Energy Transition and Fiscal Preferences: A Survey Experiment

We run an online survey experiment through Prolific that targets United States residents above the age of eighteen. The survey is designed in Qualtrics and will be administered on Prolific.

The survey has five sections (1) demographics and socio-economic background, (2) experiment 1 (Out-Migration), (3) experiment 2 (In-migration to Laramie, WY.), (4) experiment 3 (measuring loss aversion), and (5) choice of fiscal policy to counter fossil-fuel revenue decline,

In the first experiment, respondents residing in one of the ten resource-rich states (Alaska, Colorado, Kansas, Louisiana, Montana, New Mexico, North Dakota, Oklahoma, Texas, and Wyoming) are randomly assigned to one of seven scenarios and asked about their likelihood of moving out of their current state over the next three years. In the baseline scenario, respondents are simply asked about their likelihood of moving to a different state in the next three years. In the six treatment scenarios, respondents are told to imagine a 10% reduction in state government revenue due to a fall in fossil-fuel production, and that their state government responds with a different fiscal policy in each treatment: (i) raising the average property tax rate; (ii) raising the average sales tax rate; (iii) raising the average individual income tax rate; (iv) cutting public spending on education; (v) cutting public spending on healthcare; and (vi) cutting public spending on infrastructure.

In the second experiment, all respondents from all states are asked about their willingness to move to Laramie, Wyoming under varying informational treatments. In the control group, subjects are given basic innocuous information about Laramie, WY. (e.g., that it is a college town and that it is at an elevation of 7,200 ft). Treatments will provide subjects with one of three pieces of additional information about Laramie, WY., depending on treatment: i) tax policy, ii) socio-economic conditions, or iii) natural amenities. Further, each of the treatments and the control will be presented with either i) a real image of downtown Laramie, WY., or ii) a modified one that enhanced place-making modifications (including things like urban greenery and pedestrian friendly design).

The third experiment elicits respondents’ loss aversion with respect to tax increases versus spending cuts. Subjects are told to imagine that their state government experiences a 10% change in federal funding. In two “loss” scenarios, the state receives 10% less funding and balances its budget either (i) by increasing the respondent’s average tax rate, or (ii) by reducing public spending. In two symmetric “gain” scenarios, the state receives 10% more funding and responds either (iii) by reducing the respondent’s average tax rate, or (iv) by increasing public spending. Comparing evaluations across these gain and loss scenarios provides an individual level measure of loss aversion. This experiment is included so that we can use these responses to parameterize a measure of loss aversion and use it to calibrate a separate theoretical model (not included in the survey experiment paper) examining how reference-dependent preferences may shape policymakers’ fiscal decisions in fossil-fuel-dependent states.

The three experiments are shown in a randomized order to control for order effects.

2025-12-09

2025-12-10

Depending on the experiment, the outcome variable with be either:
1. Willingness to leave your current state of residence measured as a self-reported score from 0-100.
2. Willingness to move to Laramie measured as a self-reported score from 0-100.
3. Life satisfaction from moving to, and living in, Laramie, Wyoming measured as a self-reported score from 0-100.
4. (Log) time subjects spend voluntarily collecting additional information about Laramie.
5. Number of total clicks subjects make when collecting additional information about Laramie.
6. Whether subjects click on specific links to gather information about i) schools, ii) job opening, iii) housing, iv) general information in Laramie.
7. Loss aversion parameter for tax raise = mean score for tax increase / mean score for tax decrease = (loss)/(gain).
8. Loss aversion parameter for spending cut = mean score for spending cut / mean score for spending raise = (loss)/(gain).

Most of the primary outcome variables are directly measured. Time taken reading about Laramie with be transformed using the natural log function. Loss aversion will be measured as a ratio of loss response over gain response.

We will estimate heterogeneous treatment effects by interacting the treatment indicators with the following pre-specified characteristics:
1) the presence of dependents below the age of eighteen, 2) age, 3) income, 4) employment status, 5) political ideology (measured as a binary variable derived from a left/liberal and right/conservative Likert scale), voting behavior in the 2024 Presidential election, 6) the subject stated mobility (ranging from rooted, stuck, to mobile), and 7) respondents’ state of residence

For respondents in fossil-fuel-rich states, we will examine whether treatment effects vary with: 1) indicator for direct or indirect employment in the fossil-fuel industry, 2) fiscal policy preference (tax raise or spending cut) in the event of a fossil-fuel revenue decline, 3) updated fiscal preference after exposure to different categories of tax and spending options, (4) preferences over the type of taxes should increase or which income-groups should face the tax-burden, and (5) preferences over which public service categories should face spending cuts.

In the Laramie-specific experiment, we will additionally explore heterogeneities based on 1) whether subject currently resides in a top-10 fossil-fuel-rich state or in any other state, 2) resides in a state neighboring Wyoming, 3) intends to pursue higher education (given that the general information mentions the University of Wyoming), 4) holds a job that allows fully remote work, and 5) reports entrepreneurial intent, 6) is familiar with Laramie.

We will examine treatment effects for the full sample and for the subset of i) respondents who self-report that they are mobile, ii) residents in the neighboring states of Wyoming (for the experiment about moving to Laramie)

The survey uses a between-subjects experimental design. This study has two online surveys administered through Prolific. (i) an “oil-rich states” survey administered only to respondents currently residing in the top ten fossil-fuel dependent states (Alaska, Colorado, Kansas, Louisiana, Montana, New Mexico, North Dakota, Oklahoma, Texas, and Wyoming), and (ii) an “oil-poor-states” survey administered to respondents from the remaining states. The screening for the surveys will be conducted using Prolific’s pre-screening tool to restrict participation to residents of the designated states. State-of-residence will be verified again at the start of the survey.

