There is a phrase in corporate hiring that does more work than any other phrase in the English language.
Culture fit.
It appears in interview scorecards. It appears in debrief meetings where a hiring panel is trying to explain why a technically qualified candidate — one who cleared the aptitude test, the coding round, the system design deep-dive — still did not make it through. The feedback is always some version of the same feeling: "Great on paper, but something felt off." Or: "I'm not sure they'd gel with the team." Or, if the interviewer is being honest with themselves in a way most interviewers are not: "They reminded me of someone I wouldn't invite to lunch."
Nobody says caste. Nobody has to.
The Study That Said It Plainly
In 2026, economist Soumitra Shukla published a study that should have ended every argument about meritocracy in Indian hiring. He analyzed end-to-end screening data from over a thousand jobs at multinational consulting, technology, and financial-services firms recruiting at an elite Indian college — four rounds of screening, from application to aptitude test to group debate to final interview, all documented through the placement office.
He found something precise and devastating.
At every stage before the final round — application screening, written aptitude tests, group debates — the caste gap was negligible after controlling for GPA and entrance-exam rank. Oppressed-caste candidates advanced at virtually the same rate as privileged-caste candidates. The gap appeared in exactly one place: the final-round personal interview, where interviewers probed hometown, schooling, hobbies, family background, and parents' occupation — all under the banner of "cultural fit."
The word for what these interviewers were measuring is not fit. The word is caste.
And here is the detail that makes the study impossible to dismiss: the oppressed-caste candidates who made it past the culture-fit filter — the ones who slipped through — were promoted at higher rates and were less likely to leave their first job than their privileged-caste peers. They performed better on every post-hire metric the paper measured. The filter was not selecting for quality. It was selecting against a background.
The penalty was roughly 30 percentage points — almost halving an oppressed-caste candidate's chance of converting a final interview into an offer. Over 95% of the entire caste gap in hiring success concentrated in that single round. Every objective screen said these candidates were equally qualified. The personal interview said otherwise, and the personal interview is where the preference hides.
The Numbers on the Glass
In 2019, researchers Ajit, Donker, and Nofsinger published an updated study examining the boards of India's top 1,000 companies — the firms that represent four-fifths of market capitalisation on the major stock exchanges.
They calculated a Blau diversity index for caste. The median score was zero. Not low — zero. Nearly 70% of Indian corporate boards had no caste diversity whatsoever.
| Position | Oppressor caste | OBC (Bahujan) | SC/ST (Dalit + Adivasi) |
|---|---|---|---|
| Board seats | 90% | Less than 2% | 0.01% |
| Population share | ~7% | ~52% | ~25% |
Seven percent of the population holds ninety percent of corporate board seats. Twenty-five percent of the population holds one-hundredth of one percent. These are not rounding errors. These are numbers that describe an apartheid, dressed in a suit, sitting in a boardroom with a diversity statement on the wall.
The Blau diversity index ranges from 0 (complete homogeneity) to 1 (maximum diversity). A score of zero means every person in the group belongs to the same category. Nearly 70% of India's top corporate boards scored zero for caste. The glass ceiling is not a ceiling. It is a floor — and it was poured before the building was built.
The Pipeline
The question people ask when they see these numbers — the question designed to make the problem someone else's fault — is: "But what about the pipeline?"
The pipeline. The sacred word of corporate diversity theatre. We'd love to hire more diverse candidates, but the pipeline just isn't there. As if pipelines are geological formations. As if they were not built, maintained, and guarded by the same people who benefit from them.
Let me show you what the pipeline looks like.
In January 2023, Nature published an investigation into caste representation at India's top research institutions. At the five premier IITs — Delhi, Bombay, Madras, Kanpur, Kharagpur — 98% of professors were from privileged castes.
Ninety-eight percent.
At IIM Indore, 106 of 109 faculty members were from the General category. Zero SC. Zero ST. At TIFR — granted "Institution of Excellence" status, exempting it from reservation — the number was 100%. Every single professor, from a single caste bracket.
The government mandate is clear: 27% OBC, 15% SC, 7.5% ST. Across 11 IITs, there were 1,557 faculty vacancies. 415 reserved for OBCs, 234 for SCs, 129 for STs. All unfilled. Year after year after year. The positions exist. The people exist. The hiring does not happen.
Now follow the chain. Faculty who are 98% oppressor-caste produce graduates who are disproportionately oppressor-caste — through mentorship, through recommendation letters, through research opportunities that go to students who remind professors of themselves. Those graduates enter companies. They become hiring managers. They refer their friends. They interview candidates and assess culture fit.
The pipeline is not broken. The pipeline is working exactly as designed.
The Lunch Table
Caste does not announce itself in a modern office. There is no placard on the door, no sign above the water cooler. It moves through smaller signals — signals that are invisible unless you have spent your life learning to read them.
A Dalit engineer — documented in a 2022 Rest of World investigation — described the moment his caste became visible at a major tech company. It was not during a performance review. It was at lunch.
He ate meat. The team noticed. The lunches that used to include him stopped including him. The WhatsApp group for ordering food together stopped adding him. Nobody said anything. Nobody had to. The message was delivered through absence — the oldest language caste has ever spoken.
In another case, a senior manager at a different company recruited six Dalit candidates. Senior management was furious. All six were dismissed after their probation period. The reason given was performance. The performance reviews told a different story.
