CoChat
Sign Up
Cited by 158Labor market dynamics and wage inequalityIntergenerational and Educational Inequality StudiesEducation and Labor Relations

Changes in the Characteristics of American Youth: Implications for Adult Outcomes

Altonji, Joseph G., Bharadwaj, Prashant, Lange, Fabian·2012·Journal of Labor Economics
DOI 10.1086/666536
Open this paper in CoChat — free account

Abstract

We examine changes in the characteristics of American youth between the late 1970s and the late 1990s, with a focus on characteristics that matter for labor market success. The current generation is more skilled than the previous one. Blacks and Hispanics have gained relative to whites, and women have gained relative to men. However, the skill distribution has widened overall. Shifts in parental education generate many of the observed changes. We also provide speculative estimates suggesting that if recent trends in technology and the supply of human capital continue, wage inequality will increase substantially by 2025.

AI snapshot

Written from the paper’s full text.

TL;DR

Altonji, Bharadwaj, and Lange compare the skill characteristics of the NLSY79 and NLSY97 cohorts and use reweighting to ask how the later cohort’s characteristics would affect adult wages and employment under the NLSY79 wage structure. They find the 1997 cohort is more skilled overall, with narrower race and gender skill gaps but a wider overall skill distribution, and that changes in parental education account for much of the shift.

Key findings

  • The 1997 cohort is stronger on most skill dimensions: mean AFQT rises from 42.35 to 44.19, highest grade completed by age 22 rises from 12.64 to 13.02, mother’s education rises from 11.77 to 12.71 years, and father’s education rises from 12.09 to 12.81 years. The share living with both biological parents at age 14 falls from 75.23% to 52.98%.
  • The implied mean log wage increase from the skill shift is 0.058 log points in the full model, with the paper reporting gains of about 5% below the 10th percentile, 5–6% between the 20th and 85th percentiles, and about 10% in the top decile; the abstract summarizes this as about a 5% wage increase for whites and more for minorities.
  • Employment effects are small overall: the full-population counterfactual change is about −0.001, while black men’s employment rises from 84.9% to 87% and Hispanic men and women also show gains.
  • Race and gender gaps narrow: the white–black male mean log wage gap falls from 32.2% to 28.2%, the white–Hispanic male gap from 18.2% to 14.3%, and the male–female gap from 27.82% to 26%. Parental education is central: the marginal effect of family background is 0.055 log points, versus 0.002 for AFQT, 0.018 for highest grade completed, and −0.009 for work transition; the paper says more than two-thirds of the skill shift is linked to parental education.

Methods

  • The paper compares NLSY79 respondents (aged 14–22 in 1979, birth years 1957–1964) with NLSY97 respondents (aged 12–16 in 1997, birth years 1980–1984), using NLSY79 adult outcomes from 1998–2004, when that cohort was 39–47 years old.
  • It reweights the NLSY79 to match the NLSY97 distribution of characteristics using a DiNardo–Fortin–Lemieux propensity-score procedure, and also reports Blinder–Oaxaca/regression decompositions. The skill vector includes race, gender, parental education and presence, AFQT, highest grade completed/enrollment, and school-to-work transition variables.
  • Wages are log real hourly wages standardized to year 2002 and 23 years of potential experience. The main sample retains 8,822 NLSY79 respondents (91.32% of eligible) and 4,882 NLSY97 respondents (68.30% of eligible) with valid AFQT scores; a second sample includes those with missing AFQT. Propensity weights are capped at 10, and standard errors are bootstrapped.

Limitations

  • Stated: the NLSY97 base-year sample may be nonrepresentative; attrition and AFQT nonresponse are substantial; the attrition and AFQT weights do not correct for correlation between attrition and unobservables affecting wages or employment conditional on observables; the analysis assumes NLSY79 attrition after age 22 is random; Assumption A1 assumes the distribution of unobservables conditional on skills is the same across cohorts; decompositions are sensitive to variable order; AFQT comparability across paper-and-pencil and computer formats and across test-taking ages is imperfect; and the authors call for supplementing results with NAEP and CPS data.
  • Evident from the text: NLSY97 retention is much lower than NLSY79 retention (68.30% vs. 91.32%), and AFQT nonresponse is much larger in NLSY97 (about 20% vs. 6%); uncapped propensity weights reach 90.8, with 37 individuals above 10 and 8 above 20, which is why the authors cap weights at 10.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1086/666536

Cites

44 of 44 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 8 cited sources checked · 0 supported · 0 flagged

  • Findingown data
    § 3. DATA

    Whites are more likely to leave the sample prior to age 22 than are blacks.

    “For example, whites are more likely to leave the sample prior to age 22 than are blacks.”
    Quote in the paperNot a substantial claim
  • Findingnone given
    § 1. INTRODUCTION

    The fraction of individuals who lived with both parents at age 14 declined substantially between 1979 and 1997.

    “However, the fraction of individuals who lived with both parents at age 14 declined substantially between 1979 and 1997.”
    Quote in the paperNot a substantial claim
  • Findingown data
    § 3. DATA

    Among the 1979 cohort, only a slim majority (50.21%) of black youths lives with both parents.

    “Among the 1979 cohort, only a slim majority (50.21%) of black youths lives with both parents.”
    Quote in the paperNot a substantial claim
  • Findingown data
    § 3. DATA

    The gaps decreased along some but not all skill dimensions between 1979 and 1997.

    “The gaps decreased along some but not all skill dimensions between 1979 and 1997.”
    Quote in the paperNot a substantial claim
  • Findingown data
    § 3. DATA

    There are relatively small differences between the full sample and the stayers in the 1997 cohort in spite of the higher 1997 attrition rate.

    “We also find relatively small differences between the full sample and the stayers in the 1997 cohort in spite of the higher 1997 attrition rate.”
    Quote in the paperNot a substantial claim
  • Classified as probably not a substantial claim. 47
  • May combine two claims. 74
  • The quote only partly states this claim. 23

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data
    § 3. DATA

    If the analysis does not condition on observing the AFQT score, about 85% of the base sample is retained.

    “If we do not condition on observing the AFQT score, we retain about 85% of the base sample.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § 3. DATA

    By 1997 white females exceed males in both of these skill dimensions and have an education advantage of one half year.

