Poverty and IQ: what cash trials and adoption show

0 The short answer

Poverty and lower IQ scores are consistently associated, but randomized and quasi-experimental studies have tested only part of the causal question, and most of them measured school achievement, behavior or brain activity rather than IQ. In Baby's First Years, a United States randomized cash trial, 891 children were analyzed at age 4, and none of four preregistered primary outcomes differed significantly between mothers given 333 dollars a month and mothers given 20 dollars a month. A tax credit natural experiment estimated a gain of 6 percent of a standard deviation in achievement test scores per 1,000 dollars of income, child benefit studies reported positive test score effects, and a late adoption study found larger IQ gains in higher status adoptive families than in lower status ones. A null result at age 4 is not evidence that money does not matter, and none of these designs says anything about the score of one person.

1 Does Poverty Lower IQ, or Does It Only Travel With It?

Poverty and measured IQ are associated in the large samples reviewed here, and the open question is how much of that association is caused by money itself. The most detailed description of the association comes from the Twins Early Development Study. Von Stumm and Plomin followed 14,853 children who were assessed nine times on IQ between ages 2 and 16. In that sample, children from low socioeconomic status families scored on average 6 IQ points lower at age 2 than children from high status backgrounds, and by age 16 the difference had almost tripled. The authors state that their finding "does not warrant causal interpretations of the relationship between SES and the development of intelligence."

That caution is the starting point of this page. Three different explanations produce the same correlation, and each predicts something different when a study changes family income on purpose.

• Income changes development. Money buys books, stable housing, nutrition, health care and parental time, and it may reduce the strain on parents that shapes the home environment.
• A third factor produces both. Parents differ in education, health, neighborhood and inherited traits, and these can influence both family income and a child's test score.
• Cognition influences income. Ability can affect earnings across a lifetime, which is the direction covered by the page on IQ and income.

The three explanations are not exclusive, and a study that supports one does not rule out the others. A randomized trial that gives some families extra cash tests the first explanation while the third is held roughly equal by chance. A study of children adopted into different families breaks the link between a child's inherited traits and the family that raises them. A twin study asks how much of the variance in scores is attributable to shared family environment versus genes at different income levels. Each of these is a different question, and they have different answers, which is why this page is organized by design.

Poverty here means what the studies mean: low family income, low parental education, or an index that combines them. These describe household circumstances, not the people in them. The page on what IQ measures explains what a score does and does not capture, and the page on the Flynn effect shows that average scores in whole populations have changed across generations, which is one more reason the correlation cannot be read as a fixed trait of anyone.

Units matter for the questions that follow. IQ is reported on a scale with a mean of 100 and a standard deviation of 15, as the page on the standard deviation of 15 explains. Most of the studies below report effects in standard deviations of a test, and many of the tests are achievement tests or brain recordings. A gain of 0.20 standard deviations would equal 3 points on an IQ scale (our arithmetic), but a standard deviation on a school achievement test is not interchangeable with a standard deviation of IQ, so every conversion in this article is labeled as illustrative.

2 What Each Research Design Can Show

The strength of a causal claim depends on the design that produced it, and the poverty literature contains five designs with different strengths. In a widely cited review, Duncan, Magnuson and Votruba-Drzal put the point directly: in contrast to the nearly universal associations between poverty and children's outcomes in the correlational literature, impacts estimated from social experiments and quasi-experiments are more selective. They link increases in family income to increased school achievement in middle childhood and to greater educational attainment in adolescence and early adulthood. They also report that, at the time of writing in 2017, there was no experimental or quasi-experimental evidence in the United States linking child outcomes to economic deprivation in the first several years of life.

Observational designs with controls compare families with different incomes and adjust for measured differences, or follow the same family as its income changes (fixed effects). They are the most common and the most vulnerable to unmeasured differences. Cooper and Stewart's systematic review found that fixed effect studies identify smaller impacts of income than experimental designs, a point the review section below quantifies.

