A National Survey of State Legislators' Views on Health Infrastructure

Legislators will be assigned to different compensation amounts for completing the survey and to different follow-up protocols, ranging from no reminder to multiple reminder emails and, for some, a phone call. Respondents will also be assigned to one of four messages about health infrastructure that vary the argument presented and the information source. These interventions will be used to study survey participation, nonresponse bias, and how message framing affects legislators' views.

2026-07-01

2026-10-31

Survey completion.
Priority assigned to subsidizing struggling rural hospitals relative to other state priorities.
Support for subsidizing struggling rural hospitals if doing so required raising taxes.
Preferred targeting of rural hospital subsidies.
Perceived persuasiveness of arguments for and against subsidies for rural hospitals.

The primary response outcome is an indicator equal to one if a sampled legislator completes the survey. The primary vignette outcomes are measured after respondents view a randomized message about rural hospitals. These outcomes include the respondent's stated priority for subsidizing struggling rural hospitals, support or opposition to subsidies if they require raising taxes, and preferences over whether subsidies should be limited to the most financially distressed hospitals, extended to most rural hospitals with eligibility criteria, or provided to all rural hospitals regardless of financial status. Persuasiveness outcomes will be measured using respondents’ ratings of statements about rural hospital subsidies, including statements emphasizing health equity, local need, voter priorities, and concerns about cost control and efficiency.

Partial survey completion.
Time to response.
Item nonresponse.
Survey completion time.
Ranking of health care, state budget deficits, public safety, economic opportunities, and rural quality of life as state priorities.
Perceived health of district residents relative to the nation.
Perceived adequacy of district health infrastructure.
Beliefs about hospital financial challenges in the district.
Familiarity with the Rural Health Transformation Program.
Rural health infrastructure funding priorities.
Views of Medicaid and state responses to potential federal Medicaid cuts.
Trust in and reliance on different information sources.
Preferred types of evidence for health infrastructure decisions.
Nonresponse-adjusted estimates of survey means and vignette treatment effects.

Secondary outcomes will be constructed from survey responses and survey paradata. Attitudinal outcomes include respondents’ views on district health, local health infrastructure, hospital financial challenges, rural health funding priorities, Medicaid, and sources of health policy information. Evidence and trust outcomes will be based on items asking which sources respondents trust and rely on, including state health agencies, federal health agencies, local hospitals and providers, universities, news media, social media, advocacy groups, and other sources

This study is a randomized survey experiment of U.S. state legislators. Legislators will be randomly assigned to different survey outreach strategies that vary compensation and follow-up intensity. Respondents will also be randomly assigned to brief informational messages about rural hospital challenges. The study will estimate how outreach strategies affect survey participation, whether harder-to-reach respondents differ from other respondents, and how different messages affect views on rural health infrastructure.

Not available

Randomization will be conducted by computer using reproducible Stata code. Compensation amount and follow-up intensity will be randomized separately within state-party strata. The vignette experiment will be randomized by computer within the survey platform or pre-assigned in the contact-level randomization file.

The unit of randomization is the individual state legislator. Compensation amount and follow-up intensity are randomized at the individual-legislator level. Among survey respondents, vignette assignment is also randomized at the individual level.

No

Approximately 7,000 individual state legislators. Because treatment is not clustered, the number of randomized units is the same as the number of planned observations.

Approximately 7,000 state legislators.

Approximately 7,000 state legislators will be randomized.

Compensation arms:

$0: approximately 1,400 legislators
$25: approximately 1,400 legislators
$50: approximately 1,400 legislators
$75: approximately 1,400 legislators
$100: approximately 1,400 legislators

Follow-up arms:

No follow-up email: approximately 1,680 legislators
One follow-up email: approximately 1,680 legislators
Two follow-up emails: approximately 1,680 legislators
Three follow-up emails: approximately 1,680 legislators
Three follow-up emails plus one phone call: approximately 280 legislators

Vignette arms among respondents:

Access argument from a news source: approximately 25% of respondents
Access argument from a university source: approximately 25% of respondents
Economic argument from a news source: approximately 25% of respondents
Economic argument from a university source: approximately 25% of respondents

The survey will be fielded in three waves: an initial pilot wave of approximately 100 legislators, a second wave of approximately 2,000 legislators, and a final wave consisting of the remaining sampled legislators, subject to financial constraints.

For the survey outreach experiment, the planned sample includes approximately 7,000 state legislators. For compensation arms, with approximately 1,400 legislators per arm, a two-sided 5% significance level, 80% power, and a baseline response rate of 15%, the minimum detectable difference for a pairwise comparison between two compensation arms is approximately 3.8 percentage points. For comparisons between one of the four larger follow-up arms, with approximately 1,680 legislators per arm, the minimum detectable difference is approximately 3.5 percentage points. For comparisons involving the smaller phone-call arm, with approximately 280 legislators, the minimum detectable difference is approximately 6 percentage points, depending on the comparison group. Power for the vignette experiment depends on the realized survey response rate. If the response rate is between 15% and 25%, the expected number of respondents will be approximately 1,050 to 1,750. With equal assignment across the four vignette arms, this implies approximately 260 to 440 respondents per vignette arm. For the main effects in the 2 × 2 vignette design, this provides 80% power at the 5% significance level to detect effects of approximately approximately 8.6 percentage points for binary outcomes with a mean near 0.50. Detectable effects for pairwise comparisons between individual vignette arms and for interaction effects will be larger. These calculations are approximate and do not account for precision gains from covariate adjustment or randomization-strata fixed effects.