View in EDS HTML Full Text PDF Full Text

What Do Changes in State NAEP Scores Imply for Birth Cohorts' Later Life Outcomes?

Saved in:
Bibliographic Details
Title: What Do Changes in State NAEP Scores Imply for Birth Cohorts' Later Life Outcomes?
Language: English
Authors: Elena Doty, Thomas J. Kane (ORCID 0009-0008-1297-5426), Tyler Patterson, Douglas O. Staiger
Source: Journal of Policy Analysis and Management. 2026 45(1).
Availability: Wiley. Available from: John Wiley & Sons, Inc. 111 River Street, Hoboken, NJ 07030. Tel: 800-835-6770; e-mail: cs-journals@wiley.com; Web site: https://www.wiley.com/en-us
Peer Reviewed: Y
Page Count: 15
Publication Date: 2026
Document Type: Journal Articles
Reports - Research
Descriptors: Educational Change, National Competency Tests, Scores, Predictor Variables, Outcomes of Education, Mathematics Achievement, Income, Educational Attainment, Early Parenthood, Institutionalized Persons, Crime, Cohort Analysis
Assessment and Survey Identifiers: National Assessment of Educational Progress
DOI: 10.1002/pam.70018
ISSN: 0276-8739
1520-6688
Abstract: Since 1990, the National Assessment of Educational Progress (NAEP) has been the primary benchmark for tracking the progress of state education reform. The focus on math and reading achievement is motivated by the cross-sectional relationship between test scores and adult outcomes, such as earnings and college completion. But do changes in NAEP scores predict changes in long-term economic and social outcomes for future earners--or do they reflect other factors unrelated to earnings such as teaching to the test? We investigate by linking long-term outcomes by year and state of birth to NAEP scores. We find that more recent birth cohorts in states with large increases in NAEP math achievement enjoyed higher incomes, improved educational attainment, and declines in teen motherhood, incarceration, and arrest rates compared to those in states with smaller increases. In fact, the relationship between changes in NAEP achievement and cohort earnings is about two thirds the size of the cross-sectional relationship observed in prior research: a 6% to 8% rise in earnings per standard deviation rise in 8th grade math. The results are not sensitive to controls for student demographics, labor market conditions, or measures of children's health (such as low birthweight).
Abstractor: As Provided
Notes: https://doi.org/10.7910/DVN/QOR3A3
Entry Date: 2026
Accession Number: EJ1493991
Database: ERIC
Full text is not displayed to guests.
FullText Links:
  – Type: pdflink
    Url: https://content.ebscohost.com/cds/retrieve?content=AQICAHj0k_4E0hTGH8RJwT4gCJyBsGNe_WN95AvKlDbXJGqwxwHUGXaIGVa4AAHd6nXfLdHYAAAA4jCB3wYJKoZIhvcNAQcGoIHRMIHOAgEAMIHIBgkqhkiG9w0BBwEwHgYJYIZIAWUDBAEuMBEEDBLMy4qzOCkf7D3_fgIBEICBmtNyY7OD9GlhnPYw2L2j2PggqJAUd5VthA5gKezQ4IF4Fu8GT2tL7wrKjRVl49z0HvHjKj8nVw_t9sz_4oGwmxkjH4XqZPiujRAY0fEKnxMr16zezPsfMKa7ibRIGvsGmhtXuBbz5-OSIqkvB6uqvH6Tx1IdhQ3c7uo7K21YdblOH9iE7q414lmRycrop9GeZS5pnEKzMuqOgBE=
Text:
  Availability: 1
  Value: <anid>AN0190550171;jpa01jan.26;2025Dec31.04:49;v2.2.500</anid> <title id="AN0190550171-1">What do changes in state NAEP scores imply for birth cohorts' later life outcomes? </title> <sbt id="AN0190550171-2">INTRODUCTION1Between 1989 and 2024, at least 47 articles published in this journal cited data...</sbt> <p>Since 1990, the National Assessment of Educational Progress (NAEP) has been the primary benchmark for tracking the progress of state education reform. The focus on math and reading achievement is motivated by the cross‐sectional relationship between test scores and adult outcomes, such as earnings and college completion. But do changes in NAEP scores predict changes in long‐term economic and social outcomes for future earners—or do they reflect other factors unrelated to earnings such as teaching to the test? We investigate by linking long‐term outcomes by year and state of birth to NAEP scores. We find that more recent birth cohorts in states with large increases in NAEP math achievement enjoyed higher incomes, improved educational attainment, and declines in teen motherhood, incarceration, and arrest rates compared to those in states with smaller increases. In fact, the relationship between changes in NAEP achievement and cohort earnings is about two thirds the size of the cross‐sectional relationship observed in prior research: a 6% to 8% rise in earnings per standard deviation rise in 8th grade math. The results are not sensitive to controls for student demographics, labor market conditions, or measures of children's health (such as low birthweight).</p> <p>Coming out of the COVID‐19 pandemic, the loss in mean achievement on the 2022 National Assessment of Educational Progress (NAEP) captured headlines. Yet even before the pandemic, the biennial release of NAEP scores had become a major news event. Students in some states, such as North Carolina, Texas, and Florida, made large improvements in math and reading achievement between 1990 and 2015, while students in other states, such as Iowa, Oregon, and Oklahoma, saw much smaller improvements. The interest in NAEP is motivated by the positive cross‐sectional relationship between test scores and young adult outcomes, such as earnings. But do changes in NAEP scores provide a signal of students' future social and economic prospects—or do they reflect factors less valuable to career earnings, such as students' test‐taking skills or a narrowed curriculum?</p> <p>It is a timely question, given that mean 8th‐grade math achievement declined by .2 standard deviations between 2019 and 2022, erasing 40% of the progress during the prior 3 decades. Should policymakers and parents be concerned? To gain some insight into the long‐term consequences for students, we investigate the changes in later life outcomes which accompanied past changes in achievement on the NAEP tests.</p> <p>To do so, we exploit state‐level variation in achievement gains since 1990, when the NAEP first began measuring state differences in achievement. In a subset of states, NAEP scores rose sharply between 1990 and 2019. For instance, the average student in North Carolina improved by .92 standard deviations in 8th‐grade math achievement—an improvement roughly equal to the national Black–White gap in 1990 (.91 standard deviations). Although every state saw an improvement in 8th‐grade math, the gains in other states, such as Iowa, were substantially smaller (.10 standard deviations). We use these within‐state differences in the magnitude of NAEP improvements along with outcomes by year and state of birth in the American Community Survey as well as annual data on arrest rates to estimate the association between state NAEP scores and later‐life outcomes.</p> <p>We make three primary contributions: First, we measure the predictive value of state NAEP scores, a commonly used barometer for educational progress in the U.S. Despite prior research on the relationship between achievement and earnings, improvement in NAEP scores over time could reflect factors other than students' understanding of math or reading (such as test taking skills), which are not as valuable. Moreover, if schools have narrowed the curriculum to focus on math and reading, students with higher scores in those subjects could bring with them less knowledge of social science, history, and the humanities, which could lead to a diminished labor market value. Rather than adding to the cross‐sectional literature, we investigate whether past improvements in NAEP have forecasted improvements in long‐term outcomes such as earnings.</p> <p>Second, the prior literature on the relationship between achievement and earnings has relied on plausibly exogenous—but small—differences in achievement associated with one's kindergarten classroom (Chetty et al., [<reflink idref="bib4" id="ref1">4</reflink>]) or middle school teacher (Chetty et al., [<reflink idref="bib5" id="ref2">5</reflink>]) or state accountability law changes (Deming et al., [<reflink idref="bib8" id="ref3">8</reflink>]). Even if the relationship between earnings and achievement is locally linear, the linear relationship may not apply to the larger changes in achievement measured in the NAEP over time. The long‐term changes in NAEP and earnings we use provide a test of the impact of large increases in achievement.</p> <p>Third, we address a longstanding puzzle in the NAEP data, in which the sizeable progress in 4th‐ and 8th‐grade math achievement has been followed by declines in 12th‐grade reading scores and a modest improvement in 12th‐grade math (Blagg & Chingos, [<reflink idref="bib3" id="ref4">3</reflink>]). The stagnation in 12th‐grade scores is typically cited as evidence of lack of progress in U.S. K–12 education. But did the large increase in 4th‐ and 8th‐grade achievement merely represent a shift in the timing of skill acquisition with little increase in the eventual stock of student knowledge as of 12th grade? Harris and colleagues ([<reflink idref="bib16" id="ref5">16</reflink>]) argued that the trend in 12th‐grade scores is biased downward due to rising high school graduation. However, even if selection cannot explain all of the divergence, it is possible that learning the same math and reading skills earlier could allow students to learn more in other subjects in high school, such as science or higher‐level math.</p> <p>Thus, as others have done when studying test score fade‐out in earlier grades (Chetty et al., [<reflink idref="bib4" id="ref6">4</reflink>]; Deming, [<reflink idref="bib7" id="ref7">7</reflink>]), we investigate changes in long‐term outcomes by birth cohort within state. We use the student‐level microdata to adjust the NAEP scores for parental education and race/ethnicity. We also include fixed effects for state of birth, year of birth, Census division by year of birth, and adult state of residence by year. We also control for state unemployment rate at age 18, state median household income at age 13, and percent low‐birthweight for state birth cohorts—although such controls have little impact on our results. We find that a standard deviation rise in 8th‐grade math achievement is associated with an 8% rise in adult's earned income—roughly two thirds of the cross‐sectional relationship between achievement scores and young adult earnings. A standard deviation rise in 8th‐grade math achievement on the NAEP is also associated with improvements in educational attainment and declines in unemployment, teen motherhood, incarcerations, and arrests. When combined with prior research on earnings and achievement, our findings suggest that NAEP scores do provide valuable information about students' long‐term outcomes.