The Inertia Problem: Why Underperformers Rarely Get Fired, and Why Everyone Else Picks Up the Slack
Firing an underperformer looks like simple accountability from the outside. From inside the organization’s own cost structure, it is often the more expensive choice — which is why so many managers quietly choose not to make it.
Why promotion rewards the wrong signal
The real dollar cost of replacing someone
Why high performers absorb more work
The math behind institutional inertia
In almost every organization of meaningful size, the same two complaints circulate quietly, usually not to management: one employee has been visibly coasting for over a year with no consequence, and another has been quietly absorbing that person’s overflow on top of their own workload, without a corresponding raise, title, or thank-you. Both observations are usually correct. What is less obvious is that neither one is really a story about character. It is a story about arithmetic.
The common explanation for tolerated underperformance is emotional: managers are conflict-avoidant, HR is slow, people are too “nice” to have hard conversations. There is truth in that, but it is an incomplete theory, because it cannot explain the second half of the pattern — why the same organizations that let low performers coast also load their best people with more work, more scrutiny, and more risk. A purely psychological account of niceness or conflict-avoidance would predict that high performers get an easier ride too. They do not. What connects both halves of the pattern is a set of well-documented findings from labor economics, organizational psychology, and behavioral decision theory about how firms actually evaluate the cost of acting versus the cost of doing nothing. Once that cost structure is visible, the behavior of real managers in real companies stops looking mysterious and starts looking, uncomfortably, rational.
This piece traces that logic from its origins in mid-century management theory through the empirical research that has confirmed and complicated it, works through the actual numbers a manager is implicitly weighing when they choose not to act, and considers what happens to that calculus as performance data itself becomes cheaper and more automated to collect.
Where the Asymmetry Comes From
The modern performance-management apparatus — reviews, ratings, improvement plans, promotion committees — descends from Frederick Taylor’s early-twentieth-century project of measuring and standardizing labor. Taylor’s premise was that output could be quantified and optimized like any other input. That premise still underwrites most corporate performance systems, but it obscures a harder problem Taylor never had to solve: once you can measure performance, what do you do about the person at the bottom of the distribution, and who decides?
Two developments, a few years apart, gave the modern version of this problem its shape. In 1969, management researcher Laurence J. Peter and co-author Raymond Hull published The Peter Principle, a book intended partly as satire, which proposed that employees in a hierarchy are promoted based on performance in their current role until they reach a role they perform badly — at which point promotion stops, and they stay there, indefinitely, doing that job poorly. It was a joke with an uncomfortable ring of truth, and it stuck. In 1976, economists Michael Jensen and William Meckling published a far more technical paper formalizing what they called agency theory: the idea that any relationship in which one party (a principal, such as a firm) delegates decisions to another (an agent, such as a manager or employee) generates unavoidable costs, because the agent’s interests never perfectly align with the principal’s, and monitoring that gap is itself expensive. Between them, these two ideas explain most of what follows in this article: Peter named the symptom, and Jensen and Meckling explained why organizations tolerate it.
From Taylor’s stopwatch to today’s engagement dashboards — how performance became something firms measure but still struggle to act on.
The 1980s and 1990s added an institutional overcorrection: General Electric, under Jack Welch, popularized “forced ranking” or “stack ranking,” a system that required managers to sort employees into performance tiers and systematically remove the bottom ten percent every cycle. It was marketed as rigor. In practice, it optimized for a statistical assumption — that performance is normally distributed, so there will always be a definable bottom decile worth cutting — that later research would show is false for most kinds of work. Stack ranking was largely dismantled across the industry by the early 2010s, not because companies became more forgiving, but because forcing a fixed percentage of terminations regardless of actual performance turned out to be expensive, demoralizing, and legally exposed. What replaced it, in most organizations, was not a better system for identifying and removing underperformance. It was less system, and more discretion — which, as the next section explains, defaults toward inaction.
Modern performance systems inherited a measurement tradition without ever solving what to do about the people at the bottom of the measurement.
Working Definitions
A handful of terms recur through the rest of this piece. None of them require specialized training to follow, but they are precise concepts with specific research histories, so it’s worth defining them plainly before using them loosely.
Principal-agent problem
The gap between what an organization (the principal) wants and what the individuals it delegates decisions to (its agents — managers, employees) actually do, because their incentives are never perfectly aligned. A manager avoiding a difficult firing decision to protect their own comfort, at some cost to the company, is a textbook instance.
