EV or not EV: Information Frictions and Identity as Drivers of Vehicle Demand

Experiment 1:
Prior beliefs about the ZIP-specific emissions associated with several popular EV brands are elicited. Participants are then randomly divided into three groups. Group 1 is assigned information relaying the emissions of a 2024 Tesla Model 3 and the average EV. Group 2 is assigned information relaying the emissions of a similarly efficient 2024 Lucid Air Pure and the average EV. Group 3 is assigned information on price for the 2024 Lucid Air Pure and the average EV. Before each group sees the information, willingness-to-pay (WTP) to see their assigned piece of information is measured using a multiple price list. From here, the participants are shown the information they were assigned and then posterior beliefs are elicited.

Experiment 2:
Prior beliefs about the ZIP-specific emissions associated with an average EV are elicited. Participants are randomly shown either a placebo quote or treatment quotes which vary the strength of desire to be a motivated reasoner. For liberals, the treatment quote emphasizes job gains from EVs. For conservatives, the treatment quote emphasizes job losses from EVs. Moderates are randomly divided between the two treatment quotes with 50% probability. The placebo quote is unrelated to EV jobs, instead focusing on the role of socioeconomic status in driving purchases. The placebo quote does not vary by political ideology. Each participant is then randomly assigned to one of two pieces of information. For half, this information is about commuting times. For the other half, it is the true value of average EV emissions in their ZIP code. Before each group sees the information, willingness-to-pay (WTP) to see their assigned piece of information is measured using a multiple price list. From here, the participants are shown the information they were assigned and then posterior beliefs are elicited.

2025-06-28

2025-07-31

For both experiments, the outcomes of interest are the WTP to see assigned information and total belief updating upon seeing it (distance between measured prior and posterior). Prior beliefs will also be correlated with demographics and used in a demand estimation exercise using observational registration data.

Experiment 1:
We propose to collect data via a survey administered on Prolific. The survey first determines political leanings and car ownership of all participants. From here, the survey elicits participants' prior beliefs on the emissions from several EV brands, including Teslas. This allows for a baseline measurement of how informed consumers of different political beliefs are on the relative cleanliness of different EVs. These priors are correlated with demographics for use in a demand estimation exercise. Next, participants are randomly divided into three groups. Group 1 is assigned information relaying the emissions of a 2024 Tesla Model 3 and the average EV. Group 2 is assigned information relaying the emissions of a similarly efficient 2024 Lucid Air Pure and the average EV. Group 3 is assigned information on price for the 2024 Lucid Air Pure and the average EV. Before each group sees the information, willingness-to-pay (WTP) to see their assigned piece of information is measured using a multiple price list. From here, the participants are shown the information they were assigned and then posterior beliefs are elicited.

Experiment 2:
We propose to collect data via a second survey administered on Prolific. This survey is distinct from the first and disallows participants from experiment 1 from participating in experiment 2. The survey first determines political leanings and car ownership of all participants. From here, the survey elicits participants' prior beliefs on the emissions from an average EV versus an average GV. This allows for a baseline measurement of how informed consumers of different political beliefs are on the relative cleanliness of EVs relative to GVs, on average. These priors are correlated with demographics for use in a demand estimation exercise. Note, additionally, that these priors are specific to a participant's ZIP code due to electricity grids across the US having varying emissions intensity. Participants are then randomly shown either a placebo quote or treatment quotes which vary the strength of desire to be a motivated reasoner. For liberals, the treatment quote emphasizes job gains from EVs. For conservatives, the treatment quote emphasizes job losses from EVs. Moderates are randomly divided between the two treatment quotes with 50% probability. The placebo quote is unrelated to EV jobs, instead focusing on the role of socioeconomic status in driving purchases. The placebo quote does not vary by political ideology. Each participant is then randomly assigned to one of two pieces of information. For half, this information is about commuting times. For the other half, it is the true value of average EV emissions in their ZIP code. Before each group sees the information, willingness-to-pay (WTP) to see their assigned piece of information is measured using a multiple price list. From here, the participants are shown the information they were assigned and then posterior beliefs are elicited.

Randomization will be done using Qualtrics' internal algorithms.

Individual level randomization.

No

2,800 individuals (1,200 in experiment 1; 1,600 in experiment 2)

2,800 individuals (1,200 in experiment 1; 1,600 in experiment 2)

Experiment 1: 1,200 individuals divided evenly by treatment (3 treatments) = 400 individuals per arm
Experiment 2: 1,600 individuals divided evenly by treatment (4 treatments) = 400 individuals per arm