Natural experiments use a change in income that arrives for reasons unrelated to the family, such as a tax credit expansion or a payment from casino profits that does not depend on household characteristics. They are as good as the claim that the change is unrelated to other things that matter for children.

Randomized trials assign income by lottery, which makes the groups comparable on everything measured and unmeasured, apart from chance. Their limits are sample size, the size and length of the payment, attrition, and what the follow-up measures.

Adoption and twin designs do not change income. They separate the contribution of the rearing family from inherited differences, so they bear on the question of how much the family environment matters, not on money specifically.

Laboratory and field scarcity studies test whether financial concern interferes with performance in the moment. They address a short-term mechanism, not the long-run development of children.

The table summarizes how the designs differ. The page on reliability and validity explains why the measure used, an IQ test or a school achievement test, shapes what a result can mean.

3 Baby's First Years: Design and What It Measured

Baby's First Years randomized low-income mothers to a large or a nominal monthly cash payment for years and followed their children, and that design determines what its results can say. The authors describe it in their age 4 working paper for the National Bureau of Economic Research. One thousand racially and ethnically diverse mothers with incomes below the United States federal poverty line were recruited from postpartum wards in 2018 and 2019. They were randomized to receive either 333 dollars a month or 20 dollars a month for the first several years of their children's lives. The payments were unconditional, meaning that the mothers could spend the money as they chose and did not have to meet any requirement to receive it.

The comparison is therefore between a large gift and a nominal one, not between a gift and nothing. The high-cash amount came to about 4,000 dollars a year, which the authors report amounted to 18 percent of the average mother's income at baseline and led to an average 14 percent impact on net income over the first three years. Of the 1,000 mothers, 984 remained eligible for the age 4 data collection, which ran from July 2022 to August 2023, and the analysis sample for the age 4 outcomes was 891. The design divided the sample 40 percent to the high-cash group and 60 percent to the low-cash group, and the authors planned for enough participants to detect an effect of about 0.20 standard deviations with 80 percent power, assuming 20 percent attrition.

The trial preregistered four primary outcomes and three secondary outcomes, and the choice of measures matters for this page.

• Language, measured with the Receptive One Word Picture Vocabulary Test, a receptive vocabulary test.
• Executive function, measured with the Minnesota Executive Function Scale.
• Social-emotional problems, reported by the mother on the Child Behavior Checklist.
• High-frequency brain activity, a composite from resting electroencephalography (EEG).
• Visual processing and spatial perception, a secondary outcome, measured with the Matrices subtest of the Wechsler Nonverbal Scale of Ability.
• Pre-literacy, a secondary outcome, measured with a book-based task called The Reading House.
• Developmental condition diagnosis, a secondary outcome reported by the mother.

IQ was planned and then dropped. The authors state in a footnote that IQ scores on the Wechsler Nonverbal Scale of Ability, which are calculated from both a Matrices subtest and a Recognition subtest, were originally preregistered as a primary outcome. Preliminary analysis of the first 71 participants showed that 21 percent of them scored at the floor of the Recognition assessment, so the team dropped that subtest from data collection. That decision precluded calculating IQ for later participants, and the authors changed the preregistration to treat the Matrices subtest, which the authors describe as visual processing and abstract spatial perception, as a secondary outcome. In plain terms, the trial reports no IQ, and what it does report is a nonverbal matrices raw score and a set of language, executive function, behavior and brain measures. A test of this kind is related to the fluid reasoning (Gf) tasks described on the page about Matrix Reasoning, but a raw score on one task from a four-year-old is not an IQ.

Randomization is the trial's strength. The contrast is 333 against 20 dollars, so it tests the effect of roughly 313 extra dollars a month (our arithmetic), not the effect of moving a family out of poverty.

