|
Field
Experimental Design (Public)
|
Before
Online survey experiment fielded to a nationally representative sample of Spanish adults (18 years or older). The survey consists of four sequential blocks administered in a single session. Block 1 collects sociodemographic information, the respondent's relation to the university system, self-rated knowledge, and frequency of following news on higher education. Block 2 elicits prior beliefs about labour-market earnings premiums, public financing levels, and quality heterogeneity, as well as pre-treatment versions of the policy preference outcomes. Block 3 contains the randomised information treatment (one of five conditions) and an arm-specific manipulation check for the four treatment arms. Block 4 contains the post-treatment outcomes, the secondary outcomes, an attention check, self-reported survey duration, and an item on perceived demand effects. There is no follow-up wave. The fieldwork uses a representative online panel of Spanish adults provided by a commercial survey supplier. Randomisation is implemented at the individual level by the survey platform prior to Block 2. The analysis follows an analysis-of-covariance specification with the pre-treatment value of the outcome and pre-specified baseline covariates as right-hand side variables, with heteroskedasticity-robust standard errors. Multiple hypothesis testing is addressed via Bonferroni correction for four pre-specified treatment-versus-control comparisons on the primary outcomes, complemented by Romano-Wolf step-down adjusted p-values.
|
After
Online survey experiment fielded to a nationally representative sample of Spanish adults (18 years or older). The survey consists of four sequential blocks administered in a single session. Block 1 collects sociodemographic information, the respondent's relation to the university system, self-rated knowledge, and frequency of following news on higher education. Block 2 elicits prior beliefs about labour-market earnings premiums, public financing levels, and quality heterogeneity, as well as pre-treatment versions of the policy preference outcomes. Block 3 contains the randomised information treatment (one of five conditions) and an arm-specific manipulation check for the four treatment arms. Block 4 contains the post-treatment outcomes, the secondary outcomes, an attention check, self-reported survey duration, and an item on perceived demand effects. There is no follow-up wave. The fieldwork uses a representative online panel of Spanish adults provided by a commercial survey supplier. Randomisation is implemented at the individual level by the survey platform prior to Block 2. The analysis follows an analysis-of-covariance specification with the pre-treatment value of the outcome and pre-specified baseline covariates as right-hand side variables, with heteroskedasticity-robust standard errors. Multiple hypothesis testing is addressed via Bonferroni correction for four pre-specified treatment-versus-control comparisons on the primary outcomes, complemented by Romano-Wolf step-down adjusted p-values. Clarification, registered before any author has accessed the outcome data (see forecast lock, Social Science Prediction Platform study sspp-0b2fyx-T, release date 12 August 2026). The data files delivered by the survey firm on 28 July 2026 remain unopened. The PAP defined the primary analysis sample as respondents passing the post-treatment attention check (AC_PASS = 1). Following external review, and because conditioning on a post-treatment variable can compromise randomisation if treatments affect attention (Montgomery, Nyhan and Torres 2018), the primary sample is the full randomised sample; the restriction to attention-check passers becomes a pre-specified robustness check, and we will test whether attention-check pass rates differ by treatment arm. No other element of the PAP changes. (Clarification dated 10 August 2026.)
|