Information asymmetry
A situation where one party in a decision — often the employee — has more accurate information about their own effort, ability, or intentions than the other party — often the manager — who has to act on incomplete evidence.
Status quo bias
A well-documented tendency to prefer the current state of affairs over a change, even when the change would be net-positive, largely because potential losses from acting are weighted more heavily than equivalent potential gains.
Power-law (Paretian) distribution
A statistical pattern in which a small number of cases account for a disproportionate share of the total — unlike a bell curve, where most cases cluster near the average. Individual job performance, across a wide range of professions, has been shown to follow this pattern rather than a normal distribution.
Regression to the mean
The statistical tendency for an unusually strong (or weak) measurement to be followed by a more average one, simply because extreme results are partly due to circumstance, not just underlying skill — a key part of why star performers often look less exceptional in a new role.
The next section leans on some decision-theory framing and a bit of arithmetic. None of it requires math beyond addition, multiplication, and percentages — it’s the concepts, not the computation, that take a moment to sit with.
The Mechanics of Doing Nothing
Jensen and Meckling’s agency framework treats every organization as a chain of principal-agent relationships rather than a single unified actor. A CEO is an agent of the board; a manager is an agent of the CEO; an employee is an agent of the manager. At every link in that chain, the person making a decision bears a personal cost for acting that the organization as a whole does not fully share, and vice versa. Firing an underperformer is a clean illustration. The manager who initiates it absorbs the immediate, personal, visible costs — the uncomfortable conversation, the documentation, the risk of a wrongful-termination claim, the team’s morale dip, the burden of covering the role until it’s refilled. The organization absorbs the long-run, diffuse benefit — slightly better average performance, months from now, spread across a P&L the manager doesn’t personally see. That mismatch, not laziness or conflict-aversion, is the first load-bearing piece of the puzzle.
Firing an underperformer concentrates its costs (paperwork, risk, discomfort) on the manager who decides, while its benefits (better team output) are diffused across the whole organization over time. Economists call this a principal-agent problem — and it predicts inaction even when everyone agrees the underperformance is real.
The second piece is status quo bias, formalized by economists William Samuelson and Richard Zeckhauser, building on Daniel Kahneman and Amos Tversky’s earlier work on loss aversion. Their finding, replicated widely since, is that people do not weigh a potential loss and an equivalent potential gain symmetrically — a loss looms larger in the decision than a same-sized gain, even when the two are numerically identical. Applied to a firing decision: the known, chronic cost of an underperforming employee (mediocre output, some team grumbling) is psychologically discounted, because it’s familiar and already “priced in.” The uncertain, front-loaded cost of firing and replacing that person (open role, interviews, a stranger’s unknown ceiling, months of ramp-up) is weighted more heavily, because it’s new, uncertain, and reversible-feeling only in theory.
Think of it like a leaking faucet you’ve lived with for a year versus calling a plumber: the drip is annoying but familiar; the plumber visit is a known, immediate cost with an uncertain outcome. Most people delay the plumber far longer than the math of the water bill would justify.
A third piece, from research by Ernest O’Boyle and Herman Aguinis, complicates the story further: individual job performance does not follow the bell curve most performance-review systems quietly assume. Across five large studies covering over 600,000 people in a range of professions, O’Boyle and Aguinis found that performance instead follows a power-law, or Paretian, distribution — a small number of people account for a hugely disproportionate share of output, while the rest cluster much closer together than a normal curve would predict. This matters for the asymmetry described here because it means the “gap” between a mediocre performer and an average one is often smaller, and harder to justify firing over, than the gap between a great performer and everyone else — which is one reason organizations find it easier to recognize and reward excellence at the top than to act decisively on mediocrity in the middle.
The same three-stage loop repeats at every review cycle unless the underlying cost structure changes, not just the manager’s resolve.
Put these three pieces together and a manager’s implicit decision looks less like a character judgment and more like a cost comparison — one that, as the next section shows with real figures, frequently tilts toward inaction even for a manager who is neither lazy nor cowardly.
Inaction on underperformance is what rational cost-minimization looks like when the person deciding bears the risk and the organization gets the benefit.
A Manager’s Decision, Worked in Numbers
Here is the kind of back-of-envelope comparison a manager is, whether consciously or not, running when they decide whether to formally address an underperforming employee. This is deliberately simplified — a real HR decision involves legal counsel, documentation history, and team dynamics that don’t reduce to five lines of arithmetic — but the simplified version is exactly the intuitive shortcut most managers actually use, and it’s enough to show why the “obvious” choice isn’t always the cheap one.
