The most useful question to ask about workplace harassment in 2026 is not whether the #MeToo movement worked. It is which of its interventions actually changed behavior and which left the underlying conditions intact. The empirical record is uneven enough to be both encouraging and sobering. The headline harassment cases got harder to ignore; the underlying microaggression-level conduct that most workers experience daily got reframed but not, on most measures, meaningfully reduced. That is the gap this piece is about.
The argument is specific. The conventional toolkit for addressing workplace harassment — annual mandatory training, anonymous hotlines, written non-discrimination policies — has weak evidence for actually changing behavior. The interventions that do change behavior are structural: bystander training that is skills-based rather than awareness-based, the removal of forced arbitration from workplace contracts, and changes in the underlying power asymmetries that allow harassment to go unreported.
The data on what harassment actually looks like
The EEOC's 2016 Select Task Force on the Study of Harassment in the Workplace report — co-chaired by Commissioners Chai Feldblum and Victoria Lipnic — remains the most authoritative federal review. Its central finding was that roughly 75% of workers who experience harassment never report it formally, and that workplace anti-harassment training as commonly practiced "has not worked as a prevention tool" because most programs are designed to reduce employer legal liability rather than to change worker behavior.
Frank Dobbin at Harvard and Alexandra Kalev at Tel Aviv University have spent decades studying which workplace interventions actually change diversity, harassment, and inclusion outcomes. Their 2022 book Getting to Diversity (Harvard University Press) summarizes findings from over 800 companies tracked across multiple decades: mandatory anti-harassment training, taken in isolation, produces no measurable reduction in harassment complaints and in some cases produces backlash effects. Voluntary, skills-based training combined with structural accountability — diversity managers, mentoring programs, formal mentorship — produces measurable improvements.
The 2024 Pew Research Center survey on workplace experiences found that 42% of women report experiencing some form of workplace gender discrimination, with rates highest among women in male-dominated industries (technology, finance, construction, manufacturing). The 2017 Pew survey on tech-industry workplace experiences put the figure at 50% in tech specifically. These numbers have not moved meaningfully despite seven years of post-#MeToo reform.
Why the grey area is the binding constraint
The most consequential workplace conduct is rarely the overt physical or verbal harassment that triggers legal action. It is the lower-grade pattern of behaviors — credit-taking, interruption, exclusion from informal networks, undermining comments, gendered task assignment — that compound over years and shape who advances. The cleanest research on this comes from Kieran Snyder's 2014 analysis of performance reviews at tech firms, which found that women received critical feedback in 87.9% of reviews vs. 58.9% for men, with women's reviews three times more likely to contain personality criticism that did not appear in any men's reviews.
Joan C. Williams at UC Hastings has documented this pattern across multiple industries in her 2014 book What Works for Women at Work (NYU Press) and subsequent research. The patterns she identifies — "Prove It Again," "Tightrope," "Maternal Wall," "Tug of War" — have all been replicated in subsequent empirical work and form the backbone of how researchers now think about cumulative workplace bias.
The legal framework, framed around hostile work environment as defined in the Supreme Court's 1986 decision Meritor Savings Bank v. Vinson, requires conduct to be "severe or pervasive" enough that a reasonable person would find the environment abusive. That standard captures the overt cases. It does not, by design, capture the cumulative grey-area pattern that affects far more workers.
Three interventions with strong evidence
Skills-based bystander training
The most-evaluated workplace intervention is "Green Dot," the bystander-intervention program developed at the University of Kentucky and adapted to workplaces. Randomized evaluations of Green Dot in school and workplace settings have found 11–17% reductions in self-reported sexual violence and harassment over multi-year follow-ups. The mechanism is not awareness — it is teaching specific, scripted behaviors that bystanders can use to interrupt conduct in real time. The empirical case for this approach is meaningfully stronger than for any didactic training program.
Removal of forced arbitration
The 2022 Ending Forced Arbitration of Sexual Assault and Sexual Harassment Act — a rare bipartisan piece of federal labor legislation — prohibits pre-dispute arbitration agreements from covering sexual-assault and sexual-harassment claims. The empirical case for this is straightforward: mandatory arbitration moved harassment claims out of public view, reduced settlement amounts, and made repeat-offender patterns harder to detect. Subsequent state-level laws (notably California's 2019 AB 51 and New York's 2022 reforms) have extended similar logic to a broader range of workplace claims.
Structural accountability for managers
Dobbin and Kalev's research consistently finds that the single most reliable correlate of reduced workplace harassment is whether managers are accountable for their team's harassment-complaint trajectory. Where managers' performance evaluations include harassment-and-retention metrics, and where dedicated diversity officers have organizational standing, complaint rates fall. Where they don't, all the training in the world produces little durable change.
What employers can actually do this year
Three operational moves have the strongest evidence behind them. First, replace single-vendor annual training with skills-based bystander programs that include specific scripts and role-play. Second, voluntarily remove forced-arbitration clauses from all employment agreements, not just sexual-harassment-specific clauses. Third, build manager accountability into compensation and advancement — not as a one-time HR review but as a continuous metric.
For deeper coverage of the broader patterns shaping women's labor-market outcomes, see our flagship analysis Women, Work, and the Future →.
The harder structural question
The grey area persists, in the end, because workplaces are organized hierarchically and most workers cannot afford to confront the conduct of people with power over them. The interventions above address pieces of that asymmetry. They do not eliminate it. The deeper structural moves — sectoral bargaining that gives workers collective standing, real pay-transparency that reduces the financial penalty of confrontation, codetermination-style worker representation in governance — are all on the table in current policy debates and have more evidence behind them than most participants admit.
Claudia Goldin's 2014 American Economic Review paper and her Nobel-recognized body of work also bear on this: she identified opaque, individually negotiated compensation in "greedy jobs" as a structural amplifier of gender-pay gaps and, by extension, of harassment vulnerability. Closing those opaque structures reduces the leverage that harassment exploits.
The grey area in workplace harassment isn't a definitional problem. It's a power problem. The interventions that work are the ones that change the power, not the ones that ask workers to manage it better.
What gets better in the next five years
Three trajectories are visible. Forced-arbitration coverage will continue to narrow as state laws follow the 2022 federal statute. Bystander-intervention training will displace older awareness-based training in serious employers because it has better evidence. And the AI-driven hiring, promotion, and evaluation systems now spreading through enterprise HR will either accelerate or attenuate the patterns the empirical literature identifies, depending on how they are designed and audited.
That last point may be the most important. Algorithmic systems can encode the existing bias, or they can be deliberately designed to surface and counteract it. The choice is being made now, in procurement decisions most workers will never see. That is where the next generation of workplace-harassment policy will be fought — and where it most needs to be watched.
Updated May 21, 2026. This piece was substantively rewritten as part of NWLB's 2026 editorial refresh.
