📈 economics

Audit Silence and the Capacity Trap

This paper demonstrates that in a repeated inspection game, a sustained period of audit silence signals capacity constraints rather than mere inaction, triggering a "capacity trap" where strategic firms violate and functioning inspectors over-exert effort, ultimately causing enforcement deterioration and adverse sorting that highlights the critical policy importance of the distribution of enforcement capacity over its average level.

Original authors: Georgy Lukyanov

Published 2026-07-29

📖 7 min read🧠 Deep dive

CC BY 4.0

Original authors: Georgy Lukyanov

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a world where rules are enforced by a game of "Catch Me If You Can." In this game, a player (the firm) decides whether to play by the rules or cheat, and a referee (the inspector) decides how hard to look for cheating. This isn't just about sports; it's a branch of science called Game Theory, which studies how people make decisions when their outcomes depend on what others do. The paper you're about to read leans on two famous ideas: Reputation, which is the idea that your past actions change how people expect you to act in the future, and Bayesian Updating, a fancy way of saying "learning from new clues." If you see a referee checking your bag every day, you know they are on the job. But what happens if the referee never shows up? Does that mean they are lazy and don't care, or does it mean they are stuck in traffic and physically can't get there? This question matters because if we get the answer wrong, we might punish the wrong people or miss the real danger.

This paper, titled "Audit Silence and the Capacity Trap," tackles a very specific and tricky puzzle: What does it mean when an audit simply doesn't happen? The author, Georgy Lukyanov, argues that a "silent" period—where no inspection occurs—is not a neutral event. It is a loud signal, but one that is dangerously easy to misread.

The Two Faces of Silence

Imagine you are a student waiting for a teacher to check your homework. If the teacher walks in, looks at your paper, and says, "All clear," that's a Clean Audit. You know they looked, and you know you passed. But what if the teacher never walks in at all? That's Audit Silence.

The paper points out that silence has two completely opposite meanings:

The "Slacker" Reading: The teacher could have checked your work but chose not to because they are lazy or think you are trustworthy. If this is the case, you should expect them to show up soon with a sharper eye.
The "Stuck" Reading: The teacher wanted to check but couldn't. Maybe their car broke down, the school is locked, or they are overwhelmed with other work. If this is the case, no amount of "trying harder" will make them appear, because the problem isn't effort; it's a lack of capacity (resources, staff, or access).

The paper's big discovery is that in a world where we don't know which type of teacher we have, a long string of silence forces us to believe the "Stuck" story. And once we believe that, the whole system breaks down in a way that is surprisingly counterintuitive.

The Trap of "Doing Nothing"

The author sets up a repeated game where a company (the firm) and an inspector play over and over. Some companies are "Good Guys" who always follow the rules. Some are "Strategic" and will cheat if they think they won't get caught. Similarly, some inspectors are "Functioning" and can choose to work hard or slack off. But there's a third type: the "Constrained" inspector. This inspector is like a worker with a broken car; no matter how hard they try, they can only check a few files because their tools are limited.

Here is the trap:
When a "Functioning" inspector works hard, they find more problems. When a "Constrained" inspector works, they find fewer problems. But here is the kicker: Silence is the only clue we have.

If you see a long run of silence (no audits), the math in the paper shows that your belief shifts. You start to think, "The inspector must be the 'Constrained' type who can't do the job." Why? Because the "Constrained" type produces silence much more frequently than the "Functioning" type, even when the Functioning type is working hard. Crucially, the paper notes that detection is imperfect: even if a Functioning inspector works hard, they might miss a violation. A "Clean Audit" (an audit with no findings) does not prove the firm was compliant. However, a complete absence of an audit is a much stronger signal of incapacity than a clean audit is.

Once you believe the inspector is "Constrained," the Strategic Company changes its behavior. They think, "Hey, the inspector can't catch me even if they try! I might as well cheat." So, they start violating the rules.

The Paradox of Maximum Effort

This is where the story gets weird and the paper's main finding shines.

Usually, we think: "If people start cheating, the inspector should work harder to catch them."
The paper proves that in the specific "dominance region" reached after a finite run of silence, the "Functioning" inspector does work as hard as they possibly can. They exert maximum effort.

