The average IQ in India: three numbers, and what the samples show

0 Quick Answer

The average IQ in India has no single documented value, because no representative sample of Indians has ever taken an intelligence test normed for India and reported on the IQ scale; the three numbers in circulation are estimates built from different materials, and each is best read as a statement about its own dataset rather than about the country. Becker's national IQ dataset, version 1.3.5 dated October 25, 2023 and downloadable from View on IQ, puts India at 76.14 on its combined measure, computed from 20 published samples totalling 19,890 people, most of them schoolchildren tested with Raven's matrices and scored against British norms. Lynn and Vanhanen's IQ and the Wealth of Nations gave India 81 in 2002, and the 2012 revision of the same series gave 82.2, on a thinner base of the same kind. Self-selected online tests give a different picture: the International IQ Test country table that the ACIS page on average IQ by country uses shows India at 98.44 from 8,340 test takers, a sample of people with a device, a connection, working English and an interest in taking the test.

The three figures do not contradict each other, because they do not measure the same thing. The 76 is a weighted average of children in nutrition studies, epilepsy control groups, inbreeding studies and two Raven's standardization samples, converted to a British scale of a different decade. The 82 is an earlier average of a subset of the same kind of material. The 98 is the mean of whoever chose to sit a free online test, on a scale whose mean is fixed at 100 by construction. A test that has been standardized in India, such as the WAIS-IV India edition, returns 100 for the average member of its own norm sample by definition, which is why the phrase "average IQ in India" means nothing until you say whose norms and whose sample.

What the evidence does document is not an IQ at all. The fifth National Family Health Survey, which visited 636,699 households between June 2019 and April 2021, found that 36 percent of children under five were stunted and 67 percent of children aged 6 to 59 months were anaemic. The Annual Status of Education Report for 2024, which tested 649,491 rural children, found that 23.4 percent of Std III pupils in government schools could read a Std II level text. Those are conditions known to depress performance on any cognitive test, and the sections below set out how much of the gap between 76 and 100 they can and cannot account for.

1 Three Numbers, Three Objects

The three figures do not disagree about one quantity; each measures a different population with a different instrument against a different norm, and the spread between them is exactly what those differences predict. Becker's dataset is the most transparent of the three, because every row of it can be inspected. Its national table gives India four separate estimates before any averaging with school assessments: an unweighted mean of 73.94 across the twenty samples, a mean weighted by sample size of 78.42, and a mean weighted by both sample size and a quality rating of 78.71. It then converts international school assessment results into the IQ metric, which for India gives 73.57, and averages that with the weighted mean to produce the headline 76.14. On that combined measure India ranks 119th of the 157 countries with a value, between the Gaza Strip at 76.27 and Algeria at 76.00, a few tenths from Egypt and Uganda, and well below Sri Lanka at 86.62 and Pakistan at 80.05, by our reading of the same table.

Lynn and Vanhanen worked from fewer studies and a simpler rule. Their 2002 book assigned India 81, and the 2012 revision that Becker's table records as the Lynn and Vanhanen 2012 value gives 82.2. The difference between 82 and 76 is not new information about Indians; it is the addition of low scoring samples, a different weighting and a different correction for the Flynn effect, applied by different authors to overlapping literature. The page on the Flynn effect explains why the year of a norm table changes the IQ assigned to the same raw score, and that mechanism alone accounts for several of the six points.

The 98.44 belongs to a different universe. The International IQ Test publishes a country table built from its own takers, with the dataset mean fixed at 100. In that table, read on September 19, 2026 through the ACIS country page, India stands at 98.44 from 8,340 takers, the United Kingdom at 101.57 from 7,310, the United States at 101.04 from 22,605, and China at 106.48 from 229,918. Nobody in that sample was selected; everyone chose to be there, in English, on a device, and the scale is anchored to the takers rather than to any nation. The ACIS country page adopts that table for its ranking tool precisely because it is one consistent dataset with one consistent selection, and it says on the page that older national estimates are mentioned only as context for why other pages disagree.

The last row is the one most readers miss. A Wechsler scale adapted for India reports the average person in its Indian standardization sample as 100, in the same way that the American WAIS reports the average American as 100. The page on how IQ scores are normed sets out why a norm sample defines the scale rather than measuring against some universal one. The question "what is the average IQ in India" therefore has three honest answers: 100 on an Indian test, about 76 to 82 on an old British scale applied to convenience samples, and about 98 among the self-selected online takers. Only the first is a measurement in the ordinary sense, and it is uninformative by design.

