Mandatory Protection: Survey Evidence on Internalities, Externalities, and Paternalistic Preferences

The survey study addresses the following research questions:

1. Is support for regulatory interventions shaped by perceived individual costs and benefits?
1a. Costs: Is support for regulation related to instrumental and/or intrinsic valuations of choice rights?
1b. Benefits: Is support for regulation related to individuals' perceptions of internalities and/or externalities?
1c. Trade off: How is the awareness of a trade off between restriction and protection related to individual costs and benefits?

2. Is acceptance of internal and external harms (i.e., tolerance for such harms) related to instrumental and/or intrinsic valuations of choice rights?

3. Are regulatory interventions perceived as paternalistic?

4. How does support for regulation vary with the salience of internalities and externalities?

We additionally conduct exploratory analyses to examine heterogeneity in preferences and trade-offs across individuals and conditions. These analyses are labeled as exploratory.

2026-05-26

2026-08-28

Primary outcomes:
- Support for regulation (regulatory preference)
- Perceived paternalism
- Acceptance of internalities
- Acceptance of externalities
- Perceived general trade-off between protection and freedom

Key explanatory variables (mechanisms):
- Instrumental valuation
- Intrinsic valuation
- Perception of externalities
- Perception of internalities
- Scenario indicator (as control variable)

- Sociodemographic characteristics (e.g., income, education, age, gender, religious affiliation)
- General attitudes toward freedom and regulation
- General control variables (e.g., trust)

The survey consists of three parts, with Part 1 being the main component of the study:

Part 1. Participants are presented with short, hypothetical but realistic policy-relevant scenarios (vignettes) set in Germany. After each vignette, participants answer a set of questions measuring their support for regulation (regulatory preference), acceptance of internalities and externalities, perceived trade-offs between protection from internalities and externalities and freedom, and further related attitudes (see primary outcomes).

Each participant evaluates two scenarios. The order of scenarios and assignment to participants are randomized. We control for order effects in all analyses. We may restrict the analysis to the scenario on order first.

Part 2. General questions on attitudes toward freedom and regulation.

Part 3. Post-experimental questionnaire.

We include a set of scenarios covering different policy contexts (list provided below):
- Ban on cigarettes and tobacco products (physical health)
- Ban on the use of beauty filters on social media (mental health and society)
- Mandatory helmet use among cyclists (physical health and society)
- Ban on oil and gas heating systems (environment)

We implement a between-subjects design with three treatment conditions:

1. Neutral: Description of the status quo in different countries.
2. Internality: Description of the status quo plus information about potential internal harms.
3. Externality: Description of the status quo plus information about potential external harms.

Not available

Randomization done by a computer (treatment and scenario).

Individual level.

No

No clustering.

Number of total observations: 2,400 participants x 2 scenarios = 4,800 observations.

Observations per scenario (first shown):
600 scenario I (cigarettes)
600 scenario II (beauty filters)
600 scenario III (helmets)
600 scenario IV (heating)

Participants per treatment (scenario first shown):
800 Neutral
800 Internality
800 Externality

We base our power analysis on the between-subject comparison for the salience treatments, using the most conservative restriction of using only the scenario first shown. These comparisons yield the smallest number of observations. Under these assumptions, we are able to detect a minimum effect size of d = 0.3. This corresponds to a 1-point difference on our 11-point Likert scale, assuming alpha=0.05, beta=0.8, and, conservatively, an uniform distribution of participants across grid points.