    “Both groups show improvement, but by 1997 white females exceed males in both of these skill dimensions and have an education advantage of one half year.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data
    § 3. DATA

    Non-participation in the ASVAB eliminates an additional 379 NLSY79 and 1,132 NLSY97 respondents.

    “Non-participation in the ASVAB eliminates an additional 379 (1,132) respondents.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § 5. CHANGES IN THE SKILL DISTRIBUTION BETWEEN NLSY79 AND NLSY97

    This is the net result of an increase of 0.005 for males and a decrease of 0.007 for women.

    “This is the net result of an increase of 0.005 for males and a decrease of 0.007 for women.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § 3. DATA

    Black females had very large gains in AFQT scores (8.45 points) and also gained about 0.45 years in highest grade completed.

    “particularly black females, had very large gains in AFQT scores (8.45 points) and also gained about 0.45 years in highest grade completed.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § 3. DATA

    White females stagnated along the seamless transition dimension, while black and Hispanic females experienced substantial gains.

    “On the other hand, white females stagnated along this dimension, while black and Hispanic females experienced substantial gains.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § 3. DATA

    Among whites and blacks the share of individuals with HS diplomas stayed roughly constant, while the fraction of those with more than 14 years of education by age 22 increased substantially for all subgroups.

    “Rather, we observe that among whites and blacks the share of individuals with HS diplomas stayed roughly constant, while the fraction of those with more than 14 years of education by age 22 increased substantially for all subgroups.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingreasoning
    § 3. DATA

    The final sample retains 8,822 NLSY79 respondents and 4,882 NLSY97 respondents, constituting 91.32% of the total eligible NLSY79 sample but only 68.30% for the NLSY97.

    “We retain a total of 8,822 NLSY79 respondents and 4,882 NLSY97 respondents, which constitute 91.32% of the total eligible sample in the NLSY79 but only 68.30% for the NLSY97.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data
    § 5. CHANGES IN THE SKILL DISTRIBUTION BETWEEN NLSY79 AND NLSY97

    The bulge on the left tail comes from the fact that wages are bound below at $3 in 2003 values.

    “the bulge on the left tail comes from the fact the wages are bound below at $3 in 2003 values.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § 4. ESTIMATION OF PROPENSITY SCORES

    If the weights are not capped, estimates of the tails of the counterfactual wage distributions are implausible, highly sensitive to variations in the model specification, and have very large bootstrapped standard errors.

    “If we do not cap the weights, then we do get some estimates of the tails of the counterfactual wage distributions that are implausible, are highly sensitive to variations in the model specification and have very large bootstrapped standard errors.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § 3. DATA

    Hispanic females and black females gain relative to their male counterparts.

    “Hispanic females and black females gain relative to their males counterparts”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § 3. DATA

    Whites in the 1997 cohort completed about 1/2 of a year more than the 1979 cohort, while blacks only gained about 1/4th of a year along this dimension.

    “While whites in the 1997 cohort completed about 1/2 of a year more than the 1979 cohort, blacks only gained about 1/4th of a year along this dimension.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § 3. DATA

    The differences between the full population and stayers in the means of mother's education and father's education are only -0.05 and -0.03 years respectively.

    “For instance, the differences between the full population and stayers in the means of mother's education and father's education are only -0.05 and -0.03 years respectively.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data
    § 4. ESTIMATION OF PROPENSITY SCORES

    One individual (a black female with 16 years of education and an AFQT score of 87 who was enrolled in school at age 22 and did not live with either biological parent at 14) has a propensity weight of 88.

    “One individual, (a black female with 16 years of education and an AFQT score of 87 who was enrolled in school at age 22 and did not live with either biological parent at 14) has a propensity weight of 88.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § 3. DATA

    In 1997, those with valid AFQT scores have 0.67 years more education by age 22 than those without valid scores but only 0.11 years more education than the full population.

    “For instance, those with valid AFQT scores in 1997 have 0.67 years more education by age 22 than those without valid scores but only 0.11 years more education than the full population.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § 3. DATA

    Those who have AFQT scores have higher education levels by age 22 and have better educated parents.

    “Furthermore, those who have AFQT scores have higher education levels by age 22 and have better educated parents.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § 3. DATA

    Across the two cohorts, the percentage of youths living with both biological parents at age 14 declines by 21.99 percentage points for whites, 24.87 percentage points for blacks, and 9.49 percentage points for Hispanics.

    “Across the two cohorts, the percentage of youths living with both biological parents at age 14 declines by 21.99%-points for whites, 24.87%-points for blacks and 9.49%-points for Hispanics.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § 3. DATA

    By 1997 Hispanic females close the gap with blacks and Hispanic males surpass black males.

    “By 1997 Hispanic females close the gap with blacks and Hispanic males surpass black males.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § 3. DATA

    The differences in mean characteristics by AFQT availability are not uniformly larger in the NLSY97 than NLSY79, but some differences between those with or without an AFQT-score are sizable.

    “The differences in the mean characteristics by AFQT availability are not uniformly larger in the NLSY97 than NLSY79, but some of the differences between those with or without an AFQT-score are sizable.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § 3. DATA

    The decline of the traditional family structure is dramatic for whites and blacks and large for Hispanics as well.

    “The decline of the traditional family structure is dramatic for whites and blacks and large for Hispanics as well.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § 3. DATA

    Black males gained 5.6 points on the AFQT but only 0.17 years of education.

    “Black males gained 5.6 points on the AFQT but only 0.17 years of education.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § 3. DATA

    Parental education gains have been larger for minorities, reflected in larger observed increases in parental education of blacks and Hispanics relative to whites.

    “These gains have been larger for minorities, which is reflected in the larger observed increases in parental education of blacks and Hispanics relative to whites.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § 3. DATA

    Hispanics gain substantially along the schooling dimension in both absolute terms and in comparison to whites.

    “Hispanics also gain substantially along the schooling dimension in both absolute terms and in comparison to whites.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § 3. DATA

    AFQT scores rose on average by 1.48 points for white, 5.6 for black, and 4.88 for Hispanic males.

    “For example, AFQT scores rose on average by 1.48 points for white, 5.6 for black, and 4.88 for Hispanic males.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § 3. DATA

    In 1979 white males and females are about equal in average education and AFQT scores.

    “In 1979 white males and females are about equal in average education and AFQT scores.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim. May combine two claims.

