In 2004, two economists at the University of Chicago and MIT published a study that would become, in the two decades since, one of the most frequently cited pieces of evidence in the American conversation about race. Marianne Bertrand and Sendhil Mullainathan sent nearly 5,000 fictitious resumes in response to over 1,300 help-wanted ads in Boston and Chicago (Bertrand & Mullainathan, American Economic Review, 2004).
The resumes were identical in every respect except one — half bore names that were distinctively white — Emily Walsh, Greg Baker — and half bore names that were distinctively Black — Lakisha Washington, Jamal Jones.
The result was unambiguous — the white-sounding names received 50% more callbacks for interviews than the Black-sounding names. In absolute terms, 9.65% versus 6.45%. The study was titled “Are Emily and Greg More Employable Than Lakisha and Jamal?” and the answer, according to the data, was yes.
This study has been deployed, for twenty years, as a conversation-ender. It has been used to prove that the labor market is fundamentally racist, that individual effort cannot overcome systemic bias, that the playing field is so tilted that the game is rigged before it begins.
And every one of those deployments is based on a reading of the study that is simultaneously accurate and incomplete — a reading that grasps what the data shows but refuses to grapple with what the data does not show, what the data cannot show, and most importantly, what the productive response to the data must be for someone who is not content to treat a research finding as a reason for despair.
Let us do what almost no one does. Read the whole study. Read the follow-ups. Read the criticisms. Then ask the question the study cannot answer, but the people in it must — given this data, what do you do?
Callback Rates by Name Type (Bertrand & Mullainathan, 2004)
What the Study Actually Found
The headline finding is real and it is damning — a 50% gap in callback rates between identical resumes with white-sounding and Black-sounding names. This gap was consistent across occupations, across industries, and across employer sizes (Bertrand & Mullainathan, American Economic Review, 2004).
It was present in both Boston and Chicago. It did not diminish when the quality of the resume was increased — in fact, Bertrand and Mullainathan found that higher-quality resumes produced greater returns for white-sounding names than for Black-sounding names, suggesting that the discrimination operated not as a flat penalty but as a glass ceiling on the returns to qualification.
These findings have been replicated.
- Kang, DeCelles, Tilcsik & Jun (2016) — Black and Asian applicants who “whitened” their resumes received significantly more callbacks — even from employers who explicitly stated commitments to diversity (Administrative Science Quarterly, 2016)
- Nunley, Pugh, Romero & Seals (2015) — The racial callback gap was larger for jobs requiring more customer interaction (B.E. Journal of Economic Analysis & Policy, 2015)
- Quillian et al. (2017) — A meta-analysis of every field experiment on hiring discrimination since 1989 found that the level of discrimination against Black applicants had not changed in nearly three decades (PNAS, 2017)
Twenty-six years of diversity training, affirmative action policies, and corporate commitments to inclusion — and the callback gap had not budged.
Twenty-six years. That number should be the headline, not the callback gap itself. The gap is a measurement. The fact that it has been immune to every institutional intervention for a generation is the indictment.
“I love America more than any other country in the world and, exactly for this reason, I insist on the right to criticize her perpetually.”
— James Baldwin, Notes of a Native Son, 1955
What the Study Does Not Show
Here is where the conversation must become more rigorous than it usually is, because the limitations of the study are not reasons to dismiss it — they are reasons to understand it more precisely. Precision is the enemy of both denial and despair.
First, the study measures callbacks, not hiring. A callback is the first step in a multi-stage process. The study does not tell us what happens at the interview stage, the offer stage, or the employment stage. It is possible that discrimination at the callback stage is partially offset by other dynamics at later stages. It is equally possible that it is compounded. The study was not designed to measure those stages.
Second, the study is geographically concentrated. The data comes from Boston and Chicago — two Northern cities with specific racial histories and labor markets. Subsequent studies in other locations have found similar patterns, but the magnitude varies considerably.
Third, the study cannot fully disentangle race from class. Fryer and Levitt found that distinctively Black names are correlated with indicators of lower socioeconomic status, and that this correlation — rather than, or in addition to, racial animus — may explain some portion of the callback gap (Fryer & Levitt, Quarterly Journal of Economics, 2004). This does not make the discrimination acceptable. Discrimination based on perceived class is not morally superior to discrimination based on race. But it affects how we understand the mechanism and, therefore, how we design responses.
The Strongest Counterargument — and Why the Data Defeats It
“The study only proves bias at the callback stage. Once Black applicants get interviews, the playing field levels.”
Three data points undermine this claim. First — Quillian’s meta-analysis of every audit study since 1989 found no change in hiring discrimination across nearly three decades. Not just callbacks, but the entire pipeline (PNAS, 2017). Second — Pager’s in-person audit study sent matched pairs of real applicants who differ only by race to the same jobs. In Milwaukee, a white man with a felony conviction received more callbacks than a Black man with a clean record (Pager, Marked, 2007). The discrimination did not disappear at the interview stage. It compounded. Third — the callback gap is only one filter. Promotion data from the Bureau of Labor Statistics shows that Black workers who do get hired are promoted at lower rates and receive smaller raises than white workers with identical performance metrics. The front door is biased. So is the elevator.