2 The Twenty Samples Behind the 76

A national IQ is only as good as the studies it averages, and the studies behind India's figure were, with two exceptions, never designed to estimate the intelligence of Indians. The dataset lists each sample with the study it came from, and reading those studies by title is the fastest education in what the number contains. Two rows come from a 1988 study of the consequences of consanguinity on cognition, in which two outbred groups of children aged 9 to 12 were tested on the WISC-R against American norms of 1974. One row is the non inbred comparison group of a 1984 study of the effects of inbreeding on Raven's matrices, with 100 adolescents. One is a 1980 paper on modernization among Indian farmers, tested around 1968. One is the group without reading difficulties in a 2002 survey of reading problems among South Indian children. Four rows are the same 403 Pune schoolgirls from a 2014 study of zinc status and cognition, split into four plasma zinc quartiles and entered as four samples. Three rows are pupils in government aided, private aided and private unaided schools in a 1996 economics paper on school quality. One is a 2010 study of Santal children in Purulia district, explicitly about nutritional and socioeconomic status. Two rows are the epilepsy and control groups of a 1998 Kerala study of psychiatric disorder in children with epilepsy. One is the control group of a 1993 study of diabetic children. One, labelled teachers, comes from a 2010 study of low cost private schools. One is 140 medical students given the WAIS-IV in 2015. The remaining two are the Indian standardization samples reported in the 1999 Raven's manual, 8,609 children aged 8 to 15 and 2,248 adolescents aged 16 to 18, tested in 1997.

Becker rates each sample for representativeness, and on that scale, where the best rated samples anywhere in the dataset reach 0.8, fifteen of India's twenty rows score 0.1 and none scores above 0.4. By our count of the age columns, about 80 percent of the people in the twenty rows were 18 or younger, and the dataset's own national summary gives a mean age of 14.6 for India against a lowest and highest age of about 11.7 and 18.4. The figure that is presented as the intelligence of a country of 1.4 billion is, in its actual composition, a figure about children in convenience samples, most of them assembled to study something else.

None of this is a criticism of the individual studies, which measured what they set out to measure. A nutrition study that finds a 22 point gap between the lowest and highest zinc quartiles of the same schoolgirls has found something real about zinc and matrices performance. What it has not done is estimate the mean of the population, and averaging it with an epilepsy control group and a 1968 survey of farmers does not turn the collection into a census. The page on the six cognitive domains explains that a matrices score is one indicator of fluid reasoning, not a full battery, which is a second limit on what these rows can say even about the children in them.

3 Our Arithmetic on the Same Rows

Re-running the average with different but equally defensible choices moves India's figure by more than ten points, which is the clearest measure of how much the number depends on method rather than on Indians. Every calculation in this section is ours, performed on the twenty rows of the dataset as downloaded on September 19, 2026, and it is labelled as such so that nobody mistakes it for a published estimate. Weighting each row by its listed sample size, the mean of Becker's corrected IQs is 77.8, and the mean of the uncorrected IQs, before his Flynn adjustments, is 80.4. The unweighted mean of the corrected values is 73.5 and the median is 72.1. The lowest row is 60.0 and the highest is 88.7.

The two Raven's standardization samples deserve separate treatment, because they are the only rows drawn for the purpose of describing Indian performance and they hold 10,857 of the people. Weighted together they give 85.5 after correction and 88.7 before it. The five samples that used the Coloured Progressive Matrices, a shorter and easier form built for young children and for clinical use, hold 3,352 people and average 62.0. Whether the Coloured form is an appropriate basis for an adult IQ estimate is a fair question, and the page on matrix reasoning describes how the difficulty gradient of a matrices test governs what it can discriminate. If a reader took the position that only the standardization samples count, India's figure would be 85 to 89. If a reader took the position that the nutrition and clinical samples count, it would be near 62. Becker's 76 is a weighted compromise between those positions, and the compromise is a decision, not a finding.