The scenario: An employee earning $65,000 a year is, by the manager’s own honest assessment, completing roughly 65% of the volume and quality of work the role calls for — a 35% shortfall.
This worked example is a simplified heuristic, not a rigorous return-on-investment model — it ignores morale costs to the rest of the team, compounding effects across multiple years, and the fact that “35% shortfall” is rarely as cleanly measurable as it sounds. It’s presented because it approximates the mental shortcut real managers use, not because it’s a precise instrument.
Try the comparison yourself with the sliders below. Adjust the performance gap and the replacement cost assumption to see where the breakeven point actually sits — and how rarely it favors quick action.
Assumes a $65,000 salary and a four-month ramp-up period at 50% productivity for the replacement. Watch how rarely the bars cross.
Even honest, straightforward arithmetic tends to favor tolerating underperformance over the short-term cost of correcting it.
The Asymmetry in Practice
If the cost structure explains why low performers are protected by inertia, it also explains — through the same mechanism, pointed in the opposite direction — why high performers are handed more work, more scrutiny, and less patience. A 2016 Harvard Business Review study by Rob Cross and Reb Rebele, drawing on data from more than one hundred organizations, found that 20% to 35% of an organization’s value-adding collaboration comes from just 3% to 5% of its employees. Those employees become the default routing point for requests precisely because they’re reliable — and because redirecting a request to someone less reliable carries the same asymmetric risk described above: a known cost (this person is busy) against an uncertain one (will the other person actually deliver).
The best individual performer is promoted to manager because they’ll naturally be a good one.
Economists Alan Benson, Danielle Li, and Kelly Shue studied promotion decisions for sales employees across 131 firms and found that the workers most likely to be promoted into management were, on average, the same ones whose performance declined most sharply after the promotion — current-role performance is a poor predictor of managerial skill.
This is the empirical confirmation of the Peter Principle Laurence Peter proposed as satire half a century earlier — and it deepens the asymmetry rather than resolving it. High performers face a double bind: excel in place, and the organization routes more work to you without necessarily promoting you; excel enough to get promoted, and you’re statistically likely to be worse at the new job than you were at the old one, through no fault beyond the fact that the two jobs require different skills. Either path adds pressure without a matching increase in support.
How the same organizational inertia looks from opposite ends of the performance range.
| Criteria | High Performer | Low Performer |
|---|---|---|
| Visibility to leadership | High — output is easy to point to | Low — shortfalls diffuse across a team |
| Documentation needed to act | None — recognition is informal | Extensive — termination requires a paper trail |
| Career risk to the manager | Low — rewarding success is safe | High — firing carries legal and morale exposure |
| Cost if they leave | Very high — hard to replace, per HBR | Moderate — but replacement still costs 50–200% of salary |
| Typical organizational response | More responsibility, same pay | Quiet tolerance or informal reassignment |
Gallup’s most recent State of the Global Workplace research adds the macro picture: global employee engagement has fallen to roughly 21%, and the steepest declines are concentrated among managers, who account for an estimated 70% of the variance in their team’s engagement. A stressed, under-supported manager is, unsurprisingly, the least likely person to have the bandwidth for a difficult termination conversation — and the most likely to lean on whoever is already reliable. Forbes reporting on a 2026 Leadership IQ survey of nearly 700 HR executives found that they trust, on average, only 35% of their managers to handle a genuinely difficult performance conversation without HR in the room. That is not a description of widespread cowardice. It’s a description of an undertrained, underengaged managerial layer being asked to make a decision whose personal cost it bears and whose benefit it mostly doesn’t.
None of this is an argument that low performance is blameless or that high performers should simply accept more work. It’s a description of the incentive structure that produces both outcomes — a necessary first step before either can be meaningfully changed.
The same cost asymmetry that protects underperformers also quietly taxes the people already exceeding expectations.
Where This Goes Next
The clearest lever on this system is the one Jensen and Meckling identified in 1976: information asymmetry. Much of the inertia described in this piece survives because performance is expensive and ambiguous to measure, which makes the “known cost of tolerance” easy to underweight and the “unknown cost of replacement” easy to overweight. As continuous, automated performance data — output tracked directly from the tools people already use, rather than filtered through an annual review — becomes cheaper to collect, the ambiguity that protects the status quo shrinks. In principle, this should make firing decisions less personally risky for managers, because the case no longer rests on their subjective judgment alone.