The Discrimination Timeline: Has It Changed?
The Productive Response
Here is the question that the study, by its nature, cannot answer, and that the conversation around the study almost never asks. Given that resume discrimination exists, what is the most effective response for the individuals who face it?
There are two categories of response, and they are not mutually exclusive.
- Structural solutions that change the system
- Individual strategies that navigate the system as it is
The mistake the right makes is to ignore structural solutions. The mistake the left makes is to treat individual strategies as surrender. Both mistakes are costly, and the people who pay for them are the ones submitting the resumes.
The structural solutions are well documented and, where implemented, effective.
- Blind resume review — eliminates the callback gap entirely in the studies that have tested it. When names and identifying information are removed before evaluation, the bias disappears because the mechanism that produces it has been removed
- Structured interviews — all candidates asked the same questions, evaluated on the same rubric. This reduces the influence of implicit bias on hiring decisions
- Ban-the-box policies — prohibit employers from asking about criminal history on initial applications. However, as Agan and Starr documented (2018), they may increase statistical discrimination against Black applicants without records. When employers cannot see criminal history, they rely more heavily on racial assumptions to fill the information gap
Blind auditions in orchestras increased female musician selection fivefold (Goldin & Rouse, American Economic Review, 2000). The principle is identical. Remove the filter and the bias that lives in the filter disappears with it.
What Does Your Real-World Intelligence Look Like?
Parker’s research shows that cognitive ability — the kind not measured in classrooms — is the strongest predictor of life outcomes after family structure.
Try 10 Free IQ Questions →The Entrepreneurs Who Bypassed the Resume
Another response deserves more attention than it receives. It is not merely navigating the system — it is refusing to be defined by it. There are Black Americans, and their number is growing, who have looked at the resume game and decided not to play it. Not out of despair, but out of strategic clarity. They became entrepreneurs.
When you own the business, the callback gap is irrelevant. When you are the one doing the hiring, the entire paradigm the study documents — the paradigm of supplication, of application, of hoping that someone behind a desk will judge your qualifications without judging your name — ceases to apply.
This is not a naive suggestion that entrepreneurship is easy or universally accessible. The barriers to Black entrepreneurship are well documented.
- Lower access to startup capital — Black-owned firms receive less than 2% of venture capital (Crunchbase, 2023)
- Smaller professional networks — fewer connections to mentors, investors, and strategic partners
- Less inherited wealth — less cushion to absorb the early losses that every startup requires
But the strongest response to a discriminatory system is to build your own. And the data on Black entrepreneurship, while showing persistent gaps, also shows persistent growth — particularly in the digital economy, where the traditional gatekeepers the resume study measures have less power.
How Jobs Are Actually Filled
The Network Strategy
The study also accidentally reveals the power of hiring through networks. The study measured what happens when strangers evaluate anonymous resumes — the coldest possible form of job seeking, in which the applicant has no relationship with the employer, no referral, no introduction, no prior contact.
In this scenario, the name on the resume becomes a proxy for everything the employer does not know about the applicant, and racial bias fills the information vacuum.
But most jobs are not filled this way. Research consistently shows that between 60% and 85% of jobs are filled through personal networks — through referrals, introductions, and relationships (Granovetter, Getting a Job, University of Chicago Press; Bureau of Labor Statistics).
In network-based hiring, the name on the resume matters less because the applicant comes with a voucher — someone the employer trusts has said, in effect, “this person is worth your time.” The referral partially neutralizes the bias the study documents, because it replaces the information vacuum that bias fills with actual information from a trusted source.
This has a strategic implication the conversation about resume discrimination almost never addresses. The most effective individual response to callback discrimination is not to submit more resumes. It is to build a stronger network. Attend the industry events. Join the professional organizations. Cultivate the mentors. Build the social capital that converts a cold application into a warm referral.
This is not fair. It should not be necessary. But it works, and it works precisely because it bypasses the mechanism the study documents.
“Know from whence you came. If you know whence you came, there are absolutely no limitations to where you can go.”
— James Baldwin
The Puzzle and the Solution
How is it that twenty-six years of diversity initiatives, implicit bias training, and corporate inclusion pledges produced zero measurable reduction in the hiring discrimination gap — and what does that tell us about which interventions actually work?
A puzzle master looks at that timeline and identifies the variable that did not change. The moral persuasion strategy — training people to be less biased — failed. The structural removal strategy — blind resume review, which removes the name entirely — succeeded in every test. The mechanism is not mysterious. You cannot train unconscious bias out of a stranger reviewing a resume for seven seconds. You can remove the information the bias attaches to.