4 Baby's First Years Results at Ages 1 and 4

At age 1 the trial reported higher high-frequency brain activity in the high-cash group with significance that depended on the analytic specification, and at age 4 it reported no significant difference on any preregistered outcome. The first-year paper, Troller-Renfree and colleagues in the Proceedings of the National Academy of Sciences, used resting EEG at about 1 year of age in a subsample of 435 infants. The authors hypothesized more power in mid to high frequency bands and less power in a low frequency band in the high-cash group. They reported that infants in the high-cash group showed more power in high-frequency bands, that the effect sizes were similar in magnitude to many scalable education interventions, and that the significance of the estimates varied with the analytic specification. They concluded that monthly cash gifts to mothers experiencing poverty "may change infant brain activity," and they described the pattern as one that has been associated with the development of subsequent cognitive skills. That association is a statement about the pattern, not a measurement of any later cognitive score.

The age 4 results are in the NBER working paper by Noble and colleagues, which its own cover note says has not been peer reviewed. The abstract reports no statistically significant impacts on the four primary outcomes or on the three secondary outcomes. The table below gives the estimates as printed in the paper's Table 3. Effect sizes are treatment effects divided by the standard deviation of the low-cash sample, estimated with site fixed effects and baseline covariates.

Three details in that table deserve plain reporting. The authors state that only two of the six estimated coefficients associated with directional hypotheses were in the hypothesized direction. The matrices estimate points the opposite way from the hypothesis, with the high-cash group scoring slightly lower on average, and it is not statistically significant. And the sample sizes range from 637 to 886 children depending on the measure, with response rates higher in the high-cash group on all seven outcomes. The authors adjusted the estimates with weights that model nonresponse and report that results were robust, but they note that because the same outcomes could not be measured at baseline, unmeasured group differences arising from differential attrition remain possible.

The journal paper on the age 4 brain recordings, Troller-Renfree and colleagues in Developmental Cognitive Neuroscience, used 637 children with usable EEG data, whose mean age was 4.03 years. In the version of the article available through PubMed Central, the preregistered mid to high frequency composite showed an effect size of 0.12 with p = 0.11, and frontal gamma showed minus 0.09 with p = 0.266. The two papers report slightly different estimates for the composite, 0.097 in the working paper's table and 0.12 in the journal article, and both are not statistically significant. In exploratory analyses, relative alpha power showed an effect size of 0.15 with p below 0.05, while absolute alpha power showed 0.16 with p = 0.067, and theta, beta and gamma showed no significant group differences. The authors describe the alpha result as a possible effect needing replication, and an exploratory finding among many tested frequency bands is the kind of result that a confirmatory study must check.

5 Why a Null Result at Age 4 Is Not a Verdict on Money

A null result at age 4 means that this trial, with these payments, at this age, did not detect a difference on these measures, and the authors themselves list several reasons why that is not the same as no effect of income. The working paper is explicit that its conclusions are limited to what was tested, and it discusses explanations in two groups.

The first group concerns the design. Cash gifts were disbursed monthly to virtually all families in both groups across the four years, so non-receipt does not explain the result. The sample was designed with 80 percent power for an effect of about 0.20 standard deviations, and the estimates above are smaller than that, so the trial could miss a smaller true effect. Response rates differed between groups, and the authors had no baseline version of the outcomes to correct for that. The children had little prior exposure to assessment settings, and the assessors reported difficulty maintaining attention, which the authors suggest may have increased measurement error, though they say it is unclear whether that would bias the estimate.

The second group concerns the intervention and its setting. The authors note that the gifts may have been too small, since 91 percent of study families still reported household incomes, including the gifts, under twice the official poverty line when children were 4, and the two groups did not report different levels of material hardship in a paper then under review. They note that inflation reduced purchasing power by about 18 percent over the study period. They also note that the cash transfers were as large as the average Earned Income Tax Credit and welfare reform transfers that boosted child school achievement in earlier studies, which keeps the amount from being obviously too small. The pandemic overlapped the period between the children's first and third birthdays, 60 percent of mothers reported losing income because of it, and 72 percent reported receiving pandemic-related government payments when their children were 2, so many families received other transfers during the study. The authors add that the payments continue until the children's sixth birthdays, so later waves can show whether effects emerge at older ages or after longer exposure.

