Consumer Surveys and Reality

This paper investigates the usefulness of Italian consumer surveys as estimation and forecasting tool over the period 1982-2003. To this end, standard consumption equations are estimated and then compared, in terms of in-sample and out-of-sample predictive ability, with corresponding models which differ from them only because of the presence of the confidence indicator. Unlike mainstream literature, the present work focuses on the relationships between subjective and objective information at a less aggregate level. In particular, the overall sentiment index is divided into four sub-indices related to the opinion about the i) current, ii) future, iii) general, and iv) personal situation. In turn, the total private consumption is divided in five items. The idea behind is to check if one attitudinal measure is more or less informative than another, and if some outlay is more or less “sentiment sensitive” than another. It is shown that the qualitative information obtained from household surveys improves both the goodness- of-fit of consumption equations and their forecasting performances. It is noteworthy that these improvements are all the more evident when working on disaggregated data, i.e., linking a particular kind of consumption to a particular sub index. For instance, perceptions about the future help to explain consumption for services more than disbursement for non durable goods.