</p> <hd id="AN0190550171-3">RELEVANT LITERATURE</hd> <p>There is a large literature in labor economics measuring the relationship between adult earnings and pre‐labor market achievement test scores. For instance, Neal and Johnson ([<reflink idref="bib25" id="ref8">25</reflink>]) found that a standard deviation in AFQT scores was associated with a 20% difference in earnings at age 26 to 29 for both men and women. Murnane et al. ([<reflink idref="bib20" id="ref9">20</reflink>]) found that a standard deviation in 10th‐grade math scores was associated with a 12% difference in earnings at age 31. Chetty et al. ([<reflink idref="bib4" id="ref10">4</reflink>]) found that a standard deviation in scores as measured in kindergarten are associated with an 18% difference in earnings in early adulthood. More generally, test scores have been linked to a wide range of later life outcomes for individuals, including educational attainment, teen childbearing, and illegal behaviors (Goldhaber & Özek, [<reflink idref="bib14" id="ref11">14</reflink>]).</p> <p>Much of the research on the relationship between test scores and earnings has been correlational in nature. An obvious concern is that a student's score on a test of math or reading could be correlated with pre‐existing ability (such as fluid intelligence) or family background, both of which could have their own direct relationship with long‐term outcomes. Yet several studies have shown that interventions which have causal impacts on contemporaneous measures of math or reading achievement also have impacts on long‐term outcomes. For example, using long‐term outcomes for participants in the Tennessee classroom size experiment, Chetty et al. ([<reflink idref="bib4" id="ref12">4</reflink>]) found that the quality of one's randomly assigned kindergarten class has effects on both end of year achievement and earnings at age 25 to 27. Using the covariance in classroom effects on test scores and earnings, they concluded that class‐level factors (such as class size, teacher experience, and other unmeasured peer and teacher quality), which boost test scores by 1 percentile point (approximately .047 standard deviations), also raise adult annual earnings by $76.48 on average (.16% of average earnings in their sample).</p> <p>While Chetty et al. ([<reflink idref="bib4" id="ref13">4</reflink>]) used randomly assigned classrooms, Chetty et al. ([<reflink idref="bib5" id="ref14">5</reflink>]) used non‐experimental value‐added methods to measure impacts of teachers on achievement and long‐term outcomes. They found that a 1 standard deviation difference in teacher quality (as measured by contemporaneous impacts on achievement) is associated with a 1.34% increase in lifetime earnings. Given that a standard deviation in teacher quality was equivalent to .13 standard deviations in test scores, the ratio would imply a 10% rise in earnings per standard deviation of test scores.</p> <p>Deming et al. ([<reflink idref="bib8" id="ref15">8</reflink>]) estimated the effect of school accountability laws in Texas on math achievement, finding that the introduction of accountability boosted math achievement for low‐achieving students by .2 standard deviations and raised earnings at age 25 by 2.1%. If the impact of school accountability operated primarily through boosting academic achievement, the implied instrumental variable estimate would be similar to the cross‐sectional estimates, about 11% (2.1%/.2).</p> <p>There is a related literature on the relationship between international differences in economic growth and country‐level differences in educational attainment/ cognitive test scores (Barro, [<reflink idref="bib1" id="ref16">1</reflink>]; Hanushek & Woessman, 2008). One obvious challenge is the potential endogeneity of growth with educational attainment and achievement (Bils & Klenow, [<reflink idref="bib2" id="ref17">2</reflink>]). We face a similar challenge in studying the relationship between achievement growth in a state and a birth cohort's income and educational attainment. Thus, we use a number of statistical controls to address endogeneity: relying on changes in achievement within state of birth, controlling for labor market conditions by including fixed effects by state of residence by year and Census division by year of birth and controlling for race and parental education.</p> <p>As discussed by Blagg and Chingos ([<reflink idref="bib3" id="ref18">3</reflink>]), the sizeable progress in 4th‐ and 8th‐grade math achievement has not been matched with rising achievement on the 12th‐grade NAEP. Between 1990 and the peak in 2013, the average math achievement of 4th‐ and 8th‐grade students in the U.S. had improved by .9 and .6 standard deviations, respectively. Over the same time period, 12th‐grade NAEP achievement declined in reading and improved by only .15 standard deviations in math. However, Harris et al. ([<reflink idref="bib16" id="ref19">16</reflink>]) argued that such results for high school students are misleading, given that high school graduation rates have been rising and more would‐have‐been drop‐outs are reaching 12th grade. The authors used student scores on 8th‐grade math and reading achievement to impute 12th‐grade scores for students who dropped out before 12th grade in Louisiana. Using this method, they estimated that at least half of the measured decline in 12th grade NAEP score nationally between 1992 and 2012 can be explained by decreasing school drop‐out rates.</p> <p>Yet, even if the apparent contradiction of rising 4th and 8th grade scores with stagnant or declining 12th grade achievement were not completely explained by the combination of rising high school graduation rates and selection bias, students may have benefited from earlier acquisition of skills. A number of studies (including Chetty et al., [<reflink idref="bib4" id="ref20">4</reflink>], and Deming, [<reflink idref="bib7" id="ref21">7</reflink>]) have found that interventions which impact achievement in early grades (such as class size or Head Start) have long‐term benefits, even when test score impacts fade out in later grades. Such studies have typically concluded that the long term impacts operate through "non‐cognitive" skills. But, as Neal ([<reflink idref="bib24" id="ref22">24</reflink>]) argued, math and reading skills are inputs in the production of other skills, both cognitive and non‐cognitive. For example, students with better verbal skills may be better able to resolve conflicts with other students; students with better math skills may make valuable friendships in the robotics club. And improved math and reading knowledge could lead to improvements in other cognitive domains, such as if students learned more science and history and social science skills during middle school and high school as a result of their earlier development of math and literacy skills.</p> <p>Thus, rather than address the question of selection bias in 12th‐grade NAEP scores, we ask whether there was any improvement in long‐term outcomes for those born in states with large increases in 8th‐grade achievement. Should policymakers be ignoring the improvements in NAEP scores in earlier grades if 12th‐grade scores are not rising? Does the rise in 4th‐ and 8th‐grade achievement combined with flat 12th‐grade achievement simply mean that the timing of skill acquisition has shifted earlier, even though eventual mastery has not improved? The question is still relevant, regardless of the extent to which improvement in 12th grade has been understated by selection bias.</p> <hd id="AN0190550171-4">DATA</hd> <p>To measure state differences in student achievement over time, we use the Main NAEP. Since 1990, the Main NAEP has collected data from a representative sample of more than 100 public schools in each participating state, with a total of roughly 4,800 schools and 125,000 students annually. Prior to 2002, states could opt out of participating, and the number of participating states ranged from 38 to 46, depending upon the subject and grade. Since 2003, all states have participated and the NAEP has been administered every other year, in both reading and mathematics.</p> <p>For this analysis, we focus on the 8th‐grade mathematics assessment, which was administered in 1990, 1992, 1996, 2000, and biennially from 2003 to 2019. Due to the pandemic, the 2021 assessments were postponed and the tests were administered between January and March 2022. We focus on math achievement because prior research has found it to be more correlated with earnings than reading or vocabulary (Murnane et al., [<reflink idref="bib19" id="ref23">19</reflink>]). Unfortunately, 8th‐grade reading scores were not available by state before 1998, meaning we would lose the birth cohorts entering 8th grade between 1990 and 1998 for whom math scores are available. Moreover, since 4th‐grade scores were not available until 1992 (the birth cohort of 1983) in math or reading, the 8th‐grade math scores allow us to include six additional birth cohorts (helpful when measuring long‐term outcomes, such as earnings). Although we focus on 8th‐grade math achievement, we provide results for 8th‐grade reading scores in the appendix.<sups>,</sups></p> <p>Specifically, we use the mean score of 8th graders in a state as an estimate of the mean achievement of those born in the state 13 years before. We impute scores for missing years using the average of scores from years t−1 and t+1 if they are available. (If there is more than a 1‐year gap, we treat achievement as missing for the intervening years.) We standardize scores in all years using the national mean and standard deviation from 1992.