The null result at age 4 is a result about language, executive function, behavior, brain activity and a matrices task in children who were, on average, 4. It is not a result about IQ in adults, about the effect of leaving poverty, or about the school achievement and attainment outcomes that natural experiments have linked to income later in childhood. The page on chronic stress and IQ covers the adversity pathways through which income might act on cognition, and it is the right page for the mechanism question that a null on cognitive tests leaves open.

The trial did confirm that randomization balanced the groups at baseline and that payments reached the families. Whether the first-year and exploratory alpha power results lead to differences in later test scores is not measured here.

6 Natural Experiments: Tax Credits, Child Benefits and Casino Payments

Natural experiments with a tax credit, child benefits and casino-profit transfers link extra income to higher school test scores and educational attainment, and the abstracts of the three report achievement, attainment or offense outcomes rather than IQ. Each exploits a change in income that was not chosen by the families who received it.

Dahl and Lochner used large, nonlinear changes in the Earned Income Tax Credit between 1993 and 1997 as an instrument for family income. The largest of these changes raised family income by as much as 20 percent, or about 2,100 dollars. Their baseline estimates imply that a 1,000 dollar increase in income raises combined math and reading test scores by 6 percent of a standard deviation in the short run, with larger gains for children from disadvantaged families. The outcome is an achievement test. Rescaled to the IQ metric, 6 percent of a standard deviation would be 0.9 points (our arithmetic), and that conversion is illustrative only, since school achievement tests and IQ tests do not share a standard deviation.

Milligan and Stabile used differences in Canadian child benefit expansions across provinces, years and family types to study outcomes spanning test scores, mental health, physical health and deprivation measures. They report that the benefit programs had significant positive effects on test scores, maternal health and mental health, among other measures. Their abstract does not give a single standardized effect size for test scores, so none is quoted here.

Akee and colleagues studied a transfer of casino profits to households that was unrelated to household characteristics, and followed the children's long-run outcomes. The abstract reports that children in affected households had higher levels of education in young adulthood and a lower incidence of criminality for minor offenses, and the authors estimate that an additional 4,000 dollars a year for the poorest households increases educational attainment by one year at age 21 and reduces the chances of committing a minor crime by 22 percent for 16 and 17 year olds. Educational attainment and offenses are the outcomes, not test scores.

Two features of this group of studies bear on the poverty and IQ question.

• The outcomes are achievement and attainment. A school test measures what has been taught as well as general ability, so a gain on it is not automatically a gain on an IQ test. The page on IQ and academic achievement explains how the two relate.
• The changes are in the range of thousands of dollars. These natural experiments describe the effect of adding a few thousand dollars a year to a household budget, not the effect of leaving poverty in a more general sense.

The natural experiments are not contradicted by the age 4 result of Baby's First Years, or the reverse. They mostly measure school-age children on achievement and attainment, while the trial measured preschoolers on different instruments, so the results answer neighboring questions, not the same one.

7 What the Systematic Reviews and the Mexican Program Add

Systematic reviews find that most credible studies show positive effects of income on child outcomes, with larger estimates in experimental and quasi-experimental studies than in observational ones, and with a cognitive category that includes school achievement and not only IQ. The Cooper and Stewart review covered OECD country studies from 1988 to 2017 that used randomized trials, quasi-experiments or fixed effect techniques on longitudinal data. Of the 54 studies that met its inclusion criteria, 34 showed a significant and positive relation between income and each of the child outcomes they examined. The authors conclude that the results lend strong support to a positive causal effect of household income on cognitive and social-behavioral development and health, particularly in households with low income to begin with.

The review quantifies the design gap that the earlier section described. Fixed effect studies find that 1,000 dollars consistently delivers just 1 to 2 percent of a standard deviation improvement in cognitive outcomes, while studies using randomized trials or quasi-experiments find effect sizes ranging between 5 and 37 percent, in year 2000 dollars. The authors give three reasons the observational estimates might be smaller, the first of which is that experimental and quasi-experimental studies focus on the low-income part of the population while longitudinal surveys include higher-income families. Applied to the IQ metric, the review's range would be about 0.75 to 5.6 points per 1,000 dollars (our arithmetic). The same caveat applies: the review's cognitive category mixes test scores, school engagement and grade point average, and the width of the range reflects different studies and outcomes, not a precise estimate.

