Risk Preferences and Outcome Bias in the Delegation Process

Last registered on May 21, 2021

Pre-Trial

Trial Information

General Information

Title
Risk Preferences and Outcome Bias in the Delegation Process
RCT ID
AEARCTR-0007691
Initial registration date
May 20, 2021
Last updated
May 21, 2021, 9:36 AM EDT

Locations

Region

Primary Investigator

Affiliation

Other Primary Investigator(s)

PI Affiliation
University of Mannheim
PI Affiliation
University of Mannheim

Additional Trial Information

Status
In development
Start date
2021-05-21
End date
2021-06-04
Secondary IDs
Abstract
As demonstrated by the COVID-19 pandemic, many political decisions involve a high level of uncertainty. Policy outcomes are subject to a variety of factors that lie beyond the scope of influence and cognition of the decision-makers and are thus – at least from the point of view of the politician – subject to chance. However, when evaluating the past performance of politicians, voters tend to focus on policy outcomes rather than on the political decisions and the circumstances under which they were taken, leading to less than optimal voting decisions. In this study, we examine this so-called outcome bias and its effect on the delegation process. In an online experiment, subjects assigned the role of a politician decide how much to invest in a risky policy, whereas subjects assigned the role of a voter decide whether to re-elect the politicians based on the politicians’ decision and the outcome. The treatments, which vary potential reward and risk of the policy as well as information about the outcome of the risk decision, allow us to derive causally whether the probability of success affects (I) voters’ preferences for political representatives in decisions taken under uncertainty, (II) the magnitude of voters’ outcome bias, and (III) politicians’ responsiveness to voters’ risk preferences and biases. Considering the outcome bias, it may be beneficial for politicians to disregard voters’ risk preferences and focus on the sheer probability of policy success. In this manner, the study provides novel insights into the role of risk and risk preferences in representative democracy.
External Link(s)

Registration Citation

Citation
Debus, Marc, Monika Mühlböck and Manuel Schwaninger. 2021. "Risk Preferences and Outcome Bias in the Delegation Process." AEA RCT Registry. May 21. https://doi.org/10.1257/rct.7691-1.0
Experimental Details

Interventions

Intervention(s)
When evaluating the past performance of politicians, voters tend to focus on policy outcomes rather than on the political decisions and the circumstances under which they were taken. In this study, we take this so-called outcome bias as a starting point and go one step further to examine the role of risk in political delegation processes. More specifically, we aim to understand on an abstract level whether (I) subjects are more likely to vote for representatives who take the same level of risk as they take for themselves, (II) their (re-)election decision is affected by the outcome bias even if it is clear that the outcome is only determined by chance, and (III) whether representatives adapt to the risk preferences of the represented and anticipate the potential outcome bias among them.

To answer our research questions, we employ a 2x2 between-subject design, where we systematically vary the immediacy of the outcome as well as the lottery of the risky policy in an online experiment (see details under 'Experimental Design' below).

Intervention 1: Immediacy of the outcome. In the 'long-term' treatment, the allocation decision takes effect in the long run, and voters get informed about the outcome of the first terms’ lottery only after their voting decision. In the 'short-term' treatment, the allocation decision takes effect in the short run, and voters get informed about the outcome of the first terms’ lottery before their voting decision. This implies that in the long-term treatment, the voting decision can be affected by the politicians’ allocation decision (but not by the outcome of the first term’s lottery, because it is still unknown), whereas in the short-term treatment, the voting decision can be affected by the politicians’ allocation decision and the outcome of the first terms’ lottery. This enables us to measure the causal influence of the outcome on the voting decision of the represented. Furthermore, it enables us to assess whether politicians anticipate an outcome bias among voters and act differently under the short-term treatment than under the long-term treatment.

Intervention 2: Lottery of the risky policy. The safe policy always has a return of 1 and the return of the risky policy in case of a downswing is always 0. In the 'high risk-reward' treatment, the probability of an upswing is 1/3 and the return 3.5 times the allocated points, whereas the probability of a downswing is 2/3. In the 'low risk-reward' treatment, the probability of an upswing is 2/3 and the return 1.75, whereas the probability of a downswing is 1/3. Note that the expected payoff of the risky policy is in both cases 1.16 times the allocated points. The two treatments enable us to infer via a difference-in-difference analysis whether the lottery affects the allocation decision of politicians due to the anticipated outcome bias.
Intervention Start Date
2021-05-21
Intervention End Date
2021-06-04

Primary Outcomes

Primary Outcomes (end points)
Through the experimental design, we obtain two main variables of interest. First, we can observe the representatives’ allocation decision between the safe and risky policy. Second, we observe the voting decision of the represented.
Primary Outcomes (explanation)

Secondary Outcomes

Secondary Outcomes (end points)
Secondary Outcomes (explanation)

Experimental Design

Experimental Design
We programme an online experiment using otree (Chen, Schonger, and Wickens 2016). In the first and second part of the experiment, we elicit individuals’ risk preferences through allocation tasks similar to the one developed by Gneezy and Potters (1997). In these tasks, subjects have to allocate a certain amount of points between two policies, called ‘safe’ and ‘risky’ policy, which differ regarding their payoffs in case of two possible outcomes, which are called ‘upswing’ and ‘downswing’. The risky policy gives a higher return in case of an upswing and no return in case of a downswing, whereas the safe policy gives the same intermediate return in case of an upswing or downswing. The difference between the first and the second part of the experiment is that in the first part, subjects need to take one single allocation decision which takes effect for two consecutive terms (in each of which an upswing or a downswing may occur), while in the second part, subjects take two (potentially different) allocation decisions, one in the first term and one in the second term.

In the third and main part of the experiment, we randomly assign participants to the roles of politicians (incumbents and challengers) and voters. Subjects are assigned to groups of three: one voter, one incumbent, and one challenger. Both politicians must decide how to allocate the same amount of points as in the previous parts between the two policies on behalf of the voter for two consecutive terms. Additionally, we ask them which risk preferences they expect the voter to have. In the first term, the allocation decision of the incumbent is implemented. After the first term, the voter is informed about the politicians’ decision and then must decide whether to re-elect the incumbent or to elect the challenger for the second term. The politicians’ payoff depends on the voter’s voting decision. If a politician is elected for the second term, she earns a fixed payoff, otherwise, she does not receive any payoff in this part of the experiment. The voter’s payoff depends on the decisive politician’s allocation decision and the outcome of the lottery in each term.

In a final part, we gather information on personal characteristics of the subjects and ask about general and specific preferences regarding risk decisions and political representation.
Experimental Design Details
Randomization Method
randomization done in office by a computer
Randomization Unit
experimental sessions (between-subject design)
Was the treatment clustered?
No

Experiment Characteristics

Sample size: planned number of clusters
160 groups
Sample size: planned number of observations
480 subjects, 3 per group
Sample size (or number of clusters) by treatment arms
40 groups per treatment, á 40 incumbents, 40 challenger and 40 voters
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
IRB

Institutional Review Boards (IRBs)

IRB Name
GfeW (Gesellschaft für experimentelle Wirtschaftsforschung e. V.)
IRB Approval Date
2021-05-19
IRB Approval Number
jeeMHER2

Post-Trial

Post Trial Information

Study Withdrawal

There are documents in this trial unavailable to the public. Use the button below to request access to this information.

Request Information

Intervention

Is the intervention completed?
No
Data Collection Complete
Data Publication

Data Publication

Is public data available?
No

Program Files

Program Files
Reports, Papers & Other Materials

Relevant Paper(s)

Reports & Other Materials