    Findingown data
    § 4. ESTIMATION OF PROPENSITY SCORES

    There are 37 individuals with weights above 10 and 8 with weights above 20.

    “There are 37 individuals with weights above 10 and 8 with weights above 20.”
    Quote in the paperSubstantial claim

  • The quote only partly states this claim.

    Findingown data
    § Table 3.3 Summary Statistics

    For the full sample, the proportion living with father only at age 14 was 2.92 percentage points higher in 1997 than in 1979 (SE 0.319, significant at 1%).

    “2.92 (0.319)***”
    Quote in the paperNot a substantial claim

  • May combine two claims. The quote only partly states this claim.

    Findingown data
    § Table 3.3 Summary Statistics

    For white males, the mean marginal effect of family background on the 1979 log wage distribution is 0.057 (SE 0.0245), significant at the 5% level.

    “0.057 (0.0245)**”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § 6. IDENTIFYING THE CONTRIBUTION OF SUBSETS OF VARIABLES TO DIFFERENCES BETWEEN THE 1979 AND 1997 WAGE DISTRIBUTIONS

    As for the aggregate population, the marginal effects of parental education are strikingly large and consistently exceed the partial effects.

    “As for the aggregate population, we find that the marginal effects of parental education are strikingly large and consistently exceed the partial effects.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data
    § Table 3.3 Summary Statistics

    In the OLS regression of log wages, the coefficient on White Female is -0.175 (standard error 0.0180).

    “-0.175 (0.0180)***”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § 5. CHANGES IN THE SKILL DISTRIBUTION BETWEEN NLSY79 AND NLSY97

    Black females are expected to gain over the entire distribution, with gains greater than 10% for about two-thirds of the distribution.

    “Black females likewise are expected to gain over the entire distribution, with gains greater than 10% for about two-thirds of the distribution.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingnone given
    § 6. IDENTIFYING THE CONTRIBUTION OF SUBSETS OF VARIABLES TO DIFFERENCES BETWEEN THE 1979 AND 1997 WAGE DISTRIBUTIONS

    Cognitive tests scores of black males improved rapidly, while schooling of black males did not.

    “Cognitive tests scores of black males improved rapidly, while schooling of black males did not.”
    Quote in the paperNot a substantial claim

  • The quote only partly states this claim.

    Findingown data
    § Table 3.3 Summary Statistics

    For the full sample, the proportion enrolled at age 22 was 9.57 percentage points higher in 1997 than in 1979 (SE 0.714, significant at 1%).

    “9.57 (0.714)***”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § Table 3.3 Summary Statistics

    In the counterfactual LFP table, for Hispanic males, the counterfactual LFP rate exceeds the observed by 0.022 (SE 0.0078), significant at 1%.

    “| Hispanic | 0.909 (0.0093) | 0.902 (0.0096) | 0.022 (0.0078)***”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingreasoning
    § 6. IDENTIFYING THE CONTRIBUTION OF SUBSETS OF VARIABLES TO DIFFERENCES BETWEEN THE 1979 AND 1997 WAGE DISTRIBUTIONS

    The marginal effect of the shift in HGC (with AFQT excluded) is 0.017 and once HGC is included, the marginal contribution of AFQT is negative (-0.008).

    “In contrast, the marginal effect of the shift in HGC (with AFQT excluded) is 0.017 and once HGC is included, the marginal contribution of AFQT is negative (-0.008).”
    Quote in the paperSubstantial claim

  • May combine two claims. The quote only partly states this claim.

    Findingown data
    § Table 3.3 Summary Statistics

    In Table 6.1A, family background contributes 0.055 log points (SE 0.0091) to the mean difference between the 1979 and 1997 wage distributions, significant at 1%.

    “| Mean | 7.265 (0.0082) | -0.008 (0.0020)*** | 0.055 (0.0091)***”
    Quote in the paperNot a substantial claim

  • The quote only partly states this claim.

    Findingown data
    § Table 3.3 Summary Statistics

    For white females, the marginal effect of highest grade on the 1979 log wage at the 50th percentile is 0.032 (SE 0.0115), significant at the 1% level.

    “0.032 (0.0115)***”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § 7. CONCLUSION

    Significant skill gaps remain, but blacks and Hispanics have narrowed the gap in skills relative to whites.

    “Significant skill gaps remain, but blacks and Hispanics have narrowed the gap in skills relative to whites.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given
    § 10. APPENDIX C: FAMILY STRUCTURE IN THE NLSY SAMPLES AND IN THE CENSUS

    A small number of cases in both cohorts are lost due to a death prior to age 22.

    “Note that a small number of cases in both cohorts are lost due to a death prior to age 22.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingreasoning
    § 6. IDENTIFYING THE CONTRIBUTION OF SUBSETS OF VARIABLES TO DIFFERENCES BETWEEN THE 1979 AND 1997 WAGE DISTRIBUTIONS

    For the regression decomposition, the marginal effect of HGC, AFQT, and school-to-work combine to -0.002 (DFL:0.012) which is much smaller than the sum of the partial effects, which is 0.022.

    “For the regression decomposition, the marginal effect of HGC, AFQT, and school-to-work combine to -0.002 (DFL:0.012) which is much smaller than the sum of the partial effects, which is 0.022.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § Table 3.3 Summary Statistics

    For white males, the marginal effect of family background on the 1979 log wage at the 75th percentile is 0.048 (SE 0.0234), significant at the 5% level.

    “0.048 (0.0234)**”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data
    § Table 3.3 Summary Statistics

    In the OLS regression of log wages, the coefficient on White Male is 0.195 (standard error 0.0184).

    “0.195 (0.0184)***”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims. The quote only partly states this claim.

    Findingown data
    § Table 3.3 Summary Statistics

    The OLS regression of log wages has an R-squared of 0.197, F(18, 23865) = 185.48, and N = 23884.

    “R-sq = 0.197, F (18, 23865) = 185.48, N = 23884.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given
    § 6. IDENTIFYING THE CONTRIBUTION OF SUBSETS OF VARIABLES TO DIFFERENCES BETWEEN THE 1979 AND 1997 WAGE DISTRIBUTIONS

    The DFL estimate is -0.004.

    “The DFL estimate is -0.004.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingreasoning
    § 6. IDENTIFYING THE CONTRIBUTION OF SUBSETS OF VARIABLES TO DIFFERENCES BETWEEN THE 1979 AND 1997 WAGE DISTRIBUTIONS

    Parental education has not only a large direct impact on the change in mean log wages, but also a large indirect impact through other variables.

