Emotional Differences between Isomorphic Auctions

Abstract

First price and Dutch auctions are theoretically isomorphic, but previous experiments report that the institutions are not behaviorally isomorphic. This article uses facial analysis of video recordings of laboratory experiments to investigate whether these auctions invoke different emotional responses from bidders. The results indicate that bidders are angrier during the Dutch auction and that winners exhibit more contempt after the first price auction. Overall, subjects in both auctions appear to be mostly bored and depressed, but there is evidence that bidders exhibit more positive emotions as prices fall in the Dutch auction.

Keywords:

• Auctions
• Emotions
• Experiments

Notes

1 Deck and Wilson (2020) found that when subjects gained considerable experience via a near-continuous environment that allowed hundreds of auctions to be conducted in a short time, behavior in first price and Dutch auctions was similar.

2 Karmeliuk and Kocher (2021) compare individuals and teams in first price, English and second price auctions. They evaluate emotional response to learning of the results in the first price auction and find that winning is associated with greater happiness and less anger while losing is associated with increased sadness. They do not analyze emotions in second price and English clock auctions, which are theoretically isomorphic to each other, but are not isomorphic to first price and Dutch auctions.

3 In half of the sessions subjects competed in a block of 25 first price auctions and then a block of 25 Dutch clock auctions. In the other half of the sessions the order of the blocks was reversed. Subjects read institution specific computerized directions, available in the appendix, immediately before participating in the relevant block of auctions. The experiment was computerized using zTree (Fischbacher 2007).

4 Matsumoto and Willingham (2009) report that those who are blind exhibit similar facial expressions to the those who are not blind in response to these emotions.

5 Ethics approval was granted by the Institutional Review Board at Chapman University.

6 A fixed matching protocol is typical in auction experiments going back to Cox, Roberson, and Smith (1982) who found that as long as there were more than three bidders that behavior was competitive. Subjects were informed in the instructions that they were in a fixed group of four bidders.

7 A total of 2150 auctions were conducted, but two observations of first price auctions are dropped because the bids placed by the subjects exceeded the maximum value of 240 three times over.

8 Statistical significance based on regression analysis with standard errors clustered by session.

9 If a subject’s face is not visible in an image, because for example the subject looks away from the screen, no emotional evaluation can be conducted and that image is dropped from the analysis. There is no apparent relationship between auction behavior and a facial image being captured in the camera frame.

10 All results are qualitatively similar if the Value Revelation and Auction Results are reduced to one second or extended to three seconds (i.e. based on approximately 20 or 60 images per subject per auction, respectively). Because multiple groups participated at the same time, the price and profit feedback was not shown until all of the groups in a session completed their auction even though no information was shared across groups. But, the outcome of a Dutch auction is known to the bidders when the auction ends whereas the outcome is not known to the bidder in a first price auction until the feedback is given. Therefore, we rely on the first two seconds after the auction ends to evaluate responses to the results in the Dutch auction and the first two seconds after market feedback is displayed to evaluate responses in the first price auction. We also note that, even though the time one waits for results in the first price auction varies within and between groups, there is no substantive relationship between the amount of time a subject waits for the results and the subject’s emotions upon learning the results.

11 Time spent in the biding stage of the first price auction is endogenous. Analysis using a median split of the time spent in the bidding stage of a first price auction reveals no substantial differences in emotions associated with time spent biding.

12 This boredom is not unique to our specific auction experiments or driven by the duration of our experiments. Both Breaban, van de Kuilen, and Noussair (2016) and Breaban and Noussair (2018) report that subjects exhibit negative valence on average. Additionally, video recordings of our subjects reveal they exhibited negative valence (average value = -0.254) and low arousal (average value = 0.389) prior to any information pertaining to auctions being presented. Finally, minimal emotional differences are observed when comparing the first ten auctions and the last ten auctions of either format.

13 We do not believe our failure to find a price difference between first price and Dutch auctions is driven by our software, which was not specifically developed for this project, or by our subject pool, which is that of a large experimental economic laboratory generating typical behavior across a wide variety of topics. Instead, we speculate that it is a type II error.