Does loss aversion prevent redistribution? The role of rank concerns in redistribution by the poor and by the rich.

Aversion to rank reversal refers to the unwillingness to redistribute income from the rich to the poor when this results in reversing their ranks (Xie et al., 2017). Last rank aversion, consider that when asked to donate to the poor, people are less willing to do so if this results in themselves becoming poorer than the poor, that is, if this results in them being ranked last (Kuziemko et al., 2014; Martinangeli and Windsteiger, 2019).
We will examine the role of social distance and concern with rank when facing decisions about income allocations. In terms of social distance, we vary the wealth of the third party, who decides whether to re-allocate wealth from a rich to a poor recipient (the passive players). We also vary the extent of a concern with rank, by making the relative rank of the third party more salient in some treatments.

2019-06-10

2019-06-30

Our key dependent variable is the choice whether to make the proposed income redistribution or not in choices inspired by Xie et al. (2017)'s experiment. Participants decide on the allocation of bonus points for other MTurk workers. They take decision on whether to redistribute the points from other workers.

We will have 24 menus, which are 6 variations on the basic menu in Xie et al. (2017), plus 4 menus that redistribute away from the poor, so as to elicit possible extreme preferences for inequality maintenance.

We investigate the social identity of the third-party: are they rich or poor, and does their relative rank before and after redistribution matter? To vary the salience of the information we either display or not the income of the third party along to that of participants they decide the allocation of (workers).
We thus would run four treatments, in a two by two design, whereby we vary the wealth of the third party (Rich, Poor), and we vary whether the third party, when making her decisions, sees her wealth along those of workers (Salience, No Salience)

This design allows us to distinguish whether aversion to rank reversal is mainly due to thinking about others’ wealth and rank, whereby simply manipulating affiliation would have an effect, or whether it is mainly due to concern with own rank, whereby the effect would appear mainly when one’s relative rank in the income distribution is made more salient.

With reference to table 1, the statistic we monitor is the difference in the percentage of participants who are in favor of Type 1 changes and the percentage of participants who are in favor of Type 2 changes.

We assess the difference between this difference when the third party is rich vs poor, and when rank is made salient vs. when it is not.

In addition to computing those basic statistics, we will also identify parameters as in Xie et al (2017) by running regressions and interacting parameters in those regressions with our treatment variables (Salience / Wealth of third party). The likelihood to reject a transfer will be a function of the difference in perceived utility between accepting the transfer or not:

U_reject -U_accept =α x reversal+γ×transfer+β×final inequality+ϕ×original inequality

Based on this equation, we thus will be able to consider the impact of inequality and rank of the third party, and inequality and rank of the individual to whom the third party affiliates and/or compares with.

We elicit loss aversion by comparing decisions in a list of lotteries in the gain domain and decisions in a list of lotteries in a mixed domain (one outcome is a loss, the other is a gain). We will also measure social value orientation. Both those statistics work as controls for participants’ willingness to redistribute and sensitivity to reference points (gain, loss).

We will consider if own loss aversion makes third parties more unwilling to redistribute and to accept rank reversals. We will also consider if own social value orientation (competitive, altruistic, egoistic, inequality averse) interacts with decisions: competitive participants will be more concerned with own rank, those who are inequality averse will be more willing to reduce inequality even if this entails rank reversal).

We also elicit participants’ perceived closeness to other participants in the study, to control for how effective our manipulation of wealth was. We ask participants to give some feedback on the drivers of their decisions, their feeling of legitimacy in their role, and their view on the fairness of their remuneration. We use socio-demographic variable, including perceived position in society, and consider the impact of those variables on participants’ decisions.

Finally, we hypothesize a gender effect whereby:
• men are more competitive, so they will be more concerned with preservation and enhancement of their own rank.
• women are more loss averse, and more sensitive to the context, in particular the social environment, so they will be less willing to implement rank reversals.

Our design builds on Xie et al. (2017) experiment. We will run the experiment on MTurk. Participants decide on the allocation of bonus points for other MTurk workers. They take decision on whether to redistribute the points from other workers. We will have 24 menus, which are 6 variations on the basic menu in Xie et al. (2017), plus 4 menus that redistribute away from the poor, so as to elicit possible extreme preferences for inequality maintenance. Our key dependent variable is the choice whether to make the proposed income redistribution or not.

We will run the experiment on MTurk. Our design builds on Xie et al. (2017) experiment. Participants decide on the allocation of bonus points for other MTurk workers. They take decision on whether to redistribute the points from other workers.

We will have 24 menus, which are 6 variations on the basic menu in Xie et al. (2017), plus 4 menus that redistribute away from the poor, so as to elicit possible extreme preferences for inequality maintenance.

Our key dependent variable is the choice whether to make the proposed income redistribution or not.

computer

individual

No

1200 individuals

28 decisions on a range of redistribution options for each of 1200 individuals.

300 individuals per treatment

We will identify parameters as in Xie et al (2017) by running regressions and interacting parameters in those regressions with our treatment variables (Salience / Wealth of third party). The likelihood to reject a transfer will be a function of the difference in perceived utility between accepting the transfer or not: U_reject -U_accept =α x reversal+γ×transfer+β×final inequality+ϕ×original inequality Based on this equation, we thus will be able to consider the impact of inequality and rank of the third party, and inequality and rank of the individual to whom the third party affiliates and/or compares with. We wish to identify a difference of 5% in the rank-reversal coefficient, assuming baseline α=10%, γ=10%,β=5%,ϕ=5%. With 24 menus, then we need 300 participants per treatment to have power (1-p(Type 1 error)) more than 80%. This also guarantees us p(type 2 error) less than 5%. In total we will collect information on 1200 participants. The average payment per participant including show up fee is 5 Euros.