    “the parental education has not only a large direct impact on the change in mean log wages, but also a large indirect impact through other variables.”
    Quote in the paperSubstantial claim

  • May combine two claims. The quote only partly states this claim.

    Findingown data
    § Table 3.3 Summary Statistics

    In the counterfactual LFP table, for black males, the counterfactual LFP rate based on 1997 characteristics exceeds the observed 1979 LFP rate by 0.027 (SE 0.0097), significant at 1%.

    “| Black | 0.854 (0.0089) | 0.849 (0.0090) | 0.027 (0.0097)***”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § 6. IDENTIFYING THE CONTRIBUTION OF SUBSETS OF VARIABLES TO DIFFERENCES BETWEEN THE 1979 AND 1997 WAGE DISTRIBUTIONS

    The difference is mainly due to the decline in the number of 14 year olds living with their parents and the decline in the work-transition variables, while HGC and AFQT together contribute a negligible amount (-0.003).

    “The difference is mainly due to the decline in the number of 14 year olds living with their parents and the decline in the work-transition variables, while HGC and AFQT together contribute a negligible amount (-0.003).”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). The quote only partly states this claim.

    Findingown data
    § Table 3.3 Summary Statistics

    The change in mean AFQT between 1979 and 1997 is 2.367.

    “2.367”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims. The quote only partly states this claim.

    Findingown data
    § Table 3.3 Summary Statistics

    In Table 6.1A, AFQT contributes 0.002 log points (SE 0.0041) to the mean difference, not statistically significant.

    “| Mean | 7.265 (0.0082) | -0.008 (0.0020)*** | 0.055 (0.0091)*** | 0.002 (0.0041)”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § 10. APPENDIX C: FAMILY STRUCTURE IN THE NLSY SAMPLES AND IN THE CENSUS

    The results from the NLSY and from the Census are roughly consistent, even though in the NLSY we typically find more individuals living with two (social) parents.

    “The results from the NLSY and from the Census are roughly consistent, even though in the NLSY we typically find more individuals living with two (social) parents.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § Table 3.3 Summary Statistics

    In NLSY79, 83.34% of youth lived with both mother and father at age 14, compared to 68.16% in NLSY97.

    “| Social Mother and Father | Mother and Father Mother only Father only | 75.1 | 70.61 | 66.88 | 83.34 | 68.16 |”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § Table 3.3 Summary Statistics

    In NLSY79, 75.16% of youth lived with both biological parents at age 14, compared to 52.75% in NLSY97.

    “| Biological Parents only | Mother and Father Mother only Father only | .. | .. | .. | 75.16 | 52.75 |”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data
    § Table 3.3 Summary Statistics

    In the regression decomposition for females, the overall change value is 6.425.

    “| Overall Change | 6.425 | 0.039 | 0.047 (0.0093)*** | 0.047 (0.0093)*** | | (6) | (7) 0.097 (0.0091)*** | (8) 0.097 (0.0040)*** | (9”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § 6. IDENTIFYING THE CONTRIBUTION OF SUBSETS OF VARIABLES TO DIFFERENCES BETWEEN THE 1979 AND 1997 WAGE DISTRIBUTIONS

    The combined marginal effect of HGC and AFQT is 0.055 for black females and 0.06 for Hispanic females, which are about as large as the marginal effects of parental education.

    “The combined marginal effect of HGC and AFQT is 0.055 for black females and 0.06 for Hispanic females, which are about as large as the marginal effects of parental education.”
    Quote in the paperSubstantial claim

  • The quote only partly states this claim.

    Findingown data
    § Table 3.3 Summary Statistics

    For the full sample, the proportion with highest grade completed >=14 at age 22 was 8.68 percentage points higher in 1997 than in 1979 (SE 0.795, significant at 1%).

    “8.68 (0.795)***”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § Table 3.3 Summary Statistics

    For black females, the sum of marginal effects of family background, AFQT, highest grade, and work transition on the 1979 log wage at the 75th percentile is 0.15 (SE 0.0432), significant at the 1% level.

    “0.15 (0.0432)***”
    Quote in the paperNot a substantial claim

  • May combine two claims. The quote only partly states this claim.

    Findingown data
    § Table 3.3 Summary Statistics

    In Table 5.2, the counterfactual wage distribution based on 1997 cohort characteristics (Model 4.6) exceeds the actual 1979 wage distribution at the 75th percentile by 0.049 log points (SE 0.0140), significant at 1%.

    “| 75% | 7.665 (0.0086) 8.042 | 7.663 (0.0082) 8.039 | 0.0”
    Quote in the paperNot a substantial claim

  • The quote only partly states this claim.

    Findingown data
    § Table 3.3 Summary Statistics

    For white males, the mean marginal effect of highest grade on the 1979 log wage distribution is 0.088 (SE 0.0331), significant at the 5% level.

    “0.088 (0.0331)**”
    Quote in the paperNot a substantial claim

  • May combine two claims. The quote only partly states this claim.

    Findingown data
    § Table 3.3 Summary Statistics

    In Census data, the percentage of youth living with both mother and father at age 14 was 75.1% in 1980, 70.61% in 1990, and 66.88% in 2000.

    “| Social Mother and Father | Mother and Father Mother only Father only | 75.1 | 70.61 | 66.88 | 83.34 | 68.16 |”
    Quote in the paperNot a substantial claim

  • The quote only partly states this claim.

    Findingown data
    § Table 3.3 Summary Statistics

    In the OLS regression of log wages, the coefficient on graduating late is -0.208 (standard error 0.0257).

    “-0.208 (0.0257)***”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § Table 3.3 Summary Statistics

    For black females, the marginal effect of AFQT on the 1979 log wage at the 75th percentile is 0.063 (SE 0.0253), significant at the 5% level.

    “0.063 (0.0253)**”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data
    § 10. APPENDIX C: FAMILY STRUCTURE IN THE NLSY SAMPLES AND IN THE CENSUS

    The inability to distinguish one versus two adoptive parents in 1997 affects only 103 cases.