Three details from the review sharpen what it says about IQ itself. It notes that Norwegian administrative data include military records that capture IQ for men at age 18, and it reports that years of schooling, high school completion and IQ at age 18 were all found positively linked to local area income shocks between ages 1 and 16 in one Norwegian study. It also reports that the randomized evidence on cognitive outcomes comes from four studies of up to 14 United States and Canadian welfare-to-work programs. And it discusses a study of Swedish lottery wins that found no effects on grade point average or on English, math or Swedish scores, one of the few cases it reviewed without a significant income effect. The review reports a few nulls inside a positive majority, and this page does not pool them into one figure.

The Duncan, Magnuson and Votruba-Drzal review reaches a more selective conclusion about timing and outcome. It links experimental and quasi-experimental income increases to school achievement in middle childhood and educational attainment, and as of its 2017 publication it found no experimental or quasi-experimental United States evidence for the first several years of life. Baby's First Years began recruiting in 2018 and randomizes income in the first years of life, and its results so far are the ones described above.

One randomized program outside the United States adds a different kind of evidence. Fernald, Gertler and Neufeld analyzed Mexico's Oportunidades program, in which 506 low-income communities were randomly assigned to enroll immediately or 18 months later. In 2003 they assessed 2,449 children aged 24 to 68 months who had been enrolled their entire lives. Enrollment was randomized, but the amount of cash was not: the authors used regression to estimate the effect associated with a doubling of cash transfers while controlling for covariates, including measures of socioeconomic status. A doubling of cash transfers was associated with a higher height-for-age score, lower prevalence of stunting, and children doing better on a scale of motor development, three scales of cognitive development, and receptive language. The abstract does not report effect sizes for the cognitive scales, so none is given here. The study is evidence from a different setting, and the page on average IQ in India covers natural experiments on sanitation and poverty in India.

8 Adoption: What Happened to IQ After a Change of Family

A French late adoption study found larger IQ gains in children placed with higher status families than with lower status families, which supports an influence of the family environment on IQ but does not isolate income. Duyme, Dumaret and Tomkiewicz started from 5,003 files of adopted children and selected 65 deprived children, defined as abused or neglected during infancy, who were adopted between 4 and 6 years of age and who had an IQ below 86 before adoption (mean 77, standard deviation 6.3).

At adolescence, with a mean age of 13.5 years, the average IQs of the adopted children in lower and higher socioeconomic status families were 85 (standard deviation 17) and 98 (standard deviation 14.6). The authors report a significant gain in IQ that depended on the status of the adoptive families, a mean gain of 7.7 points in lower status families and 19.5 points in higher status families. They also report that IQs after adoption were significantly correlated with IQs before adoption, and that verbal IQs were significantly lower than performance IQs in adolescence.

Four limits should travel with those numbers.

• The sample is small and selected. Sixty-five children were chosen from thousands of files under strict criteria, so the study cannot be generalized to adoption or to deprivation in general.
• Randomized placement is not reported in the abstract. Which family a child joined may relate to features of the child, so the comparison of 7.7 and 19.5 points is between adoptive homes, not between randomly assigned incomes.
• Status is not income. Higher status adoptive families differ from lower status ones in education, schooling, books, expectations and many other ways, and the design cannot tell which of these produced the difference.
• Selection on low scores matters. Children were chosen because their earlier IQs were below 86, so some of the average gain would be expected from regression toward the mean even without any change in family. The contrast between the two adoptive groups is the informative part, since both were selected the same way (our reasoning, not the authors' statement).

Read narrowly, the study shows that the average measured IQ of deprived children was higher after placement in higher status homes than in lower status homes, with large individual variation, which is the pattern an environmental effect would produce. It does not show that money causes IQ, and it says nothing about children who were not deprived before adoption. The page on whether IQ is genetic explains why a family environment effect and a heritability estimate are compatible rather than rival claims.

