Are subjects making financial decisions in lab auctions or are they just gambling?

Abstract

Optimal bidding strategies in first-price and Dutch auctions are theoretically isomorphic but depend on bidder risk attitudes. However, laboratory experiments consistently find different behaviour between auction formats. This article explores whether the notion in psychology that financial and gambling risks are viewed differently can explain the discrepancy. Ultimately, the evidence does not support this hypothesis, but a bidder’s propensity to gamble is associated with how much risk he takes in both auctions whereas his propensity to take financial risks is not. The results suggest that subjects may view themselves as gambling in laboratory auctions rather than making financial decisions.

Keywords:

• multi-domain risk attitudes
• auctions
• laboratory experiments
• gambling

JEL Classification:

• C9
• D4

Notes

¹ In a related study, Deck et al. (2014) demonstrate that risk-taking behaviour can be affected by framing a decision as either a gambling task or a financial investment and that one’s propensity to undertake financial risk can partially explain within subject variation on the mathematically identical tasks.

² The complete DOSPERT battery also includes the perceived benefits and risks of engaging in an activity and covers domains such as social risk, recreational risk and ethical risk.

³ Copies of the survey and the auction directions are available upon request.

⁴ Their auctions involve four bidders with values drawn from a uniform distribution, U[0,250], so the expected profit per bidder is greater in our experiments. In our experiment, the upper bound of the support was changed to be a multiple of 15 for reasons described previously.

⁵ The actual number of observations in the Dutch auction varied depending on realized values and the behaviour of other bidders.

⁶ One subject did not fully complete the survey and another subject went bankrupt in the Dutch auction. Once a subject goes bankrupt, the experimenter has effectively lost control over the subject’s incentives as losses cannot be imposed due to institutional constraints. If the subject will receive $0 if she has experimental losses of $5 or $5000, then her incentive is to take long shot gambles regardless of risk attitude as losses or small gains do not affect her real payoff.

⁷ This ordering effect could also be related to the presence of the survey.