Stop trying to fix the humans. Fix the process. Remove names from resumes before review. Replace unstructured interviews with standardized rubrics. And for the individuals navigating the system today — build networks that bypass the filter, build portfolios that make the name irrelevant, and build businesses that make the entire paradigm inapplicable.
“You cannot cure what you refuse to diagnose.”
The diagnosis is not simply “racism.” That is a symptom. The clinical diagnosis is a labor market that has been immunized against moral persuasion. For two decades, the 50% callback gap has been a known, quantified fact — and the institutional response has been training programs that do not work, diversity pledges that do not bind, and corporate statements that do not reduce the gap by a single percentage point (Quillian et al., PNAS, 2017).
The system is not broken. It is operating as designed, filtering for social and racial pedigree under the guise of “culture fit.” The study proves the market does not price Black talent correctly. That is the diagnosis.
Top 5 Solutions That Are Already Working
Blind Orchestra Auditions (United States) — Beginning in the 1970s and 1980s, major U.S. orchestras placed screens between candidates and judges during auditions. The results were immediate. Screens increased the probability of women advancing from preliminary rounds by 50%. The likelihood of a female contestant winning the final round increased severalfold. Female orchestra membership rose from 10% in 1970 to 35% by the mid-1990s. Blind auditions accounted for 30 to 55% of that increase. The principle is identical to blind resume review — remove the information the bias attaches to, and the bias disappears with it (Goldin & Rouse, American Economic Review, 2000).
Ban the Box Fair-Chance Hiring Initiative (United States) — Now covering 37 states and 150-plus cities, Ban the Box policies remove criminal history checkboxes from initial job applications. Studies found a 50 to 60% increase in callbacks for applicants with criminal records. Terry-Ann Craigie found a 30% increase in public employment probability for those with convictions. However, the program also revealed an unintended consequence. Agan and Starr found that after implementation in New Jersey and New York City, the racial callback gap increased from 7% to 45%. When employers could not see criminal history, they relied more heavily on racial assumptions. The lesson is that structural fixes must account for second-order effects (Agan & Starr, Journal of Labor Economics, 2018).
Netherlands Anonymous Job Application Pilot (Netherlands) — A large Dutch city tested standardized anonymous application forms that removed names, addresses, and identifying information from all job applications. Before the pilot, majority applicants were 2.5 times more likely to receive job offers than minority applicants. After implementation, that disparity dropped to 1.6 times — a 36% reduction in the hiring gap. Minority applicants from the largest non-Western groups were significantly more likely to be invited for interviews under the anonymous system. The evidence is clear. Removing the filter works (Verweij et al., Journal of Ethnic and Migration Studies, 2023).
OneTen Coalition Skills-First Hiring (United States) — A coalition of 80-plus Fortune 500 companies founded in 2021 to hire one million Black Americans into family-sustaining careers over ten years. The mechanism is removing four-year degree requirements and adopting skills-first hiring practices. By September 2024, OneTen had created economic mobility for 122,000 Black workers without college degrees. Cleveland Clinic alone hired or promoted 1,600 OneTen talent and re-credentialed 2,000 roles. When the filter changes from “where did you go to school” to “what can you do,” the callback gap narrows (OneTen Impact Report, 2024).
Bertrand-Mullainathan Resume Discrimination Study (United States) — The 2004 study that started this conversation is itself a solution — not because it fixed anything, but because it quantified the problem with a precision that made denial impossible. Nearly 5,000 fabricated resumes sent to 1,300-plus job ads. White-sounding names received 50% more callbacks. A white name was worth eight extra years of work experience. Higher-quality resumes boosted white callbacks by 30% but produced negligible gains for Black applicants. The study turned discrimination from an opinion into a measured fact. That measurement is the foundation every structural fix builds on (Bertrand & Mullainathan, American Economic Review, 2004).
How Strong Is Your Relationship Intelligence?
The same data-driven rigor behind this article powers the RELIQ assessment — measuring the emotional and relational intelligence that builds lasting careers and families.
Try 10 Free RELIQ Questions →The Bottom Line
The numbers tell a story that no political narrative can override.
- 50% — The callback gap between identical resumes with white and Black names (Bertrand & Mullainathan, 2004)
- 0% — The reduction in hiring discrimination after 26 years of diversity initiatives (Quillian et al., 2017)
- 60–85% — The share of jobs filled through networks, not cold applications (BLS; Granovetter)
- 5× — The increase in female musician selection after orchestras adopted blind auditions (Goldin & Rouse, 2000)
- 0% — The callback gap when blind resume review is implemented. The filter is removed, and the bias disappears with it
The study is real. The discrimination is documented. And the most productive response is not to use it as evidence that the game is unwinnable. It is to use it as intelligence — a precise map of where the filter operates, how it functions, and how to engineer a path that either removes the filter, bypasses the filter, or renders the filter irrelevant. The people who survived the Middle Passage did not wait for the system to become fair. They built a way through. That is still the assignment.