Behavioral Incentive Compatibility and Personalized Interventions

Participants complete an online incentivized experiment consisting of 20 trials. In each trial, a participant is assigned an induced-value token, completes a multiple price list that determines a reported switch point, and then chooses between two alternatives in a subsequent allocation stage. The intervention is the treatment-specific rule that maps the participant’s token value and reported switch point into the two alternatives offered in the allocation stage. Participants are randomly assigned to one of seven between-subjects treatment arms and face the same rule throughout the experiment. Across treatments, the allocation rule varies whether reporting has no future consequence, creates incentives to misreport upward or downward, or introduces uncertainty in how reports affect later options, including cases in which incentive compatibility is preserved. This design allows the study to isolate how anticipated downstream consequences of reporting affect behavior in an otherwise standard incentivized measurement task. Refer to the Analysis Plan, Section 2: Experimental Design, for additional information.

2026-04-20

2026-10-20

The primary outcomes are two measures of reporting behavior in the incentivized trials: the difference between the participant’s reported switch point and the assigned token value, and an indicator for whether the participant’s reported switch point exactly matches the assigned token value. Both are measured for each participant in each of the 20 incentivized trials. Refer to the Analysis Plan, Sections 3.2: Data, Outcomes and Analysis Overview, for additional information.

Secondary outcomes include the absolute distance between the participant’s report and the treatment-specific payoff-maximizing report, an indicator for whether the participant exactly submits that payoff-maximizing report, participant-level summary measures of reporting behavior, specifically each participant’s average deviation from the assigned token value, average matching rate, average distance from the treatment-specific payoff-maximizing report, average rate of exact optimal reporting across incentivized trials and measures of responses after costly versus non-costly misreports. Additional secondary outcomes include quiz-based participant categories, classifications based on backward induction, narrow framing, and lying aversion, survey responses on consideration of future consequences, practice-question performance and dominance violations. Refer to the Analysis Plan, Sections 2 and 3, for additional information.

The participant-level summary outcomes are constructed by averaging trial-level behavior across incentivized rounds for each participant. These include the participant’s average signed deviation between the reported switch point and the assigned token value, the share of incentivized rounds in which the participant exactly matches the token value, the average absolute distance between the participant’s report and the treatment-specific payoff-maximizing report, and the share of incentivized rounds in which the participant exactly submits that payoff-maximizing report. Costly-misreport response measures are constructed by identifying rounds in which the participant’s report differs from the token value and the realized BDM draw makes that misreport payoff-relevant, and then tracking next-round adjustment.

Quiz-based categories are constructed from the first block of quiz questions based on which of the questions on total incentives, Measurement Stage incentives, and Allocation Stage incentives are answered correctly. The narrow-framing indicator is constructed from whether the participant gives different responses to two equivalent lottery problems presented in different formats. The backward-induction indicator is constructed from whether the participant reveals the payoff boxes and chooses the action consistent with backward induction. The lying-aversion indicator is constructed by comparing the participant’s reported roulette outcome to the realized draw. The consideration-of-future-consequences responses are constructed from the average answers to six survey items asked at the end of the study. Practice performance is based on the number of attempts required to answer practice questions correctly. Dominance violations are constructed by counting cases in which a participant chooses the lower monetary option in the Allocation Stage. Refer to the Analysis Plan, Sections 2 and 3, for additional information.

This study is an online, between-subjects experiment conducted on Prolific. Participants are randomly assigned at the individual level to one of seven treatment arms and face the same treatment throughout the study. After consent and a bot check, participants read instructions with comprehension questions and complete a practice trial with open-response practice questions that must be answered correctly before they can continue. The main experiment consists of 20 incentivized trials. In each trial, the participant is assigned an induced-value token, completes a multiple price list that determines a reported switch point, chooses between two alternatives in an allocation stage, and then receives feedback on trial earnings. The intervention is the treatment-specific allocation rule that determines how the token value and reported switch point affect the alternatives offered in the allocation stage. Following the trials, participants complete an incentivized post-experimental quiz and a non-incentivized exit survey. Participants receive a fixed completion payment, additional earnings for correct first-attempt practice answers, and a probabilistic bonus based on either one randomly selected trial or one randomly selected quiz question. Refer to the Analysis Plan, Section 2: Experimental Design, for additional information.

Not available

Randomization done by computer at the individual level, with approximately equal assignment across seven treatment arms. Refer to the Analysis Plan, Section 3.4.2: Power and Sample Size, for additional information.

The unit of randomization is the individual participant. Each participant is randomly assigned to one treatment arm and remains in that treatment for the full experiment.

No

700 individual participants.

700 individual participants.

Approximately 100 individual participants per treatment arm, with equal planned assignment across the 7 treatment arms; realized completed sample sizes may be slightly above or below 100 per arm due to attrition. Refer to the Analysis Plan, Section 3.4.2: Power and Sample Size, for additional information.

Power calculations are based on simulation using the pilot data and the planned analysis specification, including repeated observations per participant, participant-level clustering, token fixed effects, round fixed effects, and treatment-by-token interactions. The study is powered for the two primary outcomes: mean deviation from matching and the fraction of matching reports. Mean deviation from matching is measured in switch-point units, and the fraction of matching reports is measured in percentage points. Because the planned inference is based on joint Wald tests in the full clustered design, I do not use a single closed-form minimum detectable effect size; instead, sample size is chosen to achieve approximately 80% power for the planned treatment-versus-control tests. Refer to the Analysis Plan, Section 3.4.2: Power and Sample Size, for additional information.