The Ad Can Change While the Hiring Pattern Stays the Same
This chapter examines the strongest direct evidence in the supplied material: what happened to actual hiring after large employers removed degree requirements from job postings. The aim is to read that evidence correctly—especially its denominators, comparisons and limits—then use the results to distinguish firms that changed hiring from firms that changed only their public language. The conclusion is more precise than either “nothing changed” or “skills-first hiring worked”: the effect was real in some roles and firms, but small and uneven across the wider system.
What the study measured—and what it did not
Counting job postings can tell us that an employer changed a stated requirement. It cannot tell us whether a hiring manager changed the candidate comparison that followed. To examine that difference, the Burning Glass Institute and Harvard Business School followed roles at large firms for which two conditions were met: the employer removed a degree requirement from the role, and enough hiring could be observed for at least one year before and at least one year after the change. The unit of analysis was a role at a particular employer, not an announcement or an entire company.
The researchers then linked those roles to career histories and education information in a database covering more than 65 million US workers. The executive summary describes about 11,300 roles; the methodological appendix gives the more precise figure of 11,332 roles matched directly from the earlier analysis. The study also compared the roles with similar roles at similar firms, matching on occupation and industry. That design gives the result more substance than a tally of corporate promises because it asks what happened to the people who were actually hired into the affected roles.
Choosing the role as the unit of analysis matters. A company-wide statistic could combine thousands of occupations, different hiring volumes and different policy changes into one average. It might show that a firm hired more non-degree workers overall without showing whether the roles advertised as degree-free were the ones that changed. A role-level comparison keeps the policy change attached to the work it was supposed to affect. It asks whether the hiring profile for a particular occupation at a particular employer moved after that role’s requirement changed.
The comparison also preserves a useful baseline. The relevant question is not whether the post-change workforce looks different from an abstract ideal, but whether it differs from the pattern that existed before the change and from similar roles that did not experience the same change at the same time. This makes the result more informative than a simple count of people without degrees. It also limits the claim. A change in the share of non-degree hires shows that the composition moved; it does not reveal whether the new hires had stronger skills, whether the role became easier to perform, or which managerial decisions produced the movement.
It is still important to understand the design as an observation of outcomes, not a complete recording of the hiring process. The study identifies a change in the posting and observes the education profile of hires before and after it. It does not, from the evidence supplied here, show every change made to sourcing, screening, interview practice, assessment or manager incentives. Those hidden steps are precisely where a degree proxy can survive. The study can show that the realised pattern changed or failed to change; it cannot by itself attribute the change to one internal intervention.
The headline result is a conditional one. In roles where the degree requirement was removed, the share of workers hired without a bachelor’s degree increased by about 3.5 percentage points on average. Put simply, for every 100 opportunities in a role that removed the requirement, roughly four more non-degree workers were hired than would have been expected before the change. That is a meaningful movement in the affected roles. It is not a claim that four out of every 100 hires across the labour market became non-degree hires.
The wider denominator changes the picture. The researchers found that only about 3.6% of the roles in the relevant period removed degree requirements. Applying the average change in those affected roles to the broader US hiring system produced an estimated 0.14% increase in incremental hiring of candidates without degrees. The report translates that estimate into approximately 97,000 additional workers a year out of an estimated 77 million annual hires—fewer than one in 700 hires. The calculation is an extrapolation, not a headcount of every person who benefited, but it makes the scale distinction visible: a sizeable conditional effect can coexist with a small system-wide effect when the intervention reaches few roles.
There are several denominators here, and each answers a different question. The 3.5 percentage points describe the average change inside roles that actually lost the requirement. The 3.6% describes how widespread that policy change was among the roles in the study’s earlier universe. The 0.14% describes the estimated incremental effect after the conditional change is placed against the much larger hiring system. It is not the percentage of all hires who lacked a degree, and it is not a measure of the share of postings that became degree-free. Confusing these quantities can make a limited reform sound either transformative or pointless.
The system estimate also shows why reach matters. A firm could make a substantial change in a small set of roles and produce a visible local result while leaving most of its hiring untouched. Another firm could remove the requirement broadly but change little in candidate selection. To assess the practical importance of either move, a reader needs both intensity—how much the affected roles changed—and coverage—how many roles and hires were actually exposed to the change.
This sequence explains why the study is useful and why it cannot answer every question. The posting change is the identifiable event; the hiring record is the outcome. Between them sits the operational funnel, including decisions that the study does not directly observe. A firm can therefore receive credit for changing eligibility language while leaving the mechanism that ranks candidates largely untouched. Conversely, a firm can make additional changes that are not visible in the wording but show up in who is hired.
The study’s comparison is also not a random experiment. Firms choose which roles to change, and the roles with enough observed hiring may differ from roles that are harder to track. Large employers are not representative of every business, and career histories may not capture every hire or every relevant skill. Matching similar occupations and industries helps with comparison, but it does not eliminate all differences in labour-market conditions, candidate pools, management or the timing of the policy. Those limitations do not make the result useless. They define the conclusion it can support: among observable roles at large firms that removed a requirement, the average hiring mix shifted modestly, and the shift varied sharply across firms.