9 Gene by Socioeconomic Status Interaction: Turkheimer and the Scarr-Rowe Hypothesis

Twin research asks how much of the variance in IQ is attributable to genes and to shared family environment at different income levels, which is a question about variance and not about why group averages differ. Turkheimer and colleagues analyzed Wechsler Intelligence Scale for Children scores of 7-year-old twins from the National Collaborative Perinatal Project, a sample in which a substantial proportion of the twins were raised in families living near or below the poverty level. Their models allowed the proportions of variance attributable to genes and to shared and nonshared environment to change with socioeconomic status measured as a continuous variable. The proportions varied nonlinearly with status. In the models, in impoverished families 60 percent of the variance in IQ was accounted for by the shared environment and the contribution of genes was close to zero, while in affluent families the result was almost exactly the reverse.

The idea behind the finding is that genetic influences on intelligence are suppressed under socioeconomic privation and more fully expressed under advantage. It is often called the Scarr-Rowe hypothesis, and replication has been uneven. Tucker-Drob and Bates performed a meta-analysis of tests of gene by status interaction on intelligence and academic achievement scores, stratified by country, with tests for publication bias and between-studies heterogeneity. In United States studies they found clear support for moderately sized gene by status effects. In studies from Western Europe and Australia, where social policies ensure more uniform access to high-quality education and health care, the effects were zero or reversed.

What this does and does not mean can be stated in three points.

• It is evidence that context shapes how much family circumstances matter. The United States pattern, and its absence in settings with more uniform access to education and health care, is consistent with the idea that the environment sets how far inherited differences can show up. The authors tie the cross-national difference to social policy, which is an interpretation and not a tested result.
• It does not describe averages. A heritability estimate describes how variation divides in a given sample and setting, and not how much of any person's score is inherited. A low genetic share in one income range says nothing about whether average scores differ between income ranges or why.
• It does not identify poverty as the cause of low scores. If shared environment accounts for more of the variance where resources are scarce, that supports taking environments seriously, and it fits a money explanation as well as many others.

The page on what heritability actually means gives the general explanation of these quantities, which is a prerequisite for reading any twin study that splits variance by status.

10 Scarcity Studies in Brief

Experiments on financial scarcity test whether concern about money interferes with performance in the moment, a short-term mechanism distinct from the long-run development of children. Mani, Mullainathan, Shafir and Zhao reported two studies. In the first, they experimentally induced thoughts about finances and found reduced cognitive performance among poor but not well-off participants. In the second, they examined farmers over the planting cycle and found diminished performance before harvest, when poor, compared with after harvest, when rich. They suggest that poverty-related concerns consume mental resources. Press coverage translated the task differences into IQ points, as in a Phys.org headline that read "translates to as many as ten IQ points", but the Science abstract describes performance on spatial and reasoning tasks, not a change in IQ.

Wicherts and Scholten reported that a reanalysis without dichotomization of income failed to corroborate those findings, and they pointed to spurious interactions between income and the experimental manipulation due to ceiling effects caused by short and easy tests. They suggest that effects of financial worries are not limited to the poor.

This page does not reproduce the full account of that exchange, which belongs to the page on what a low IQ means and the page on chronic stress and IQ. The point that bears on the present question comes from the design: these are measures of performance under a manipulated or seasonal condition. They do not show that anyone's underlying ability changed, and a score that differs between two conditions is a reminder that a single score is a snapshot.

11 What This Does Not Say About Your Own Score

Nothing in the poverty literature can tell you what your own score means, because every estimate above describes an average across many children or households and none describes one person. A study that finds an average difference of a few points between groups of families leaves a large overlap between the groups, and the range of individual scores within any circumstance is far wider than the gap between averages. The page on how rare an IQ score is and the page on IQ score versus percentile show how wide that range is on a normal curve.