The dataset's own version history makes the same point at world scale. The release table on the View on IQ page, read on September 19, 2026, lists a global mean of 81.98 for version 1.3.2 in March 2019 and 86.72 for version 1.3.3 three months later, a rise of 4.7 points in the average IQ of the planet from a change of method, by our subtraction. Nobody's intelligence changed between March and June 2019. What changed was how the samples were weighted and corrected.

Read on the ordinary scale, 76 is 1.6 standard deviations below the British mean of 100, which corresponds to roughly the 5th percentile of that reference population, as the percentile calculator will confirm for any score. The page on the standard deviation of 15 explains the unit, and the page on score versus percentile explains why a 24 point gap sounds larger as a percentile than as a score. A gap that size between two populations would be an extraordinary claim if it rested on comparable samples and comparable norms. It rests on neither, and the rest of this page is about why.

4 British Norms, the Flynn Effect and the Ceiling of a Children's Test

A Raven's raw score becomes an IQ only after comparison with a norm table, and every norm table in India's rows belongs to another country and another decade. Seventeen of the twenty rows are Raven's matrices scored against British standardizations of 1979, 1992 or 2007; two are the WISC-R against American norms of 1974; one is the WAIS-IV. When a child in Kerala in 2002 is scored against British children of 1979, the number that comes out encodes two things at once: how the Kerala child performed, and how much British performance rose between 1979 and 2002. The second component is the Flynn effect, and it is large. Trahan and colleagues' meta-analysis of 285 studies covering 14,031 people put the rise at 2.31 points per decade, and 2.93 for modern Wechsler and Stanford-Binet tests. Pietschnig and Voracek's century of global IQ gains, across 271 samples and almost four million people in 31 countries, put full scale gains at 0.28 points per year and gains on fluid tests, which is what Raven's is, at 0.41 per year.

Becker corrects for this, which is why his corrected values are usually below the uncorrected ones. The larger standardization sample falls from 88.4 to 84.6 in the correction, and the Kerala reading survey rises slightly, from 64.0 to 65.1, because its norm year was later than its test year. The corrections are principled, and they are also assumptions: they apply a British rate of gain to Indian children, and they assume that the gain in the norm country is the right quantity to remove. Wicherts and colleagues showed in a study of Raven's performance in Africa that the same test can behave differently across populations, that a Flynn effect is visible within the low scoring populations themselves, and that the psychometric properties of the test, not only the mean, need to be checked before a mean is interpreted. None of that checking is available for most of India's rows.

The choice of form adds a further problem. The Coloured Progressive Matrices, used in five of the rows, has 36 items intended for children and for people who cannot be tested on the Standard form. It has a low ceiling and a compressed range, and its norms cover young children and, in some editions, older adults and clinical groups, but not the general adult population. The page on Raven's 2 describes the current edition and its age coverage. Converting a Coloured form score into an adult IQ requires an extrapolation the manual does not license, and the five Coloured form rows are also the five lowest in the table. A reader who wants to know whether the low values reflect Indian children or the instrument has no way to tell from the published means, and neither has Becker, which is why his quality ratings for those rows are at the bottom of the scale.

5 What a National Figure Would Require

India runs national surveys that reach hundreds of thousands of households; none of them administers an intelligence test, and that absence is the single most important fact about the average IQ in India. The fifth National Family Health Survey, published by the International Institute for Population Sciences and the Ministry of Health and Family Welfare, gathered information from 636,699 households, 724,115 women and 101,839 men in two phases between June 17, 2019 and April 30, 2021, with anthropometry and haemoglobin measured in the home. The Annual Status of Education Report for 2024 reached 649,491 children in 17,997 villages across 605 rural districts, and gave each child a one to one reading and arithmetic task with a method unchanged since 2006. Those are what representative measurement looks like: a sampling frame, stratification by state and by rural and urban residence, a standard protocol, and a published sampling design.

A national IQ for India would need the same apparatus and something more. The instrument would have to be adapted and normed across the languages of instruction, of which the Eighth Schedule of the Constitution recognizes 22, with items shown to function equivalently across them. The International Test Commission's guidelines for translating and adapting tests set out what equivalence requires, and no such evidence has been published for any Indian population sample. The norm sample would have to be drawn by probability from the adult population, not assembled from schools, clinics and volunteers. And the scale would have to be defined on that sample, at which point the average Indian would score 100 and the question would have changed into a different one: how a probability sample of Indians performs on a foreign scale, which is the comparison the national IQ literature actually attempts.

