Demand for Anti-Trust: An Experimental Investigation

Participants in a U.S.-based survey are randomly assigned to one of five recent antitrust cases involving Google, Luxottica, Apple, Live Nation, or Meta. Within each case, participants are further randomized to either a control condition, in which they receive no additional information, or to one of four treatment conditions, each of which provides different information about the case. We measure effects on beliefs and preferences in the baseline survey and again in a follow-up survey to be conducted one month later.

2026-05-04

2026-12-31

The full survey instruments for both the baseline and follow-up surveys are attached to this application. The primary outcome is respondents’ preferences for antitrust enforcement. This outcome is measured first using a support scale capturing respondents’ level of support for the specific antitrust action in the case to which they are assigned. We also collect alternative and complementary outcome measures.

As shown in the survey instrument, the alternative and complementary outcome measures include respondents’ support for remedies regarding the case they were assigned to (e.g., break down the company), general antitrust policy proposals, donations to a pro-antitrust nonprofit, willingness to sign an antitrust petition, willingness to write an email to a representative in support of stronger antitrust enforcement, and a willingness-to-pay measure capturing whether they like the company.
To improve the precision of the estimates, in addition to estimating effects on specific outcomes, we will also construct an index for each outcome category.

To shed light on the underlying mechanisms, the survey elicits both prior and posterior beliefs about four objects: the company’s perceived market share, the fairness of competition in the relevant market, the extent of consumer harm, and the respondent’s overall perception of the company. In addition to company-specific posterior beliefs, we include similar questions to measure posterior beliefs about dominant companies in general. The follow-up survey also asks respondents to represent the government side (for example, the DOJ) in a debate against an AI lawyer representing the company. This exercise allows us to measure whether the treatment affects the types of arguments respondents use as well as the overall persuasiveness of those arguments.

Participants are randomly assigned to one of five recent antitrust cases involving Google, Luxottica, Apple, Live Nation, or Meta. Within each case, participants are further randomized to either a control condition, in which they receive no additional information, or to one of four treatment conditions, each of which provides different information about the case. We measure effects on beliefs and preferences in the baseline survey and again in a follow-up survey to be conducted one month later.

Not available

Randomization is implemented automatically by the survey software Qualtrics at the individual respondent level.

Individual respondent

No

4,000 respondents

4,000 respondents

The 4,000 respondents are first randomly assigned, with equal probability, to one of five antitrust cases.
These same respondents are then randomly assigned, with equal probability, to one of the five treatment groups.