A third case: a Dalit employee was identified as a "reservation candidate" after posting on social media about a Dalit student's suicide. The post was shared internally. The phrase "reservation guy" followed him through the office like a smell. He described the feeling in one sentence: "Like a fish out of water."
These are not aberrations. They are the texture of a system that never needed laws to maintain itself — only proximity, and the unspoken agreement that certain people belong in certain rooms and others do not.
The Atlantic Crossing
If you thought this was an Indian problem that stayed in India, the Cisco case should end that assumption.
In June 2020, California's Department of Fair Employment and Housing filed a lawsuit against Cisco Systems. Two managers at the San Jose headquarters — both IIT graduates — had discovered that a colleague was Dalit. The discovery happened through IIT alumni networks. The networks that are supposed to be about professional connection and mentorship were, in this case, a caste-identification system operating six thousand miles from the country where caste was supposed to have been abolished.
The Dalit engineer's career was blocked. He was isolated. When he complained, he was retaliated against. The case was filed under the Civil Rights Act of 1964, arguing that caste discrimination constitutes discrimination based on race and ancestry. The case against Cisco as a company remains ongoing.
He was not the only one. Two hundred and fifty Dalit employees across Google, Meta, Microsoft, Apple, and Netflix came forward with reports of discrimination, bullying, ostracisation, and harassment by dominant-caste colleagues. Thirty female Dalit engineers from the same companies filed a joint statement describing how caste networks replicate bias through referrals and performance reviews — the same mechanisms that operate in Bangalore, transplanted cleanly into Cupertino.
Google canceled an internal talk on caste bias after employees called it "anti-Hindu." The organising manager resigned. Sixteen hundred workers demanded caste be added to the code of conduct. The company did not respond.
Apple was the only major tech company to explicitly add caste to its anti-discrimination policy. One company. In an industry that employs millions of South Asian workers.
The Legal Vacuum
Here is the fact that makes all of this possible:
India has no anti-discrimination employment law covering the private sector.
Over 90% of Indian employment is in the private sector. There is no reservation mandate. No affirmative action requirement. No caste diversity reporting. The constitution bans caste discrimination. The private sector is not required to do anything about it.
The SC/ST Prevention of Atrocities Act exists — but it is criminal law, not employment law. It requires a police FIR, a preliminary inquiry, a magistrate's approval before arrest. It was designed for mob violence and public humiliation, not for the quiet architecture of a hiring committee that looks at a surname and feels something off.
In the US, Seattle banned caste discrimination in 2023 — the first jurisdiction outside South Asia to do so, led by council member Kshama Sawant. California passed a similar bill through both legislative chambers. Governor Newsom vetoed it, arguing existing law already covers caste under ancestry. Whether that is true has not been tested. It will be tested when someone files suit and a court decides whether culture fit is a protected phrase or an admission.
What DEI Refuses to Say
Indian corporate diversity programmes talk about gender. They talk about disability. Some, cautiously, talk about sexuality. Almost none talk about caste.
The reason is simple: gender diversity means hiring more women from the same caste bracket that already dominates the organisation. It adds diversity along one axis while preserving homogeneity along the axis that actually structures Indian society. A boardroom that is 50% women and 100% oppressor-caste is not diverse. It is the same room with a better photograph.
Oxfam-Newslaundry's 2022 report on Indian media found 90% of leadership positions across print, TV, and digital occupied by General category individuals. Zero Dalits or Adivasis heading any mainstream media outlet. The people who tell India's stories come from 7% of India's population. The stories reflect the teller.
The Wage Gap That Discrimination Built
The Oxfam India Discrimination Report (2022) measured the earnings gap between General category and SC/ST workers in regular urban employment: 33%. A third less, for the same work, in the same cities.
| Sector | Wage gap | % attributable to discrimination |
|---|---|---|
| Regular urban employment | 33% | Significant after controlling for education |
| Self-employment | ₹5,000/month | 41% due to discrimination |
| Rural casual labour | Large | 79% due to discrimination |
Seventy-nine percent. In rural casual labour, nearly four-fifths of income inequality between castes is not explained by education, experience, or skill. It is explained by the fact of being born into an oppressed caste. The gap is not a market outcome. It is a sentence.
What the Filter Actually Filters
Shukla's study stays with me because it answers the question so precisely.
The filter does not filter for competence. The technical rounds already measured that. The filter does not filter for communication skills. The group discussions already measured that. The filter does not filter for anything that correlates with job performance — Shukla proved this by showing that the oppressed-caste candidates who cleared the filter outperformed their privileged-caste peers on promotion and retention, meaning the bar was set higher for them, not lower.
The filter filters for familiarity. For the feeling of recognition that passes between two people who grew up in the same kind of house, ate the same kind of food, pronounce certain words the same way, carry certain names, know certain references. It filters for the comfort of interviewing someone who reminds you of yourself.
And since the people conducting the interviews are overwhelmingly from a single caste bracket — because the pipeline delivered them there, because their professors mentored them there, because their networks referred them there — the filter reproduces the room. Endlessly. Without malice. Without paperwork. Without anyone ever having to say the word.
That is how caste works in 2026. Quietly. In glass buildings. Over coffee. In the seventeen seconds between a handshake and a feeling.
Previously in this series: What Ambedkar Actually Said — The gap between what people quote and what he actually wrote.