It is worth being equally clear about ACIS here, because this page is published by an assessment. The ACIS adult reference frame described in the technical manual contains 3,243 English speaking records aged 16 to 90, with the technical tables computed on 2,750 complete records. That is an English speaking online reference frame, not a national sample of any country, and ACIS makes no claim to estimate the mean of India or of anywhere else. The page on reliability and validity states what the frame supports, which is the interpretation of an individual profile against a defined reference group, and that is the only use this page recommends.

6 The Self-Selected Online Numbers

An online test with a mean fixed at 100 cannot tell you the average IQ of any country, because its takers choose themselves and its scale is anchored to them, not to any population. The International IQ Test table is the largest such dataset that is publicly broken down by country, and the ACIS country page uses it with that warning printed beside the tool. In it, India's 8,340 takers average 98.44, within a point and a half of the United Kingdom and the United States and eight points below China and Japan, whose takers number in the hundreds of thousands. Nothing about that spread describes populations. It describes who arrives at an English language puzzle site in each country, and in India the arriving group is a narrow slice of a very large and very unequal population.

ASER 2024 offers a rare glimpse of how narrow. Among rural children aged 14 to 16, close to 90 percent reported having a smartphone at home and 82.2 percent said they could use one, which sounds like near universal access. Yet 57 percent of those who could use a phone had used it for an educational activity in the previous week, only 27 percent of 14 year olds owned a phone of their own, and the same survey found that more than half of Std V children in government schools could not read a Std II level text. Sitting a timed English reasoning test on the open web selects heavily on schooling, language and interest, and the 98 is the mean of the selected. The page on free versus validated IQ tests explains why a self-selected mean of 100 is not a norm, and the page reviewing the International IQ Test examines that site on its own terms.

The right conclusion is not that 98 is the true figure and 76 a slander, nor the reverse. It is that the two numbers are answers to different questions. If the question is how Indians who take online tests compare with Britons who take online tests, the answer is that they are within two points, and the reader should remember that the two groups are selected in similar ways from populations that are unlike each other in schooling and income. If the question is how a representative sample of Indian adults would perform on a battery normed for them, no dataset in existence answers it. The page on whether online IQ tests are accurate treats the general problem of self-selection, and everything it says applies with more force to a country whose online population is so different from its whole.

7 Indian Norms: Why an Indian Test Gives 100 by Construction

A test standardized in India reports 100 for the average member of its own norm sample by definition, so the choice of that sample decides what an Indian IQ means, and the published critique of the best known Indian adaptation is that its sample was drawn from the top of the country. Pearson Clinical India lists a WAIS-IV India edition on its product page, read on September 19, 2026, as an edition of the fourth Wechsler adult scale adapted and standardized for India. In 2024, Porrselvi, Irani, Shah and Sunderaraman published a critical review of the WAIS-IV India in Archives of Clinical Neuropsychology, comparing its manual with the American and British editions. Their reading of the manuals found that the normative sample consisted of fluent English speakers with high education levels, that the cultural and linguistic adaptation of the subtests was minimal, and that item selection, demographic composition, sample selection and psychometric properties were not reported clearly enough to judge reliability and validity. Their conclusion was that the edition may not adequately represent India's population.

The consequence for the phrase "average IQ in India" is direct. On the WAIS-IV India, an educated English speaking adult in a city is compared with a norm sample of educated English speaking adults and scores near 100, which tells a clinician something useful about that adult against peers, and tells nobody anything about India. A rural Hindi or Tamil speaking adult tested on the same edition would be compared with a sample unlike them, and the number that came out would be an artefact of the sample as much as a measurement of the person. The page on whether IQ tests are biased explains the technical meaning of bias, which is a difference in what a score predicts across groups rather than a difference in means, and the page on culture fair tests explains why no instrument has yet escaped the problem entirely.

Older Indian instruments carry the same limits in different forms. Adaptations of Western scales for Indian children have existed since the 1930s, and the Malin's adaptation of the Wechsler children's scale from the 1960s is still used in clinics; their norms are old, regional and, for the most part, unpublished in the peer reviewed literature. The page on average IQ by language shows what happens to verbal subtests when the language of testing is not the language of the home, and the page on verbal IQ tests describes how heavily a Wechsler full scale score depends on those subtests. An Indian test with a national probability norm, in the languages of the population, with published invariance evidence, would produce a mean of 100 by construction and a distribution that could finally be compared with anything. It does not yet exist.

