Economists’ Narratives, Economic Literacy, and Support for Rent Control

The survey experiment manipulates two dimensions, independently at the individual level.
(1) Economic narrative. After respondents are informed about the status quo of rent regulation in Germany and the proposed stricter policy, and after an open-ended (unstructured) elicitation of their prior mental model, respondents are randomly assigned to one of three conditions:
• Control: no narrative is presented.
• Economists’ Economic Narrative: respondents read a short text summarizing the economic assessment of stricter rent regulation drawn from the 2024/25 annual report of the German Council of Economic Experts („Wirtschaftsweise“).
• Economists’ Economic Narrative with Disagreement Signal: respondents read the same narrative, and, in addition, they receive information that the assessment is shared by four of the five Council members and that one member has issued a dissenting minority opinion.
(2) Post-treatment elicitation. Orthogonally to the narrative treatment, respondents are randomly assigned to either a No-Elicitation branch (no structured elicitation of the post-treatment mental model) or a Structured-Elicitation branch (structured elicitation of the post-treatment mental model along a fixed set of eight outcome dimensions). This second randomization allows us to assess whether structured elicitation itself primes particular mechanisms and shapes policy-support responses. Both dimensions are assigned in a single draw, so that each respondent is allocated to one of six equally likely conditions (see Randomization Method).

2026-08-03

2026-11-30

(i) Individuals’ support for the proposed stricter rent-control policy; and (ii) individuals’ mental models about the effects of stricter rent control.

Support is measured on a Likert scale after the narrative treatment and again in the follow-up survey. Mental models are elicited in an open-ended (unstructured) format before the narrative treatment (all respondents) and again in the follow-up survey (all follow-up respondents), and, in the structured-elicitation branch only, in a structured format immediately after the narrative treatment; open-ended responses are coded into pre-specified indices (see the pre-analysis plan). Economic literacy is the primary moderator of both outcomes; it is measured before treatment using a 10-item test and enters as a standardized continuous score, with a median-split indicator reported for interpretability and as a robustness check (see Secondary Outcomes and the pre-analysis plan). More details can be found in the experimental design section.

i) Individuals’ mental models about the effects of stricter rent control, including mechanism uptake, the residual and total breadth of the model, confidence in the elicited model, a low-effort-response indicator, and the individual dimensions of the structured elicitation battery; ii) support for other housing policies, relative support for rent control, and support for other price-control policies; and iii) trust in and perceived competence of economists and beliefs about disagreement among economists.

Mental-model sub-measures are constructed from the open-ended and structured elicitations: mechanism uptake counts the narrative's causal channels that also appear in a respondent's response; residual and total breadth count, respectively, the causal channels invoked that are outside the narrative and in total; confidence is elicited after each mental-model measure; the low-effort indicator flags responses that are empty, off-topic, or disengaged; and the individual dimensions are the five signed and three unsigned items of the structured battery, analyzed separately as well as summarized in the primary directional-alignment index. Absolute support for all policies is measured on a Likert scale, and relative support for housing policies is measured with a ranking question. All policy support questions are measured after the narrative treatment and again in the follow-up survey. Trust in and perceived competence of economists and beliefs about disagreement among economists on rent control are measured on Likert scales in the follow-up survey (trust in economists is also measured pre-treatment). See the pre-analysis plan for full definitions.

We conduct the survey experiment with a quota-representative sample of approximately 7,500 German adults. The survey is programmed and hosted using the Tivian license of the ifo Institute; respondents are provided by the online panel provider Talk Online Panel, subject to representativeness quotas (gender, age, federal state, education, employment status, renter vs. non-renter, and municipality size). Participants answer questions autonomously on their own digital devices. Randomization is performed by the survey software. The panel provider also supplies survey weights; the main analyses are unweighted, and weighted estimates are reported in robustness checks.
The survey proceeds in the following order: (1) landing page and a willingness-to-write screening item ; (2) quota demographics; (3) pre-treatment moderators, including housing situation and self-interest items, expectations about local rents and house prices, economic values, interest in and consumption of economic news, trust in and perceived competence of economists, self-assessed economic understanding, a 10-item economic-literacy test, and beliefs about own test performance; (4) an attention check; (5) information on rent increases since 2010 and an open-ended diagnosis of their perceived causes; (6) a description of the rent-control status quo and the proposed stricter policy; (7) an open-ended (unstructured) elicitation of the prior mental model and confidence in it; (8) randomization, in a single draw, into one of six conditions defined by the three narrative conditions and the two elicitation branches; (9) policy-support questions; (10) additional demographics and cognitive controls; and (11) an outro. The follow-up survey is fielded two to four weeks later and includes a reminder of (6) and repeats (7), (9), trust in economists, and additionally includes questions on perceived disagreement on rent control among economists, as well as two questions on whether people remember the treatment content from the main survey

Not available

Randomization is carried out by the randomizer of the survey software (Tivian). Each respondent is assigned in a single draw to one of six equally likely conditions, defined by the cross of the three narrative conditions (Control; Economic Narrative; Economic Narrative with Disagreement Signal) and the two elicitation branches (No Elicitation; Structured Elicitation). This is equivalent to assigning the narrative condition and the elicitation branch independently at the individual level. Assignment probabilities do not depend on respondent characteristics.

Individual level.

No

Approximately 7,500 individuals.

Approximately 7,500 individuals in the main survey; approximately 6,000 individuals in the follow-up survey (an 80% recontact rate).

Randomization is into six equally sized cells defined by the cross of the three narrative conditions and the two elicitation branches: approximately 1,250 respondents per cell. This implies approximately 2,500 respondents per narrative condition (Control; Economic Narrative; Economic Narrative with Disagreement Signal) and approximately 3,750 respondents in each of the No-Elicitation and Structured-Elicitation branches. Outcomes that are observed only in the Structured-Elicitation branch (the structured mental model) are therefore available for approximately 1,250 respondents per narrative condition, and the Control condition contains approximately 1,250 respondents in each elicitation branch.

For pairwise comparisons of narrative conditions on standardized outcomes, minimum detectable effect sizes are on the order of 0.07–0.08 SD at 80% power (n ≈ 2,500 per narrative condition). The corresponding figures are approximately 0.10–0.11 SD for comparisons within an economic-literacy subgroup or within an elicitation branch (n ≈ 1,250), and approximately 0.135–0.16 SD for interactions with the economic-literacy score (n ≈ 2,500 per narrative condition). See the pre-analysis plan, Section 5.