The Consequence Gap
Companies got harsher with everyone except the people the harshness was for.
Same violation. A different rulebook for the top performers.
A November 2026 national experiment with 2,000 US workers on how an employee's performance level changes the way managers judge identical misconduct, and what coworkers expect once they report it.
A GCheck Compliance for Good® Report, part of The Trust Recession research series.
of managers backed a policy exception for a top performer after a confirmed violation, against 27% for a typical performer.
of managers given a top performer were sure an ambiguous charge was "very likely" an honest mistake, against 24%.
recommended formal action for a low performer, against 68% for a typical one: a premium for stars, not a penalty for strugglers.
of managers say policy should apply equally. 59% say stars get more leniency where they work.
of workers and managers have personally seen a star avoid consequences or a struggler face harsher ones.
Executive Summary
Managers were asked to decide one case: an employee submits the same expense twice. Each manager saw one version, and the versions differed only in how much was known and how the employee was described. No manager judged more than one employee, so any gap between versions comes from the description, not from the manager.
49% of managers given a confirmed, knowingly submitted duplicate expense supported an exception for a top performer, against 27% for a typical performer and 25% for a low performer (sensitivity sample; on all completes the figures are 44%, 33% and 29%). Investigation and discipline barely moved. Roughly two in three managers recommended formal warning, discipline or escalation whoever the employee was, and investigation ran at 95% to 96% in every confirmed version. The gap opened in one decision: whether this case, for this person, deserved an exception.
Before the facts were settled, managers given a top performer were also half again as likely to be sure the charge was an honest mistake (37% vs 24% “very likely”). The mirror image never appeared: a low performer received the same outcomes as a typical one. What changed was the explanation. 41% of managers deciding a low performer’s case pointed to reputation, against 24% who pointed to the incident itself. And managers know the standard they are meant to hold: 81% say policy should apply equally regardless of performance; 59% say high performers get more leniency where they work.
Every figure names its question, base and label. INCIDENCE means something the respondent has done or seen; PERCEPTION means a judgment or likely action. Experimental results are stated intentions on a hypothetical case and show association with the described performance level, not proof of how any manager acts. Where the sensitivity sample differs materially from all completes, both appear. Methodology and the findings that cut against this report are at the end of the page.
“The rules held. The exceptions bent.”
The premise
Rule-breaking at work is ordinary. 81% of respondents have seen at least one of six kinds of violation in a current or past workplace. Expense violations, the case used in this study’s experiment, sit near the bottom of that list, which makes the scenario a concrete, lower-frequency test: easy to picture, rarely routine.
Managers report more of almost everything, most sharply expense violations (25% vs 11%), consistent with managers being the people who approve the claims. The standard story says a rule is a rule. This study tested the part of that story nobody audits: whether the same rule reads the same way when the person who broke it is the team’s best performer.
The tendency for an employee's performance level to shift how managers interpret and respond to identical conduct. It names a pattern in the data, never a label for any individual manager or employee.
Violations personally witnessed in a current or past workplace (select all), managers beside non-managers.
How the study was run
GCheck surveyed 2,000 employed US adults on Pollfish on November 5, 2026, weighted by age, gender and region. 893 supervise at least one direct report (45% weighted). Every respondent was randomly assigned one version of the same case: an employee submits a duplicate expense claim. Only two things vary: how much is known, and how the employee performs.
No manager saw more than one employee. That is what makes the gaps readable.
Would they investigate, document, discipline or grant an exception, and what drove that choice?
Q11 to Q19Would they report it, and would they expect management to handle it the same way for anyone?
Q9, Q1032% of managers misread whether their version was ambiguous or confirmed, most often in the ambiguous versions. Following the pre-registered rule, every experimental result was computed twice: on all completes, and on a sensitivity sample that sets aside responses flagged for speeding, straight-lining or a failed comprehension check (524 of 893 managers retained). The sensitivity figure leads a headline only where the two differ materially and its base reaches n=100.
The benefit of the doubt
When it was not yet clear whether the duplicate was a mistake, managers given a top performer were far more certain that it was. The facts were identical across both versions; only the description of the employee changed.