The two surveys share common demographic and background questions, but have different experimental modules. The experiment about out-migration is only administered in the oil-rich survey, and the experiments about in-migration and measuring loss aversion are distributed in both the surveys.

In the first experiment, respondents living in the top ten fossil-fuel dependent states are randomly assigned to one of three hypothetical scenarios describing how their state might respond to a large decline in fossil-fuel revenue. After reading the assigned scenario, respondents report how likely they would be to move to a different state in the next three years. The scenarios are:

Baseline: Respondents are simply asked how likely they are to move to a different state in the next three years.

Treatment 1 (Property Tax Increase) : Respondents are told to imagine a 10% reduction in state government revenue due to a fall in fossil-fuel production, and that their state government responds by mandating a rise in the average local property tax rate.

Treatment 2 (Sales Tax Increase) : Respondents are told to imagine a 10% reduction in state government revenue due to a fall in fossil-fuel production, and that their state government responds by increasing the average sales tax rate.

Treatment 3 (Individual Income Tax Increase) : Respondents are told to imagine a 10% reduction in state government revenue due to a fall in fossil-fuel production, and that their state government responds by increasing the average individual income tax rate.

Treatment 4 (Spending Cut on Education) : Respondents are told to imagine the same 10% revenue decline, but that their state government responds by cutting public spending on education.

Treatment 5 (Spending Cut on Healthcare) : Respondents are told to imagine the same 10% revenue decline, but that their state government responds by cutting public spending on healthcare.

Treatment 6 (Spending Cut on Infrastructure) : Respondents are told to imagine the same 10% revenue decline, but that their state government responds by cutting public spending on infrastructure.

All respondents, regardless of state of residence, participate in a second experiment that elicits their willingness to move to Laramie, Wyoming under different informational and visual treatments. Each respondent is shown images of downtown Laramie either original images or modified images that visually incorporate placemaking improvements (e.g., added urban greenery, pedestrian-friendly streetscape design). After viewing the assigned information and images, respondents report how likely they would be to move to Laramie in the next three years.

The experiment consists of seven scenarios:
Baseline (General Information - Original Images): Respondents are shown general information about Laramie along with original photographs.

Treatment 1 (Tax Environment - Original Images) : Along with same general information as the baseline scenario, respondents receive additional information describing Laramie’s tax environment (e.g., absence of state income tax, relative tax burden). They are shown original images.

Treatment 2 (Tax Environment - Enhanced Images) : Identical informational treatment as in treatment 1, but respondents are shown visually enhanced images with placemaking improvements.

Treatment 3 (Socio-Economic Factors - Original Images) : Along with same general information as the baseline scenario, respondents receive information about cost of living, unemployment rates, safety, and community amenities in Laramie, accompanied by original downtown images.

Treatment 4 (Socio-Economic Factors - Enhanced Images): Identical socio-economic information as in treatment 3, but respondents are shown visually enhanced images with placemaking improvements.

Treatment 5 (Natural Amenities - Scenic Image) : Along with same general information as the baseline scenario, respondents receive information about Laramie's outdoor amenities, accompanied by a photograph of the mountain landscape surrounding Laramie

Treatment 6 (General Information - Enhanced Images): Respondents are shown general information about Laramie along visually enhanced images of downtown Laramie with placemaking improvements.

We will estimate treatment effects by regressing respondents’ self-reported scores on indicators for the treatment scenario they were randomly assigned to. We report results both without controls and with a set of demographic and socioeconomic controls. Outcome variables will be treated as continuous variables in baseline regressions. In conditional regressions, covariates include subject i) age, 2) gender, 3) dependents below the age of eighteen, 4) current living arrangement (owns their home, rents, or stays with friends and family), 5) household income, 6) employment status, 7) current work arrangement (fully remote, fully in-person or hybrid), 8) indicator for direct or indirect employment in the fossil-fuel industry, 9) entrepreneurial intent, 10) education, 11) intent to pursue higher education, and12) political ideology.

The third experiment is also shown to all respondents regardless of their state of residence. This experiment elicits the respondents loss aversion to tax raises over spending cuts. In all scenarios, subjects are told to imagine that their state government experiences a 10% change in funding from the federal government. In two “loss” scenarios,

Scenario 1 (Tax Increase) - the state receives 10% less funding than usual and balances its budget by increasing the respondent’s average tax rate

Scenario 2 (Spending Decrease) - the state receives 10% less funding than usual and balances its budget by reducing public spending

Scenario 3 (Tax Decrease) - the state receives 10% more funding than usual and responds either by reducing the respondent’s average tax rate

Scenario 4 (Spending Increase) - the state receives 10% more funding than usual and responds either by increasing public spending

Comparing respondents’ evaluations across these gain and loss scenarios will provide an individual level measure of loss aversion.

These three experiments are shown in a random order to the respondents. Suppose experiment 1 is A, experiment 2 is B and experiment 3 is C. For the oil rich survey, there can be 6 orders - ABC, ACB, BAC, CAB, BCA, CBA. For the oil poor survey, there can be two orders - AB, BA.

Randomization using Qualtrics.

Individual

No

NA

Oil Rich Survey: 1500 Prolific users Oil Poor Survey: 1500 Prolific users

Oil Rich Survey (Experiment 1 + Experiment 2 + Experiment 3): For experiment 1 and 2, we will have 214 respondents in each treatment arm. For experiment 3, we will have roughly 375 respondents in each arm.

Oil Poor Survey (Experiment 2 + Experiment 3): For experiment 2, we will have roughly 214 respondents in each arm. For experiment 3, we will have roughly 375 respondents in each arm.