Most corporate anti-harassment programs do not work, and the field's own researchers have been saying so for almost two decades. That is the uncomfortable starting point any honest 2026 essay on workplace harassment has to acknowledge. The Equal Employment Opportunity Commission's 2016 Report of the Co-Chairs of the Select Task Force on the Study of Harassment in the Workplace, led by Commissioners Chai Feldblum and Victoria Lipnic, reviewed 30 years of training research and concluded plainly that there is "no evidence" most workplace harassment training reduces harassment. The policies are real. The trainings happen. The behavior persists. The gap is not a communications problem — it is a structural one.
The argument here: zero-tolerance language is part of the problem, not the solution. The companies that actually reduce harassment do four specific things, and none of them appear on a typical compliance dashboard.
Why "zero tolerance" backfires
Frank Dobbin (Harvard) and Alexandra Kalev (Tel Aviv) have spent two decades studying which corporate diversity and harassment interventions actually move outcome data. Their 2022 book Getting to Diversity (Harvard University Press) draws on personnel records from 829 U.S. firms over 30 years. The headline finding: mandatory harassment training, grievance procedures with binary outcomes, and "zero tolerance" framings often increase the share of women and minorities who leave the firm. The reason is mechanical. When the only available response to a complaint is termination, supervisors and HR officers raise the evidentiary bar so high that most reports get dismissed. The complainant, not the harasser, becomes the cost center.
The empirical pattern is consistent with what the EEOC itself reports. Of the roughly 81,000 charges of workplace discrimination filed with the agency in fiscal year 2023, fewer than 18% resulted in a finding of "reasonable cause." The remainder — around 67,000 cases — are dismissed or closed administratively. That is not because most complaints are frivolous. Lilia Cortina's research at the University of Michigan, summarized in her 2018 Annual Review of Organizational Psychology paper, estimates that 70% of harassment incidents are never formally reported at all, primarily because employees correctly anticipate retaliation or non-action.
What the evidence does support
Dobbin and Kalev's data point to a different intervention stack — one that companies rarely highlight because it is harder to put on a poster.
Graduated response, not binary outcomes
Firms that offer a range of responses — coaching, mediation, role separation, performance management, and termination as a last resort — see higher reporting rates and lower repeat-offense rates. The EEOC task force called this the "menu of responses" model. It works because it lets HR act on credible but unprovable complaints rather than dismissing them.
Manager accountability tied to compensation
The single most consistent predictor in Dobbin and Kalev's dataset of increased diversity is whether managers' bonuses depend on the retention and promotion outcomes of women and minorities on their teams. The same logic applies to harassment: when a supervisor's variable pay depends on team turnover, exit-interview themes, and engagement scores, complaints stop being a career risk for the manager to suppress and start being a signal they want to surface early.
Ombuds and informal-resolution channels
An ombuds office — an independent, confidential intake function — captures reports the formal process never sees. The International Ombudsman Association's 2023 practitioner survey found that roughly 60% of issues brought to corporate ombuds are resolved without formal complaint. That is not a feature of weak enforcement; it is what catches the 70% Cortina's data says never get reported at all.
The legal floor is not the moral ceiling
Most U.S. workplace harassment law still flows from Meritor Savings Bank v. Vinson (1986) and Faragher v. Boca Raton (1998), the Supreme Court decisions that established hostile-work-environment doctrine and the affirmative defense that lets employers escape liability if they show they had a policy and a reporting channel. The unintended consequence is that the U.S. compliance industry optimized for the affirmative defense — policy, training, hotline — rather than for actually reducing harassment. The policy manual exists because it gets you out of court, not because it gets harassment out of your hallways.
The legal frame is changing, slowly. The federal Speak Out Act (2022) voided pre-dispute NDAs covering sexual harassment and assault claims, and the Ending Forced Arbitration of Sexual Assault and Sexual Harassment Act of 2021 ended mandatory arbitration for the same. State-level whistleblower protections continue to expand. But none of that touches the underlying behavioral problem. It only changes who learns about it.
What a serious culture looks like in 2026
The companies with measurably lower harassment incidence share a profile. They publish annualized turnover and exit-interview data broken out by demographic. They run climate surveys at least twice yearly and treat the results as a leading indicator. They have a written policy on workplace romantic relationships with explicit reporting requirements when there is a power differential. They give first-line managers eight or more hours of behaviorally specific manager training, not 45 minutes of compliance video. And they report harassment metrics — number of investigations opened, substantiation rate, range of outcomes — to their board's audit committee, not just to their general counsel.
None of that is mysterious. It is also not what most American employers actually do. McKinsey and LeanIn.Org's Women in the Workplace 2024 report found that while 90% of companies provide harassment training, only 36% track the outcomes of harassment complaints, and only 21% report those outcomes to senior leadership. The instrumentation is missing. Until it is built, "zero tolerance" remains a slogan that mostly protects the company.
For the broader argument about how labor power is being rebuilt in this decade — including the role of NDAs, arbitration, and worker voice — see NWLB's The New Labor Movement → framework.
The American harassment-policy regime was optimized for the affirmative defense in Faragher, not for fewer victims. That is why the policies multiply and the behavior does not change.
The good news, such as it is: the interventions that actually work are knowable, measurable, and inexpensive compared to litigation. The bad news is that they require a board willing to treat harassment as a governance metric, not a PR exposure. That is the line between a company with a harassment policy and a company with a harassment culture, and in 2026 it is still where most employers are choosing to stand.
Updated May 21, 2026. This piece was substantively rewritten as part of NWLB's 2026 editorial refresh.