    “This problem arises only for a small number of cases (103)”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § 6. IDENTIFYING THE CONTRIBUTION OF SUBSETS OF VARIABLES TO DIFFERENCES BETWEEN THE 1979 AND 1997 WAGE DISTRIBUTIONS

    For white males the total increase in log wages is 0.028, while the marginal effect of parental education suggests an increase of 0.072.

    “For white males the total increase is log wages is 0.028, while the marginal effect of parental education suggests an increase of 0.072.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § 6. IDENTIFYING THE CONTRIBUTION OF SUBSETS OF VARIABLES TO DIFFERENCES BETWEEN THE 1979 AND 1997 WAGE DISTRIBUTIONS

    This pattern holds for white males, white females, and Hispanic males.

    “This pattern holds white males, white females, and Hispanic males.”
    Quote in the paperNot a substantial claim

  • May combine two claims. The quote only partly states this claim.

    Findingown data
    § Table 3.3 Summary Statistics

    In Table 5.2, the counterfactual wage distribution based on 1997 cohort characteristics (Model 4.6) exceeds the actual 1979 wage distribution at the 50th percentile by 0.063 log points (SE 0.0131), significant at 1%.

    “| 50% | 7.268 (0.0089) | 7.265 (0.0089) | 0.063 (0.”
    Quote in the paperNot a substantial claim

  • The quote only partly states this claim.

    Findingown data
    § Table 3.3 Summary Statistics

    For the full sample, the proportion living with both mother and father at age 14 was 22.25 percentage points lower in 1997 than in 1979 (SE 0.731, significant at 1%).

    “-22.25 (0.731)***”
    Quote in the paperNot a substantial claim

  • May combine two claims. The quote only partly states this claim.

    Findingown data
    § Table 3.3 Summary Statistics

    For white females, the mean marginal effect of highest grade on the 1979 log wage distribution is 0.025 (SE 0.0086), significant at the 1% level.

    “0.025 (0.0086)***”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § 6. IDENTIFYING THE CONTRIBUTION OF SUBSETS OF VARIABLES TO DIFFERENCES BETWEEN THE 1979 AND 1997 WAGE DISTRIBUTIONS

    For black and Hispanic females the observed marginal effects of parental education are as large as those observed for whites of both gender and for minority males, but black and Hispanic females have also made large gains in HGC and AFQT conditional on family background.

    “For black and Hispanic females the observed marginal effects of parental education are as large as those observed for whites of both gender and for minority males, but black and Hispanic females have also made large gains in HGC and AFQT conditional on family background.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § Table 3.3 Summary Statistics

    For black females, the marginal effect of AFQT on the 1979 log wage at the 50th percentile is 0.043 (SE 0.0258), significant at the 10% level.

    “0.043 (0.0258)*”
    Quote in the paperNot a substantial claim

  • May combine two claims. The quote only partly states this claim.

    Findingown data
    § Table 3.3 Summary Statistics

    In Table 5.2, the counterfactual wage distribution based on 1997 cohort characteristics (Model 4.6) exceeds the actual 1979 wage distribution at the 25th percentile by 0.049 log points (SE 0.0131), significant at 1%.

    “| 25% | 6.846 (0.0089) | 6.841 (0.0086) | 0.049 (0.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingnone given
    § 6. IDENTIFYING THE CONTRIBUTION OF SUBSETS OF VARIABLES TO DIFFERENCES BETWEEN THE 1979 AND 1997 WAGE DISTRIBUTIONS

    The means of both HGC and AFQT increase between 1979 and 1997.

    “The means of both HGC and AFQT increase between 1979 and 1997.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § 5. CHANGES IN THE SKILL DISTRIBUTION BETWEEN NLSY79 AND NLSY97

    Gains for white women are small near the bottom of the distribution but increase along the entire distribution.

    “Gains for white women are small near the bottom of the distribution but increase along the entire distribution.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § Table 3.3 Summary Statistics

    In the counterfactual LFP table, for Hispanic females, the counterfactual LFP rate exceeds the observed by 0.024 (SE 0.0138), significant at 10%.

    “| Hispanic | 0.821 (0.0117) | 0.819 (0.0117) | 0.024 (0.0138)*”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § Table 3.3 Summary Statistics

    For black females, the marginal effect of AFQT on the 1979 log wage at the 90th percentile is 0.058 (SE 0.0293), significant at the 10% level.

    “0.058 (0.0293)*”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). The quote only partly states this claim.

    Findingown data
    § Table 3.3 Summary Statistics

    The change in mean highest grade completed between 1979 and 1997 is 0.443.

    “0.443”
    Quote in the paperNot a substantial claim

  • The quote only partly states this claim.

    Findingown data
    § Table 3.3 Summary Statistics

    For the full sample, the proportion living with mother only at age 14 was 17.09 percentage points higher in 1997 than in 1979 (SE 0.6803, significant at 1%).

    “17.09 (0.6803)***”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data
    § Table 3.3 Summary Statistics

    In the OLS regression of log wages, the coefficient on Black Male is 0.028 (standard error 0.0191).

    “0.028 (0.0191)”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § Table 3.3 Summary Statistics

    For black females, the sum of marginal effects of family background, AFQT, highest grade, and work transition on the 1979 log wage at the 50th percentile is 0.12 (SE 0.0421), significant at the 1% level.

    “0.12 (0.0421)***”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data
    § Table 3.3 Summary Statistics

    In the OLS regression of log wages, the coefficient on Black Female is -0.131 (standard error 0.0180).

    “-0.131 (0.0180)***”
    Quote in the paperNot a substantial claim

  • May combine two claims. The quote only partly states this claim.

    Findingown data
    § Table 3.3 Summary Statistics

    In Table 5.2, the counterfactual wage distribution based on 1997 cohort characteristics (Model 4.6) exceeds the actual 1979 wage distribution at the mean by 0.058 log points (SE 0.0121), significant at 1%.

    “| Mean | 7.265 (0.0082) | 7.261 (0.0081) | 0.058 (0.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § Table 3.3 Summary Statistics

    For white males, the marginal effect of family background on the 1979 log wage at the 10th percentile is 0.049 (SE 0.0230), significant at the 5% level.

    “0.049 (0.0230)**”
    Quote in the paperNot a substantial claim

  • May combine two claims. The quote only partly states this claim.

    Findingown data
    § Table 3.3 Summary Statistics

    For white females, the mean sum of marginal effects of family background, AFQT, highest grade, and work transition on the 1979 log wage distribution is 0.053 (SE 0.0198), significant at the 1% level.