The gap sits in conviction, not direction. Managers extended good faith broadly: three in four leaned toward an honest mistake either way (77% vs 74%), a difference too small to read as real. A top performer simply got that good faith with more certainty, and in the sensitivity sample the “very likely” gap roughly doubles (48% vs 23%).
Experience did not close it. The 13 to 14 point gap held among managers with under 4, 4 to 9, and 10+ years of experience. Tenure tracks age closely in this sample (average age 36, 43 and 50 across the bands), so the experience result is not read as a pure experience effect.
How likely is it this was an honest mistake? Managers in the ambiguous versions.
“The facts never changed between versions. Only the reputation did.”
The exception
Once a review confirmed the duplicate was submitted knowingly, managers still investigated and documented almost universally. Performance showed up in one place: the willingness to set the policy aside.
The exception row is the largest, most consistent difference in the study. On all completes the gap is 11 points (44% vs 33%). In the sensitivity sample it widens to 22, which makes a top performer 1.8 times as likely to be granted an exception for the same knowing violation.
Managers given a confirmed, knowingly submitted duplicate expense, sensitivity sample. On all completes: typical 33%, top 44%, low 29%.
The gap grows with experience: the top-performer advantage in support for an exception was +3 points among managers with under 4 years of experience, +14 points at 4 to 9 years, and +21 points at 10+ years (Q15 · Q6, n=95 to 139 per band, directional, age-confounded). Seasoned managers endorse equal enforcement at the same 81% as newer ones.
Investigation, documentation and discipline barely moved across the confirmed versions.
| Manager response to the confirmed violation | Typical (C) | Top (D) | Low (E) |
|---|---|---|---|
| Very or somewhat likely to investigate further (Q12) | 96% | 96% | 95% |
| Very likely to investigate further (Q12) | 62% | 65% | 71% |
| Very or somewhat likely to formally document (Q13) | 94% | 87% | 97% |
| Recommend formal warning, discipline or escalation (Q14) | 68% | 64% | 70% |
| Support an exception to the usual policy (Q15) | 33% | 44% | 29% |
| Support an exception, sensitivity sample (Q15) | 27% | 49% | 25% |
Q12 to Q15 · PERCEPTION · n=173 / 167 / 179 (sensitivity n=105 / 100 / 121). The documentation gap (87% vs 94%) disappears in the sensitivity sample and is treated as fragile.
“The rules held. The exceptions bent toward the stars.”
The correction
The study was built to test a second idea too: that the same written policy lands harder on a struggling employee. If that were true, the low performer in version E should have drawn more discipline and fewer exceptions than the typical performer in version C.
None of the outcome differences is distinguishable from chance. The single hint of extra scrutiny is intensity: more managers would be “very likely” to dig further into a low performer’s claim, a gap that reaches significance only once flagged responses are set aside.
The pre-registration committed to reporting this outcome plainly, without hunting for a smaller effect to replace it. The finding stands as registered: a premium for top performers, not a penalty for low performers.
What managers recommended in the confirmed versions: low performer (E) beside typical performer (C).
“Performance moved outcomes in one direction only.”
What drives it
Managers who decided the confirmed top-performer and low-performer cases were each asked for the main factor behind their decision, from lists written for each direction. No single reason reached a majority in either case.
For the top performer, leniency is explained more often through trust than through fear of losing someone (30% vs 21%), though one in five managers names retention directly. For the low performer, reputation outweighed the incident (41% vs 24%), and that holds in the sensitivity sample (44% vs 33%).
This is the study’s quiet tension. Managers deciding a low performer’s case say reputation shaped their judgment, yet their recommended outcomes match a typical performer’s. 78% of managers who decided a low performer’s case say overall performance influenced their decision a moderate amount or a great deal (Q19, n=346).
Each manager named one main factor, chosen from a list written for the top-performer case.
Each manager named one main factor, chosen from a list written for the low-performer case.
“Reputation shows up in how a decision is explained before it shows up in the decision.”
The trust gradient
Before any violation enters the picture, managers already trust their best people far more than the rest: 83% trust the judgment of their highest performer a great deal or quite a bit, against 61% for a typical performer and 32% for their lowest. Day-to-day review and discretion do not follow that slope.