</p> <p>In Figure 1, the height of the bars portrays the change in mean 8th‐grade math scores by state between 1990 and 2019. Every state's mean score in 2019 exceeded its 1990 mean score, although the magnitude of the improvement varied—from .10 s.d. in Montana to .92 s.d. in North Carolina. The black arrows portray the change in scores between 2019 and 2022—so that the difference between the height of the bar and the corresponding black arrow is the net decline in 8th‐grade math scores that occurred during the pandemic. For example, the 2019 to 2022 losses were nearly as large as the improvement over the prior 3 decades in Oklahoma, New Mexico, and Missouri. Students in five states (Oregon, North Dakota, Maine, Iowa, and Montana) actually scored below their 1990 counterparts in 2022. The negative impacts on achievement were not merely transitory. As of 2024, U.S. students had barely begun to recover, with 8th‐grade math achievement falling by an additional .025 standard deviations among public schools nationally.</p> <p> <img src="https://imageserver.ebscohost.com/img/embimages/rdk/JPA/01jan26/pam70018-fig-0001.jpg?ephost1=dGJyMNHX8kSepq84v%2bvlOLCmsE6epq5Srqa4SK6WxWXS" alt="pam70018-fig-0001.jpg" title="1 Changes in 8th‐grade math achievement before and after pandemic by state.Notes: Based on mean 8th‐grade math achievement in the NAEP. For the seven states that had a valid score for 1992, but not 1990 (Massachusetts, Maine, Missouri, Mississippi, South Carolina, Tennessee, and Utah), we report the change since 1992. Six states had neither a 1990 nor a 1992 NAEP score (Alaska, Kansas, Nevada, South Dakota, Vermont, and Washington) and are excluded from the graph. Mean achievement had increased in all states over the 29 years preceding the pandemic. Every state lost ground between 2019 and 2022, with five states scoring below their 1990 (or 1992) means. On average, the U.S. forfeited 41% of the pre‐pandemic improvement between 2019 and 2022." /> </p> <p></p> <p>We interpret changes in NAEP scores as measuring changes in student knowledge across birth cohorts. Of course, improvements in achievement could be accompanied by other factors which could have their own direct effect on earnings—such as demographic changes (i.e., improved parental education or changes in race/ethnicity) or improving household income or child health. When feasible, we attempt to control for such factors directly, either using the student‐level measures of household characteristics to adjust NAEP scores or measures of state unemployment when each cohort turned 18, median household income and percent of births that were low birth weight in the year the cohort was born. We also include fixed effects by state and birth cohort and by Census division by birth cohort to control for other unmeasured factors. We also control for current labor market conditions using fixed effects by state of residence and year. We assume that the remaining variation in within‐state changes in achievement is due to improved school quality.</p> <p>Accordingly, we use the student‐level microdata from the NAEP to adjust state achievement measures for changes in race/ethnicity and parental education of tested students. To do so, we estimate the following model separately for each NAEP cohort: 1 <ephtml> <math display="block" altimg="urn:x-wiley:02768739:media:pam70018:pam70018-math-0001" xmlns="http://www.w3.org/1998/Math/MathML"><semantics><mrow><msub><mi>S</mi><mi>ijc</mi></msub><mo linebreak="badbreak">=</mo><msub><mi>β</mi><mn>0</mn></msub><mo linebreak="goodbreak">+</mo><msub><mi>Race</mi><mi>i</mi></msub><msub><mi>β</mi><mn>1</mn></msub><mo linebreak="goodbreak">+</mo><msub><mi>ParentEd</mi><mi>i</mi></msub><msub><mi>β</mi><mn>2</mn></msub><mo linebreak="goodbreak">+</mo><msub><mi>δ</mi><mrow><mi>j</mi><mi>c</mi></mrow></msub><mo linebreak="goodbreak">+</mo><msub><mi>ε</mi><mi>ijc</mi></msub></mrow><annotation encoding="application/x-tex">$$\begin{equation} {{S}_{\textit{ijc}}} = {{\beta }_0} + \textit{Race}_{i}{{\beta }_1} + \textit{ParentEd}_{i}{{\beta }_2} + {{\delta }_{jc}} + {{\varepsilon }_{\textit{ijc}}} \end{equation}$$</annotation></semantics></math> </ephtml> where <ephtml> <math display="inline" altimg="urn:x-wiley:02768739:media:pam70018:pam70018-math-0002" xmlns="http://www.w3.org/1998/Math/MathML"><semantics><msub><mi>Race</mi><mi>i</mi></msub><annotation encoding="application/x-tex">$\textit{Race}_{i}$</annotation></semantics></math> </ephtml> is a vector of indicators of student race/ethnicity and <ephtml> <math display="inline" altimg="urn:x-wiley:02768739:media:pam70018:pam70018-math-0003" xmlns="http://www.w3.org/1998/Math/MathML"><semantics><msub><mi>ParentEd</mi><mi>i</mi></msub><annotation encoding="application/x-tex">$\textit{ParentEd}_{i}$</annotation></semantics></math> </ephtml> is a vector of indicators of student‐reported parental education. We refer to the state by cohort effects from equation (<reflink idref="bib1" id="ref24">1</reflink>), <ephtml> <math display="inline" altimg="urn:x-wiley:02768739:media:pam70018:pam70018-math-0004" xmlns="http://www.w3.org/1998/Math/MathML"><semantics><msub><mover accent="true"><mi>δ</mi><mo>̂</mo></mover><mrow><mi>j</mi><mi>c</mi></mrow></msub><annotation encoding="application/x-tex">${{\hat{\delta }}_{jc}}$</annotation></semantics></math> </ephtml> , as "adjusted math achievement."</p> <p>To measure later life outcomes, we use the 1% sample from the 2000 Decennial Census and the American Community Survey (ACS) from 2001 through 2019, obtained via IPUMS (Ruggles et al., [<reflink idref="bib27" id="ref25">27</reflink>]). In addition to providing data on earned income, occupation, educational attainment, college enrollment, teen motherhood, Food Stamp (SNAP) receipt, home ownership, and incarceration, the Census and ACS provide information on year and state of birth.</p> <p>We use state of birth as a proxy for the state where the student was residing in grade 8. As a measure of achievement, we assign each birth cohort the state mean 8th grade achievement from the year they were 13 years old. In the 2000 Census, 80% of U.S.‐born 13‐year‐olds were living in their state of birth. Because the District of Columbia was an outlier (21% of those born in DC were residing in the district at age 13), we drop those born in DC. Prior work using the ACS to link later life outcomes to state policies has found that using state of birth introduces little bias (Lovenheim & Willén, [<reflink idref="bib17" id="ref26">17</reflink>]). We also drop observations with Census‐allocated values for race, age, birthplace, and sex.</p> <p>In addition to the Census and American Community Survey, we use estimates of violent and property crime arrests by state of occurrence, year, and age from the the FBI Uniform Crime Report Arrest Master File (Donohue, [<reflink idref="bib9" id="ref27">9</reflink>]; Donohue & Levitt, [<reflink idref="bib10" id="ref28">10</reflink>]).</p> <p>We also use occupation‐level measures of the level of mathematics required for a job from the U.S. Department of Labor's Occupational Information Network survey (O*NET), a provider of occupational information.</p> <p>For covariates in our analyses of later‐life outcomes, we use several additional data sources. We calculate the percentage of parents with at least a 2‐year college degree by state and year using the 1990 and 2000 Census and the 2009 ACS via IPUMS. We calculated median household income for the years in which each cohort was 12 to 14 years old by state for families with children using the CPS Annual Social and Economic Supplement via IPUMS (Flood et al., [<reflink idref="bib13" id="ref29">13</reflink>]). We use state‐year level unemployment rates from BLS Local Area Unemployment Statistics. Finally, we calculate the percentage of low birth weight births by state and year using data from the National Vital Statistics System of the National Center for Health Statistics, obtained from the NBER.</p> <hd id="AN0190550171-6">CHANGES IN ACHIEVEMENT AND LATER LIFE OUTCOMES BY STATE OF BIRTH</hd> <p>In Figure 2, we report scatterplots of changes in outcomes against changes in math achievement by state of birth. For each outcome, we compare the earliest two NAEP cohorts (those born between 1977 and 1979) against the latest two NAEP cohorts for which we observe at least 2 years of the outcome measure in our data. For instance, in the top left panel, we report the change in mean log earned income (which we measure for those aged 28 and above) against the change in achievement between those born in 1977 to 1979 (age 13 in 1990 to 1992) and those born in 1987 to 1990 (age 13 in 2000 to 2003); for the 1990 cohort, we observe 2 years of income data in IPUMS in which the respondents are at least 28 (2018 and 2019). The bivariate slope coefficient on the long difference is 0.078, only slightly smaller than the cross‐sectional estimates summarized above.</p> <p> <img src="https://imageserver.ebscohost.com/img/embimages/rdk/JPA/01jan26/pam70018-fig-0002.jpg?ephost1=dGJyMNHX8kSepq84v%2bvlOLCmsE6epq5Srqa4SK6WxWXS" alt="pam70018-fig-0002.jpg" title="2 Within‐state changes in birth cohort outcomes and corresponding changes in 8th‐grade math achievement.Notes: We use the longest span in birth year possible for each of the outcomes, averaged across two NAEP cohorts and conditional on being able to observe at least 2 years of the outcome measure. For each outcome measure, the earliest birth cohorts, the 8th graders in 1990 and 1992, were born in 1977 to 1979. For instance, we measure log earnings for those aged 28 and above. The x‐axis of the chart on the top left shows each state's change in average NAEP score between 1990 to 1992 and 2000 to 2003; the y‐axis shows each state's change in average log earnings between those born in 1977 to 1979 (who would have been 8th graders in 1990 to 1992) and those born in 1987 to 1990 (who would have been 8th graders in 2000 to 2003, and who have at least 2 years of being 28+ in our IPUMS data ending in 2019). The cohorts used for high school completion (which we measure at ages 23 to 42) were born between 1977 and 1979 or between 1992 and 1994; college enrollment (which we measure between ages 19 and 24) was measured for the 1977 to 1979 and 1996 to 1998 birth cohorts; teen motherhood (which we measure between ages of 18 to 30) includes the 1977 to 1979 and 1998 to 2000 birth cohorts; institutionalization (which we measure between ages 18 and 24) includes the 1977 to 1979 and 1998 to 2000 birth cohorts; violent crime arrests (which we measure for those aged 15 to 24, but for which our data ends in 2014) includes the 1977 to 1979 and 1996 to 1998 birth cohorts." /> </p> <p></p> <p>In the remaining panels of Figure 2, we report similar graphs for five additional outcomes: the change in high school graduation rates (for those aged 23 and above), the change in college enrollment (for those aged 18 to 24), the change in teen motherhood (for females aged 18 to 30), the change in rates of institutionalization (for males between the ages of 18 and 24), and the change in arrest rates for violent crimes (for males aged 15 to 24). The states with the largest improvements in achievement also saw larger improvements in high school graduation and college enrollment, as well as bigger decreases in teen motherhood, rates of institutionalization, and arrests for violent crimes.