ABSTRACT
In the monthly surveys of US consumers conducted in the United States by the Survey Research Center of the University of Michigan, individuals are asked questions about their perceptions and expectations of general business conditions, their perceptions and expectations about their own household’s financial situation, and their expectations about a range of economic indicators, including unemployment, borrowing costs, and prices. One aim of this chapter is to determine which of these general economic indicators really matters to consumers, and whether uncertainties about unemployment, interest rates and inflation are as important as their expected values. Parallel to this, we are interested in the reliability of the survey data, and in particular the robustness of estimates of the mean and dispersion of expectations derived in different ways from qualitative survey questions. For the most part, the surveys do ask qualitative questions. For example the Michigan SRC survey asks: during the next 12 months, do you think that prices in general will go up, or go down, or stay where they are now? The questions asked about unemployment, interest rates, past and future general business conditions, and the past and future financial situation of the respondent’s household are similar. Responses are reported in the form of balance statistics – the difference between the percentages of respondents expecting prices to go up, and to go down, suggesting that this gives information about the inflation expectations of a typical consumer. Balance statistics enter press reports, broader indices of consumer confidence, and in some cases indices of leading economic indicators. Under strong assumptions, it is possible to translate the qualitative survey responses into quantitative estimates of the mean and standard deviation of the distribution of expectations across survey respondents. The pioneering paper is the Carlson and Parkin (1975) study of UK Gallup poll data. Their key assumptions are that the distribution of expectations is normal, that the range of price changes which elicit a no-change response is constant over time and symmetric around zero, and that expectations are unbiased. As it happens, the Michigan SRC also asks for quantitative estimates of future inflation, so for this variable it is possible to obtain direct measures of the mean and standard deviation of the
expectations distribution, and test the reasonableness of the Carlson-Parkin assumptions. Quantified mean expectations have been used to test whether consumer expectations are efficient or technically rational, and as inputs into econometric models. The standard deviation of individual expectations looks as if it should contain information about the degree of uncertainty of a typical respondent, and uncertainty about employment, interest rates and inflation should in turn affect consumer decisions. However, the conditions under which the dispersion of individual expectations correlates with subjective uncertainty prove to be quite restrictive, and, perhaps because of this, less use has been made of data on the dispersion of expectations in economic modelling. An exception is Batchelor and Dua (1992) where the dispersion of expectations about inflation and household financial position are found to be significant in the US consumption function. This chapter sets out in Section 2 a framework for analysing qualitative survey responses and clarifying the relationship between dispersion and uncertainty; then in Section 3 examines how this framework can be used to interpret aggregate survey responses, and discusses the statistical properties of quantified expectations and dispersion measures from the SRC surveys; and finally in Section 4 assesses which expectations and uncertainties really matter for economic well-being, as perceived by US consumers. We also, crucially, test whether the inflation expectations obtained from qualitative data lead to the same inferences as the directly measured inflation expectations. We find that mean expectations of unemployment, interest rates and inflation are all statistically significant determinants of consumer expectations about general business conditions. Unemployment and inflation also impact significantly on expectations about households’ own financial position. Consumers seem subject to some fallacy of composition in framing these expectations, however, with general business conditions expected to behave in a regressive way (with bad times expected to follow good times), while their own financial situation is simply extrapolated (with bad times followed by more bad times). Regarding the quality of the survey-based expectations, measures of mean inflation expectations are fairly robust, with similar results obtained for both qualitative and quantitative measures. The same is not true of the dispersion of expectations. Measures of the standard deviation of inflation expectations differ significantly between quantitative and qualitative surveys, and produce contradictory and counterintuitive results in our models of consumer expectations formation.
