Predictive behavior: An experimental study (Q1312635)

Within six chapters the author investigates the predictive behavior of individuals in an experimental study. To proceed along this line the author distinguishes between empirical and experimental studies, each category subdivided into a direct and an indirect forecasting approach. Empirical studies use survey data whereas experimental studies use data collected in laboratory experiments. The direct approach asks for forecasts reported by the repondents whereas the indirect approach asks for economic decisions based on expectations. After reviewing the literature on this topic (chapter 0), the author introduces the concept of rational expectations (chapter 1) and describes the experimental design (chapter 2), consisting of fourteen different models with different information content to the respondents. Fourteen experiments were run, each experiment with about twelve respondents who were not allowed to participate in more than one experiment. Each experiment was performed 80 to 100 times. In chapter 3 the author evaluates the outcomes of the experiments at an aggregated level by looking at averages and median forecasts. Judged by various criteria the author concludes that the test result rejects the assumption of rationality. Introducing two types of least-squares learning and a delta method for forecasting, the author concludes that the mean absolute deviation from rational predictions is smaller than the mean absolute forecast error, and deviations from delta predictions are less than the deviations from rational predictions. Chapter four repeats the results of the foregoing chapter on an individual base and in chapter 5 the author discusses features of individual forecasts. A brief summary is given in the final part of the book. Summing up, the book presents a thoroughly worked through study on predictive behavior and should be recommended to all who are interesting in evaluating forecasting behavior.