielded conflicting results when applied to foreign exchange markets. Here a two-stage procedure which does not assumej risk-neutrality is derived from a model of a utility-maximizing importer who incurs an adjustment cost if he changes his import order. Although a short-run test (based on aggregate imports) leads to rejection of the REH in the majortity of cases, a longer-ren test (based on manchinery imports) is more favourable to the null huypthesis. The observed shot-run tendency to, in effect, ignore low levels of real exchange rate variation is found to be more likely when bilateral rates have remained relatively stable, when the importingj economy is relatively closed, or when governments have announced policies of intervention to stabilize bilateral rates.