Heller, Blake; Lyon, Melissa.
2025. "Ballot Language and School Bond Support." AEA RCT Registry. AEARCTR-0015329.
https://doi.org/10.1257/rct.15329-1.0
Download:
BibTeX
·
RIS— for reference managers without a browser
translator. Zotero and most
others pick up this page's citation metadata directly.
This study aims to understand how ballot language influences public support for school bonds. We are specifically interested in understanding if adding language that increases the salience of property tax obligations to hypothetical school bond proposals influences stated support for school bonds and general investment in public schools in a national sample of survey respondents. Topics and questions to be explored include:
RQ1: Does an explicit reference to property tax obligations in school bond referenda impact individuals' willingness to support increased in public investment in public schools?
H1: Individuals who are presented with school bond proposals that include explicit references to property tax obligations are less likely to support school bond proposals compared to those who are presented with the same bond proposals with no reference to property taxes.
H2: Individuals who are presented with school bond proposals that include explicit references to property tax obligations are less likely to support increasing general investments in public schools compared to those who are presented with the same bond proposals with no reference to property taxes.
Intervention
All respondents will be presented with a hypothetical school bond proposal. The treatment group's proposal will conclude with the text "This is a property tax increase.", mimicking the language required by HB-3, a law passed by the Texas legislature in 2019. The control group's proposal will be identical, except that it will omit this final sentence. The size of the proposed bond will be based upon the respondent's reported county size, where residents of larger counties evaluate larger hypothetical bonds and residents of smaller counties evaluate smaller hypothetical bonds.
Experimental design
This is a simple two-armed randomized experiment where all respondents will be presented with a hypothetical school bond proposal, and the treatment group's proposal will conclude with the text "This is a property tax increase." The control group's proposal will be identical, except that it will omit this final sentence. The size of the proposed bond will be based upon the respondent's reported county size, where residents of larger counties evaluate larger hypothetical bonds and residents of smaller counties evaluate smaller hypothetical bonds.
Design details
Randomization method
Randomization embedded in Qualtrics; approximately 50% Treatment and 50% Control.
Randomization unit
Individual
Clusters
A national sample of approximately 1000 respondents will be recruited via Harvard DLABSS.
Observations
Our target sample is 1000 individuals.
Arms
Our target sample of 1000 individuals will be evenly divided between the treatment and control groups:
500 Treatment
500 Control
Primary outcomes
The primary outcome of interest is a binary measure of support for the bond. Respondents will be asked "How would you vote on [the hypothetical school bond] if it were on your local ballot in the next general election?" where FOR is coded as 1 and AGAINST is coded as 0. Our primary analysis will test whether there are differences in the average rate of bond support between the treatment and control groups using ordinary least squares regression, controlling for county size, state, political orientation, and a set of demographic covariates to increase precision.
Our primary analysis will examine whether our treatment impacts overall rates of support for school bonds. Exploratory analyses will assess heterogeneity in effects by subgroup, including by political affiliation, gender, race/ethnicity, population, and region.
Secondary outcomes
The first secondary outcome of interest is a measure of general support for school spending. Respondents will be asked: "State legislatures must make choices when making spending decisions on important state programs. Would you like your legislature to increase or decrease spending on education?" Where the choices are "Greatly Increase", "Slightly Increase", "Maintain", "Slightly Decrease", and "Greatly Decrease." We will estimate the impact of the treatment on support for increasing school spending by combining the first two categories ("Greatly increase" and "Slightly Increase") into a single binary measure of support for increasing school spending vs. decreasing or maintaining. We will also examine whether the treatment influences support for decreasing school spending by combining the last two categories ("Greatly Decrease" and "Slightly Decrease") into a single binary measure of support for decreasing school spending vs. increasing or maintaining. Our analysis of their secondary outcome will test whether there are differences in the average rate of support for school spending between the treatment and control groups using ordinary least squares regression, controlling for county size, state, political orientation, and a set of demographic covariates to increase precision.
A second set of secondary outcomes are measures of which factors respondents considered as they decided whether to support the school bond. These outcome variables will be constructed from responses to the question, "Which of the following was on your mind as you decided whether to support or oppose the school bond proposition?" Respondents can select all that apply from the following choices: Teacher salaries, school building condition, academic achievement, technology in schools, property taxes, sales taxes, current school spending, overall government spending or the deficit, inflation, my local school board or district leadership, my state's department of education, and something else (write-in response).
Our exploratory analysis will assess whether and how our treatment changes the factors that respondents report considering as they decided whether to support the school bond measure (using ordinary least squares regression as described above). Additionally, we will use assignment to treatment as an instrument for considering a particular factor to estimate the direct effect of increasing consideration of/the salience of a given factor (e.g., property taxes, teacher salaries, district leadership, etc.) on support for the bond (and overall spending) via two-stage-least-squares instrumental variables regression.
Cited by
No citation data available for this registration yet.