Wading Through Sludge: Evidence from Title IX Reporting Systems

The information intervention presents administrators responsible for Title IX reporting systems with a summary of empirical research concerning the relationship between administrative burden in reporting systems and reporting outcomes. The active-control condition presents a summary of unrelated research concerning professional training and does not provide information about Title IX reporting processes. Before the information assignment, respondents complete one of two randomized elicitation modules concerning workload and capacity trade-offs or intake-form preferences.

2026-09-14

2026-12-31

Two co-primary outcomes, measured immediately after the information intervention in the same survey:

An indicator for recommending the lowest-burden intake-form design (Design A).
An indicator for recommending the lowest-burden reporting landing-page design (Design A).

The intake-form outcome uses question Q56 (internal identifier QID56), and the landing-page outcome uses Q88 (QID88). For each outcome, Design A is coded 1 and either Design B or Design C is coded 0. Coding follows design identity, not screen position. An unanswered item is missing and is not coded 0.

Design A in the intake task allows optional contact information and a short optional narrative, requires no supporting documents, and requests only basic incident information. Design A in the landing-page task provides a prominent reporting button and direct access to the form without login or intermediate steps.

The primary estimand is the effect of information assignment on the probability of recommending Design A, averaged equally across the two elicitation modules. We estimate the evidence-minus-control difference within each module and average those two differences with equal weights. We will report effect estimates and confidence intervals for both co-primary outcomes and apply Holm’s procedure to their two-sided p-values to control the family-wise error rate at 5%.

The full distribution of intake-form and landing-page recommendations across Designs A, B, and C; expected consequences of the recommended designs; perceived deterrence of truthful and false reports and legal defensibility of the chosen intake form; confidence in recommendations; designs preferred under alternative stated objectives; reported reasons for recommendations; and willingness to contribute to an institutional consultation.

The full design-choice distributions use Q56 and Q88. Expected consequences use the individual 1–7 responses in Q61 and Q92. Perceived characteristics of the chosen intake form use Q62 (internal identifier QID113), with each item retained separately. Confidence uses Q68 and Q93. Alternative-objective choices use Q110–Q112 and Q114–Q116. Closed-ended reasons use Q124 and Q91, and open-ended explanations use Q87 and Q109. Consultation willingness to pay uses Q119, measured from $0 to $25. Free-text explanations will be exploratory unless a coding protocol is specified before analysis.

The reporting-rate and false-reporting beliefs in Q21–Q24, workload measures and toolkit willingness to pay in the workload module, and the four discrete-choice tasks are collected before information assignment. They describe baseline beliefs and preferences relative to the information intervention and will not be interpreted as outcomes of that intervention. Perceived consequences of a chosen design combine possible changes in the chosen design and changes in its evaluation; comparisons do not identify belief updating about a fixed design alone.

This is an individually randomized online survey experiment with a two-by-two design. The target population is Title IX coordinators and officers currently working at U.S. colleges and universities. We plan to survey 500 eligible administrators through the ATIXA ListServ. Eligibility requires a current role in a U.S. college or university Title IX office and informed consent. Participants’ eligibility to participate in the survey will be confirmed through our partner organization and at the beginning of the survey.

All respondents first complete common questions about their professional background, office constraints, and baseline reporting beliefs. They are then randomly assigned to one of two elicitation modules. The workload and capacity module elicits the importance of workload, a workload trade-off threshold, predictions about peers’ responses, and willingness to pay for a workload toolkit instead of an access toolkit. The discrete-choice module presents four paired choices between intake forms whose attributes vary, followed by questions about confidence and design priorities.

Within each module, respondents are subsequently randomly assigned to either an evidence summary about the relationship between burden in Title IX reporting systems and reporting outcomes, or an active-control summary about knowledge checks and spaced review in professional training. Thus, the four cells are workload/control, workload/evidence, discrete choice/control, and discrete choice/evidence. The target allocation is approximately one quarter of participants to each cell.

All respondents then recommend an intake-form design and a landing-page design, report their reasons and expected consequences, and state willingness to contribute part of their participation payment toward an institutional consultation. The principal comparison is assignment to evidence versus active control, averaged across the two elicitation modules. All survey outcomes are collected during the same session. Recruitment will run until 500 participants have completed the survey. If we have not achieved our targeted number of surveys by the intervention end date, we will run the analysis on the sample available to us.

Not available

Computerized randomization within Qualtrics. A first Survey Flow randomizer presents one of the two elicitation modules, using the “Evenly Present Elements” setting. A separate randomizer within each module then assigns one of the two information conditions using the same setting. The target allocation is 1:1 between modules and 1:1 between evidence and active control within each module. Final completed-response counts may differ because of attrition.

Within the discrete-choice module, respondent-specific form profiles are generated by computer across five attributes: narrative requirements; contact-information requirements; date, time, and location requirements; requirements to name the accused; and supporting-document requirements. The generator redraws identical pairs and pairs ordered in the same direction on every coded attribute, subject to its programmed retry limit. These within-respondent profile draws are distinct from assignment to the four experimental cells.

The individual survey respondent (Title IX administrator) for both module assignment and information assignment. Within the discrete-choice module, attributes of the two form profiles vary across four tasks for each respondent. These repeated tasks do not constitute additional participants.

No

Not applicable. There is no group-level treatment assignment. The planned number of individual randomization units is 500 Title IX administrators.

500

The target allocation is approximately 125 administrators in each of four cells: workload/control; workload/evidence; discrete choice/control; and discrete choice/evidence. Pooling across modules gives approximately 250 administrators assigned to evidence and 250 assigned to active control. These are target allocations; realized completed-response counts may differ.