Why adoption is not a binary result
The firm-level results make the average more revealing. Among the firms studied, approximately 37% followed through and were classified as Skills-Based Hiring Leaders. In the years after changing their requirements, these firms hired nearly 18% more non-degreed workers on average in the relevant roles. The number is a relative change in the share of non-degree hires, not a claim that 18 additional people entered every 100-person workforce. The important point is that some employers did change the realised profile of their hires.
The largest group, about 45%, was classified as “In Name Only.” These firms removed degree requirements from postings but showed no meaningful difference in the share of degree-holders they hired into those roles. Their public rule changed while the observed outcome remained broadly similar. That outcome does not prove that managers acted in bad faith. It could reflect unchanged sourcing, a candidate pool that was already heavily degree-holding, selection criteria that were not captured in the posting, or a belief that the degree remained useful even if it was no longer mandatory. The study does not identify one explanation. It does show why the language change alone cannot settle the question.
The final group, approximately 18%, were Backsliders. They initially increased the share of non-degree hires but later returned to the previous pattern; over the longer period, they ended up hiring a smaller share of non-degree workers. This is a different failure from stopping at the announcement. It suggests that a pilot or a moment of executive attention can produce movement without changing the routines that sustain it. When attention shifts, the organisation may return to the familiar signals that feel safer or cheaper to use.
These categories also warn against treating industry labels as explanations. The report finds firms from multiple sectors in all three groups. It observes some sector patterns—for example, government and accommodation and food-service firms were particularly represented among leaders, while technology firms were overrepresented in the “In Name Only” group—but it also says that more study is needed to explain the archetypes. The differences could involve occupational mix, labour-market pressure, the depth of the commitment, manager practice or other factors. A company should use the categories as questions about its own process, not as a ranking of industries.
The categories are best understood as different points of failure in the translation from policy to practice. An “In Name Only” result says that the visible rule changed without a detectable shift in the hiring mix. A backslider result adds a time dimension: the organisation did show movement, but the movement did not survive. A leader result shows sustained change in the observed hiring profile, but it does not tell us which internal practices made that change possible. The labels are descriptive, not explanations. They help a firm ask whether its intervention reached the candidate pool, the screening stage, the hiring manager and the routines that remain after the launch.
That distinction matters for evaluation. If a firm measures only the number of postings edited, all three trajectories can look like success. If it measures the candidate mix immediately after launch, a backslider may also look successful. A more demanding evaluation checks the affected roles over time and examines what changed around them. The HBS study can identify the trajectory; it cannot observe every cause. A responsible employer would therefore treat the classification as a prompt to investigate its process rather than as proof that a particular remedy will work.
The benefits reported among firms that made a meaningful change are part of the evidence, not an afterthought. For non-degreed workers who entered roles that had previously required a degree, the study reports an average salary increase of about 25%, or more than USD 12,400 in additional annual earnings. At Skills-Based Hiring Leader firms, non-degreed workers had a two-year retention rate of 58%, compared with 48% for degree-holder colleagues in the relevant roles. That is a ten-percentage-point difference, described in the report as a 20% increase in retention. The comparison gives employers a plausible business reason to care: widening access may also reduce the costs of turnover.
Those figures need the same careful reading as the hiring estimate. The salary result concerns workers moving into roles that had previously required a degree, not every non-degree worker. The retention result concerns colleagues in relevant roles at firms classified as leaders, not a universal advantage for non-degree hires. The supplied evidence does not establish that removing the degree requirement alone caused the salary increase or retention difference. Firms that become leaders may also change onboarding, management, development or pay practices. The outcomes are consistent with a win-win case for genuine implementation, but the mechanism is broader than the wording of the posting.
The OECD report cites a separate, more recent study that found only a two-percentage-point increase in the share of non-degree hires after companies removed degree requirements. That figure should not be blended with the HBS study’s 3.5-point estimate as if the two were measurements of the same population. The studies may differ in sample, time window, role definition or method. Their magnitudes are not directly resolved by the supplied material. They do, however, point in the same qualitative direction: average change after degree removal is limited, and the existence of a changed posting is not evidence that the hiring system has been rebuilt.
The most defensible interpretation is therefore neither cynical nor celebratory. Degree removal can matter. In the roles and firms where the change was followed through, it was associated with a different hiring mix and with reported pay and retention benefits for the workers who entered through that route. But the majority of firms in the study did not show a sustained substantive change, and the intervention covered a small portion of the potential role universe. A statement about access becomes a statement about hiring only when the organisation can show what happened to candidates after they encountered the new rule.
The evidence leaves one practical distinction in view. The study tells us that the gap between an announcement and an outcome exists, how large it was in the observed sample and how it varied over time. It does not tell us which hidden decision created each gap. To make that difference actionable, we need to trace the layers between policy language and realised hiring: the role definition, the screening funnel, the evidence managers trust and the systems that keep a change from reversing. That is where “skills-first” becomes a process rather than a label.