8 Nutrition and Early Growth: What Is Documented

The best documented facts about cognitive development in India are not IQ scores; they are growth and nutrition figures from a survey that measured children in more than 600,000 households, and they describe conditions with known effects on any cognitive test. NFHS-5 reports that 36 percent of children under five were stunted, 19 percent wasted and 32 percent underweight in 2019 to 2021, down from 38 and 21 percent for stunting and wasting in 2015 to 2016. Anaemia among children aged 6 to 59 months rose over the same interval from 59 to 67 percent. Stunting was 37 percent in rural areas against 30 percent in urban ones, and 46 percent among children born to mothers with no schooling against 26 percent among children of mothers with twelve or more years. Those gradients are the ones that matter for a country mean, because they show that the average conceals a population in which early nutrition varies enormously with schooling, wealth and place.

The link from those conditions to test performance is established in the child development literature rather than in the IQ literature. Grantham-McGregor and colleagues' Lancet estimate that 219 million children under five in developing countries were failing to reach their developmental potential in 2004, through stunting, poverty or both, placed India first by absolute number with 65 million, ahead of Nigeria at 16 million and China at 15 million, and put the share of disadvantaged children in South Asia at 52 percent. The companion paper by Walker and colleagues on risk factors for adverse outcomes set out the evidence for stunting, iodine and iron deficiency, and inadequate stimulation as causes of poorer cognitive development. Within India, Spears used the India Human Development Survey, a representative sample of 40,000 households that matched anthropometry to learning tests, to show that the height achievement slope among Indian children is more than twice as steep as in the United States, and that a child one standard deviation taller was 5 percentage points more likely to be able to write, falling to 3.4 points after a long list of contemporary and early life controls.

The zinc study that supplies four of Becker's rows is itself evidence of the same kind. In Chiplonkar and Kawade's sample of 403 Pune schoolgirls, 72 percent had low plasma zinc, and plasma zinc correlated 0.39 with the Raven's score after adjustment for socio demographic factors and haemoglobin. A 22 point span across the zinc quartiles of one school sample is a measurement of nutrition, and any national mean that contains those rows contains that nutrition gradient inside it. The page on what IQ measures is careful to say that a score reflects the performance on the day, and on the days in question a large fraction of the children were undernourished by the survey's own definitions.

9 Sanitation and Poverty: Two Indian Natural Experiments

Two Indian studies with designs strong enough to support causal language show that conditions of early life and of the moment move cognitive test performance, and both are cited far less often than the national IQ tables they help to explain. Spears and Lamba examined India's Total Sanitation Campaign, a government programme that encouraged local bodies to build latrines and promote their use, and published the results in the Journal of Human Resources in 2016 after a 2013 World Bank working paper. Using the timing and intensity of the programme across districts, they found that six year olds who had been exposed to it in their first year of life were more likely to recognize letters and simple numbers than children who had not. The mechanism they proposed is that open defecation spreads disease that stunts growth and development in infancy, so a sanitation programme reaches cognition through the same channel that NFHS-5 measures as stunting.

Mani, Mullainathan, Shafir and Zhao followed 464 sugarcane farmers in Tamil Nadu through a harvest cycle and reported in Science that the same farmers performed worse on Raven's matrices items and on a cognitive control task before the harvest, when they were poor, than after it, when they had been paid. The authors could not explain the difference by time available, nutrition, work effort or measured stress, and concluded that financial concern itself consumes mental resources. Wicherts and Scholten challenged the study on statistical and psychometric grounds, and the authors replied that the pre harvest deficit survived a continuous income measure, was not driven by ceiling effects and was robust to learning. A reader should hold both the finding and the challenge; what survives either way is that a matrices score taken from a poor farmer before harvest is not the same measurement as one taken after, and the 1968 farmers in Becker's table were tested at an unknown point in that cycle.

Both studies matter for the national figure in a specific way. They show within India, with Indian participants and Indian instruments, that the quantity a Raven's score captures moves with sanitation in infancy and with money in the month of testing. A mean computed from children in nutrition studies and farmers in the lean season is therefore not a mean of some fixed national trait; it is a mean of test performances under conditions that the same literature shows to depress them. The page on IQ and income reviews the correlational literature that runs in the other direction, and the two directions are not in conflict: the association between income and score is produced by causes running both ways, and India supplies unusually clear evidence for the one that runs from poverty to score.