But here is the twist: Even though the inspector is working their hardest, the actual number of audits keeps dropping.

Why? Because the "Constrained" inspectors (who can't do much) are now the ones you think are in charge. The "Functioning" inspector is working hard, but their hard work is drowned out by the fact that the "Constrained" ones are still stuck in their traffic jams. The result is a Capacity Trap: The system is in a state where the good inspector is trying their best, but the overall enforcement is getting worse and worse because the "bad luck" of the constrained type is dominating the signal.

The "Survivor" Problem

The paper also looks at what happens to the relationships that survive this chaos. Imagine you are an observer looking at a list of companies that have never been caught cheating. You might think, "Wow, these must be the best, most honest companies!"

The paper says: No, that's a trap.

Because the "Constrained" inspectors are the ones who can't catch the cheaters, the Strategic Companies (the cheaters) are most likely to survive if they are paired with a Constrained Inspector. The "Good Guys" survive too, but the "Bad Guys" paired with the "Stuck" inspectors are the ones who slip through the cracks and stay in the game the longest.

So, if you look at a list of "long-standing, trouble-free" relationships, you aren't seeing a list of the best companies. You are seeing a list of the worst pairings: Cheaters who got lucky enough to be matched with inspectors who physically couldn't catch them. The longer a relationship lasts without an audit, the more likely it is to be a disaster in disguise.

The Solution: A Higher Floor

The paper ends with a lesson for the people who design these inspection systems (the regulators).

Usually, people think, "We just need to increase the average number of audits." The paper says that's not enough. If you have a system where some inspectors are super-efficient (checking 90% of files) and others are broken (checking 0%), a long silence will destroy trust very quickly.

The solution is to raise the floor. Instead of having a few super-inspectors and many broken ones, you should ensure that every inspector has a minimum level of capacity (a "floor"). Even if you have to lower the peak performance of the super-inspectors to do it, it's worth it.

By making sure no inspector is "broken" (raising the minimum audit rate from, say, 2% to 20%), you make silence less scary. A period of silence no longer screams "We are broken!" because even the worst inspector can still do something. This delays the moment where companies decide to cheat and keeps the system from falling into the trap.

The Bottom Line

This paper proves that silence is not empty. It is a signal that changes how we see the world. If we don't understand the difference between "not trying" and "not being able to," we will misinterpret silence as weakness, which leads companies to cheat, which leads inspectors to work harder but fail more, and eventually, the whole system collapses into a state where the worst pairings survive the longest.

The authors show that to fix this, we shouldn't just aim for a high average; we need to make sure the worst case scenario is still good enough to keep the game fair. It's a reminder that in the game of enforcement, how you distribute your resources matters just as much as how many you have.

Technical Summary: Audit Silence and the Capacity Trap

Problem Statement
The paper investigates the informational content of "audit silence"—the public absence of an audit—in a repeated inspection game. Unlike a "clean audit" (an audit that finds no violation), silence admits two distinct interpretations: either the authority chose not to act (a statement about effort) or it could not act due to capacity constraints (a statement about implementation technology). The central problem is that these two readings are observationally equivalent in the data but lead to opposite strategic inferences. If silence signals a lack of capacity, a regulated firm may expect continued silence and adjust its behavior accordingly, potentially leading to a self-reinforcing deterioration of enforcement. The paper asks how a run of missing audits affects compliance, enforcement effort, and the composition of surviving firm-inspector relationships when the authority's capacity is a persistent, unknown type.

Methodology
The analysis employs a repeated inspection game with two-sided incomplete information and sequential equilibrium as the solution concept.