    “0.053 (0.0198)***”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Findingnone given
    § 7. CONCLUSION

    A cursory look at the NAEP scores for the relevant years indicates that they move in the same direction as the AFQT scores.

    “A cursory look at the NAEP scores for the relevant years indicate that they move in the same direction as the AFQT scores”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims. The quote only partly states this claim.

    Findingown data
    § Table 3.3 Summary Statistics

    In Table 6.1A, highest grade contributes 0.018 log points (SE 0.0047) to the mean difference, significant at 1%.

    “| Mean | 7.265 (0.0082) | -0.008 (0.0020)*** | 0.055 (0.0091)*** | 0.002 (0.0041) | 0.018 (0.0047)***”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § Table 3.3 Summary Statistics

    For black males, the mean sum of marginal effects of family background, AFQT, highest grade, and work transition on the 1979 log wage distribution is 0.051 (SE 0.0241), significant at the 5% level.

    “0.051 (0.0241)**”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Study design and conductown data
    § 5. CHANGES IN THE SKILL DISTRIBUTION BETWEEN NLSY79 AND NLSY97

    The employment results by race and gender are somewhat sensitive to the specification, but the findings for the full population and for males and females separately are robust to the choice of specification.

    “The employment results by race and gender are somewhat sensitive to the specification, but the findings for the full population and for males and females separately are robust to the choice of specification.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim. May combine two claims.

    Study design and conductown data
    § 3. DATA

    The study addresses attrition and non-response to the AFQT by constructing two alternative sets of weights: one adjusting only for attrition by age 22 (used with sample 2) and one adjusting for both attrition and missing AFQT responses (used with sample 1).

    “We address the problem of attrition and non-response to the AFQT by constructing two alternative sets of weights. The first adjusts only for attrition by age 22 and is used with”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data
    § 5. CHANGES IN THE SKILL DISTRIBUTION BETWEEN NLSY79 AND NLSY97

    Generally results are quite robust across specifications.

    “Generally results are quite robust across specifications.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Premisereasoning
    § 1. INTRODUCTION

    The DFL procedure is a generalization of the Blinder-Oaxaca decomposition procedure.

    “The DFL procedure can be interpreted as a generalization of the Blinder-Oaxaca decomposition procedure, since it is not based on an outcome equation that is linear in parameters.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Dinardo, Fortin, and Lemieux (1995) source

  • Classified as probably not a substantial claim.

    Study design and conductown data
    § 3. DATA

    Both sets of weights are estimated using probit specifications based on race, sex, parental presence at age 14, parental education, birth-year indicators, urban and SMSA residence status, and variables describing attitude towards the interview.

    “Both sets of weights are estimated using probit specifications based on race, sex, parental presence at age 14, parental education, birth-year indicators, urban and SMSA residence status as well as variables describing the attitude towards the interview.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisereasoning
    § 3. DATA

    The NLSY79 includes immigrants who arrive after age 16 while the NLSY97 does not.

    “Consequently, the NLSY79 includes immigrants who arrive after age 16 while the NLSY97 does not.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Study design and conductnone given
    § 4. ESTIMATION OF PROPENSITY SCORES

    The study estimates the propensity score using Probits based on various sets of skill measures.

    “We estimate the propensity score using Probits based on various sets of skill measures.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: The error term in the latent index function is normal.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data
    § 3. DATA

    The result on the magnitude of the fraction of individuals living outside traditional family structures was confirmed using Census data.

    “We were sufficiently surprised by the magnitude of the fraction of individuals living outside of traditional family structures that we have confirmed this result using Census data.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given
    § 4. ESTIMATION OF PROPENSITY SCORES

    All specifications estimated on sample 1 are estimated with weights adjusting for both attrition by age 22 and AFQT non-response using the same variables.

    “All specifications estimated on sample 1, which in addition excludes individuals without valid AFQT scores, are estimated with weights adjusting for both attrition by age 22 and AFQTnon-response using the same variables.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Study design and conductnone given
    § 4. ESTIMATION OF PROPENSITY SCORES

    The models are estimated by MLE-Probit after pooling the data from both samples.

    “We estimate the models by MLE-Probit after pooling the data from both samples.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: The error term in the latent index function is normal.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductown data
    § 5. CHANGES IN THE SKILL DISTRIBUTION BETWEEN NLSY79 AND NLSY97

    Results from the tails are consistent with the findings but noisy.

    “Results from the tails are consistent with our findings here, but noisy.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Study design and conductreasoning
    § 3. DATA

    The study exploits the overlap in test-taking age across both cohorts by applying an equipercentile procedure on each cohort with the population of test takers who were 16 year olds when taking the test.

    “We exploit the overlap in the test-taking age across both cohorts by applying a equipercentile procedure on each cohort with the population of test takers who were 16 year olds when taking the test.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Premisereasoning
    § 1. INTRODUCTION

    The DFL approach does not estimate parameters that relate outcomes to characteristics and that can potentially be interpreted.

    “The main disadvantage is that the DFL approach does not estimate parameters that relate outcomes to characteristics and that can potentially be interpreted.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Dinardo, Fortin, and Lemieux (1995) source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given
    § 4. ESTIMATION OF PROPENSITY SCORES

    The base year sample weights are adjusted for attrition by age 22 based on observable characteristics in both cohorts for specifications estimated on sample 2.

    “The base year sample weights are adjusted for attrition by age 22 based on observable characteristics in both cohorts for specifications estimated on sample 2, which includes all individuals observed at age 22.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationown data
    § 3. DATA

    The finding that the coefficients of the regression of age-adjusted test scores on individual characteristics do not vary with test-taking age supports the assumption that the AFQT-score is not informative about skills of an individual during adulthood once we condition on the rank of an individual within his/her age group.

    “The finding that the coefficients of the regression of age-adjusted test scores on individual characteristics do not vary with testtaking age supports this assumption.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: The AFQT-score is not informative about skills of an individual during adulthood once we condition on the rank of an individual within his/her age group.

  • Probably not a substantial claim: may not pass S4 (based).

    Premisenone given
    § 3. DATA

    The NLSY79 and NLSY97 surveys vary sufficiently to pose challenges to achieving comparability.