10 Schooling: What ASER and PISA 2009 Plus Measure

Schooling raises measured intelligence by one to five points for each additional year, and India's schooling data show gaps in basic skills that make a child's test score as much a record of the school as of the child. Ritchie and Tucker-Drob's meta-analysis of 142 effect sizes from 42 datasets and more than 600,000 participants, using designs that control for earlier ability, exploit compulsory schooling reforms or use school entry cut offs, found consistent gains of about one to five IQ points per year of education, persisting across the lifespan and across all broad ability categories studied. Brinch and Galloway's Norwegian reform study found 3.7 points per year on a conscription test. In a country where the length and the quality of schooling vary as much as they do in India, those effects are not a footnote to a national mean; they are a large part of what the mean measures.

ASER 2024 gives the scale of the variation. Nationally, 23.4 percent of Std III children in government schools could read a Std II level text, the highest figure since the survey began but still fewer than one in four. Among Std V children the figure was 44.8 percent in government schools and 59.3 percent in private schools. In Std VIII, 67.5 percent of government school pupils could read that text and 45.8 percent could do basic arithmetic. Two thirds of Std I and Std II classrooms in primary schools were multigrade. Only 7.9 percent of 15 and 16 year olds were not enrolled, so the children who do not attend are now a small minority; the larger issue the survey documents is what attendance delivers. Becker's two school economics rows illustrate the point: in the 1994 sample, pupils in private unaided schools scored 78.4 against 65.4 in government aided schools on the same test in the same city, a 13 point gap between school types that a national average silently blends.

India's only international assessment result comes from PISA 2009 Plus, in which Himachal Pradesh and Tamil Nadu took part alongside eight other economies. The report by Walker for ACER records a mean reading score of 337 for Tamil Nadu and 317 for Himachal Pradesh against an OECD average of 493, with 83 and 89 percent of students respectively below the baseline proficiency level; mathematics means of 351 and 338; and a science mean of 325 in Himachal Pradesh, the lowest observed in the round along with Kyrgyzstan. The report also notes that both Indian states did not meet all of the PISA sampling standards and that their results should be read with caution. Rindermann's analysis of international comparisons found that national means on student assessments and on intelligence tests correlate at about 0.85 and load on a single factor, which is why Becker folds school assessments into his figure; the PISA 2009 Plus result is, in effect, the second half of India's 76. The page on IQ and academic achievement explains why the two kinds of measure track each other so closely at the individual level, and the page on average IQ by education shows the size of the gradient within a single country.

11 The Cross-National Critique: Selection, Invariance and What the Correlations Show

The strongest objections to national IQ tables were made about Africa, and they apply to India's rows point for point: the samples are selected, the tests have not been shown to measure the same thing across populations, and the correlations with national outcomes are real but do not identify a cause. Wicherts, Dolan and van der Maas' systematic review of the literature on sub-Saharan Africa found that Lynn's estimate of an average below 70 depended on the inclusion of unrepresentative samples and on how the Flynn effect was handled; applying explicit inclusion rules to the same literature, they arrived at about 82 against United Kingdom norms, and about 76 on Raven's tests specifically in the companion study. The lesson is not that the corrected number is right; it is that the procedure moves the answer by ten or more points, which is the same sensitivity our arithmetic finds for India.

The second objection concerns measurement invariance, the property that a test measures the same construct in the same units across groups. Without it, a difference in means cannot be read as a difference in ability, because it may be a difference in familiarity with the format, the language of instruction, or the meaning of the items. The African Raven's study documented departures from invariance and a Flynn effect inside the African samples themselves, and the paper on why national IQs do not support evolutionary theories by Wicherts, Borsboom and Dolan set out why a table of national means, whatever its accuracy, cannot carry the weight of causal claims about populations. No invariance study exists for the Indian samples in Becker's table, and the WAIS-IV India review found that even the flagship adaptation did not report what would be needed.

