Extension to “Market Interaction and Efficient Cooperation” (Brandts and Riedl, 2018)

We call the new treatment OSDG-IT, where IT stands for individual task. This treatment is an exact copy of the MP treatment except that the double-auction market is replaced by the individual task.

In the experiment participants will earn experimental currency units (ECU) which are converted to Euro at the end of the experiment as described in Brandts and Reidl (2018). The individual task will consist of 18 rounds in which participants will be able to earn ECU. Each of the 18 rounds will last for 95 seconds, which is the average length of a market round in the MP treatment. In each of these rounds participants will face the task of adjusting six sliders in the well-known slider task of Gill and Prowse (2012). The task is to position each slider at 50. Each correctly adjusted slider will yield ECU to the participants.

In the individual task, participants will be divided into two groups which differ regarding the comensation per correctly positioned slider. One group, low-earners, will receive 3.15 ECU and the other group, high-earners, will receive 24.74 ECU. These compensations are chosen such that when all sliders in all rounds are positioned correctly, the (rounded) total earnings are equal to the average earnings of respectively buyers and sellers (340 and 2672, respectively) in the MP treatment. Participants will be informed of their own compensation before the start of the slider task.

To maximize the likelihood that all participants position all sliders correctly there will be two unpaid trial rounds before the first round of the individual task. In the first trial round there will be no time limit. The reason for having this trial round is to let participants learn about the individual task. In the second trial round there will be a time limit of 95 seconds, just as in the subsequent paid rounds of the individual task. This second trial is meant to familiarize participants with the time pressure under which they will have to work in the paid rounds of the individual task. In addition, during the individual task, participants will receive a warning message when they did not position correctly all 6 sliders in a round.

Mirroring the MP treatment, before the start of the individual task, participants will be informed about the first six rounds of the OSDG, which take place immediately after the individual task is finished. Here participants are also informed that in the OSDG they will be paired with one other participant to play the OSDG and that this other participant will have either (a) the same individual task compensation, (b) a better individual task compensation, or (c) a worse individual task compensation. After the slider task and before the 6 rounds of the OSDG this information will be repeated together with a summary of the most important elements of the OSDG (in the same way as in the MP treatment).

The exact value of the paired participant’s slider task compensation will not be communicated to participants. This information condition is parallel to the one we used in the social dilemma game of the MP treatment, where participants were informed of whether they were paired with a buyer or a seller. In the double-auction market of the MP treatment earnings were private information. Nevertheless, from the market dynamics buyers and sellers may have been able to infer that buyers were in a worse earnings position than sellers.

The described matching of participants leads to three pairings: pairs of participants with low compensation (low-low), pairs of participants with high compensation (high-high), and pairs of participants with one participant having a low and the other participant having a high compensation (low-high). In total it is planned to collect data of 174 participants (approximately) equally divided across the three different pairings, giving a targeted 29 pairs per compensation pairing (low-low, high-high, low-high). As some participants may not be reaching the maximum earnings in the individual task we expect to get about 20-25 pairs with (close to) maximum income per compensation pairing.

Hypotheses

Our hypotheses focus on the first 6 rounds of the OSDG that take place immediately after respectively the double auction market in the MP treatment and the individual task of the OSDG-IT treatment. The hypotheses are based on our previous results and we hypothesize that the directions of the differences between the contributions to the social-dilemma game in the MP treatment and in our new OSDG-IT treatment will be the same as between the MP treatment and the other OSDG-MP treatment reported in the main text of Brandts and Riedl (2018). That is, our hypotheses are guided by Result 2 in Brandts and Riedl (2018).

Hypothesis 1a (Buyer-buyer matchings in the MP treatment vs. low-low matchings in the OSDG-IT treatment.):
Contributions in the buyer-buyer matchings are lower than contributions in the low-low matchings.

Hypothesis 1b (Seller-seller matchings in the MP treatment vs. high-high matchings in the OSDG-IT treatment):
Contributions in the seller-seller matchings are (weakly) lower than contributions in the high-high matchings .

Hypothesis 1c (Buyer-seller matchings vs. low-high matchings):
Contributions in the buyer-seller matchings are equal to contributions in the low-high matchings.

Analysis Plan

We plan to use only pairs were both participants have achieved at least 90% of the maximum possible earnings in the slider task. This is to keep earnings in the individual task of the new treatment in keeping of the average earnings achieved in the markets of the MP treatment.

To test the hypotheses we will use non-parametric Mann-Whitney tests as well as parametric Tobit regressions with censoring of contributions at 0 and 50 where we will control for period effects using period dummies. See Brandts and Riedl (2018) for details.

