Experiments on the determinants of asset price bubbles

Abstract

Price ‘bubbles’, which refer to sustained overvaluation in an asset, represent a serious threat to the stability of markets. This dissertation contributes to the understanding of the determinants of bubbles in experimental asset markets.

The first study investigates whether bubbles emerge in experimental markets because the assets used are endowed rather than earned, thus diminishing their ‘legitimacy’. It takes a new methodological approach to this question by requiring participants in some markets to earn their initial allocation. The results suggest that asset legitimacy is not likely to be a serious threat to the validity of existing results, as the frequency, severity, and duration of bubbles does not noticeably differ between markets where the initial allocation is earned versus endowed.

The second study examines how relative-performance based compensation (‘tournament’ incentives) and its composition impacts price behaviour. Existing studies suggest that tournament incentives exacerbate bubbles, and that this worsens with experience. In contrast to the existing studies, which use single-asset markets, this study implements a two-asset market, allowing for more natural risk-taking. Mispricing in tournaments is found to diminish with experience, while compelling evidence of a difference in price behaviour between tournament and absolute-performance based compensation (‘normal’ incentives) is not detected. This outcome suggests the conclusions of earlier studies are likely driven by the single-asset nature of their markets. Furthermore, in markets containing inexperienced traders, adding penalties for underperformance is associated with less trading activity, but also larger and longer-lived bubbles compared to reward-only tournaments. This result is largely more consistent with herding-driven price behaviour.

The third study explores whether peer effects driven by relative performance feedback explain the price behaviour observed in tournaments. In this study, normally-incentivised participants traded in markets while receiving periodic feedback about the average trader’s performance. Results from these markets, assessed in conjunction with the results from the second study indicate that when traders are experienced and compensated under normal incentives, supplying relative performance feedback reduces mispricing. In contrast, introducing tournament compensation when relative feedback is already provided magnifies bubbles, especially under rank-based compensation. These results suggest that information on ‘benchmarks’ may aid market efficiency.