Players and Types: A regulated firm is either "committed" to compliance (always complies) or "strategic" (chooses between compliance and violation). An inspector is either "functioning" (strategic, choosing between low and high effort) or "capacity-constrained" (a persistent technological bottleneck that generates audits at a fixed, low rate $q$ regardless of effort).
Signals: The public signal space distinguishes between three outcomes: $N$ (no audit), $A_0$ (audit with no finding), and $D$ (detected violation, ending the relationship). Crucially, $N$ is separated from $A_0$ ; $N$ is informative about the inspector's capacity, while $A_0$ is informative about the firm's compliance and the inspector's detection accuracy.
Belief Dynamics: The model relies on a factorization property (Lemma 3.1, Lemma B.1) where the posterior distributions of the firm's type and the inspector's type remain independent after any public history, even when players condition on their own private action histories. This allows the state space to be reduced to two marginal probabilities: $\lambda$ (probability the firm is committed) and $\mu$ (probability the inspector is constrained).
Key Assumption (Capacity Separation): The model assumes $q < \alpha < \bar{\alpha}$ , where $q$ is the audit rate of a constrained inspector, and $\alpha, \bar{\alpha}$ are the low and high effort audit rates of a functioning inspector. This ensures that even a functioning inspector exerting minimum effort audits more frequently than a constrained one.

Key Contributions and Results

The Capacity Trap and Global Entry:
The paper demonstrates that a finite run of audit silence forces the game into a "capacity trap" (Theorem 5.1). Under the capacity separation assumption, each silent period increases the posterior odds of the inspector being capacity-constrained by a factor bounded away from one.
- Mechanism: After a bounded number of consecutive silent periods ( $K$ ), the aggregate audit probability falls below a threshold $\pi_\delta$ (the firm's dominance threshold), even if the functioning inspector exerts maximum effort.
- Outcome: Once this threshold is crossed, it becomes strictly optimal for the strategic firm to violate in every period. Simultaneously, the resulting high violation rate makes it strictly optimal for the functioning inspector to exert maximum effort.
- Paradox: Despite the functioning inspector exerting maximum effort, the realized audit rate continues to deteriorate because the posterior probability of the inspector being constrained ( $\mu$ ) rises. The deterioration is driven by composition (the mix of types) rather than slack in effort.
Adverse Survivor Selection:
The paper analyzes the composition of relationships that survive long enough to be observed (Section 7). While long duration selects for committed firms (who never trigger detection), it also selects for constrained inspectors.
- Result: Strategic firms matched with constrained inspectors are less likely to be audited and thus less likely to be detected. Consequently, among long-lived relationships, there is a negative covariance between firm compliance and inspector capacity. The "survivor sample" is disproportionately composed of harmful pairings (strategic firms with constrained inspectors), making duration a non-monotone measure of relational quality.
Capacity Design and Distributional Policy:
The paper argues that the distribution of enforcement capacity, not just its mean, is a policy object (Section 6).
- Variance Reduction: Holding the current maximum-effort audit rate fixed, reducing the variance of the capacity distribution (by raising the "floor" $q$ and lowering the "ceiling" $h$ ) delays the entry into the capacity trap.
- Welfare: A "mean-preserving contraction" of the capacity distribution strictly lowers expected social loss during a silence spell (Proposition 6.6), provided specific conditions on detection and exit costs are met. This suggests that investing in reliability (staffing floors, redundant systems) is distinct from and potentially more valuable than investing in peak investigative capacity.

Significance and Claims
The paper claims to fill a gap in the literature on dynamic enforcement and reputation games by distinguishing between monitoring outcomes (finding or not finding a violation) and monitoring incidence (whether an audit occurs at all).

Distinction from Literature: Unlike models where a "slippery slope" arises from failed detection (finding nothing), this mechanism relies on the absence of an audit. This allows the model to separate implementation capacity from detection accuracy.
Robustness: The entry into the capacity trap is a robust dominance result valid in every sequential equilibrium, regardless of the strategies played before the threshold is reached. It does not rely on patient-player limits or specific equilibrium selection.
Policy Implication: The results suggest that a declining audit rate is ambiguous: it could signal a regulator "going soft" (reducing effort) or a regulator "running out of capacity." The paper argues that without measuring administrative capacity separately, observers may draw the wrong inference. Furthermore, policy interventions should focus on the distribution of capacity (raising the floor) rather than just the mean, as this delays the onset of the capacity trap and reduces the adverse selection of harmful relationships.

The paper maintains that its results are contingent on the "capacity separation" assumption and a specific range of discount factors where stage-payoff dominance holds, but within these bounds, the mechanism of audit silence driving a capacity trap is a general feature of the environment.