</p> <hd id="AN0190550171-8">INCOME SPECIFICATIONS</hd> <p>We investigate the robustness of these bivariate relationships by introducing a variety of statistical controls. We start by estimating the following linear model of log of earned income <ephtml> <math display="inline" altimg="urn:x-wiley:02768739:media:pam70018:pam70018-math-0005" xmlns="http://www.w3.org/1998/Math/MathML"><semantics><mrow><mi>L</mi><mi>n</mi><mspace width="0.33em" /><msub><mi>Y</mi><mi>ijct</mi></msub></mrow><annotation encoding="application/x-tex">$Ln\ {{Y}_{\textit{ijct}}}$</annotation></semantics></math> </ephtml> : 2 <ephtml> <math display="block" altimg="urn:x-wiley:02768739:media:pam70018:pam70018-math-0006" xmlns="http://www.w3.org/1998/Math/MathML"><semantics><mrow><mi>L</mi><mi>n</mi><mspace width="0.33em" /><msub><mi>Y</mi><mi>ijct</mi></msub><mo linebreak="badbreak">=</mo><msub><mi>γ</mi><mi>o</mi></msub><mo linebreak="goodbreak">+</mo><msub><mi>γ</mi><mn>1</mn></msub><msub><mover><mi>Score</mi><mo>¯</mo></mover><mrow><mi>j</mi><mi>c</mi></mrow></msub><mo linebreak="goodbreak">+</mo><msub><mi>γ</mi><mn>2</mn></msub><msub><mi>Gender</mi><mi>i</mi></msub><mo linebreak="goodbreak">+</mo><msub><mi>Race</mi><mi>i</mi></msub><msub><mi>γ</mi><mn>3</mn></msub><mo linebreak="goodbreak">+</mo><msub><mi>θ</mi><mi>j</mi></msub><mo linebreak="goodbreak">+</mo><msub><mi>μ</mi><mi>c</mi></msub><mo linebreak="goodbreak">+</mo><msub><mi>π</mi><mi>age</mi></msub><mo linebreak="goodbreak">+</mo><msub><mi>τ</mi><mi>t</mi></msub><mo linebreak="goodbreak">+</mo><msub><mi>ε</mi><mi>ijt</mi></msub></mrow><annotation encoding="application/x-tex">$$\begin{equation} Ln\ {Y}_{\textit{ijct}} = {\gamma }_{o} + {\gamma }_{1}{\overline{\textit{Score}}}_{jc} + {\gamma }_{2}\textit{Gender}_{i} + \textit{Race}_{i}{\gamma }_{3} + {\theta }_{j} + {\mu }_{c} + {\pi }_{\textit{age}} + {\tau }_{t} + {\varepsilon }_{\textit{ijt}} \end{equation}$$</annotation></semantics></math> </ephtml> Where i subscripts the individual, j the state of birth, c the year of birth and t the year of the outcome is measured. <ephtml> <math display="inline" altimg="urn:x-wiley:02768739:media:pam70018:pam70018-math-0007" xmlns="http://www.w3.org/1998/Math/MathML"><semantics><msub><mover><mi>Score</mi><mo>¯</mo></mover><mrow><mi>j</mi><mi>c</mi></mrow></msub><annotation encoding="application/x-tex">${\overline{\textit{Score}}}_{jc}$</annotation></semantics></math> </ephtml> is the mean score on the 8th grade math assessment. Equation (<reflink idref="bib2" id="ref30">2</reflink>) includes controls for gender and race, as well as fixed effects for state of birth, birth cohort, year, and year of age. Standard errors are calculated while clustering at the state of birth by cohort level. Observations with Census‐allocated values of earned income are dropped. Table 1 reports results from a number of regressions using the log of earned income as a dependent variable. In column 1, estimated based on equation (<reflink idref="bib2" id="ref31">2</reflink>), the coefficient on mean 8th‐grade math score is .112.</p> <p>1 TABLE The relationship between log income and math achievement by state birth cohort.</p> <p> <ephtml> <table><thead><tr><th /><th>(1)</th><th>(2)</th><th>(3)</th><th>(4)</th><th>(5)</th><th>(6)</th><th>(7)</th><th>(8)</th></tr></thead><tbody><tr><td>Math achievement</td><td>0.112<ext-link /><sup>***</sup></td><td /><td /><td /><td /><td /><td /><td /></tr><tr><td /><td>(0.023)</td><td /><td /><td /><td /><td /><td /><td /></tr><tr><td>Adjusted math achievement</td><td /><td>0.096<ext-link /><sup>***</sup></td><td>0.073<ext-link /><sup>**</sup></td><td>0.062<sup>*</sup></td><td>0.080<ext-link /><sup>**</sup></td><td>0.068<sup>*</sup></td><td>0.087</td><td>0.078<ext-link /><sup>**</sup></td></tr><tr><td /><td /><td>(0.025)</td><td>(0.024)</td><td>(0.024)</td><td>(0.029)</td><td>(0.032)</td><td>(0.057)</td><td>(0.029)</td></tr><tr><td>Either parent college graduate</td><td /><td /><td>0.180<ext-link /><sup>***</sup></td><td>0.185<ext-link /><sup>***</sup></td><td>0.104</td><td>0.111<sup>*</sup></td><td>0.039</td><td>0.103</td></tr><tr><td /><td /><td /><td>(0.038)</td><td>(0.039)</td><td>(0.058)</td><td>(0.056)</td><td>(0.061)</td><td>(0.058)</td></tr><tr><td>Fixed effects:</td><td /><td /><td /><td /><td /><td /><td /><td /></tr><tr><td>Gender</td><td>X</td><td>X</td><td>X</td><td>X</td><td>X</td><td>X</td><td>X</td><td>X</td></tr><tr><td>Race/Ethnicity</td><td>X</td><td>X</td><td>X</td><td>X</td><td>X</td><td>X</td><td>X</td><td>X</td></tr><tr><td>Age</td><td>X</td><td>X</td><td>X</td><td>X</td><td>X</td><td>X</td><td>X</td><td>X</td></tr><tr><td>Year</td><td>X</td><td>X</td><td>X</td><td>X</td><td>X</td><td>X</td><td>X</td><td>X</td></tr><tr><td>Year of birth</td><td>X</td><td>X</td><td>X</td><td>X</td><td>X</td><td>X</td><td>X</td><td>X</td></tr><tr><td>State of birth</td><td>X</td><td>X</td><td>X</td><td>X</td><td>X</td><td>X</td><td>X</td><td>X</td></tr><tr><td>State of residence</td><td /><td /><td /><td>X</td><td>X</td><td>X</td><td>X</td><td>X</td></tr><tr><td>Division of birth X year of birth</td><td /><td /><td /><td /><td>X</td><td>X</td><td>X</td><td>X</td></tr><tr><td>State of resid. X year</td><td /><td /><td /><td /><td /><td>X</td><td /><td /></tr><tr><td>Trends by year of birth and state of birth</td><td /><td /><td /><td /><td /><td /><td>X</td><td /></tr><tr><td>State‐level macroeconomic variables</td><td /><td /><td /><td /><td /><td /><td /><td>X</td></tr><tr><td>N</td><td>922,421</td><td>922,262</td><td>922,262</td><td>922,262</td><td>922,262</td><td>922,262</td><td>922,262</td><td>922,262</td></tr></tbody></table> </ephtml> </p> <p>1 <emph>Notes</emph>: Column 1 is estimated using equation (<reflink idref="bib2" id="ref32">2</reflink>), and the remaining columns are variations using different covariates. The dependent variable is natural log of income for those aged 28 and above. The sample is drawn from the 2000 Decennial Census (1% sample) and the 2001–2019 American Community Survey from IPUMS.org (Ruggles et al., [<reflink idref="bib27" id="ref33">27</reflink>]). We measure math achievement for each state birth cohort using the mean achievement of 8th graders in the National Assessment of Education Progress 13 years later. "Adjusted math achievement" is mean achievement by state and year of birth, adjusted for student race‐ethnicity and parental education using the NAEP micro‐data. (See equation 1 in the text.) Column 5 reflects our preferred specification. Standard errors, reported in parentheses, are calculated after clustering by year and state of birth.</p> <p>2 Significance levels are (*) <emph>p</emph> < 0.05, (**) <emph>p</emph> < 0.01, (***) <emph>p</emph> < 0.001.</p> <p>Although we are conditioning on each sample member's race/ethnicity in equation (<reflink idref="bib2" id="ref34">2</reflink>), neither the ACS nor the Census provides a measure of family background (such as parental education or family income) for adults. Yet, if family resources were rising and had a direct positive effect on both achievement and income, the coefficient on mean achievement could be biased upward. Thus, in column 2 of Table 1, we replace the mean score, <ephtml> <math display="inline" altimg="urn:x-wiley:02768739:media:pam70018:pam70018-math-0008" xmlns="http://www.w3.org/1998/Math/MathML"><semantics><msub><mover><mi>Score</mi><mo>¯</mo></mover><mrow><mi>j</mi><mi>c</mi></mrow></msub><annotation encoding="application/x-tex">${\overline{\textit{Score}}}_{jc}$</annotation></semantics></math> </ephtml> , with the state‐year means ( <ephtml> <math display="inline" altimg="urn:x-wiley:02768739:media:pam70018:pam70018-math-0009" xmlns="http://www.w3.org/1998/Math/MathML"><semantics><msub><mover accent="true"><mi>δ</mi><mo>̂</mo></mover><mrow><mi>j</mi><mi>c</mi></mrow></msub><annotation encoding="application/x-tex">${{\hat{\delta}}_{jc}}$</annotation></semantics></math> </ephtml> from equation 1) which were adjusted by race and parental education. The coefficient on adjusted math achievement is somewhat smaller, .096.</p> <p>While test scores have been adjusted for parental education, a student's parental education may have had a direct effect on later life outcomes. Moreover, a larger concentration of college graduates among parents may have broader effects on the cohort—e.g., leading to investments by parents in social networks or other opportunites (e.g., extra‐curricular activities) not reflected in test scores. Thus, in column 3, we add a control for the percent of students in the birth cohort and state who had a resident parent with a college degree (associate, bachelor's, or higher). The coefficient falls slightly to .073.</p> <p>In column 4, we add controls for state of residence, to complement the controls for state of birth. The coefficient on adjusted math achievement remains statistically significant at .062.</p> <p>To ensure that the results are not simply being driven by differential trends for students born in different regions of the country (for example, perhaps economic development was improving both achievement and incomes in the South), we add controls for nine Census divisions interacted with single year of birth in column 5. (Column 5 will serve as our preferred specification for later outcomes.) Rising achievement was not driven solely by regional differences. For example, some southern states, such as North Carolina, saw much larger improvements in achievement than others, such as Alabama. The coefficient on adjusted test scores remains significant at .080.