In the Mann-Whitney tests the unit of observation will be strictly independent observations. That is, for the MP treatment the unit of observation will be the average contributions over the six rounds taken across pairs in a market and in the OSDG-IT treatment the unit of observation will be the average contributions over the six rounds in a pair. For details see Brandts and Riedl (2018).

In the Tobit regressions we will us average contributions within a pair as the unit of observation and correct for data dependency by clustering on markets and pairs, respectively. In the regressions we will control for period effects using period dummies. For details see Brandts and Riedl (2018).

Hypotheses 1a and 1b are directed and accordingly, we will apply one-sided test statistics. Hypothesis 1c is undirected and accordingly we will use two-sided test statistics.

As we are testing three sub-hypothesis we will report test statistics without and with correction for multiple (three) comparisons using the false discovery rate correction proposed by Benjamini and Hochberg (1995).

References

Benjaminin, Y. and Y. Hochberg (1995). Controlling the false discovery rate: A practical and powerful approach to multiple testing. Journal of the Royal Statistical Society, 57(1):289–300.

Brandts, J. and A. Riedl (2018). Market competition and efficient cooperation. (Revised version of CESifo Working Paper Series No. 5694. Available at SSRN: https://ssrn.com/abstract=2731096).

Gill, D. and V. Prowse (2012). A structural analysis of disappointment aversion in a real effort task. American Economic Review, 102(1), 495-503.

2018-12-14

2018-12-15

Contribution levels in the social dilemma game.

We call the new treatment OSDG-IT, where IT stands for individual task. This treatment is an exact copy of the MP treatment except that the double-auction market is replaced by the individual task.

In the experiment participants will earn experimental currency units (ECU) which are converted to Euro at the end of the experiment as described in Brandts and Reidl (2018). The individual task will consist of 18 rounds in which participants will be able to earn ECU. Each of the 18 rounds will last for 95 seconds, which is the average length of a market round in the MP treatment. In each of these rounds participants will face the task of adjusting six sliders in the well-known slider task of Gill and Prowse (2012). The task is to position each slider at 50. Each correctly adjusted slider will yield ECU to the participants.

In the individual task, participants will be divided into two groups which differ regarding the compensation per correctly positioned slider. One group, low-earners, will receive 3.15 ECU and the other group, high-earners, will receive 24.74 ECU. These compensations are chosen such that when all sliders in all rounds are positioned correctly, the (rounded) total earnings are equal to the average earnings of respectively buyers and sellers (340 and 2672, respectively) in the MP treatment. Participants will be informed of their own compensation before the start of the slider task.

To maximize the likelihood that all participants position all sliders correctly there will be two unpaid trial rounds before the first round of the individual task. In the first trial round there will be no time limit. The reason for having this trial round is to let participants learn about the individual task. In the second trial round there will be a time limit of 95 seconds, just as in the subsequent paid rounds of the individual task. This second trial is meant to familiarize participants with the time pressure under which they will have to work in the paid rounds of the individual task. In addition, during the individual task, participants will receive a warning message when they did not position correctly all 6 sliders in a round.

Mirroring the MP treatment, before the start of the individual task, participants will be informed about the first six rounds of the OSDG, which take place immediately after the individual task is finished. Here participants are also informed that in the OSDG they will be paired with one other participant to play the OSDG and that this other participant will have either (a) the same individual task compensation, (b) a better individual task compensation, or (c) a worse individual task compensation. After the slider task and before the 6 rounds of the OSDG this information will be repeated together with a summary of the most important elements of the OSDG (in the same way as in the MP treatment).

The exact value of the paired participant’s slider task compensation will not be communicated to participants. This information condition is parallel to the one we used in the social dilemma game of the MP treatment, where participants were informed of whether they were paired with a buyer or a seller. In the double-auction market of the MP treatment earnings were private information. Nevertheless, from the market dynamics buyers and sellers may have been able to infer that buyers were in a worse earnings position than sellers.

The described matching of participants leads to three pairings: pairs of participants with low compensation (low-low), pairs of participants with high compensation (high-high), and pairs of participants with one participant having a low and the other participant having a high compensation (low-high). In total it is planned to collect data of 174 participants (approximately) equally divided across the three different pairings, giving a targeted 29 pairs per compensation pairing (low-low, high-high, low-high). As some participants may not be reaching the maximum earnings in the individual task we expect to get about 20-25 pairs with (close to) maximum income per compensation pairing.

In summary: we propose new treatments in the experimental, which will be compared with treatments that were run in the past. The treatments we ran in the past can be seen as "controls".

Random assignment by computer to different conditions.

N/A

No

No clustering

192 students

192 students