Introduction

Economists have increasingly used the language and symbols of probability through the last three decades. The earlier reliance on multiple regression and correlation — with a small menu of methods for estimation of structural components in probability models, for drawing statistical inferences and making numerical or qualitative predictions — has widened to deal with a host of additional complications. New interests include the recognition of limited ranges for some economic variables, awareness of their latency and incomplete measurement, the cointegration of economic variables over time and concern over difficulties stemming from the inherent structural instability of socioeconomic phenomena. Besides this widening and rapid growth of complexity, empirical research seems to have been overwhelmed by the flood of statistical testing (to the extent that we are accused of'star-gazing' by sceptics who doubt both the foundations of these tests and their awkward interpretation by economists).