Toxic Behaviour and Tolerated Underperformance — Costs and Consequences

The cost of keeping the wrong person on a team is real, but it almost never arrives as a single, itemized bill. It shows up downstream — as the strong people quitting, as a severance liability that grows every year a decision is deferred, and as performance problems that were never named and so never fixed. This page sets out what the evidence actually supports about each of those consequences, and links every claim to the source behind it.

It is written with the source quality kept in view. The numbers in this area range from one carefully-designed study to a widely-repeated figure that traces back to an unpublished dissertation reported on the radio. Each is reported faithfully to its origin, and the confidence in each claim is stated section by section rather than flattened into a single reassuring headline. The dollar figures here are directional anchors that illustrate the structure of the cost; they are not numbers to paste into a Kitchener–Waterloo business case.

What does a single toxic employee actually cost?

The single best-sourced dollar figure comes from Housman and Minor’s 2015 Harvard Business School working paper, which puts the induced-turnover cost of a toxic worker at about $12,500 — the expense of replacing the additional coworkers who quit because that person is on the team — and finds that avoiding a toxic worker is worth roughly twice as much to the firm as adding a top-1% “superstar.”

The estimate comes from data on over 58,000 hourly frontline service workers across 11 firms. The roughly $12,500 figure is the induced turnover cost: the expense of replacing the additional coworkers who quit in response to that person being on the team. The authors call it a lower bound — it excludes litigation, regulatory penalties, lost morale, and the new-hire learning curve. They compare it to the value of a top-1% “superstar,” which they put at about $5,300 in cost-savings. The headline that follows is that avoiding a toxic worker is worth roughly twice as much to the firm as adding a superstar.

What was measured matters as much as the number. The authors did not survey attitudes or self-reported personality. They defined a toxic worker as one actually terminated for an egregious policy violation — harassment, violence, fraud, or falsifying documents. About 1 in 20 workers was ultimately terminated this way. Because workers were quasi-randomly placed across workgroups, the authors argue the estimate is causal, not merely correlational — a genuinely strong design feature for this kind of question. They also found that toxic workers are faster — more productive in raw speed — which helps explain why managers tolerate them in the first place: the cost is paid through the people who leave, not in the toxic worker’s own output.

The caveats matter for a small or mid-sized Ontario employer. This is a single, non-peer-reviewed working paper, drawn from US large-enterprise, high-turnover, hourly call-centre data. The $12,500 figure should be treated as a directional anchor that illustrates the structure of the cost — contagion-driven turnover dominates — not a number to transplant into a Kitchener–Waterloo business case. The generic per-departure replacement dollars are a separate matter (the firm’s “Rule of Three” frames replacement cost in its own terms). This study is the quantified spine of the “Saints and Vampires” idea: the measurable price of a vampire is paid mostly through the saints who leave because of them.

Source: Housman & Minor, Toxic Workers, Harvard Business School Working Paper 16-057 (2015).

Confidence: single-source. One rigorous, causally-designed working paper on US large-enterprise hourly data; the dollar figure is a directional anchor, not a transferable local number.

Does one bad apple really spoil the team — and drive the good people out?

The peer-reviewed evidence supports the mechanism even though the famous percentage does not. One persistently negative member degrades group functioning, and coworkers’ job-search behaviour and low embeddedness predict an individual’s own quitting over and above their own attitudes — so tolerating a corrosive team member predictably raises the odds that others, including strong performers who leave most easily, disengage and quit.

Answering this question well requires separating two evidence streams that are usually blurred together.

The peer-reviewed core. Felps, Mitchell and Byington (2006), “How, When, and Why Bad Apples Spoil the Barrel,” is a conceptual review and integrative model — not an experiment. It argues that one persistently negative member — withholding effort, expressing negative affect, or violating interpersonal norms — triggers states in teammates such as perceived inequity, negative feelings, and reduced trust, which together degrade group functioning. Separately, Felps and colleagues (2009), “Turnover Contagion,” is a genuine empirical multilevel study (45 bank branches and 1,038 hospitality departments) that found coworkers’ job-search behaviour and low embeddedness predict an individual’s own quitting, over and above that person’s own attitudes. That is the documented contagion channel: quitting behaviour spreads. It is observational field data rather than a controlled experiment, but the effect is real and measured.

The famous “30–40%” figure. The claim that one bad apple cuts team performance by 30 to 40 per cent comes from Will Felps’s University of Washington PhD dissertation experiment (around 2007), in which an actor named “Nick” played a Jerk, a Slacker, or a Downer inside 40 student groups. This experiment was never published as a standalone peer-reviewed article. The number reached the public through NPR’s This American Life (2008) and Daniel Coyle’s The Culture Code (2018). It should be labelled unpublished-dissertation data popularized via media — compelling and directionally consistent with the peer-reviewed literature, but not a replicated, peer-reviewed effect size. There is also a near-universal online error to avoid: many sources call the 2006 review “the experiment.” It is not; the review is conceptual and the experiment is the unpublished dissertation.

For a 20-to-200-person Ontario employer, the practical reading is robust even where the precise percentage is not. Tolerating a corrosive team member predictably raises the odds that others — including strong performers, who leave most easily — disengage and quit. That is the substance under “Saints and Vampires”; the exact percentage is the least reliable part of it.

Sources: Felps, Mitchell & Byington, How, When, and Why Bad Apples Spoil the Barrel (a conceptual review, not the experiment), Research in Organizational Behavior 27 (2006); Felps, Mitchell, Hekman, Lee, Holtom & Harman, Turnover Contagion, Academy of Management Journal 52(3), pp. 545–561 (2009); Coyle, The Culture Code (reproduces the unpublished “Nick” experiment figure), coverage.