The trust gap is the one a manager carries into an ambiguous case, and it matches where the experiment found its effect: certainty that a top performer meant no harm. Review runs nearly flat, and only the lowest performer is reviewed somewhat more. Discretion drops from the highest to the typical performer, then stops falling. 36% of managers review their top performer more closely than peers, against 22% who review them less closely (Q22, n=893).
Because these questions followed the scenario, results were checked by version. The trust slope held in all five; review and discretion moved more from version to version.
Managers who trust each performer's judgment a great deal or quite a bit.
Highest performer beside typical performer: how managers review work and delegate how it is done.
“Managers do not watch their stars less. They trust them more.”
The coworker's view
Non-managers read the same five versions as a coworker who noticed the duplicate. They did not hold back on top performers: 65% said they were very or somewhat likely to report it or raise it with a manager across all five versions (Q9, n=1,107).
The registered expectation, that coworkers would be less willing to report a star, came back reversed in the ambiguous versions and null in the confirmed ones. What changed was confidence in the outcome. 38% of coworkers reading the confirmed top-performer version did not say they expected equal handling.
That gap matters because unreported problems are already the norm. The Equal Employment Opportunity Commission found that roughly three in four people who experience workplace harassment never report it, most often out of fear of disbelief, inaction or retaliation (EEOC, 2016). A coworker who doubts the outcome has one more reason to stay quiet next time.
Non-managers: would report it or raise it with a manager, and whether they expect management to handle it the same way for anyone.
“Coworkers will raise it. They are less sure it will count.”
Two rulebooks
Ask managers what fair enforcement looks like and they agree almost unanimously. 81% agree company policy should be enforced the same way regardless of performance. 59% agree that at their own organization, high performers receive more leniency when they break a rule; 21% disagree. Half of all managers (50%) agree with both.
The admission is highest among managers at organizations of 50 to 249 employees (66%) and lowest under 50 employees (49%) (Q25 · Q3, n=181 to 296 per size band).
Leniency is reported more often than harshness. That lived experience runs ahead of the experiment, which found a premium for stars but no penalty for strugglers in stated decisions. Real cases are rarely as clear-cut as one duplicate expense, and perceived unfairness is easy to remember.
Share of managers agreeing with each statement.
Personally seen a star avoid consequences or a struggler face harsher ones.
“Managers endorse one rulebook and describe a second.”
The ask
Nothing on this list asks for gentler standards or closer watching. Managers lean toward training and documentation; non-managers lean toward a safe channel and leadership example. Read against the experiment, the most direct lever is the documentation row: a written reason for every exception, applied in both directions, puts a record behind the one decision where performance measurably moved the outcome. SHRM’s culture research places the responsibility where respondents do: 76% of employees say their manager sets the culture of their workplace (SHRM, 2019).
Share selecting each option (multi-select).
An exception granted without a written reason leaves the next person with the same conduct and a weaker record on the other side of that comparison, the kind of inconsistency comparator evidence in employment disputes turns on. This report describes that risk category; it is not legal advice.
“Workers are not asking for softer rules. They are asking for one rulebook.”
The bottom line
The comfortable version of this story is that a rule is a rule. The data tells a narrower one. Investigation held. Discipline held. Most managers acted the same way whoever the employee was, and a low performer was not punished harder. What moved was a single decision: whether this case, for this person, deserved an exception. After a confirmed, knowing violation, a top performer was 1.8 times as likely to get one.
The difference starts earlier, in how much certainty the same facts earn, and it travels to the people watching. Coworkers will still report a star, but fewer expect it to count. Managers already know the standard. Four in five say policy should apply equally, and six in ten describe a workplace where it does not. Workers named the fix, and it is not a softer rulebook or a closer watch. It is one rulebook, with a written reason behind every exception.