</p> <p>In column 6, we add dummies for state of residence interacted by year to account for changing labor markets. Again, the coefficient on adjusted test scores remains significant at .068.</p> <p>In column 7, we add separate time trends for each state of birth. Since the adjusted scores vary by state of birth and year, the only remaining variation is in the differing timing of achievement increases. The point estimate remains similar, at .087, but the standard errors nearly double. This reflects the fact that much of the variation identifying our estimates comes from differential state trends in NAEP scores (the long‐difference variation seen in Figure 2) rather than cohort to cohort divergence from the trend.</p> <p>Apparently, much of the variation we are using is based on differing long‐term trends in achievement by state. Interestingly, increases in state NAEP scores do not seem to be proxying for longer term improvements in social and economic conditions facing families. For those who were born between 1977 and 1991 (the 8th graders in 1990 and 2004), the state‐level correlation between changes in 8th‐grade math scores and changes in median incomes of households with children was −.25. There was a positive correlation between increases in 8th‐grade test scores and change in percent low‐birthweight in the years the cohorts were born (.29) and changes in the state unemployment rate when the cohort was 18 (.23). In column 8 of Table 1, we add controls for all three state level covariates, and the coefficient on math achievement remains significant at .078.</p> <hd id="AN0190550171-9">OTHER LIFE OUTCOMES FROM THE AMERICAN COMMUNITY SURVEY</hd> <p>In Table 2, we report the relationship with math achievement for additional outcomes in the Census and American Community Survey. Rather than replicate all the columns in Table 1, we report results for our preferred specification (column 5, although the results are similarly robust to alternative specifications). The first row reports the coefficient on 8th‐grade achievement using log annual earnings, .080. In the second row, we report an insignificant relationship between occupational prestige and state math scores. As a check on the face validity of the finding regarding math achievement and earnings, we incorporate data on the importance of mathematical skill by occupation from O*NET. For the 55% of observations in our analysis data set with non‐missing occupations that could be directly matched to O*NET occupations, we find a positive relationship between changes in state math scores and the mathematical content of occupations.</p> <p>2 TABLE The relationship between other later life outcomes and math achievement by state birth cohort.</p> <p> <ephtml> <table><thead><tr><th /><th>Adjusted math score</th><th>% either parent college graduate</th><th>Age when observed</th><th>Birth cohorts used</th><th>Sample size</th></tr></thead><tbody><tr><td>Ln(earned income)</td><td>0.080<ext-link /><sup>**</sup></td><td>0.104</td><td>28‐42</td><td>1977‐1991</td><td>922,262</td></tr><tr><td /><td>(0.029)</td><td>(0.058)</td><td /><td /><td /></tr><tr><td>Occupational prestige score</td><td>0.247</td><td>0.278</td><td>28‐42</td><td>1977‐1991</td><td>1,090,897</td></tr><tr><td /><td>(0.350)</td><td>(0.625)</td><td /><td /><td /></tr><tr><td>Occupational math level</td><td>0.947<ext-link /><sup>*</sup></td><td>0.140</td><td>28‐42</td><td>1977‐1991</td><td>603,105</td></tr><tr><td /><td>(0.446)</td><td>(0.999)</td><td /><td /><td /></tr><tr><td>Educational attainment:</td><td /><td /><td /><td /><td /></tr><tr><td>HS +</td><td>0.013<ext-link /><sup>*</sup></td><td>0.012</td><td>23‐42</td><td>1977‐1996</td><td>2,291,061</td></tr><tr><td /><td>(0.006)</td><td>(0.012)</td><td /><td /><td /></tr><tr><td>Some college +</td><td>0.016</td><td>0.064<ext-link /><sup>**</sup></td><td>23‐42</td><td>1977‐1996</td><td>2,291,061</td></tr><tr><td /><td>(0.010)</td><td>(0.024)</td><td /><td /><td /></tr><tr><td>BA +</td><td>0.011</td><td>0.111<ext-link /><sup>***</sup></td><td>23‐42</td><td>1977‐1996</td><td>2,291,061</td></tr><tr><td /><td>(0.010)</td><td>(0.024)</td><td /><td /><td /></tr><tr><td>Graduate degree</td><td>0.005</td><td>0.032<ext-link /><sup>**</sup></td><td>23‐42</td><td>1977‐1996</td><td>2,291,061</td></tr><tr><td /><td>(0.005)</td><td>(0.011)</td><td /><td /><td /></tr><tr><td>Unemployed</td><td>−0.019<ext-link /><sup>*</sup></td><td>0.006</td><td>28‐42</td><td>1977‐1991</td><td>1,049,603</td></tr><tr><td /><td>(0.008)</td><td>(0.016)</td><td /><td /><td /></tr><tr><td>Receiving Food Stamps/SNAP</td><td>0.027</td><td>−0.011</td><td>28‐42</td><td>1977‐1991</td><td>1,262,220</td></tr><tr><td /><td>(0.014)</td><td>(0.027)</td><td /><td /><td /></tr><tr><td>Usual hours worked</td><td>0.686<ext-link /><sup>*</sup></td><td>−0.848</td><td>28‐42</td><td>1977‐1991</td><td>1,019,076</td></tr><tr><td /><td>(0.309)</td><td>(0.686)</td><td /><td /><td /></tr><tr><td>Ever married</td><td>−0.023</td><td>0.016</td><td>28‐42</td><td>1977‐1991</td><td>1,258,292</td></tr><tr><td /><td>(0.015)</td><td>(0.031)</td><td /><td /><td /></tr><tr><td>Own home</td><td>−0.014</td><td>0.006</td><td>28‐42</td><td>1977‐1991</td><td>1,213,923</td></tr><tr><td /><td>(0.016)</td><td>(0.033)</td><td /><td /><td /></tr><tr><td>Teenage mother</td><td>−0.018<ext-link /><sup>*</sup></td><td>0.035<ext-link /><sup>*</sup></td><td>18‐30</td><td>1977‐2001</td><td>1,583,294</td></tr><tr><td /><td>(0.009)</td><td>(0.014)</td><td /><td /><td /></tr><tr><td>Institutionalized/incarcerated</td><td>−0.017<ext-link /><sup>***</sup></td><td>0.010</td><td>18‐24</td><td>1977‐2001</td><td>1,076,101</td></tr><tr><td /><td>(0.004)</td><td>(0.006)</td><td /><td /><td /></tr><tr><td>Enrolled in college</td><td>0.053<ext-link /><sup>***</sup></td><td>0.041</td><td>19‐24</td><td>1977‐2000</td><td>1,748,070</td></tr><tr><td /><td>(0.013)</td><td>(0.024)</td><td /><td /><td /></tr><tr><td>Working FT/enrolled in college</td><td>0.028<ext-link /><sup>**</sup></td><td>0.016</td><td>19‐24</td><td>1977‐2000</td><td>1,657,001</td></tr><tr><td /><td>(0.009)</td><td>(0.018)</td><td /><td /><td /></tr></tbody></table> </ephtml> </p> <ulist> <item>3 <emph>Notes</emph>: All of the above specifications also include fixed effects for age, race, gender, state of birth, state of residence, census division of birth by year of birth and year—the same independent variables as column 5 of Table 1. The sample is drawn from the 2000 Decennial Census (1% sample) and the 2001–2019 American Community Survey, downloaded from IPUMS (Ruggles et al., [<reflink idref="bib27" id="ref35">27</reflink>]). Standard errors, reported in parentheses, are calculated after clustering by state and year of birth.</item> <item>4 Significance levels are (*) <emph>p</emph> < 0.05, (**) <emph>p</emph> < 0.01, (***) <emph>p</emph> < 0.001.</item> </ulist> <p>In the remainder of Table 2, we find that those born in states with larger increases in 8th‐grade math achievement saw larger increases in high school graduation and an increase in college enrollment. With regard to labor force participation, birth cohorts in those states were less likely to be unemployed and worked longer hours. They were more likely to be either working full time or enrolled in college when observed between the ages of 18 and 24. Girls born in those states were less likely to become teen mothers; boys were less likely to be incarcerated at age 18 to 24.</p> <p>The relationships for several other outcomes were not statistically significant (specifically, BA degree completion, home ownership, occupational prestige, and marital status). BA degree completion was the outcome most related to parental education.</p> <hd id="AN0190550171-10">ARRESTS</hd> <p>In Table 3, we report the relationship, estimated with OLS, between the log of the number of arrests by single year of age, state, and year and the measure of math achievement adjusted for parental education and race ( <ephtml> <math display="inline" altimg="urn:x-wiley:02768739:media:pam70018:pam70018-math-0010" xmlns="http://www.w3.org/1998/Math/MathML"><semantics><msub><mover accent="true"><mi>δ</mi><mo>̂</mo></mover><mrow><mi>j</mi><mi>c</mi></mrow></msub><annotation encoding="application/x-tex">${{\hat{\delta }}_{jc}}$</annotation></semantics></math> </ephtml> ). As in the previous analyses, we include in the regressions a measure of the percentage of parents in the birth cohort who were college graduates. We also include fixed effects for states, age by year, and geographic Census division by year of birth. Unlike the ACS and Census, arrests are only available by state of occurrence, not by state of birth. Thus, unlike in Tables 1 and 2, we assign math achievement by state of occurrence and limit the outcome to young adults, aged 15 to 24. A 1 standard deviation increase in 8th‐grade math achievement is associated with a 30% decline in violent crime arrests and a somewhat smaller (18%) decline in property crime arrests.</p> <p>3 TABLE Age specific arrests and 8th grade math achievement.</p> <p> <ephtml> <table><thead><tr><th /><th>Violent crime</th><th>Property crime</th></tr></thead><tbody><tr><td>Adjusted Math Achievement</td><td>−0.357<ext-link /><sup>***</sup></td><td>−0.201<ext-link /><sup>*</sup></td></tr><tr><td /><td>(0.088)</td><td>(0.087)</td></tr><tr><td>Either parent college graduate</td><td>0.098</td><td>−0.116</td></tr><tr><td /><td>(0.177)</td><td>(0.162)</td></tr><tr><td>State</td><td>X</td><td>X</td></tr><tr><td>Age X year</td><td>X</td><td>X</td></tr><tr><td>Geographic division X year of birth</td><td>X</td><td>X</td></tr><tr><td>N</td><td>4,130</td><td>4,130</td></tr></tbody></table> </ephtml> </p> <ulist> <item>5 <emph>Notes</emph>: The dependent variable is the log of arrests by state, age, and year. The data on arrests were provided by Donohue and Levitt ([<reflink idref="bib10" id="ref36">10</reflink>]). "Adjusted Math Achievement" consists of state by year fixed effects, after using student‐level data from the Main NAEP to adjust for parental education and student race‐ethnicity. (See equation 1.) Standard errors, reported in parentheses, are calculated clustering by state and year of birth.