Salience Effects in Portfolio Selection

A large experimental literature on choices between two reduced (or simple) lotteries has documented two robust facts: First, if both lotteries are symmetric, then subjects are typically risk averse; that is, with symmetric risks a large majority of subjects chooses the option with less variance in outcomes. Second, if at least one of the lotteries is skewed, subjects often reveal a preference for positive skewness, which can even result in risk-seeking behavior; that is, a large majority of subjects chooses a sufficiently positively skewed risk even if it is the option with more variance in outcomes.

In this study, we ask how robust the observed patterns in these simple choices are to increasing the complexity of the problem along two dimensions: we increase the number of available options and/or replace the reduced lotteries by portfolios (i.e., convex combinations of reduced lotteries), where the latter means that the final outcomes are not stated, but have to be inferred. The experiment is designed to test the predictions of salience theory of choice under risk (Bordalo et al., 2012) for such portfolio selection problems against those of naive decision rules like the 1/N-heuristic (e.g., Benartzi and Thaler, 2001), which predict that subjects diversify naively by investing equal shares into the available reduced lotteries.