Reviews of Important Papers on Forecasting,
1985-1995 Reviews
Review of:

Hoch, Stephen J.(1988), "Who do we know: Predicting the interests and opinions of the American consumer," Journal of Consumer Research, 15, 315-324.

Managers often rely on intuition. For example, the preliminary screening of ideas for new products may be based on their intuition. Consider that you were asked to answer the following question based upon unaided judgment: "Of all married people in the U.S., how many women would report that they went fishing at least once in the past year?" [What is your answer?__________ ] Hoch asked 21 such questions of three groups of novices: 93 MBAs, 94 people who responded to an invitation to participate in a focus group interview, and 141 members of a consumer panel. Responses to the same questions were also received from 83 manageria1 level marketing experts, and from 103 people in marketing research positions. Overall, predictive levels were low. MBAs were the least accurate. This is probably because they were not similar to the target market, and they had little other information on which to make these judgments. The line managers did better, presumably because they had more relevant information, but they did not do so well as the consumers. The consumers were most like the target population, and this proved to be helpful. Finally, the research managers were most accurate, although the margin of superiority was modest. One might expect that the research managers would be able to recall more relevant information for this exercise. The subjects were then asked for predictions as to the direction of the historical trend for some of the items. For example, "Has the percentage of women who went fishing gone up or down in the recent past?" Although the subjects had some degree of confidence, their predictions were not correlated with the true changes. This suggests that individuals’ unaided subjective predictions about future changes in consumer interests and opinions are likely to be of little value. That is, if they cannot predict the past after having been there, how well could they predict the future?

This is a good study for researchers to keep on file in order to justify their existence. One obvious way to improve the estimates about consumers is to use research to obtain objective data. When this is not feasible, methods may be used to improve the judgmental forecasts. For example, decomposed questions will improve judgments in cases where uncertainty is high. Furthermore, group means would be more accurate than individual estimates. The use of such guidelines for structuring subjective judgments produced impressive results in MacGregor et al. (1988). I suspect that these procedures would produce significant improvements in the type of questions asked by Hoch.

Overall then, Hoch’s study reinforces the substantial body of literature that found unaided opinions to be a poor way to make predictions. Incidentally, the percentage of married women who said that they went fishing at least once was 34%.


MacGregor, Donald, S. Lichtenstein, and Paul Slovic, 1988, "Structuring knowledge retrieval: An analysis of decomposed quantitative judgments," Organizational Behavior and Human Decision Processes 42, 303-323.