The Electoral Effects of Green Industrial Policies
How to cite
Maissen, Patricia; Fesenfeld, Lukas. 2026. "The Electoral Effects of Green Industrial Policies." AEA RCT Registry. AEARCTR-0017636. https://doi.org/10.1257/rct.17636-1.0
Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.
* Primary investigator
Abstract
This pre-registration describes a survey experiment designed to complement a qua-si-experimental difference-in-difference analysis examining the electoral consequences of green industrial investments under the U.S. Inflation Reduction Act (IRA), distinguishing between investments into renewable energy (RE) deployment (e.g., wind farms, solar farms) and investments into low-carbon technology manufacturing (e.g., EV plants, battery plants). The main electoral channel is expected to operate through symbolic (identity-based) and material (economic) channels. The survey experiment is designed to test these mechanistic pathways.
Intervention
Respondents will be randomized into different groups. Randomization occurs in three stages:
• Stage 1 (domain level): Each respondent is randomly assigned to either the RE Deployment arm or the Low-carbon Technology Manufacturing arm (50:50 split).
• Stage 2 (framing level): Within each domain, respondents are randomly as-signed to control group, neutral framing group or partisan framing group (33:33:33 split).
• Stage 3 (technology level): Within each treated arm, respondents are ran-domly assigned to one of two possible technologies (50:50 split):
o RE deployment arms: solar power park vs. wind power park
o Low-carbon Technology Manufacturing arms: EV manufacturing plant vs. battery manufacturing plant
All respondents within a domain are presented with a general information (“Across the United States, new clean [energy generation/technology manufacturing] pro-jects have been built or announced in many areas."). In addition, treated individuals are informed about a recent, salient project of their respective technology in their state. The treated respondents are informed with a short text, which includes the specific project, as well as a mock newspaper article. The mock newspaper article differs slightly depending on (a) the respondent’s framing arm (neutral vs. partisan), (b) the technology they were sampled into and (c) the state they live in.
Here is an example for the treatment information about the technology, here a wind power park (Ohio), shown before the neutral or partisan-framed newspaper article:
“Across the United States, new clean energy generation projects have been built or announced in many areas, such as wind power parks.
In Ohio, such parks have recently been opened. For example, the Martin Marietta Wind Project is located in Sandusky County. Wind power parks have also been built in other counties in Ohio.
Here is an illustrative example of media reporting on such projects:”
After this technology and project information, dependent on the randomized treatment arm and respondent’s state, mock newspaper articles are shown.
Randomization is conducted by the survey platform using simple random assign-ment without stratification. Randomization is independent of all measured covari-ates. We will verify balance on key pre-treatment covariates (e.g. age, gender, edu-cation, party identification) in the final data and report any imbalances in the analysis.
Experimental design
The full sample is randomly divided into two parallel survey arms, each focusing on one investment domain:
• Domain 1: RE Deployment Projects. Respondents answer questions about RE deployment projects (e.g., wind power parks, solar power parks).
• Domain 2: Low-carbon Technology Manufacturing Projects. Respondents answer questions about low-carbon technology manufacturing projects (e.g., EV manufacturing plants, battery manufacturing plants).
Within the domain, respondents in the treatment group get an information treat-ment—with a neutral or a partisan framing—on different domain technologies (cf. “Interventions”), but the questions before and after the treatment are the same for everyone within the respective domain (e.g., “Overall, how much do you support or oppose clean energy generation projects?”, and not “Overall, how much do you support or oppose solar energy generation projects?”). Between the two domains, the questions are identical except for the domain mentioned as well as the specific technologies mentioned in individual questions (e.g., differing project types for the question “How much do you support or oppose the following specific types of clean [energy generation/technology manufacturing] projects?”).
Before treatment occurs, respondents are asked about their demographic character-istics; political preferences and past voting behaviors; salience of, opinion on and perceived exposure to the projects from the respective domain.
In the primary analysis, the two respective technologies will be analyzed at the do-main level (RE deployment/low-carbon technology manufacturing), and not at the technology level (solar, wind, EV, battery). Additional explorative analyses allow us to investigate specific technology-level differences.
Design details
Cited by
No citation data available for this registration yet.