CAT78
Earned97.5
+ 19.5 bonus
Unify the Lit · The Earned Percentile Reservation,
Reservation,
in reverse.
No merit list in India is built on a level field, not CAT, not NEET, not JEE. Here's the math that makes the tilt visible.
02
Caste isn't a fact about ability.
Caste isn't a fact about ability. It's a fact about the environment in which ability is measured.
In 2004, World Bank economists Karla Hoff and Priyanka Pandey ran an experiment in rural Uttar Pradesh. 642 schoolboys, half high-caste, half low-caste, solved mazes for cash incentives. When caste was hidden, both groups performed identically. When caste was announced, low-caste performance dropped 25%. In segregated conditions, it dropped 39%. High-caste performance, meanwhile, rose 19%.
The experiment proved something we already suspected and could finally measure: caste isn't a fact about ability, it's a fact about the environment in which ability is measured.
The 25% figure is the conservative end of their finding. The same experiment has been replicated in China for rural hukou status, and in the Slovak Republic for Roma identity. Across societies, when stratified identity is made salient, performance drops by 25-39%.
03
Why 25%? Six reasons.
Every number in this formula is a choice. The 25% maximum reduction is the most defensible number we could find. Here is what it is anchored to.
It's the conservative end of the empirical range
Hoff & Pandey measured a 25% performance drop when caste was announced, and 39% in segregated conditions. We anchor to the lower bound, the least extreme finding in their data. Anyone arguing the math is too aggressive must argue against the most cautious figure in the original paper.
It mirrors existing reservation arithmetic
SC and ST candidates currently receive a 20-25 percentile point relaxation in CAT cutoffs at most IIMs. Applying the same magnitude as a discount on caste-advantaged scores, and as a bonus on caste-burdened ones, produces mathematical symmetry. The fairness already exists in one direction; we are naming the other.
It's cross-validated globally
The same magnitude effect has been replicated in China (rural hukou holders living in Beijing) and in the Slovak Republic (Roma vs majority students). The 25-39% range is not an Indian outlier, it is a structural constant of stratified societies, documented across continents.
It connects to foundational stereotype threat research
Steele & Aronson's seminal 1995 study documented a similar 25-30% performance gap when stereotyped identity was made salient, for African American students taking SAT-style verbal tests. Our number plugs Indian caste research into a 30-year, peer-reviewed global literature.
It's published in a top-five development economics journal
Hoff & Pandey's 2014 paper appeared in the Journal of Development Economics, one of the top-five journals in the field, indexed by Elsevier, cited hundreds of times in subsequent research. This is not advocacy data. It is the consensus of the field.
It's bounded, not unlimited
No matter how caste-aware a candidate is, the formula can never adjust their CAT score by more than 25%. There is a floor and a ceiling. A 99 percentile scorer can fall to 74 in the worst case, significant, but never erased. We are recalibrating, not deleting.
Five peer-reviewed papers.
Verify the math yourself.
- Making up people, The effect of identity on performance in a modernizing society
The primary anchor. 642 schoolboys, rural UP. 25% performance drop when caste was announced; 39% in segregated condition.
- Belief Systems and Durable Inequalities: An Experimental Investigation of Indian Caste
The original working paper, open access. Establishes the experimental methodology.
- Caste System (Policy Research Working Paper 7929)
World Bank synthesis paper. Documents replication of the maze experiment in China and Slovak Republic.
- Stereotype Threat and the Intellectual Test Performance of African Americans
The foundational stereotype threat paper. The global framework Hoff-Pandey extended to India.
- Twenty Years of Stereotype Threat Research: A Review of Psychological Mediators
Meta-review of two decades of research. Confirms the 25-35% performance gap as robust across populations.
04
Two numbers. One reveal.
Your CAT percentile, multiplied by a Caste Multiplier. The multiplier comes from two things: whether you carried caste privilege into the test, and whether caste pride or shame made it salient, the exact mechanism Hoff & Pandey measured.
If you are privileged caste
Earned=CAT×(1−adj)
Pride lifted your score. adj = 25% if proud of caste, 12.5% if not openly proud.
If you are marginalised caste
Earned=CAT×(1+adj)
Shame suppressed your score. adj = 25% if you hide caste, 12.5% if you don't.
The adjustment adj is anchored to Hoff & Pandey: caste salience produced a 25% performance gap. Full pride or shame applies the full figure; caste merely acknowledged applies half.
Caste positionPride / shameEffectMultiplier
Privileged casteProud of caste −25% × 0.750
Privileged casteNot openly proud −12.5% × 0.875
Marginalized casteDoesn't hide caste +12.5% × 1.125
Marginalized casteHides caste / shame +25% × 1.250
Caste pride and caste shame are two faces of one hierarchy. The same structure that lets a surname sit proudly on a shopfront forces another to be hidden. The Earned Percentile prices both.
05
The 20-point gap vanishes.
Reservation policy gives Dalit candidates a 20-25 percentile point boost to compensate for structural penalty. The Earned Percentile applies the same magnitude in two directions, discounting privileged candidates' scores for the entitlement that lifted them, and crediting Dalit/Bahujan candidates for the stereotype threat that suppressed them.
CAT98.43
Earned73.82
− 24.61 discount
CAT98.43
Earned86.13
− 12.3 discount
CAT78
Earned87.75
+ 9.75 bonus
A 78 percentile Dalit candidate who hid their surname to survive, carrying caste shame into every exam, climbs to 97.50. A 98.43 percentile candidate whose surname sat proudly on the shopfront falls to 73.82. The 20-point CAT gap doesn't just close, it inverts. The math reveals what the field was hiding.
06
What this is not.
Not a punishment
This isn't about shaming individuals. It's about measuring merit honestly. The 25% reduction is the lower bound of peer-reviewed research on structural penalty, we use the most conservative figure.
Not a binding ranking
IIMs do not (yet) use this formula. This is a moral proposal, a public statement of what fair merit measurement would look like if we factored in consciousness of unearned advantage.
Not a substitute for the work
Read Babasaheb. Sit with the number. The math is the doorway, not the destination. It is meant to provoke a longer conversation, not end one.