This paper examines the predictive value of the University of Michigan Survey Research Centers Index of Consumer Sentiment (ICS) and the Index of Consumer Expectations (UME), and the Conference Boards Index of Consumer Confidence (ICI) and their measure of consumer expectations (CBE). I find the logic behind the ICS and the ICI to be confusing because they combine questions about the respondents own situation (e.g., the ICS asks "Would you say that you and your family are better off or worse off financially than they were a year ago?") with questions about their predictions for the economy, (e.g., "Looking ahead which would you say is more likely: that in the country as a whole well have continuous good times during the next 5 years or so, or that we will have periods of widespread unemployment or depression, or what?"). One would expect that the respondents have little expertise for the latter type of question. If the idea is that these beliefs will affect peoples plans, why not ask directly about these plans? The ICC also combines questions about the respondents situation and their outlook for the economy. (These criticisms refer to the original surveys, and not to the paper by Huth et al.)
The indexes have been prepared on a quarterly basis from 1961 through 1977, and on a monthly basis after that. Given their widespread use, it is interesting to determine whether these indexes produce better forecasts. Huth et al. regressed the monthly indexes from 1978 to 1991 against various economic measures 6 months later. They concluded that the indexes were significantly related (statistically speaking) to six measures of consumer activity (e.g. automobile sales), seven measures of business activity (e.g., single family housing starts), and eight measures of economic activity (e.g., Dow Jones average). The correlations were almost always in the proper direction for consumer purchases. Considering the four indexes with attitude and expectation components, 45 out of 48 correlations were in the expected direction and 16 were significant at the 0.05 level. It is difficult to explain the pattern of results. For example, there is no reason to expect that the Dow Jones average should be related, yet the expectations measures from each of the four indexes had t statistics ranging from 3.1 to 4.8. Also, the results do not tell us the extent to which these measures can reduce forecast error. Finally, the results do not show how the indexes would add value to standard ways of predicting the various measures. The procedures followed by Diebold and Rudebusch (1991) would have been useful in addressing this last issue. What are we left with? While I found this paper to be quite difficult to read, the results look promising. However, more research is needed to determine whether the indexes are useful for forecasting.