Morals in multi-unit markets: Study II

We intend to shed further light on the mechanisms driving the results in Ziegler et al. (2021). We do so by: (i) eliciting subjects beliefs about their probability of being pivotal; (ii) measuring the effect of market selection by comparing groups with homogeneous moral costs to groups with heterogeneous moral costs; (iii) studying a setting based on a simultaneous double-auction with private schedules.

2021-10-25

2022-01-31

Most analysis follow the earlier analysis in Ziegler et al. (2021). Primary outcomes are:

• Individual valuations of a charity donation, directly elicited from participants in individual choices or inferred from bid and ask prices and accepted offers in markets. In these markets, trading produces an externality by reducing donations to a charity.

• Traded quantities in the markets.

• [NEW]: Beliefs about pivotality in markets

For a detailed explanation of our outcome variables, see the experimental design and the analysis plan.

We compare social norms about appropriate behavior in- and outside of markets and subjects' beliefs about the median valuation of the donation between treatments.

We run six treatments: HOM, HET, BEL-MULTI, BEL-FULL, SPEC-FULL and PRIV.

In HOM and HET we run markets identical to FULL where matching in groups is determined by IDM evaluations. In HOM we will form homogeneous groups including the 50% of the subject population that is closest to the median evaluation. In HET we will form heterogeneous groups, including the 50% of subjects that are most distant from the median.

We further run two market treatments, BEL-MULTI and BEL-FULL, which are identical in all aspect to, respectively, MULTI and FULL, except for the introduction of a belief elicitation task during trading, where subjects report their belief that trade will continue for the next unit as well as the number of active traders they expect there to be. Here, before the trade of unit 10, 12, 13 and 15, subject reports beliefs about: (1) how many other traders (other than themselves) will be active in the trading of the next unit; (2) the probabilities that: (i) their choice to trade is consequential for trade to happen; (ii) trade stops irrespective of what they do; (iii) trade continues irrespective of what they do.

In SPEC-FULL, subjects will only act as observers of the market and report beliefs. That is, each subject in SPEC-FULL is linked to a subject of a previous BEL-FULL matching group. The "observer" subject monitors how play unfolded in the linked matching group. At identical moments as in BEL-FULL, the "observer" subject reports beliefs about: (1) how many other traders (other than the linked trader) will be active in the trading of the next unit; (2) the probabilities that: (i) the linked trader's choice to trade is consequential for trade to happen; (ii) trade stops irrespective of what the linked trader does; (iii) trade continues irrespective of what the linked trader does.

In PRIV, we run a market treatment with private demand and supply schedules which are assigned such that aggregate demand and supply are equal to the MULTI/FULL market setup.

Because the subject pool of the University of Amsterdam is not large enough to run all treatments, we will run approximately 65% of the sessions of each treatment (excluding PRIV) at Tilburg University. The exception will be for PRIV, that we intend to compare to the previous treatments FULL and MULTI. Given that the previous treatments were all run at the University of Amsterdam, we will do the same for PRIV.

***
For completeness, we report below the experimental design details from Ziegler et al. (2021) that remain common features of the treatments in the present study.

All markets consist of 10 participants, 5 per market side (buyers/sellers). In BEL-MULTI and PRIV each participant can trade up to 3 units; in HOM, HET, and BEL-FULL, each participant can trade up to 15 units. Across all markets (excluding PRIV), there is a common supply/demand schedule, which means that costs/values for trading a unit only depend on the aggregate number of units traded previously. In all treatments (except SPEC-FULL) participants first participate in the elicitation of their valuation of charity donations. Afterwards, they participate in the markets; lastly, they face the identical individual elicitation as before the start of the market.

Across all treatments (except SPEC-FULL) , the experiment will conclude with elicitations of: i) the median of the distribution of valuations of charity donations; ii) social norm elicitation about behavior in markets and in individual decision-making; iii) risk preferences.

Computerized randomization

We randomize at the level of the experimental session.

Subjects are first randomly allocated to one of the following treatments or treatment groups: (i) BEL-MULTI, (ii) BEL-FULL, (iii) PRIV, (iv) treatment group [HOM+HET] (v) SPEC-FULL.

For treatments BEL-MULTI, BEL-FULL, and PRIV, subjects are then randomly allocated to a trading group.

For treatment group [HOM+HET], subjects are allocated to one or the other treatment based on their evaluation elicited in the IDM task. Specifically, the 50% of subjects whose evaluation is closest to the median of the whole are allocated to treatment HOM. The remaining 50% of subjects is allocated to treatment HET. Within these treatments allocation into trading groups is performed following a stratification procedure.

In treatment SPEC-FULL, subjects participate without interaction. They are linked to a subject of a previous BEL-FULL treatment.

Yes

8 groups with 10 subjects each per treatment for BEL-FULL, BEL-MULTI, HOM, HET, PRIV, 40 subjects for SPEC-FULL.

440 subjects

80 subjects each in BEL-FULL, BEL-MULTI, HOM, HET, PRIV, 40 in SPEC-FULL