The third point is what the tables can legitimately show. National means on cognitive tests correlate with national income, health and schooling, and Hanushek and Woessmann's review in the Journal of Economic Literature argues that measured cognitive skills, rather than years of schooling, are what predicts growth across countries. Daniele and Ostuni's study of the burden of disease found that infectious disease burden is among the strongest correlates of national IQ estimates. Those findings sit comfortably with the Indian evidence above: a country with 36 percent stunting, 67 percent child anaemia and a quarter of Std III pupils reading at grade level will produce low means on any cognitive test, and the low means will predict its economic outcomes, because they are partly measuring the same deprivation from two sides. The page on what heritability actually means explains why a heritability figure estimated within a population says nothing about the cause of a difference between populations, and that distinction is where most misreadings of the Indian figure begin.

12 The Evidence by Design

Read by design rather than by headline, the evidence about India sorts into a small number of representative surveys of conditions, two well identified studies of causes, and a large stack of convenience samples of test scores, and the table below is the part of this page a careful reader should keep. The column that matters most is the third: what population each study can speak for.

Two things stand out from the table. The studies that can speak for India measure conditions, not intelligence, and the studies that measure intelligence cannot speak for India. Everything that is said about the average IQ in India lives in the gap between those two rows, and the honest way to fill the gap is to say that the conditions are measured, the effect of those conditions on test scores is established in general and in India specifically, and the mean itself is not measured.

13 What the Evidence Supports, Stated Narrowly

Stated as narrowly as the designs allow, the evidence supports five sentences about the average IQ in India, and none of them is the sentence the ranking tables imply. First, samples of Indian children and adolescents tested on Raven's matrices between 1968 and 2015 scored, on average, well below the British norm groups of the same period, with the best sampled rows at about 85 to 89 after Flynn correction and the convenience rows lower. Second, no representative sample of Indians has been tested on any intelligence test, so no mean for the population exists, and the 76 is a weighted summary of unrepresentative rows whose composition is dominated by children. Third, the conditions documented by representative surveys, including 36 percent stunting, 67 percent child anaemia and reading levels at which most Std V government school pupils cannot read a Std II text, are conditions with established effects on cognitive test performance, and Indian studies with strong designs show those effects operating within the country. Fourth, self-selected online samples of Indians score within about two points of self-selected online samples from Britain and the United States, and that fact says nothing about either population. Fifth, a test normed in India returns 100 for the average member of its norm sample by construction, and the best known Indian adaptation was normed on educated English speakers.

What the evidence does not support is any statement about the abilities of Indians as a group relative to other groups, because the comparisons on offer are between unlike samples on tests not shown to be invariant. It does not support the claim that the 98 corrects the 76, nor that the 76 corrects the 98. It does not support the claim that the figure is fixed, because the same literature shows schooling, nutrition and sanitation moving scores, the Flynn effect operating in every population in which it has been sought, and ASER's reading figures rising between 2022 and 2024. And it does not support the reading of a national mean, from any source, as a fact about any person. The page on average IQ explains what a mean of 100 means for the population that defines it, and the page on what a good IQ is explains why the individual question is a different question with a different answer.

The narrow statement is less satisfying than a number, and it is the only statement the evidence licenses. A future survey that drew a probability sample of Indian adults, tested them in their own languages on an instrument with published invariance evidence, and reported the distribution rather than a single figure would change this page entirely. Until it exists, every average IQ in India is a statement about the dataset that produced it.

14 What This Does Not Say About You

A national mean, even a well measured one, carries no information about an individual beyond what a population base rate carries, and the Indian means are not well measured, so the number says nothing at all about any one reader. The spread of scores within any country is far wider than the gap between any two country means. On a scale with a standard deviation of 15, two thirds of a population lies within 15 points of its mean and 95 percent within 30, and the 24 point gap between 76 and 100 is smaller than the range that separates ordinary siblings. An Indian reader who scores 120 on a normed test is at the 91st percentile of that test's reference frame, and the fact that some table lists their country at 76 changes that percentile by nothing. The page on the IQ score chart shows the whole distribution, and the page on Full Scale IQ explains what a composite score does and does not summarize.

The direction of inference matters too. A national figure built from children in nutrition studies does not describe an adult who was not undernourished, did not attend a multigrade classroom and was not tested against a norm from a different country and decade. It does not describe the reader's reasoning, memory or speed, which are separate domains that a single matrices mean never touched. The page on fluid versus crystallized intelligence explains why a matrices score, the basis of seventeen of Becker's twenty rows, is one domain among several, and why a full profile can be strong in domains a matrices test never samples.