</item> <item>6 Significance levels are (*) <emph>p</emph> < 0.05, (**) <emph>p</emph> < 0.01, (***) <emph>p</emph> < 0.001.</item> </ulist> <hd id="AN0190550171-11">CONCLUSION</hd> <p>Within 2 weeks of the release of the 2022 NAEP results, 522 media outlets reported on the losses of 4th‐ and 8th‐grade students during the pandemic. U.S. Secretary of Education, Miguel Cardona, as well as governors, mayors, and superintendents were compelled to comment. Even before the pandemic, there was a similar outpouring of interest surrounding the release of state NAEP results, with news stories and editorials citing the results as evidence of the efficacy or inefficacy of federal, state, and local reform efforts. But do the NAEP results say anything about the future prospects of students? Do they deserve all the attention they receive?</p> <p>We find that they do—or, at least, they have forecasted future earnings in the past. In fact, the improvement in earnings for birth cohorts in states with rising math achievement was only slightly smaller (8% per standard deviation increase in 8th‐grade math achievement) than cross‐sectional estimates of the effect of earnings on achievement (approximately 12%). Given that the average math achievement of U.S. 8th graders improved by just over half a standard deviation (20 points) between 1990 and 2009 (the year today's 28‐year‐olds would have been in 8th grade), our results imply that the annual incomes of U.S. 28 year olds are 4.4% higher as a result, while teen motherhood rates and male incarceration rates have declined by 1.0 and 0.9 percentage points, respectively. The implied earnings improvement in states, such as North Carolina, with the largest increases in achievement would have been even larger (over 7%).</p> <p>Over the past 30 years, the NAEP has been sending a mixed message on the most basic question in U.S. education policy: are the skills of U.S. K–12 students improving—or not? The answer has depended on the subject and grade level of the student. In 12th grade, mean reading achievement <emph>declined</emph> by .12 standard deviations between 1992 and 2013. Twelfth grade math achievement also showed little improvement, increasing by .16 standard deviations between 1990 and 2000 (when the math standards changed) and by just .06 standard deviations between 2005 and 2015. The 12th‐grade scores are frequently cited to characterize U.S. education as "stagnant."</p> <p>However, the story has been very different in elementary and middle school grades: The average math achievement of 4th and 8th grade students in the U.S. had improved by .9 and .6 standard deviations, respectively, between 1990 and the peak in 2013—with much larger increases in some states than in others. Even on international assessments, such as the Trends in International Math and Science Survey (TIMSS), the achievement of 4th‐ and 8th‐grade students in the U.S. improved by .25 and .26 standard deviations, respectively, between 1995 and 2015 (Martin et al., [<reflink idref="bib18" id="ref37">18</reflink>]). As with the NAEP, it has been on the international assessments of high school students that U.S. achievement has been lagging. For example, on the PISA exam of 14‐year‐olds, average achievement in math declined slightly between 2003 and 2018 (National Center for Education Statistics, [<reflink idref="bib22" id="ref38">22</reflink>]).</p> <p>Our results suggest that the fade‐out in 4th‐ and 8th‐grade achievement gains on the 12th‐grade NAEP may be analogous to the fade‐out of effects of kindergarten and pre‐school interventions. Long‐term outcomes seem to improve, even if much of the gain in achievement has faded by 12th grade.</p> <p>During the pandemic (between 2019 and 2022), mean 8th‐grade math achievement declined by .2 standard deviations, meaning U.S. students have forfeited 40% of the improvement since 1990. The decline in 4th‐grade math was similar, .16 standard deviations. Thus, if such losses were to become permanent, our results imply that a .2 standard deviation decline in achievement would be associated with a 1.6% decline in earnings for both cohorts of students.</p> <hd id="AN0190550171-12">DATA AVAILABILITY STATEMENT</hd> <p>The data that support the findings of this study are available in "Replication Data for: What do changes in state NAEP scores imply for birth cohorts' later life outcomes?" at https://doi.org/10.7910/DVN/QOR3A3 (Doty et al., [<reflink idref="bib12" id="ref39">12</reflink>]). These data were derived from a number of resources, both publicly available and restricted‐use, as detailed in the "readme" file of the repository. Restrictions apply to the use of some of these resources. See the "readme" file and the websites of the data providers for more information.</p> <hd id="AN0190550171-13">ACKNOWLEDGMENTS</hd> <p>This research was supported by a grant from the Walton Family Foundation.</p> <p>GRAPH: Supporting Information</p> <ref id="AN0190550171-14"> <title> Footnotes </title> <blist> <bibl id="bib1" idref="ref16" type="bt">1</bibl> <bibtext> The federal government invests more than $175m annually on the assessments (National Academies of Sciences, Engineering and Medicine, [21]).</bibtext> </blist> <blist> <bibl id="bib2" idref="ref17" type="bt">2</bibl> <bibtext> The nonlinear relationships between percentiles and standard deviations and between earnings and log earnings make it difficult to compare this ratio to the results of log wage specifications cited above.</bibtext> </blist> <blist> <bibl id="bib3" idref="ref4" type="bt">3</bibl> <bibtext> In contrast to the Long‐term Trend NAEP, which tracks national trends, the Main NAEP was designed to yield state‐level estimates.</bibtext> </blist> <blist> <bibl id="bib4" idref="ref1" type="bt">4</bibl> <bibtext> Public school enrollment represented 90.2% of K–12 enrollment in the U.S. in 2019. See: https://nces.ed.gov/programs/digest/d21/tables/dt21_205.10.asp?current=yes</bibtext> </blist> <blist> <bibl id="bib5" idref="ref2" type="bt">5</bibl> <bibtext> Unfortunately, there are no state‐level measures available in the 12th grade NAEP.</bibtext> </blist> <blist> <bibl id="bib6" type="bt">6</bibl> <bibtext> With data for only four of the birth cohorts between 1985 and 1991, the coefficient on 8th grade reading scores in predicting log earnings is imprecisely estimated (actually negative, −.043, but with a standard error of .062). Using a broader range of birth cohorts, state reading scores were negatively related to institutionalization/incarceration rates and positively related to college enrollment (both significantly different from zero). However, higher reading scores were also positively (and statistically significantly) related to Food Stamp receipt, violent crime arrest rates, and teen motherhood.</bibtext> </blist> <blist> <bibl id="bib7" idref="ref7" type="bt">7</bibl> <bibtext> All appendices are available at the end of this article as it appears in JPAM online. Go to the publisher's website and use the search engine to locate the article at <ulink href="http://onlinelibrary.wiley.com">http://onlinelibrary.wiley.com</ulink>.</bibtext> </blist> <blist> <bibl id="bib8" idref="ref3" type="bt">8</bibl> <bibtext> We believe this to be a reasonable assumption given that the typical 8th grader enters at age 13 and 80% of 13‐year‐olds were still living in their state of birth in 2000.</bibtext> </blist> <blist> <bibl id="bib9" idref="ref27" type="bt">9</bibl> <bibtext> For the seven states that had a valid score for 1992 but not 1990 (Massachusetts, Maine, Missouri, Mississippi, South Carolina, Tennessee, and Utah), we report the change between 1992 and 2019.</bibtext> </blist> <blist> <bibtext> In Iowa, the mean score declined by .05 s.d. from 1992 to 2019, though it improved by .15 s.d. from 1990 to 1992.</bibtext> </blist> <blist> <bibtext> National Center for Education Statistics ([23]). For a more detailed description of the 2022 to 2024 recovery in math and reading by state and district and a discussion of the role of federal pandemic relief and chronic absenteeism, see Dewey et al. ([11]).</bibtext> </blist> <blist> <bibtext> White, Black, Hispanic, Asian/Pacific Islander, Native American/Alaskan Native, and Other/unclassified.</bibtext> </blist> <blist> <bibtext> We included a set of indicators for mother's and father's education. For each parent, students were asked to choose from five categories: did not finish high school, high school graduate, some education after high school, graduated from college, and "I don't know." The mother's education indicators and father's education indicators are fully interacted with one another.</bibtext> </blist> <blist> <bibtext> In the 2021 ACS, 98% of the 18‐ to 24‐year‐old population that were in institutional group quarters were in adult correctional facilities or juvenile detention. As a result, we use an indicator of residence in institutional group quarters as a proxy for incarceration (U.S. Bureau of the Census, Table S2603: https://data.census.gov/chart?q=S2603). We adjusted income for inflation using price index data from the OECD (Organization for Economic Co‐operation and Development, [26]).</bibtext> </blist> <blist> <bibtext> See https://<ulink href="http://www.onetonline.org/find/descriptor/result/2.A.1.e">www.onetonline.org/find/descriptor/result/2.A.1.e</ulink>. O*NET provides a number between 0 and 100 for mathematics "Level" and "Importance." We use the "Level." The two measures have a correlation of 0.94.</bibtext> </blist> <blist> <bibtext> See https://<ulink href="http://www.bls.gov/lau/data.htm">www.bls.gov/lau/data.htm</ulink>. We use file "la.data.2.AllStatesU" from download.bls.gov ‐ /pub/time.series/la/.</bibtext> </blist> <blist> <bibtext> See https://<ulink href="http://www.nber.org/research/data/vital‐statistics‐birth‐data‐nber">www.nber.org/research/data/vital‐statistics‐birth‐data‐nber</ulink></bibtext> </blist> <blist> <bibtext> Note that for this graph the 1977 cohort and 1990 cohort are observed at different ages in the ACS, but this is the same for every state so that the long differences still capture how changes in cohort earnings are associated with changes in cohort scores. In the regressions we control for age at which the cohort was observed.