    “Nevertheless, the surveys vary sufficiently to pose challenges to achieving comparability.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § 3. DATA

    Heckman and Lafontaine (2007) show using multiple data sources that high school graduation rates have stagnated over the last 20 years.

    “They show using multiple data sources that high school graduation rates have stagnated over the last 20 years”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Heckman and Lafontaine (2007) source

      Checked before — its earlier verdict is held back until a new check runs.

  • Classified as probably not a substantial claim.

    Study design and conductnone given
    § 4. ESTIMATION OF PROPENSITY SCORES

    The base year sample weights provided by the NLSY for both surveys are used in the estimation.

    “The base year sample weights provided by the NLSY for both surveys are used in the estimation.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Study design and conductnone given
    § 4. ESTIMATION OF PROPENSITY SCORES

    The study uses flexible functional forms for the latent index of the Probit model so as not to restrict the changes in the skill distributions across cohorts unduly.

    “We use flexible functional forms for the latent index of the Probit model so as not to restrict the changes in the skill distributions across cohorts unduly.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationreasoning
    § 3. DATA

    Attrition after age 22 in the NLSY79 affects both the actual wage distribution and the counterfactual one, so it probably has only a second order effect on the difference between the two.

    “We choose not to construct an additional weight to adjust for this because attrition after age 22 in the NLY79 affects both the actual wage distribution and the counterfactual one. Consequently, it probably has only a second order affect on the difference between the two, which is our main interest.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: Attrition from NLSY79 after age 22 is random.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationcitation
    § 3. DATA

    The results for the two cohorts are broadly consistent with those of Heckman and Lafontaine (2007).

    “Our results for the two cohorts are broadly consistent with those of Heckman and Lafontaine (2007).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Heckman and Lafontaine (2007) source

      Checked before — its earlier verdict is held back until a new check runs.

  • Classified as probably not a substantial claim. May combine two claims.

    Study design and conductown data
    § 3. DATA

    The attrition models are estimated for the NLSY79 and the NLSY97 separately and the weights are applied throughout the analysis.

    “We estimate these attrition models for the NLSY79 and the NLSY97 separately and apply the weights throughout the analysis as applicable.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Study design and conductreasoning
    § 3. DATA

    The summary statistics in Table 3.3 are computed using NLSY cross-sectional weights adjusted for attrition by age 22 and AFQT non-response.

    “These statistics are supposed to describe the population in 1979 and 1997 and are therefore computed using the NLSY cross-sectional weights adjusted for attrition by age 22 and AFQT non-response.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim. May combine two claims.

    Study design and conductown data
    § 10. APPENDIX C: FAMILY STRUCTURE IN THE NLSY SAMPLES AND IN THE CENSUS

    The family structure question in 1979 is retrospective and refers to family structure at age 14, while the family structure measures for the 1997 cohort are obtained during the screening interview and refer to age at screening interview (12-16 years of age).

    “Note that the family structure question in 1979 is retrospective and refers to family structure at age 14. The family structure measures for the 1997 cohort are obtained during the screening interview and refer to the age at the screening interview, ie 12-16 years of age.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductreasoning
    § 10. APPENDIX C: FAMILY STRUCTURE IN THE NLSY SAMPLES AND IN THE CENSUS

    In the 1979 cohort, everyone is expected to be observed at age 22; in the 1997 cohort, since the last year of interview is 2005, only people born on or before 1983 are expected to reach age 22 in the data.

    “In the 1979 cohort we expect to observe everyone at age 22. in the 1997 cohort, since the last year of interview is 2005, we only expect people born on or before 1983 to reach the age of 22 in the data.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data
    § 10. APPENDIX C: FAMILY STRUCTURE IN THE NLSY SAMPLES AND IN THE CENSUS

    The reported statistics are weighted using attrition-adjusted weights generated by the authors to account for attrition by age 22 in both years.

    “Reported statistics are generated for groups defined by observed AFQT test score and weighted using the attrition adjusted weights generated by the authors to account for attrition by age 22 in both years.”
    • Also stated in: § Table 3.3 Summary Statistics · plain
      “Weighted means presented. Weights used are attrition-afqt adjusted weights created by the authors.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation
    § 10. APPENDIX C: FAMILY STRUCTURE IN THE NLSY SAMPLES AND IN THE CENSUS

    The IPUMS allows one to generate a measure of social parenthood but not biological parenthood.

    “The IPUMS allows one to generate a measure of social parenthood but not biological parenthood.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites www.ipums.org

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Interpretationreasoning
    § 6. IDENTIFYING THE CONTRIBUTION OF SUBSETS OF VARIABLES TO DIFFERENCES BETWEEN THE 1979 AND 1997 WAGE DISTRIBUTIONS

    Nonlinearities among the various skill components clearly matter, and it is not sufficient to simply decompose the means with a simple additively separable linear regression to get an accurate description of the variation in skills between 1979 and 1997.

    “Nonlinearities among the various skill components clearly matter, and it is not sufficient to simply decompose the means with a simple additively separable linear regression to get an accurate description of the variation in skills between 1979 and 1997.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationown data
    § 10. APPENDIX C: FAMILY STRUCTURE IN THE NLSY SAMPLES AND IN THE CENSUS

    The results in Table C provide additional evidence for the break-up of the traditional family.

    “The results in Table C provide additional evidence for the break-up of the traditional family.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Study design and conductreasoning
    § 9. APPENDIX A: VARIABLES USED AND THEIR CONSTRUCTION

    The NLSY97 self-created variable that mimics the DOD AFQT is not comparable to the AFQT in 1979.

    “It is not comparable to the AFQT in 1979.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationreasoning
    § 5. CHANGES IN THE SKILL DISTRIBUTION BETWEEN NLSY79 AND NLSY97

    The narrowing gender gap reflects a larger increase for women than men in education and a somewhat larger increase in AFQT.

    “The narrowing gender gap reflects a larger increase for women than men in education and a somewhat larger increase in AFQT.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data
    § 10. APPENDIX C: FAMILY STRUCTURE IN THE NLSY SAMPLES AND IN THE CENSUS

    In 1997, the survey cannot distinguish individuals living with one versus two adoptive parents if they do not have biological parents.

    “The ambiguity arises because in 1997 we can not distinguish individuals living with one vs. two adoptive parents if they do not have biological parents.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § 10. APPENDIX C: FAMILY STRUCTURE IN THE NLSY SAMPLES AND IN THE CENSUS

    The IPUMS social parenthood measure includes step-parents and adoptive parents.