A score also does not carry its history. A Full Scale IQ reports how performance on a particular set of tasks, on a particular occasion, compares with a reference group of the same age. It does not report why the person scored as they did, which educational opportunities came before the test, or what the person could do with different preparation. The page on how IQ scores are normed explains what the reference group is, and the page on reliability and validity explains how much a single score can move from measurement error alone.

Backgrounds also differ in how well they match what tests ask. Subtests that draw on vocabulary and general knowledge are generally described as depending more on exposure to schooling and reading than tasks built on novel patterns, though no format is free of experience, and the pages on whether IQ tests are biased and on culture-fair tests describe what has and has not been shown. For adults, the tasks of the six CHC domains that ACIS reports, such as fluid reasoning (Gf), crystallized knowledge (Gc), working memory (Gwm) and processing speed (Gs), are described on the page on the CHC model. The report shows a profile across the six domains as well as one overall score.

ACIS does not measure poverty, income or childhood circumstances, and nothing in an ACIS report estimates how a person's background affected a score. It is an online, unsupervised, self-administered assessment in English, and it is not a clinical or diagnostic instrument and not designed for hiring, school accommodations or admission to high IQ societies. The report gives percentiles and a 95 percent confidence interval, and the technical manual holds the documentation. Prices were read on October 6, 2026 and can change: Quick is 15 dollars, Optimized is 30 dollars and Full Scale is 50 dollars, each a one time payment, and the free trial needs no card. To choose among options, see the page on how to choose a test and read the result.

12 What the Evidence Supports, Stated Narrowly

Across designs, the evidence supports a context-dependent link between added family income and children's school achievement, leaves the effect of income on IQ itself largely unmeasured, and shows no significant effect on any preregistered outcome in the randomized preschool trial reviewed here. The table puts every study in this article side by side, with the design, the sample, the measure and the result as the sources print it. Where an abstract did not give a number, the cell says so.

Stated as narrowly as the sources allow, the evidence supports six statements.

1. Poverty and lower average IQ scores are associated in a large twin sample, and the authors of that study decline to read the association as causal.
2. Added income has been linked to higher school achievement in natural experiments (6 percent of a standard deviation per 1,000 dollars in one of them) and in a review (5 to 37 percent in experimental and quasi-experimental studies).
3. The randomized cash trial in preschool-age children found no significant effects at age 4, in a working paper that has not been peer reviewed, and its planned IQ measure was dropped, so it does not speak to IQ.
4. An adoption study found larger IQ gains in higher status adoptive homes, in a small selected sample without described randomization.
5. The role of inheritance differs with context. Twin research shows more variance tied to shared environment at low status in the United States, and not in Western European or Australian samples.
6. Experiments on scarcity show that concern about money can affect test performance in the moment, though the original result has been contested.

Several things are not supported by any study in this article. No design here shows that poverty lowers the IQ of a person. No design shows that a particular amount of money moves a child's IQ by a particular number of points. No study here shows that a null result at age 4 rules out effects at older ages. And none of the results concerns the direction from IQ to income, which the page on IQ and income covers, or the related questions of how IQ relates to education and how it relates to success.

The framework for reading any score responsibly is the Standards for Educational and Psychological Testing (AERA, APA and NCME, 2014), which is used to judge whether the evidence supports a proposed interpretation and use of a score. The APA Ethical Principles of Psychologists and Code of Conduct, Standard 9.06, asks psychologists who interpret assessment results to take into account the purpose of the assessment and characteristics of the person assessed, including situational, personal, linguistic and cultural differences, and to indicate any significant limitations of their interpretations. A score shaped by years of unequal opportunity and a score shaped by different circumstances can look the same, and a responsible reading states what the number cannot show. Citing these documents does not imply that their authors endorse ACIS or any other product.

13 Sources Behind This Page

These sources separate randomized trials, natural experiments, reviews, adoption, twin and scarcity studies, so that a figure from one design is not read as a result of another. Effect sizes, sample sizes and p values are given as each source prints them. The age 4 trial paper is an NBER working paper that its cover note says is not peer reviewed. Conversions to IQ points are labeled as our arithmetic and are illustrative only. Sources were opened or checked through Crossref and PubMed on October 6, 2026.

