Model-based Analysis of Individual Decision Making in the DNB Household Survey

Abstract: Human financial decisions are known to deviate from ‘rational’, particularly under uncertainty and if time delays are involved. In the Nobel-winning prospect theory, Tversky and Kahneman showed that deviations from rational behavior were not only robust but also quite predictable, if considering average behavior in a population. They found that people consider gains and losses not in absolute terms, but relative to a reference point. People react stronger to losses than gains of the same magnitude, and they tend to overweight small probabilities and underweight large ones. It has also been shown that future outcomes are usually discounted using a hyperbolic discount function. In this thesis, I applied prospect theory to model risk attitudes and inter-temporal choices based on three different types of questions from the Dutch DNB household survey: probability equivalences, risky choices, and inter-temporal choices. Individual decision-making behavior was analyzed based on model parameters, such as loss and risk sensitivities, probability weighting, reference points, and the discounting rates. As different types of empirical measures (delays, probabilities and choices) were combined in my model, I estimated individually best-fitting parameters by minimizing the χ2-error of goodness-of-fit (Press et al. 1992). The parameter estimation results indicated slightly concave utility functions, strong loss aversion, relatively small discounting and minor influence of reference points, consistent with other model-based studies that addressed some of these aspects. Correlations between the estimated parameters indicated that individuals with more concave utility functions also discounted future less and were less affected by framing (i.e. had lower reference points). This result suggests that these different aspects of valuation may share common cognitive mechanisms, which could be investigated in future behavioral economics studies.