</bibtext> </blist> <blist> <bibtext> Similarly, in analyses of other outcome variables reported in the next section, Census‐allocated values of the outcome of interest are dropped.</bibtext> </blist> <blist> <bibtext> We estimate parental education in the state and cohort using the 1990 and 2000 decennial censuses and the 2009 ACS, which provided large samples of students in each birth cohort in each state. We normalize scores for each birth cohort, since older cohorts with older parents may report additional schooling.</bibtext> </blist> <blist> <bibtext> The nine census divisions include New England (CT, ME, MA, NH, RI, VT), Mid‐Atlantic (NJ, NY, PA), East North Central (IN, IL, MI, OH, WI), West North Central (IA, KS, MN, MO, NE, ND, SD), South Atlantic (DE, DC, FL, GA, MD, NC, SC, VA, WV), East South Central (AL, KY, MS, TN), West South Central (AR, LA, OK, TX), Mountain (AZ, CO, ID, NM, MT, UT, NV, WY), and Pacific (AL, CA, HI, OR, WA).</bibtext> </blist> <blist> <bibtext> The R<sups>2</sups> of state math scores on the state by year of birth trends and the other covariates in column 7 is .981.</bibtext> </blist> <blist> <bibtext> This is the same regression as reported in column 5 of Table 1. The other reported regressions in Table 2 differ only in the dependent variable.</bibtext> </blist> <blist> <bibtext> While both the Census Bureau and O*NET base their occupation classifications on the Standard Occupational Classification (SOC) system, each classification differs somewhat from SOC. The Census Bureau does provide an associated SOC code for each occupation, but in many cases the Census occupation is not a perfect match for a single SOC occupation, in which case it provides an aggregate SOC code representing multiple occupations. We do not use these cases in the analysis. Additionally, the SOC system changed in 2010 and 2018; we use crosswalks to map the SOC code in the ACS data to a 2018 SOC code. Finally, not all 2018 SOC occupations have a math level reported by O*NET.</bibtext> </blist> <blist> <bibtext> The coefficient estimated for Food Stamp (or SNAP) benefit receipt had an anomalous (positive) sign, though it was not quite statistically significant at a 5% level.</bibtext> </blist> <blist> <bibtext> In the 2009 ACS, 73% of U.S.‐born individuals between the ages of 18 and 24 were still living in their state of birth.</bibtext> </blist> <blist> <bibtext> If <ephtml> <math display="inline" altimg="urn:x-wiley:02768739:media:pam70018:pam70018-math-0011" xmlns="http://www.w3.org/1998/Math/MathML"><semantics><mi>β</mi><annotation encoding="application/x-tex">$\beta $</annotation></semantics></math> </ephtml> is the coefficient on test scores and the dependent variable is the natural log of arrests, then the implied percentage change in arrests per unit change in test scores is <ephtml> <math display="inline" altimg="urn:x-wiley:02768739:media:pam70018:pam70018-math-0012" xmlns="http://www.w3.org/1998/Math/MathML"><semantics><mrow><msup><mi>e</mi><mi>β</mi></msup><mo>−</mo><mn>1</mn></mrow><annotation encoding="application/x-tex">${{e}^\beta } - 1$</annotation></semantics></math> </ephtml> . We translated the coefficients in Table 3 accordingly. We did not do so in discussions of Tables 1 and 2, since <ephtml> <math display="inline" altimg="urn:x-wiley:02768739:media:pam70018:pam70018-math-0013" xmlns="http://www.w3.org/1998/Math/MathML"><semantics><mrow><msup><mi>e</mi><mi>β</mi></msup><mo>−</mo><mn>1</mn><mspace width="0.33em" /><mo>≅</mo><mspace width="0.33em" /><mi>β</mi></mrow><annotation encoding="application/x-tex">${{e}^\beta } - 1\ \cong \ \beta $</annotation></semantics></math> </ephtml> when <ephtml> <math display="inline" altimg="urn:x-wiley:02768739:media:pam70018:pam70018-math-0014" xmlns="http://www.w3.org/1998/Math/MathML"><semantics><mrow><mo>−</mo><mo>.</mo><mn>1</mn><mo><</mo><mi>β</mi><mo><</mo><mo>.</mo><mn>1</mn></mrow><annotation encoding="application/x-tex">$ -.1 < \beta <.1$</annotation></semantics></math> </ephtml> .</bibtext> </blist> <blist> <bibtext> The 2022 NAEP results were released on October 24, 2022. A search on the news aggregation site, Factiva, of the keywords "National Assessment of Educational Progress" and "NAEP" between October 24, 2022, and November 7, 2022, yielded 522 unique stories.</bibtext> </blist> <blist> <bibtext> Murnane et al. ([20]) estimated that a standard‐deviation increase in test scores led to an approximately 12% increase in age‐31 earnings.</bibtext> </blist> <blist> <bibtext> For example, in the chapter on education describing the Heritage Foundation's Project 2025 Proposals, Lindsey Burke wrote, "After trillions spent since 1965 on the collective programs now housed within the walls of the department, student academic outcomes remain stagnant" (Dans & Groves, [6], p. 319).</bibtext> </blist> <blist> <bibtext> The decline in 8th‐grade math was 8 points, with a 2019 s.d. of 40 points. The decline in 4th‐grade math was 5 points, with a 2019 s.d. of 32 points.</bibtext> </blist> </ref> <ref id="AN0190550171-15"> <title> REFERENCES </title> <blist> <bibtext> Barro, R. J. (1991). Economic growth in a cross section of countries. Quarterly Journal of Economics, 106 (2), 407 – 443. https://doi.org/10.2307/2937943</bibtext> </blist> <blist> <bibtext> Bils, M., & Klenow, P. J. (2000). Does schooling cause growth? American Economic Review, 90 (5), 1160 – 1183. https://doi.org/10.1257/aer.90.5.1160</bibtext> </blist> <blist> <bibtext> Blagg, K., & Chingos, M. M. (2016). Varsity blues: Are high school students being left behind? Urban Institute. https://<ulink href="http://www.urban.org/sites/default/files/publication/80251/2000773‐varsity‐blues‐are‐high‐school‐students‐being‐left‐behind%5f3.pdf">www.urban.org/sites/default/files/publication/80251/2000773‐varsity‐blues‐are‐high‐school‐students‐being‐left‐behind%5f3.pdf</ulink></bibtext> </blist> <blist> <bibtext> Chetty, R., Friedman, J. N., Hilger, N., Saez, E., Schanzenbach, D. W., & Yagan, D. (2011). How does your kindergarten classrooom affect your earnings?: Evidence from Project Star. The Quarterly Journal of Economics, 126 (4), 1593 – 1660. https://doi.org/10.1093/qje/qjr041</bibtext> </blist> <blist> <bibtext> Chetty, R., Friedman, J. N., & Rockoff, J. E. (2014). Measuring the impacts of teachers II: Teacher value‐added and student outcomes in adulthood. American Economic Review, 104 (9), 2633 – 2679. https://doi.org/10.1257/aer.104.9.2633</bibtext> </blist> <blist> <bibtext> Dans, P., & Groves, S. (2023). Mandate for leadership: The conservative promise. The Heritage Foundation. https://static.project2025.org/2025_MandateForLeadership_FULL.pdf</bibtext> </blist> <blist> <bibtext> Deming, D. J. (2009). Early childhood intervention and life‐cycle skill development: Evidence from Head Start. American Economic Journal: Applied Economics, 1 (3), 111 – 134. https://doi.org/10.1257/app.1.3.111</bibtext> </blist> <blist> <bibtext> Deming, D. J., Cohodes, S., Jennings, J., & Jencks, C. (2016). School accountability, postsecondary attainment and earnings. Review of Economics and Statistics, 98 (5), 848 – 862. https://doi.org/10.1162/REST_a_00598</bibtext> </blist> <blist> <bibtext> Donohue, J. J. (2020). Replication package: The impact of legalized abortion on crime over the last two decades. <ulink href="http://works.bepress.com/john%5fdonohue/192/">http://works.bepress.com/john%5fdonohue/192/</ulink></bibtext> </blist> <blist> <bibtext> Donohue, J. J., & Levitt, S. D. (2020). The impact of legalized abortion on crime over the last two decades. American Law and Economics Review, 22 (2), 241 – 302. https://doi.org/10.1093/aler/ahaa008</bibtext> </blist> <blist> <bibtext> Dewey, D. C., Fahle, E., Kane, T. J., Reardon, S. F., & Staiger, D. O. (2025). Pivoting from pandemic recovery to long‐term reform: A distict‐level analysis [Working paper]. Harvard Center for Education Policy Research. https://educationrecoveryscorecard.org/wp‐content/uploads/2025/02/Pivoting‐from‐Pandemic‐Recovery‐to‐Long‐Term‐Reform‐A‐District‐Level‐Analysis.pdf</bibtext> </blist> <blist> <bibtext> Doty, E., Kane, T. J., Patterson, T., & Staiger, D. O. (2025). Replication data for: What do changes in state NAEP scores imply for birth cohorts' later life outcomes? Harvard Dataverse. https://doi.org/10.7910/DVN/QOR3A3</bibtext> </blist> <blist> <bibtext> Flood, S., King, M., Rodgers, R., Ruggles, S., Warren, J. R., Backman, D., Chen, A., Cooper, G., Richards, S., Schouweiler, M., & Westberry, M. (2023). IPUMS CPS: Version 11.0 [dataset]. IPUMS. https://doi.org/10.18128/D030.V11.0</bibtext> </blist> <blist> <bibtext> Goldhaber, D., & Özek, U. (2019). How much should we rely on student test achievement as a measure of success?. Educational Researcher, 48 (7), 479 – 483. https://doi.org/10.3102/0013189X19874061</bibtext> </blist> <blist> <bibtext> Hanushek, E. A., & Woessmann, L. (2008). The role of cognitive skills in economic development. Journal of Economic Literature, 46 (3), 607 – 668. https://doi.org/10.1257/jel.46.3.607</bibtext> </blist> <blist> <bibtext> Harris, D. N., Jackson, C. K., & Pariguana, M. (2022). Is high school achievement really declining? An analysis of NAEP sample selection and the role of rising high school graduation rates. Presentation at the Association for Public Policy and Management, Washington, DC, Nov. 19, 2022.