Coaching fees are the visible cost of preparing for an Indian entrance examination. They are rarely the largest cost, and they are almost never the reason a candidate stops.
Coaching is sold at a single price for a package whose parts cost wildly different amounts to produce. Printed material and a fixed timetable reproduce at almost nothing. A room of similarly motivated peers costs the provider little and is worth a great deal. The component that resists cheap reproduction is a teacher's attention on an individual candidate's work, and it is rationed hardest in the classrooms charging least per head. Because the price is quoted whole, families cannot buy only the parts they need, and the entry price sits well above the cost of what most candidates consume.
Tuition is the quoted figure. Around it sit relocation to a coaching town, rent, food and, for many candidates, a second or third attempt at the same paper. Households rarely budget for the later attempts, because the first is sold with the implication that it will be enough.
The largest line is one no institute quotes. Full-time preparation withdraws a young adult from paid work for a period measured in years. Where those earnings would be load-bearing, the fee is the smaller half of the decision. The binding constraint is the number of unpaid years a family can absorb, which is why a scholarship covering tuition alone widens access less than its sponsors expect.
A good deal of what commands a premium is a description of the examination itself, and the examination describes itself in public. Question papers are released, and assembling them is clerical work. A verified archive of JEE Main papers from 2002 to 2025 contains 8,267 authentic questions, and counting across it needs no access a district-town teacher lacks.
The counts are informative. Matrices and determinants account for 8.8 per cent of recent JEE Main Mathematics, and 275 questions across the full archive. Kinematics in one and two dimensions accounts for 6.6 per cent of recent Physics, 180 questions. Coordination compounds accounts for 8.7 per cent of recent Chemistry, 171 questions. Across the three papers, 117 topics carry enough questions for their share to describe a pattern rather than a run of coincidence: 48 in Physics, 41 in Chemistry, 28 in Mathematics.
The concentration itself matters more than any single percentage. The paper returns to a limited set of areas with enough regularity to be described from public documents, and a family paying a premium for that description is paying for the labour of counting.
Some of preparation is genuinely expensive, and for good reason. The scarce good is a competent reader of a candidate's wrong answer, able to tell a careless slip from a missing prerequisite from a misunderstanding that will recur across a whole area. That work does not scale with class size. Benjamin Bloom's 1984 comparison of one-to-one tutoring with conventional classroom teaching reported an advantage large enough to become a byword in education policy, though later work treats its size as an upper bound rather than a dependable estimate.
The interventions with the strongest evidence behind them cost almost nothing to adopt. Roediger and Karpicke's 2006 experiments on retrieval practice found that testing oneself produces markedly better long-term retention than rereading, though rereading feels more productive at the time. Cepeda and colleagues' 2006 review of distributed practice found the same for spacing. A candidate working through authentic past questions at spaced intervals is doing what the research supports.
One example shows how cheap the missing knowledge can be. The accuracy at which guessing breaks even is the penalty divided by the sum of penalty and reward: 25 per cent for UPSC Prelims, 20 per cent for NEET, 20 per cent for JEE Main. A blind pick from four options succeeds a quarter of the time, so guessing is neutral in expectation under the UPSC scheme and mildly positive under the other two. That is no argument for answering at random. It does mean the familiar instruction to leave every uncertain question blank is wrong at two of the three papers, and that eliminating a single option moves the arithmetic clearly in a candidate's favour. Coached candidates are drilled in this. Others give up marks to a division nobody showed them.
A precise share of aspirants priced out would be useful, and it is not available. India keeps no register of household spending on entrance coaching. The sector runs from national chains through neighbourhood tuition to informal arrangements with a local graduate, much of it unaudited, and surveys disagree on what counts as coaching at all. Figures in circulation are best read as estimates resting on definitions their authors seldom state.
Exclusion is graduated rather than a threshold. Some candidates never enter, having priced the attempt before paying anything. Some buy fragments: a test series without the teaching, or teaching without correction. Some can fund one attempt and not a second, which in examinations where successful candidates commonly sit more than once amounts to exclusion by delay. The most revealing proxy is not the fee but the attempt count, because the capacity to wait is what the system quietly selects on.
The honest position is a floor rather than a promise of equality. Anything derivable from public documents at near-zero marginal cost should not carry a price: the structure of the paper, the areas it returns to, the arithmetic of its marking scheme, and a supply of authentic past questions with worked reasoning. Fees belong to the part that stays expensive, which is sustained attention on one candidate's particular errors.
That distinction will not close the gap. A family unable to afford a year without income is not rescued by free material. It does stop the gap being widened for no reason, which is the least an institution holding the counts can do.
The full set, with progress tracking and five agent perspectives per question, is in the JupiteX app — browse the exam catalogue or browse the Learn library.