Procurement under Imperfect Verification: How Audit Rules Shape Effort and Reporting

Last registered on June 15, 2026

Pre-Trial

Trial Information

General Information

Title
Procurement under Imperfect Verification: How Audit Rules Shape Effort and Reporting
RCT ID
AEARCTR-0018579
Initial registration date
June 05, 2026

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published
June 15, 2026, 1:44 PM EDT

First published corresponds to when the trial was first made public on the Registry after being reviewed.

Locations

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information

Primary Investigator

Affiliation
CAU Kiel

Other Primary Investigator(s)

PI Affiliation
Bard College Berlin
PI Affiliation
CAU Kiel

Additional Trial Information

Status
In development
Start date
2026-06-16
End date
2026-09-15
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
Abstract
Many procurement environments require buyers to allocate contracts based on suppliers’ self-reported claims about costly, imperfectly observable effort. Given a limited capacity to verify these claims, scarce audits must be allocated optimally to encourage high effort and honest reporting. This study investigates how the precision of audit targeting - ranging from random to highly targeted audits - affects suppliers’ behavior and buyers’ efficiency in a procurement contest experiment. In this experiment, five suppliers choose a costly effort level and then submit a performance claim (based on their unobserved effort level). The buyer receives these claims, after which an automated audit mechanism verifies the true effort of exactly one supplier. Afterward, the buyer selects three suppliers. The game is repeated for 15 periods with the same group of five suppliers and one buyer (but feedback order does not allow the feedback order does not allow for individual reputation). Three audit rules are compared: "random auditing'' (each message is equally likely to be audited), "competitive auditing'' (messages with higher values are more likely to be audited), and "highest-message auditing'' (the highest message is always audited). The question is whether these three audit rules, representing increasing levels of audit targeting precision, lead to higher supplier effort and truth-telling, and to overall higher buyer efficiency. Theoretically, under standard equilibrium assumptions, the three audit rules are outcome-equivalent: suppliers pool on the same message, so targeted audit rules become equivalent to random auditing on the equilibrium path. However, the behavior of buyers and suppliers may be affected by different motives, adding to the competition between suppliers, it may lead to different performances of the audit rules. One relevant application is in environmental economics, where a buyer may ask suppliers to report the carbon footprints of their products.
External Link(s)

Registration Citation

Citation
Requate, Till, Aurel Stenzel and Israel Waichman. 2026. "Procurement under Imperfect Verification: How Audit Rules Shape Effort and Reporting." AEA RCT Registry. June 15. https://doi.org/10.1257/rct.18579-1.0
Experimental Details

Interventions

Intervention(s)
Intervention Start Date
2026-06-16
Intervention End Date
2026-09-15

Primary Outcomes

Primary Outcomes (end points)
The primary outcome variables are 1. suppliers’ actual effort decision, 2. suppliers' reported effort 3.Misreporting (the difference between supplier reported effort and actual effort. 4. buyers purchase decision among the firms (conditional on available information), 5. buyers' efficiency (conditional on the decision of the buyer whom with which suppliers to work and their actual effort level).
Primary Outcomes (explanation)
variables explained above

Secondary Outcomes

Secondary Outcomes (end points)
Secondary Outcomes (explanation)

Experimental Design

Experimental Design
The experiment studies procurement decisions under uncertainty and imperfect verification.

The experiment will be conducted online, using an experimental economics laboratory participant pool. They are invited online at a specific time (with typically 18 or 12 participants). At the beginning of the experiment, participants are then divided into groups. Each group consists of one buyer and five suppliers. The suppliers compete for contracts with the buyer (in each period, the buyer selects three of five suppliers). There are 15 periods of these fixed groups. Each period of the experiment is as follows: The first stage starts with the suppliers who make two consecutive decisions: 1. What actual effort to have, and 2. how much effort to report to the buyer. Regarding decision 1: the higher the supplier’s effort is, the higher the buyer’s payoff, but effort is costly, so it lowers the supplier's payoff. Regarding decision 2: the reported efforts may affect the buyer's decision which supplier to purchase from. In the second stage, an automated audit rule is applied, auditing exactly one firm. After the buyer receives the audit value from one supplier, it decides which 3 of its 5 suppliers to purchase from. After each period, feedback is given in the following way. Suppliers see the reported effort of all suppliers and are informed which one was audited. The buyer receives this information about the actual effort of the audited firm. Each firm is also given feedback on its payoff in the period. There are 15 repeated periods (but where suppliers labels are changed so precise individual reputation cannot be formed). Payoff is determined by 3 random periods selected by the computer at the end of the experiment.

The experiment consists of three treatments, with three different automated audit rules. 1. RAM (“random auditing”, where each message is equally likely to be audited), 2. HMM (“highest-message auditing”, where the highest message is audited). 3. CAM (“competitive auditing”, where a message with higher values is more likely to be audited). Theoretically, under standard equilibrium assumptions, the three audit rules are outcome-equivalent: suppliers pool on the same message, so targeted audit rules become equivalent to random auditing on the equilibrium path. However, the behavior of buyers and suppliers may be affected by different motives, and given the competition among suppliers, the three rules may lead to different performances (actual effort, reported effort, and buyers’ decisions).

Our a priori hypotheses are based on symmetric equilibrium under standard money maximizing assumptions:
H1a: The distribution of suppliers’ effort choices does not differ across the RAM, HMM, and CAM treatments.
H1b: In all treatments, effort choices e_i is low (e_i≤17) because beyond this level the cost a supplier saves by shirking (i.e., effort choice of 0) exceeds the expected rent she loses from being caught and excluded by the audit.
H2a: The distribution of suppliers’ effort reports does not differ across the RAM, HMM, and CAM treatments.
H2b: Suppliers will always report the highest effort (m_i=50) under RAM, HMM, and CAM.
H3: Buyer’s efficiency (measured by buyers’ payoff: which takes into account the suppliers' effort choices and the buyer’s selection decision) does not differ across RAM, HMM, and CAM treatments.
Experimental Design Details
Not available
Randomization Method
Randomization is conducted by computer within the experimental software. Participants are randomly assigned to experimental sessions and treatment conditions. Within each decision situation, any stochastic elements of the audit mechanism, including random auditing and tie-breaking, are also implemented automatically by the computer according to the pre-specified rules.
Randomization Unit
The primary unit of randomization is the experimental session. Each session is assigned to one of three audit rule treatment. Participants within a session are randomly assigned to groups and roles (buyer, suppliers) by the experimental software. Any stochastic audit selection and tie-breaking within the game are randomized automatically at the decision-situation level by the software.
Was the treatment clustered?
Yes

Experiment Characteristics

Sample size: planned number of clusters
8-13 sessions per treatment time 3 treatments (i.e., say 10 x 3 sessions overall)
Sample size: planned number of observations
we aspire to have about 25 groups x 3 treatments= 450 participants. * Note that the randomization is done in the session level (each session with a given audit rule includes 2-3 groups). However, each group is an independent observations.
Sample size (or number of clusters) by treatment arms
For each treatment about 8-13 sessions (typically with 12-18 participants) to reach a target of 150 participants per treatment.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
considering a difference in effort of say 10 points between the most precise treatment (competitive auditing) and the least precise treatment (random auditing) with a std. dev of 12, and with alpha probability of 0.95 and power of 0.80 the sample size is 25 groups per treatments. We will try to get to this figure.
IRB

Institutional Review Boards (IRBs)

IRB Name
Central Ethics Committee of the University of Kiel
IRB Approval Date
2026-06-05
IRB Approval Number
ZEK- 33/26