</bibtext> </blist> <blist> <bibtext> Lovenheim, M. F., & Willén, A. (2019). The long‐run effects of teacher collective bargaining. American Economic Journal: Economic Policy, 11 (3), 292 – 324. https://doi.org/10.1257/pol.20170570</bibtext> </blist> <blist> <bibtext> Martin, M. O., Mullis, I. V. S., Foy, P., & Hooper, M. (2016). TIMSS 2015 international results in science. Boston College, TIMSS & PIRLS International Study Center. <ulink href="http://timssandpirls.bc.edu/timss2015/international‐results/">http://timssandpirls.bc.edu/timss2015/international‐results/</ulink></bibtext> </blist> <blist> <bibtext> Murnane, R. J., Willett, J. B., & Levy, F. (1995). The growing importance of cognitive skills in wage determination. Review of Economics and Statistics, 77 (2), 251 – 266. https://doi.org/10.2307/2109863</bibtext> </blist> <blist> <bibtext> Murnane, R. J., Willett, J. B., Duhaldeborde, Y., & Tyler, J. H. (2000). How important are the cognitive skills of teenagers in predicting subsequent earnings? Journal of Policy Analysis and Management, 19 (4), 547 – 568. https://doi.org/10.1002/1520‐6688(200023)19:4%3C547::AID‐PAM2%3E3.0.CO;2‐%23</bibtext> </blist> <blist> <bibtext> National Academies of Sciences, Engineering, and Medicine. (2022). A pragmatic future for NAEP: Containing costs and updating technologies. The National Academies Press. https://doi.org/10.17226/26427</bibtext> </blist> <blist> <bibtext> National Center for Education Statistics. (2020). Highlights of U.S. PISA 2018 results web report (NCES 2020–166 and 2020–072). U.S. Department of Education, Institute of Education Sciences. https://nces.ed.gov/surveys/pisa/pisa2018/index.asp</bibtext> </blist> <blist> <bibtext> National Center for Education Statistics. (2025). Nation's report card: Mathematics. U.S. Department of Education. Institute of Education Sciences. https://<ulink href="http://www.nationsreportcard.gov/reports/mathematics/2024/g4%5f8/?grade=8">www.nationsreportcard.gov/reports/mathematics/2024/g4%5f8/?grade=8</ulink></bibtext> </blist> <blist> <bibtext> Neal, D. A. (2018). Information, incentives, and education policy. Harvard University Press.</bibtext> </blist> <blist> <bibtext> Neal, D. A., & Johnson, W. R. (1996). The role of premarket factors in black‐white wage differences. Journal of Political Economy, 104 (5), 869 – 895. https://doi.org/10.1086/262045</bibtext> </blist> <blist> <bibtext> Organization for Economic Co‐operation and Development. (2024). Consumer price indices (CPIs, HICPs), COICOP 1999: Consumer price index: Total for United States [USACPIALLAINMEI]. Retrieved from FRED, Federal Reserve Bank of St. Louis. https://fred.stlouisfed.org/series/USACPIALLAINMEI</bibtext> </blist> <blist> <bibtext> Ruggles, S., Flood, S., Goeken, R., Schouweiler, M., & Sobek, M. (2022). IPUMS USA: Version 12.0 [dataset]. IPUMS. https://doi.org/10.18128/D010.V12.0</bibtext> </blist> </ref> <aug> <p>By Elena Doty; Thomas J. Kane; Tyler Patterson and Douglas O. Staiger</p> <p>Reported by Author; Author; Author; Author</p> <p></p> <p>Elena Doty is a software engineer at Google, 355 Main Street, Cambridge, MA 02142 (email: elena.doty@gmail.com).</p> <p>Thomas J. Kane is Walter H. Gale Professor of Education and Economics at Harvard Graduate School of Education, Harvard‐CEPR, 50 Church Street 4th Floor, Cambridge MA 02138 (email: tom_kane@gse.harvard.edu).</p> <p>Tyler Patterson is a PhD student in the Kenneth C. Griffin Department of Economics at the University of Chicago, 1126 E. 59th Street, Chicago, IL 60637 (email: tpatterson@uchicago.edu).</p> <p>Douglas O. Staiger is John Sloan Dickey Third Century Professor of the Social Sciences in the Department of Economics at Dartmouth College, 6106 Rockefeller Hall, Dartmouth College, Hanover, NH 03755 (email: doug.staiger@dartmouth.edu).</p> </aug> <nolink nlid="nl1" bibid="bib16" firstref="ref5"></nolink> <nolink nlid="nl2" bibid="bib25" firstref="ref8"></nolink> <nolink nlid="nl3" bibid="bib20" firstref="ref9"></nolink> <nolink nlid="nl4" bibid="bib14" firstref="ref11"></nolink> <nolink nlid="nl5" bibid="bib24" firstref="ref22"></nolink> <nolink nlid="nl6" bibid="bib19" firstref="ref23"></nolink> <nolink nlid="nl7" bibid="bib27" firstref="ref25"></nolink> <nolink nlid="nl8" bibid="bib17" firstref="ref26"></nolink> <nolink nlid="nl9" bibid="bib10" firstref="ref28"></nolink> <nolink nlid="nl10" bibid="bib13" firstref="ref29"></nolink> <nolink nlid="nl11" bibid="bib18" firstref="ref37"></nolink> <nolink nlid="nl12" bibid="bib22" firstref="ref38"></nolink> <nolink nlid="nl13" bibid="bib12" firstref="ref39"></nolink>
Header DbId: eric
DbLabel: ERIC
An: EJ1493991
AccessLevel: 3
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 0
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: What Do Changes in State NAEP Scores Imply for Birth Cohorts' Later Life Outcomes?
– Name: Language
  Label: Language
  Group: Lang
  Data: English
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AR" term="%22Elena+Doty%22">Elena Doty</searchLink><br /><searchLink fieldCode="AR" term="%22Thomas+J%2E+Kane%22">Thomas J. Kane</searchLink> (ORCID <externalLink term="https://orcid.org/0009-0008-1297-5426">0009-0008-1297-5426</externalLink>)<br /><searchLink fieldCode="AR" term="%22Tyler+Patterson%22">Tyler Patterson</searchLink><br /><searchLink fieldCode="AR" term="%22Douglas+O%2E+Staiger%22">Douglas O. Staiger</searchLink>
– Name: TitleSource
  Label: Source
  Group: Src
  Data: <searchLink fieldCode="SO" term="%22Journal+of+Policy+Analysis+and+Management%22"><i>Journal of Policy Analysis and Management</i></searchLink>. 2026 45(1).
– Name: Avail
  Label: Availability
  Group: Avail
  Data: Wiley. Available from: John Wiley & Sons, Inc. 111 River Street, Hoboken, NJ 07030. Tel: 800-835-6770; e-mail: cs-journals@wiley.com; Web site: https://www.wiley.com/en-us
– Name: PeerReviewed
  Label: Peer Reviewed
  Group: SrcInfo
  Data: Y
– Name: Pages
  Label: Page Count
  Group: Src
  Data: 15
– Name: DatePubCY
  Label: Publication Date
  Group: Date
  Data: 2026
– Name: TypeDocument
  Label: Document Type
  Group: TypDoc
  Data: Journal Articles<br />Reports - Research
– Name: Subject
  Label: Descriptors
  Group: Su
  Data: <searchLink fieldCode="DE" term="%22Educational+Change%22">Educational Change</searchLink><br /><searchLink fieldCode="DE" term="%22National+Competency+Tests%22">National Competency Tests</searchLink><br /><searchLink fieldCode="DE" term="%22Scores%22">Scores</searchLink><br /><searchLink fieldCode="DE" term="%22Predictor+Variables%22">Predictor Variables</searchLink><br /><searchLink fieldCode="DE" term="%22Outcomes+of+Education%22">Outcomes of Education</searchLink><br /><searchLink fieldCode="DE" term="%22Mathematics+Achievement%22">Mathematics Achievement</searchLink><br /><searchLink fieldCode="DE" term="%22Income%22">Income</searchLink><br /><searchLink fieldCode="DE" term="%22Educational+Attainment%22">Educational Attainment</searchLink><br /><searchLink fieldCode="DE" term="%22Early+Parenthood%22">Early Parenthood</searchLink><br /><searchLink fieldCode="DE" term="%22Institutionalized+Persons%22">Institutionalized Persons</searchLink><br /><searchLink fieldCode="DE" term="%22Crime%22">Crime</searchLink><br /><searchLink fieldCode="DE" term="%22Cohort+Analysis%22">Cohort Analysis</searchLink>
– Name: SubjectThesaurus
  Label: Assessment and Survey Identifiers
  Group: Su
  Data: <searchLink fieldCode="SU" term="%22National+Assessment+of+Educational+Progress%22">National Assessment of Educational Progress</searchLink>
– Name: DOI
  Label: DOI
  Group: ID
  Data: 10.1002/pam.70018
– Name: ISSN
  Label: ISSN
  Group: ISSN
  Data: 0276-8739<br />1520-6688
– Name: Abstract
  Label: Abstract
  Group: Ab
  Data: Since 1990, the National Assessment of Educational Progress (NAEP) has been the primary benchmark for tracking the progress of state education reform. The focus on math and reading achievement is motivated by the cross-sectional relationship between test scores and adult outcomes, such as earnings and college completion. But do changes in NAEP scores predict changes in long-term economic and social outcomes for future earners--or do they reflect other factors unrelated to earnings such as teaching to the test? We investigate by linking long-term outcomes by year and state of birth to NAEP scores. We find that more recent birth cohorts in states with large increases in NAEP math achievement enjoyed higher incomes, improved educational attainment, and declines in teen motherhood, incarceration, and arrest rates compared to those in states with smaller increases. In fact, the relationship between changes in NAEP achievement and cohort earnings is about two thirds the size of the cross-sectional relationship observed in prior research: a 6% to 8% rise in earnings per standard deviation rise in 8th grade math. The results are not sensitive to controls for student demographics, labor market conditions, or measures of children's health (such as low birthweight).
– Name: AbstractInfo
  Label: Abstractor
  Group: Ab
  Data: As Provided
– Name: Note
  Label: Notes
  Group: Note
  Data: https://doi.org/10.7910/DVN/QOR3A3
– Name: DateEntry
  Label: Entry Date
  Group: Date
  Data: 2026
– Name: AN
  Label: Accession Number
  Group: ID
  Data: EJ1493991
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=eric&AN=EJ1493991
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.1002/pam.70018
    Languages:
      – Text: English
    PhysicalDescription:
      Pagination:
        PageCount: 15
    Subjects:
      – SubjectFull: Educational Change
        Type: general
      – SubjectFull: National Competency Tests
        Type: general
      – SubjectFull: Scores
        Type: general
      – SubjectFull: Predictor Variables
        Type: general
      – SubjectFull: Outcomes of Education
        Type: general
      – SubjectFull: Mathematics Achievement
        Type: general
      – SubjectFull: Income
        Type: general
      – SubjectFull: Educational Attainment
        Type: general
      – SubjectFull: Early Parenthood
        Type: general
      – SubjectFull: Institutionalized Persons
        Type: general
      – SubjectFull: Crime
        Type: general
      – SubjectFull: Cohort Analysis
        Type: general
      – SubjectFull: National Assessment of Educational Progress
        Type: general
    Titles:
      – TitleFull: What Do Changes in State NAEP Scores Imply for Birth Cohorts' Later Life Outcomes?
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Elena Doty
      – PersonEntity:
          Name:
            NameFull: Thomas J. Kane
      – PersonEntity:
          Name:
            NameFull: Tyler Patterson
      – PersonEntity:
          Name:
            NameFull: Douglas O. Staiger
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 01
              M: 01
              Type: published
              Y: 2026
          Identifiers:
            – Type: issn-print
              Value: 0276-8739
            – Type: issn-electronic
              Value: 1520-6688
          Numbering:
            – Type: volume
              Value: 45
            – Type: issue
              Value: 1
          Titles:
            – TitleFull: Journal of Policy Analysis and Management
              Type: main
ResultId 1