At the dinner table, no one says the word. At the job interview, no one says it either. Yet something shapes who sits where, who gets called back, who is considered “the right fit.” Caste has not left these rooms. It has simply learned to speak in a different grammar.

This is the argument that needs making plainly: the gradual disappearance of the word “caste” from polite conversation does not mark the erosion of caste. It marks its sophistication. When a system no longer needs to announce itself, it has achieved something more durable than legality — it has achieved naturalness.

The mechanism is not mysterious. Caste travels through proxies. “Merit” does the work when merit is measured by institutions that caste already shaped. “Culture” does the work when culture is used to defend practices that enforce hierarchy. “Network” does the work when access flows through communities whose boundaries were drawn by jati. The word changes. The logic does not.

Take “merit” first, because it is the proxy that does the most work while claiming to do none. In 2007, the economists Sukhadeo Thorat and Paul Attewell sent matched sets of job applications to openings advertised in major English-language dailies — identical qualifications, different names signaling caste and religion. Low-caste and Muslim applicants had to be measurably better qualified than upper-caste applicants just to reach the same callback rate, and even added credentials, fluency, foreign experience, extracurriculars, failed to close the gap. The employers making those calls were not applying a caste test. They believed they were applying a merit test. That is exactly the mechanism: merit did not fail to notice caste. Merit was the instrument caste used.

“Network” launders the same logic on the other side of the hiring desk, and nowhere more plainly than in marriage, where “network” is rarely even disguised. India’s largest matrimonial platforms, and dedicated sites like ReddyMatrimony, KapuMatrimony, and BrahminMatrimony, still let users filter or self-sort by caste and sub-caste as an ordinary, undisputed search field, the way another site might filter by city. Nobody calls this a caste system. It is called, and experienced, as compatibility.

“Culture” and “economics” do the equivalent work in housing. Thorat’s later audit studies of India’s rental market found landlords in Delhi responding to upper-caste applicants roughly 35 percent of the time against 22 percent for Muslim applicants, with Dalit home-seekers facing comparable, well-documented rates of refusal regardless of their ability to pay the asking rent — brokers steering the “wrong” callers toward other listings before a landlord ever has to say why. The screening is described, when it is described at all, as protecting a neighborhood’s “culture” or “comfort,” not as a caste filter, even though ability to pay is precisely the variable economics predicts should matter and caste is the variable that keeps overriding it.

Same rental market, different response rate
Delhi landlord callback rate
Upper-caste applicants35%
Muslim applicants22%

Dalit home-seekers face comparable, well-documented refusal rates in the same studies, regardless of ability to pay.

Source: Thorat et al., rental-market audit studies of Delhi's housing market. Compiled by The Dalit Project.

Workplace homogeneity gets the softest euphemism of all: the pipeline problem. A 2023 study of Chennai’s IT sector found dominant-caste networks concentrated at senior technical and hiring-adjacent roles, reproducing themselves through referral hiring, informal mentorship, and the same alumni and caste-community ties that shape who even hears about an opening before it is posted. Call it a pipeline problem and the fix is a training program. Call it what the data shows and the fix is who gets to do the hiring.

Family preference, culture, economics, pipeline: four different words, one function. Each explanation is partial. Each deflects from the structural. Together, they form a language that makes caste illegible precisely where it is most active.

The counterargument exists: that these patterns are changing, that each generation is less caste-conscious than the last. Perhaps. But changed consciousness is not the same as changed structure. A person may stop thinking in caste while still inheriting its advantages and passing them forward.

What survives when the word disappears is the architecture itself — the sorted neighborhoods, the closed caste-nexus networks, the accumulated capital of centuries. To dismantle it, the first task is to see it again, in the language it is using now.