• Troller-Renfree S V, Costanzo M A, Duncan G J, Magnuson K, Gennetian L A, Yoshikawa H, Halpern-Meekin S, Fox N A and Noble K G. The impact of a poverty reduction intervention on infant brain activity. Proceedings of the National Academy of Sciences, 2022, volume 119, issue 5, article e2115649119.
• Noble K, Magnuson K, Duncan G, Gennetian L A, Yoshikawa H, Fox N A, Halpern-Meekin S, Troller-Renfree S and colleagues. The Effect of a Monthly Unconditional Cash Transfer on Children's Development at Four Years of Age: A Randomized Controlled Trial in the U.S. NBER Working Paper 33844, National Bureau of Economic Research, May 2025.
• Troller-Renfree S and colleagues. The impact of a monthly unconditional cash transfer on child brain activity: A 4-year follow-up. Developmental Cognitive Neuroscience, 2026, volume 78, article 101673.
• Dahl G and Lochner L. The Impact of Family Income on Child Achievement: Evidence from the Earned Income Tax Credit. American Economic Review, 2012, volume 102, issue 5, pages 1927 to 1956.
• Milligan K and Stabile M. Do Child Tax Benefits Affect the Well-being of Children? Evidence from Canadian Child Benefit Expansions. American Economic Journal: Economic Policy, 2011, volume 3, issue 3, pages 175 to 205.
• Akee R, Copeland W, Keeler G, Angold A and Costello E. Parents' Incomes and Children's Outcomes: A Quasi-Experiment Using Transfer Payments from Casino Profits. American Economic Journal: Applied Economics, 2010, volume 2, issue 1, pages 86 to 115.
• Cooper K and Stewart K. Does Household Income Affect Children's Outcomes? A Systematic Review of the Evidence. Child Indicators Research, 2021, volume 14, issue 3, pages 981 to 1005.
• Duncan G, Magnuson K and Votruba-Drzal E. Moving Beyond Correlations in Assessing the Consequences of Poverty. Annual Review of Psychology, 2017, volume 68, pages 413 to 434.
• Fernald L, Gertler P and Neufeld L. Role of cash in conditional cash transfer programmes for child health, growth, and development: an analysis of Mexico's Oportunidades. The Lancet, 2008, volume 371, issue 9615, pages 828 to 837.
• Duyme M, Dumaret A and Tomkiewicz S. How can we boost IQs of "dull children"? A late adoption study. Proceedings of the National Academy of Sciences, 1999, volume 96, issue 15, pages 8790 to 8794.
• Turkheimer E, Haley A, Waldron M, D'Onofrio B and Gottesman I. Socioeconomic Status Modifies Heritability of IQ in Young Children. Psychological Science, 2003, volume 14, issue 6, pages 623 to 628.
• Tucker-Drob E and Bates T. Large Cross-National Differences in Gene by Socioeconomic Status Interaction on Intelligence. Psychological Science, 2016, volume 27, issue 2, pages 138 to 149.
• von Stumm S and Plomin R. Socioeconomic status and the growth of intelligence from infancy through adolescence. Intelligence, 2015, volume 48, pages 30 to 36.
• Mani A, Mullainathan S, Shafir E and Zhao J. Poverty Impedes Cognitive Function. Science, 2013, volume 341, issue 6149, pages 976 to 980.
• Phys.org. Researcher finds that poverty's 'cognitive cost' translates to as many as ten IQ points. Phys.org, August 2013 (news article, cited only for its headline).
• Wicherts J and Scholten A. Comment on "Poverty Impedes Cognitive Function". Science, 2013, volume 342, issue 6163, page 1169.

The interpretive framework is the Standards for Educational and Psychological Testing (AERA, APA and NCME, 2014) and the APA Ethical Principles of Psychologists and Code of Conduct, effective January 1, 2017, Standard 9.06.

14 Frequently Asked Questions