When Transparency Moves Faster Than Culture

The subject of this series is what happens to venture-capital-backed startups when the law makes harassment-related silence less enforceable. The final question is not whether disclosure rights are desirable in principle; it is why the evidence shows a short-run fall in female hiring, and whether the internal changes that follow could eventually make startups more accountable. The most defensible answer is conditional: transparency changes the firm’s exposure quickly, while culture changes only when the organization has credible ways to prevent harm, hear reports, protect workers, and act on what it learns.

The short-run chain is coherent—but it is not the whole verdict

The legal starting point is narrower than the phrase “NDA ban” suggests. The Speak Out Act limits the judicial enforceability of predispute nondisclosure and nondisparagement clauses for covered sexual-assault and sexual-harassment disputes. It preserves protection for trade secrets and proprietary information, allows more-protective state law to remain in force, and does not turn an allegation into a proven fact. The earlier state reforms studied by the NBER paper varied in scope, but shared a direction: they made it less reliable for an employer to use a predispute or settlement-related confidentiality arrangement to suppress discussion of harassment-related conduct.

That change serves a clear public purpose. If a worker cannot safely describe misconduct, other employees may not learn that the same manager has harmed people before. Investors, applicants, regulators, and the public may see only isolated disputes. A legal rule that removes one formal barrier to disclosure can make patterns more visible and increase the chance that people with power face accountability. Congress’s findings describe silence as a condition that can shield perpetrators and allow illegal conduct to continue.

But the rule also changes the firm’s expected exposure before anyone reports anything. A startup that previously assumed a future covered dispute could be contained by a predispute clause must now account for the possibility that the worker can speak, file a claim, or make the conduct known through other channels. The expected cost can include litigation, reputational damage, employee departures, investor concern, and the loss of trust that makes recruiting harder. This is not a claim that every startup has committed misconduct. It is a claim about how an altered information environment can change decisions made under uncertainty.

The startup study’s main result fits that chain. Across a panel of more than 50,000 U.S. venture-backed startups observed from 2014 through 2022, the reported baseline estimate implies about 7.8% fewer female hires per year after a headquarters state adopted an NDA-weakening law. The NBER landing page summarizes the result as an approximately 8% annual reduction. Male hiring does not show a comparable change. The female share of hires falls by 1.2 percentage points from an average of about 34%, and the comparison of female with male hires implies a relative reduction of roughly 5–6% for women.

The reduction is also visible at two different points in the hiring process. Among startups that hire at least one woman, the number hired falls by about 3%. Among startups that had not yet hired a woman, the chance of making a first female hire falls by 2.8 percentage points within one year and by 4.9 points within four years. These are not two versions of the same statistic. One describes the depth of hiring once a firm has entered the market for female workers; the other describes whether the firm crosses the entry threshold at all. Together, they show why an annual flow can matter to a young company’s future composition.

The distribution of the effect is more revealing than the headline alone. Hiring falls across high- and low-skilled female workers, with no statistically significant difference between the skill groups. The sharper gradient is seniority: junior female hiring falls by roughly 10%, while the estimate for senior female hiring is smaller and statistically insignificant. Seniority maps onto the supervisor-subordinate relationship, where authority and dependence can make harassment more consequential. The result does not say that junior women are risky hires. It says that a firm responding defensively may reduce recruitment where its own hierarchy creates the greatest perceived exposure.

The same logic appears across firms. The estimated decline is about 12.4% in small startups and about 4.5% in large ones. In firms with low prior female representation, the decline is about 10.9%; among female-friendly firms, the estimate is close to zero. The study reports particularly strong effects among startups with fewer than ten employees and no detectable effect in publicly listed technology firms. Size and prior female share are imperfect proxies, not direct measures of a reporting channel or a safe culture. Even so, the pattern makes the transparency-capacity gap visible. Where internal safeguards appear weak, the law can arrive as exposure without an established way to manage it.

This evidence favors a firm-demand explanation, but it does not eliminate a worker-supply explanation. Women may reasonably become less willing to join startups when they learn that harassment can be disclosed or that a firm lacks credible protection.