For an Indian adult who wants a number that describes them, the practical position is the one this page has applied to every dataset above: ask whose norms, whose sample, and what the standard error is. An ACIS report scores an individual against an English speaking adult reference frame of 3,243 records aged 16 to 90, prints the standard error of every index, and states in the technical manual that verbal, reading and knowledge subtests depend on exposure to the English register the items sample. That frame is not India, and an Indian reader whose schooling was not in English should read the verbal indices with that caveat and weight the nonverbal and speed indices accordingly; the page on nonverbal IQ tests explains which indices depend least on language. In this site's own search data, exported on September 19, 2026, India was the sixth largest source of visitors by country, and the sentence those visitors most need is the one the ranking tables never print: the country figure is not about you, and the only score that is about you is one you sat, on a stated scale, with a stated error.

15 Sources Behind This Page

Every figure above is traceable to one of the following, and each is linked at the point where it is used. Dataset values for India, the United Kingdom and the comparison countries were read from the NIQ dataset file downloaded from View on IQ on September 19, 2026, and the calculations described as ours were performed on that file. Survey figures are quoted from the NFHS-5 final report and the ASER 2024 national findings as published.

• Becker D. The NIQ dataset, version 1.3.5, released October 25, 2023, with the release table of earlier versions. viewoniq.org, read September 19, 2026.
• Lynn R and Vanhanen T. IQ and the Wealth of Nations. Praeger, 2002. doi.org.
• International IQ Test. IQ by country table, as used by the ACIS country ranking page. international-iq-test.com, read September 19, 2026.
• International Institute for Population Sciences and ICF. National Family Health Survey (NFHS-5), 2019 to 2021: India report. dhsprogram.com, 2022.
• ASER Centre. Annual Status of Education Report (Rural) 2024: national findings. asercentre.org, January 2025.
• Walker M. PISA 2009 Plus Results: Performance of 15-year-olds in reading, mathematics and science for 10 additional participants. ACER, 2011.
• Grantham-McGregor S, Cheung Y B, Cueto S, Glewwe P, Richter L and Strupp B. Developmental potential in the first 5 years for children in developing countries. The Lancet, 2007, volume 369, issue 9555, pages 60 to 70.
• Walker S P, Wachs T D, Gardner J M, Lozoff B, Wasserman G A, Pollitt E and Carter J A. Child development: risk factors for adverse outcomes in developing countries. The Lancet, 2007, volume 369, issue 9556, pages 145 to 157.
• Spears D. Height and cognitive achievement among Indian children. Economics and Human Biology, 2012, volume 10, issue 2, pages 210 to 219.
• Spears D and Lamba S. Effects of early-life exposure to sanitation on childhood cognitive skills: Evidence from India's Total Sanitation Campaign. Journal of Human Resources, 2016, volume 51, issue 2, pages 298 to 327.
• Mani A, Mullainathan S, Shafir E and Zhao J. Poverty impedes cognitive function. Science, 2013, volume 341, issue 6149, pages 976 to 980.
• Chiplonkar S A and Kawade R. Linkages of biomarkers of zinc with cognitive performance and taste acuity in adolescent girls. International Journal of Food Sciences and Nutrition, 2014, volume 65, issue 4, pages 399 to 403.
• Porrselvi A P, Irani F, Shah U and Sunderaraman P. Cultural considerations in intelligence test adaptations: a critical review of the WAIS-IV India and its U.K. and U.S. counterparts. Archives of Clinical Neuropsychology, 2024, volume 39, issue 8, pages 1354 to 1362.
• Wicherts J M, Dolan C V and van der Maas H L J. A systematic literature review of the average IQ of sub-Saharan Africans. Intelligence, 2010, volume 38, issue 1, pages 1 to 20.
• Pietschnig J and Voracek M. One century of global IQ gains: A formal meta-analysis of the Flynn effect (1909 to 2013). Perspectives on Psychological Science, 2015, volume 10, issue 3, pages 282 to 306.
• Ritchie S J and Tucker-Drob E M. How much does education improve intelligence? A meta-analysis. Psychological Science, 2018, volume 29, issue 8, pages 1358 to 1369.

16 Frequently Asked Questions