Privatizing gains and socializing risks: Does responsibility affect risky decision making for others?

2021-07-05

2021-12-05

The key outcome variable is the amount invested, i.e. either 0, 1000, 2000, ... 10,000. We test whether this one varies with the group size (within-subjects) and compare it across treatments.

We will estimate the functional relationship between group size and the amount invested. Economic theory would have no prediction in treatment FLAT and full investment in treatment CONVEX. However, due to a responsibility effect, we expect a change of investment levels with an increase in group size in CONVEX and less so in FLAT. We further test for the order effect but have no a-priory prediction.

In addition, we elicit subjects' characteristics that have been shown to influence risk-taking and elicit risk preferences using the Dohmen et al (2011) risk elicitation task,

We will expand the estimated functional relationship between group size and the amount invested by including subjects' characteristics to control for these when estimating the effect of the group size and to establish the impact of these effects on their own and their effect on the functional shape of the impact of group size.

Hidden

We answer the research question – Does the number of principals affect the agent’s decisions for the principals? – by considering a principle-agent relationship in which we ask subjects to take investment decisions for a varying number of principals. We consider a within-subject design where subjects decide varying group sizes (1,5,10,25,50,100). We consider a between-subject design for 1) the order of group sizes, i.e. either starting 1, 5,… (LOW) or 100, 50,… (HIGH) and 2) the incentive structure, i.e. either FLAT or CONVEX. Hence, we have the treatments LF, HF, LC, and HC. In FLAT, the agents get a flat payment. In CONVEX, the agents earn the flat payment or a proportional bonus when the investment was successful. Hence, we consider a classical 2x2 factorial design.

We consider a situation, in which an agent (A) invests the monetary endowments of n principals (P) in a high-risk project with negative expected returns. We amend the investment game introduced by Gneezy and Potters (1997): The N principals are endowed with identical amounts E=10,000 ECU. The agent invests the same amount X (between 0 and E) for each of the four principals. With a probability of 1/6 the project is successful and returns 4.5 times the investment. With a probability of 5/6, however, the investment is lost. The principals keep all money that is not invested. Note that any amount invested has a negative expected return and therefore a risk-averse investor would not invest. The agents do not participate in the gains or losses of the investment.

Randomization will be implemented as an algorithm.

We invite people from Prolific and randomly apply treatments in equal size until we get the number of observations.

No

No clusters.

204 observations for each treatment, four treatments in total

204 observations for each treatment, four treatments in total

Comparing two treatments with alpha=0.05 and power of 0.80 the detectable effect size is 0.2780 (G*Power 3.1.9.7 - Sensitivity for t-tests)