Methodology and honesty notes
| Design element | Specification |
|---|---|
| Sample | 2,000 employed US adults (1,620 full-time, 380 part-time), via Pollfish, fielded November 5, 2026, weighted by age, gender and region. Non-probability opt-in panel. |
| Design | Between-subjects randomized experiment, five versions (A to E). Managers n=893 answered as decision-maker; non-managers n=1,107 answered as coworker. No role reweighting. |
| Sensitivity sample | Flags: under 90 seconds (managers) or 45 seconds (non-managers), identical trust and review matrices, or failed comprehension check. 369 managers and 62 non-managers flagged; 524 managers retained. |
| Question IDs | All IDs reference the fielded Pollfish instrument (survey 395606961). |
| Labels | Every figure carries question ID, base n, and an INCIDENCE or PERCEPTION tag. |
| Discipline | No composite scores or indexes; matrix rows and multi-selects reported individually. Race and ethnicity collected for description only. Subgroups under 100 directional, under 30 not reported. |
| Deviations | Plan assumed two surveys of 750 with a manager quota; fielded as one survey of 2,000 at the natural manager share. Coworkers saw five versions instead of three. No registered direction changed. |
Most managers acted the same way whoever the employee was: investigation stayed at 95% to 96% and formal discipline within 64% to 70% (Q12, Q14, n=519). Seriousness of the issue was the most cited reason in both directions (19% and 24%). Repeat behavior mattered more than reputation: 67% would respond more harshly to a second incident, at similar rates for every performer (Q16, n=519). Top performers are not reviewed less day to day (69% vs 69%, Q21). 27% of respondents have never witnessed unequal treatment, and 7% are unsure (Q26). About a third of managers misread the scenario; for the exception and benefit-of-the-doubt findings, setting them aside makes the effect larger, not smaller.
Every experimental measure is a stated intention in response to a hypothetical case. People tend to understate behavior that reflects unevenly on their own fairness, so the measured gaps are best read as a floor. Differences between versions are reported as association with the described performance level, not proof of how any manager acts. The named terms (The Performance Privilege, The Performance Privilege Gap, The Exception, Two Rulebooks) describe patterns in the data, never individuals, and appear as the name of no statistic. No monitoring or surveillance remedy appears anywhere in this report. The full questionnaire and dataset documentation are available on request.
Frequently asked questions
The Performance Privilege is the tendency for an employee's performance level to shift how managers interpret and respond to identical conduct. It names a pattern in the data, never a label for any individual manager or employee.
After a confirmed, knowingly submitted duplicate expense, 49% of managers supported an exception to the usual policy for a top performer, against 27% for a typical performer and 25% for a low performer (sensitivity sample). That makes a top performer 1.8 times as likely to get an exception. On all completes the figures are 44%, 33% and 29%.
No. A low performer received the same outcomes as a typical performer: 70% vs 68% of managers recommended formal warning, discipline or escalation, and 29% vs 33% supported an exception. The study found a premium for top performers, not a penalty for low performers.
Yes, with more certainty. When it was not yet clear whether a duplicate expense was a mistake, 37% of managers given a top performer were sure it was "very likely" an honest mistake, against 24% for a typical performer. Three in four managers leaned toward good faith either way (77% vs 74%).
81% of managers agree company policy should be enforced the same way regardless of performance, yet 59% agree that at their own organization high performers receive more leniency when they break a rule. Half of all managers (50%) agree with both statements.
Mostly yes. 65% of non-managers said they were very or somewhat likely to report a duplicate expense or raise it with a manager across all five versions. But fewer expected equal handling once they did: 62% for a top performer vs 72% for a typical performer.
Two in three have seen it. 66% of workers and managers have personally seen a star avoid consequences or a struggler face harsher ones, and 32% have seen both. 27% have never seen either.
Workers ranked clear written criteria applied the same way to everyone first (57%), then a safe way to raise concerns about inconsistent treatment (52%), manager training on separating trust from accountability (45%), documentation required for any exception (44%), and leadership modeling consistent standards (41%).
The report recommends four practice changes: require a written rationale, reviewed outside the reporting line, for any exception to a conduct or expense policy; record the same incident details for every employee whatever their rating; give coworkers a channel to raise inconsistent treatment and close the loop; and train managers to name what is driving a decision.
GCheck surveyed 2,000 employed US adults on Pollfish on November 5, 2026, weighted by age, gender and region; 893 supervise at least one direct report. Each respondent was randomly assigned one of five versions of the same duplicate-expense case, varying only how much was known and how the employee performs. Results were also computed on a sensitivity sample of 524 managers.