Confidence: industry-consensus on the mechanism (one corrosive member degrades functioning and spreads quitting); the 30–40% figure is unpublished-dissertation data via media and is not a peer-reviewed effect size.

What does tolerating underperformance cost — and why does delay make termination more expensive?

The direct dollar cost of a tolerated poor performer is thinly evidenced and largely practitioner commentary, but one legal principle is firm: in Ontario, statutory and common-law termination entitlements generally rise with length of service (and age and role), so every additional year a justified exit is delayed predictably increases the eventual payout — and weakens the documentary record for asserting cause.

Two compounding costs sit behind a tolerated underperformer.

The productivity and morale cost (directional, thin evidence). That a poor performer drags down output and shifts burden onto stronger colleagues is intuitive and widely asserted, but the direct dollar evidence is largely practitioner and vendor commentary — SHRM articles and consultancy blogs — not rigorous primary research. The defensible, sourced version of the morale claim routes through the turnover-contagion literature covered above: the real risk of tolerating a low performer is that high performers — who carry the slack and notice the double standard — disengage or leave. Any specific “X% productivity loss” figure in this area should be treated as uncorroborated unless it can be traced to a primary study.

The rising legal exposure (firmer). Under Ontario’s Employment Standards Act, 2000, statutory termination and severance entitlements generally accrue with length of service — roughly a week per year of service, within statutory caps, with severance pay applying to qualifying employees at larger payrolls. Separately, common-law reasonable notice — driven by the Bardal factors of age, length of service, character of employment, and availability of comparable work — has no fixed cap, and for long-service senior employees it can approach the roughly 24-month range. The unifying principle is that entitlements generally grow with tenure (and age and role), so every additional year a justified exit is delayed increases the eventual payout. The exact figures change and are fact-specific; they should always be verified against the live government source and the compliance material rather than frozen in place.

For a Kitchener–Waterloo small or mid-sized business, the takeaway is concrete: procrastination is not free. The longer a justified termination is deferred, the larger the severance liability becomes — and the weaker the documentary record for asserting cause, which is the subject of the next section.

Source: Government of Ontario, Your guide to the Employment Standards Act: Severance pay / Termination of employment, ontario.ca. For the statutory mechanics and the common-law overlay, see Ontario Employment Standards: Termination and Severance.

Confidence: mixed. The productivity-and-morale dollar cost is directional and thinly evidenced; the rising-entitlements-with-service principle is firm, while the specific figures are volatile and must be confirmed at the source.

What does withholding honest feedback cost — to the person and the organization?

The cost of feedback not given is poorly quantified in dollars, but the established literature shows people systematically suppress bad news (the MUM effect), employees report wanting corrective feedback they are not getting, and the organizational price is prolonged underperformance and stalled, mistrustful careers — with a specific legal edge, since undocumented, unaddressed performance problems undermine any later “just cause” position.

The honest starting point is that the cost of withheld feedback is far less directly quantified than the toxic-worker figures above. There is no clean primary dollar figure. What exists is a chain of well-established findings that together support a qualitative cost claim.

The suppression is real and documented. The MUM effect — the robust tendency to avoid transmitting bad news — is a 50-year-old, replicated finding (Rosen and Tesser, 1970; Tesser and Rosen, 1975). People delay, soften, or omit unfavourable information, partly out of concern for the recipient and partly out of self-protection. (Why leaders specifically avoid these conversations is a separate, practice-side question.)

Employees say they want the feedback they aren’t getting. Zenger and Folkman’s survey of more than 2,500 employees found that 57% preferred corrective feedback to praise, and 72% believed their performance would improve with more of it. This must be labelled vendor/practitioner data — a consultancy’s proprietary self-assessment, not peer-reviewed research. It suggests demand for feedback, not proof of an organizational cost. (Whether feedback actually improves performance belongs to the research evidence on feedback and performance.)

So the organizational cost is framed, not priced. When corrective feedback is withheld, the problem behaviour is not named, so it persists and compounds; the employee’s development stalls; and trust erodes when the eventual reckoning — a bad review, a termination — arrives without warning. For a small Ontario employer this carries a specific legal edge: undocumented, unaddressed performance problems undermine any later “just cause” position and make a defensible exit harder. The withheld conversation is not a kindness; it is a deferred cost borne by the person, in the form of a stalled career, and by the organization, in the form of prolonged underperformance.

Sources: Rosen & Tesser, On Reluctance to Communicate Undesirable Information (the MUM effect), Sociometry 33, pp. 253–263 (1970); Tesser & Rosen, The Reluctance to Transmit Bad News, Advances in Experimental Social Psychology 8 (1975); Zenger & Folkman, Your Employees Want the Negative Feedback You Hate to Give (vendor data), Harvard Business Review (2014).

Confidence: directional. The MUM-effect suppression rests on industry-consensus literature; the demand figures are vendor/practitioner data; the organizational cost is framed qualitatively, not priced.

---

This page is general information, not legal or professional advice. The figures discussed here are reported faithfully to their sources, including where those sources are single working papers, unpublished dissertation data, or vendor estimates that should not be treated as established fact; verify any number against its primary source before relying on it. The Ontario point that termination and severance entitlements rise with length of service is a general principle only — the exact figures are fact-specific and change over time, so confirm them against the live government source and obtain advice before acting on any planned termination.