Abstract:

The Index of Consumer Sentiment (ICS) can be used to explain and forecast changes in the aggregate course of the economy. Its explanatory as well as its predictive power appears to be relatively low. Partly this might be caused by some methodological problems inhearent to the way in which the questions (upon which the ICS is based) are asked and in which the ICS is computed. Some of these issues already occurred in the literature to a certain extent. In this paper some new evidence is presented to underline the methodological weakness of the index. This is done by means of time series analysis of the Dutch ICS in the period 1974-1983, as well as by means of analysis of one cross section referring to a questionnaire we did ourselves in the spring of 1984. There appear to be for instance serious problems with respect to the interdependences between the answers on the questions and over time, the unidimensionality of the scales, the homogeneity of the answers given. Moreover the ICS differs among various subgroups of the population and especially among different subgroups facing different developments in for instance income.