The Role of Status Quo Bias and Emotions in Elections

We will use the online-work platform Amazon Mechanical Turk in conjunction with Cloud Research to implement an incentivized common value election game. Participants are grouped with other participants in an electorate. They are first asked to perform a simple real-effort task in which they have to add up three sums.

Each sum that the participants see is drawn randomly by the computer from a larger list of 1000 sums. There are two types of sums on this list: there are sums that have an even answer and there are sums that have an odd answer. Exactly 550 of the sums have an answer that is one type and the other 450 have an answer that is the alternative type. However, the participants do not know which type of sum is more common. Nonetheless, the three sums that the participants receive in the first task can be interpreted as an indicator of the true underlying distribution of the larger list, and thus the first task also acts as an informative signal. The participants are told they will receive a payment for each sum they add correctly, and that this payment will depend on a particular payment plan, which pays more for either odd or even correct answers. In the baseline condition, the participants are not told which payment plan will be applied in the first task, although they are told what the two possibilities are. However, in the status quo bias treatment, they are told that they are being paid according to a specific pre-determined payment plan which establishes that payment plan as the status-quo default.

Following the completion of the first real-effort task, emotion inductions are conducted, which entail watching a three-minute video that is either neutral, sadness-, or anxiety-inducing. The participants are then told they will be asked to perform second real-effort task. This time the participants add up seven sums but apart from that, everything else is identical to the first task. However, this time participants are given the option to vote on which of the two payment plans will determine their payment for the second task. As mentioned above, the number of even and odd sums received in the first task may act as an informative signal to participants in this voting stage because the underlying distribution or the larger list remains unchanged throughout the entire experiment. In all treatments, the collective vote of all the group members then determines which payment plan is applied for the second task.

The election is followed by incentivized tasks to measure risk aversion and loss aversion (standard Holt and Laury tasks). Overall, we are planning on nine conditions using a three-by-three design. We vary baseline, status quo bias, and status quo bias (with the second payment plan as the status quo) treatments across three emotion inductions: neutral (baseline), sadness, and anxiety.

2024-10-07

2024-10-31

(1) Probability of voting for the status quo default option by status quo frame treatment and (2) probability of voting for the status quo default option by emotion induction.

Primary outcome (1) will be the predicted probability of voting for the status quo default by each treatment (this will be known as the status quo effect). For this we will test observed behaviour (voting for the status quo default) using a one-tailed comparison of proportions across treatment groups.

Primary outcome (2) will be the predicted change in proportions voting for the status quo default by each emotion induction. For this we will take observed behaviour (voting for the status quo default) and estimate a difference-in-difference model, where the coefficient of the interaction term between the status quo effect (mentioned above) and the emotion inductions will be the variable of interest.

Predicted probability of voting for the status quo default in relation to loss aversion and risk aversion.

For this we will take observed behaviour (voting for the status quo default) and the results of the loss aversion and risk aversion tasks and estimate a linear probability model or a logit or probit model.

Our nine proposed conditions involve varying baseline, status quo bias, and status quo bias (with the other payment plan as the status quo) treatments across three emotion inductions: neutral (baseline), sadness, and anxiety. By using status quo bias treatments, we explore if status quo bias has an impact on common value elections in an incentivized economic game context (an area that has been previously unexplored). By varying the emotion inductions, we explore if particular negative emotional states increase this status quo bias effect.

In the status quo bias treatments, participants are told that a certain payment plan is applied for their first real-effort task. Then in the voting game, they can either vote to keep or change the plan (or to abstain). This framing reinforces the pre-determined payment plan as the status quo. The literature suggests that such framing will increase voting preferences for the status quo option. We also introduce another status quo treatment where the status quo default is reversed (i.e., a different payment plan is established as the status quo default) in order to control for any potential unexpected asymmetries. In the baseline condition, the same voting options are provided but without any such status quo framing. Furthermore, participants are not told which of the two payment plans was applied in the first task. This creates a voting environment devoid of a status quo.

In the emotion induction treatments, participants are shown short clips from movies which are designed to induce specific emotions immediately prior to voting. Showing short film clips is an established method for inducing emotions from the field of experimental economics and experimental psychology. Following the previous literature, a movie clip depicting a boy’s mentor dying from the movie ‘The Champ’ is used to induce sadness, a scary scene from the movie ‘Silence of the Lambs’ is used to induce anxiety, and a weather news report was used to induce a neutral emotion for the baseline condition. Standard mood and anxiety Likert scales are then used to collect self-reported emotional states from participants following the movie-clips.

We use fully automatic randomization via the survey tool Qualtrics. When a participant clicks on our listing in mTurk then Qualtrics automatically assigns the participant to a treatment at random.

Randomization on the individual level

No

2160 individuals

2160 individuals

240 individuals per treatment. We have a 3 x 3 design (nine treatments).

We aim to collect 240 in each of the 9 treatments because in simulation that gives us 80% power to detect a difference-jn-difference (DiD) effect size with odds ratio of 3 for our primary outcome at the 5% significance level, using a linear probability regression model. Our primary DiD effect of interest is the difference in the status quo bias (i.e. the increase in rates of voting for an option when it is established and framed as the status quo) after emotional induction. For the purposes of this simulation, we assume that the baseline rate of voting for the first payment plan is 54% and the rate when it framed as the status quo is 73%, based on the outcome of a small (n=80) pilot study already completed.