In practice, it introduces a new principal-agent problem rather than resolving the old one. Automated performance data has to be built, weighted, and interpreted by someone — a new layer of agents whose incentives may not align with the organization’s, or the employee’s, any better than a manager’s did. An algorithm trained on flawed proxies for “performance” (time logged, messages sent, tickets closed) can encode the same power-law-blind assumptions that made stack ranking fail, just with a faster feedback loop and less room for a manager’s contextual judgment to intervene. The 2022 mainstreaming of terms like “quiet quitting” and “quiet firing” — employees doing the minimum defensible work, and managers making conditions unpleasant enough that people leave voluntarily rather than being formally terminated — is best read as a preview of this dynamic: informal, low-documentation responses to a system that has made the formal, high-documentation response too costly for either side to reach for.
Better data reduces information asymmetry, which should lower the personal risk of acting on underperformance — but only if someone is also auditing the data itself, or the system just relocates the same principal-agent problem one level up.
The more durable fix isn’t technological — it’s structural. Organizations that have made real progress on this asymmetry have generally done so by moving the cost of the decision off the individual manager: shared accountability for performance-management outcomes, standardized processes that reduce a single manager’s personal legal exposure, and — as SHRM’s own research on underperformance conversations has repeatedly found — direct manager training, since the discomfort driving avoidance is often a skills gap as much as a personality trait. None of this makes the underlying trade-off disappear. It just moves who bears it, which is, in the end, the entire subject of this article.
Better data can lower the cost of acting on underperformance, but only if a new information asymmetry doesn’t simply take its place.
The Real Trade-off
It is tempting to read all of this as a cynical excuse for bad management, or as proof that effort simply doesn’t matter. Neither conclusion follows. What the research actually shows is narrower and, in a way, more useful: organizations are not running a coherent meritocracy that occasionally fails. They are running a cost-minimization heuristic under uncertainty, and merit is one input among several — alongside documentation burden, legal exposure, replacement cost, and a manager’s own bandwidth — that heuristic happens to weigh imperfectly.
Inertia is not a moral failure of management; it is what rational cost-minimization looks like from the outside.
That reframing matters because it changes where the useful intervention sits. Telling a manager to “just have the hard conversation” treats a structural cost problem as a personal courage problem, and courage alone rarely survives contact with a genuinely mismatched incentive structure. The organizations that actually shift this pattern are the ones that change the arithmetic itself — lowering the personal cost of documentation, shortening replacement ramp-up time, distributing accountability so no single person absorbs all the risk of acting. Recognizing the mechanism doesn’t excuse the outcome. It just means the fix has to be aimed at the incentive, not the individual standing inside it.
For the person who has spent the last year quietly picking up someone else’s unfinished work, that reframing offers something short of consolation but more useful than blame: the unfairness is real, it is not a reflection of your value being miscalculated by people who don’t see it, and it is not, in most cases, a decision anyone consciously made against you. It is the visible residue of a much older, much more mechanical problem — one that has had a name since 1976, and a great deal of supporting data since.
Not really. Being tolerated is not the same as being valued, and most people who coast at work know it — the research on disengagement suggests it correlates with lower personal wellbeing, not higher. The organization isn’t rewarding low effort on purpose; it’s declining to pay the cost of correcting it.
They’re related but distinct. Quiet firing is a manager’s deliberate tactic to push someone out informally. The inertia described here is closer to the opposite — a default toward doing nothing at all, whether the person deserves to stay or not.
It can reduce the ambiguity that protects the status quo, but as Section 6 discusses, it introduces a new question of who audits the data itself — the same principal-agent problem, one level up.
Documentation and due-process requirements exist to prevent wrongful or discriminatory termination — a legitimate goal. This article isn’t arguing the legal friction is wrong, only that it’s a real, measurable part of the cost calculus managers are weighing.
That’s a personal and workplace-specific question this piece isn’t positioned to answer with a formula. What the research does support is naming the pattern explicitly to a manager — making the invisible routing of extra work visible tends to be more effective than either silently absorbing it or silently resenting it.
Dr. Julian Mercer is a Science & Technology Research Writer at Infinite Loop, covering artificial intelligence, computational theory, and the systems that govern how organizations make decisions under uncertainty.