    “This measure includes step-parents and adoptive parents.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites www.ipums.org

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data
    § 10. APPENDIX C: FAMILY STRUCTURE IN THE NLSY SAMPLES AND IN THE CENSUS

    Among the full sample, attriters had 0.85 fewer years of highest grade completed at age 22 than stayers (SE 0.1087, significant at 1%).

    “-0.85 (0.1087)***”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning
    § 6. IDENTIFYING THE CONTRIBUTION OF SUBSETS OF VARIABLES TO DIFFERENCES BETWEEN THE 1979 AND 1997 WAGE DISTRIBUTIONS

    The concern about the recent failure of closing the test score gap further is ever more relevant.

    “It also makes the concern about the recent failure of closing the test score gap further (Neal, 2006) ever more relevant.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Neal, 2006 source
120 of 120

  • The authors state this as an assumption.

    The distribution of unobserved skills u conditional on observed skills z is the same for the 1979 and 1997 cohorts.

    “Assumption A1 : The distribution of u conditional on z is the same for 1979 and 1997 cohorts:”

  • The authors state this as an assumption.

    The relationship between characteristics and adult outcomes is the same for the 1997 cohort as it has been for the 1979 cohort.

    “we assess what the adult outcomes of the 1997 cohort will be if the relationship between characteristics and adult outcomes turns out to be the same for the 1997 cohort as it has been for 1979.”

  • The authors state this as an assumption.

    The cross-sectional relation between family background variables and education and ability in 1979 is causal in the sense that changes in the distribution of parental background result in changes in the individual variables.

    “The decomposition therefore implicitly assumes that the crosssectional relation between family background variables and education and ability in 1979 is causal in the sense that changes in the distribution of parental background result in changes in the individual variables.”

  • The authors state this as an assumption.

    The relation between parental background variables and individual characteristics is the same in 1979 and 1997.

    “This order of the decomposition assumes that the relation between parental background variables and individual characteristics is the same in 1979 and 1997.”

  • The authors state this as an assumption.

    The cross-sectional relation between parental background, individual education and ability scores and the speed of transition into the workforce in 1979 is causal.

    “Similar assumptions are made regarding the relation between parental background, individual education and ability scores on the one hand and the variables describing the speed with which individuals transition into the workforce on the other hand.”

  • The authors state this as an assumption.

    The available data, after use of survey weights and adjustments for attrition prior to age 22 and for missing data on AFQT, are representative of the 1997 and 1979 populations.

    “Weproceed under the assumption that the available data, after use of survey weights and adjustments for attrition prior to age 22 and for missing data on AFQT, are representative of the 1997 and 1979 populations, with the obvious caveat that our results will be affected if they are not.”

  • The authors state this as an assumption.

    Attrition from NLSY79 after age 22 is random.

    “The results presented below assume that attrition from NLSY79 after age 22 is random.”

  • The authors state this as an assumption.

    The AFQT-score is not informative about skills of an individual during adulthood once we condition on the rank of an individual within his/her age group.

    “By matching percentiles across ages we implicitly assume that the AFQT-score is not informative about skills of an individual during adulthood once we condition on the the rank of an individual within his/her age group.”

  • The authors state this as an assumption.

    The error term in the latent index function is normal.

    “The error ε i in the latent index function is assumed to be normal.”

  • The authors state this as an assumption.

    The conditional distribution of schooling and AFQT is constant when attributing the shift to parental background.

    “under the assumption that the conditional distribution of schooling and AFQT constant.”

  • The authors state this as an assumption.

    School completion occurs in June of the given year.

    “School completion is assumed to occur in June of given year.”

  • The authors state this as an assumption.

    The percentile a person is placed in at age 14 is the same as the percentile the person would have been placed in at age 16.

    “We assume that the percentile a person is placed in at (say) age 14 is the same as the percentile the person would have been placed in at age 16.”

  • The authors state this as an assumption.

    The wage process faced by the NLSY79 cohort in their prime age years persists for the 1997 cohort.

    “If the wage process faced by the NLSY79 cohort in their prime age years persists”

  • The authors state this as an assumption.

    The conditional distribution of adult wages for the 1997 cohort is similar to that of the 1979 cohort.

    “if the conditional distribution of adult wages for the 1997 cohort turns out to be similar to that of the 1979 cohort”

  • The authors state this as an assumption.

    The relative rankings of individuals in the AFQT distribution on average do not depend on when they took the test.

    “This procedure assumes that the relative rankings of individuals in the AFQT-distribution on average do not depend on when they took the test.”

  • The authors state this as an assumption.

    The level of cognitive skills in adulthood associated with the qth percentile in the age 16 test taker distribution is the same as that for the qth percentile in the age a distribution.

    “It also assumes that the level of cognitive skills in adulthood associated with the qth percentile in the age 16 test taker distribution is the same as that for the qth percentile in the age a distribution.”

  • The paper relies on this without support.

    NLSY79 sample members were between 15 and 23 years old when they took the test, while NLSY97 test takers were between 12 and 18 years old and thus typically younger.

    “Second, NLSY79 sample members were between 15 and 23 years old when they took the test. Test takers in the NLSY97 were between 12 and 18 years olds and thus typically were younger than their NLSY79 counterparts.”
17 of 17

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • NLSY79 test administration year differs

    Minor

    Section 3.5 dates the NLSY79 paper-and-pencil test to 1981, but the appendix calls it the 1980 test.

  • NLSY79 AFQT-status shares total only 90%

    Minor

    Table 3.2 partitions respondents observed at age 22 into AFQT-missing and AFQT-not-missing groups, but its NLSY79 total row gives 5.92% and 84.08%. The latter would need to be 94.08% for the shares to total 100%.

    Reported 84.08 · recomputed 94.08

  • Bootstrap replication count differs

    Minor

    The methods footnote says the bootstrap procedure was repeated 300 times, whereas the results-table notes specify 200 repetitions.

3 of 3

Cites

Flagged first, then by citations

Research works printed but not linked: 10

  1. Busso Matias, Finite sample properties of semiparametric estimators of average treatment effects. Unpublished paper
  2. Busso Matias, New evidence on the finite sample properties of propensity score reweighting and matching estimators. Unpublished paper
  3. Edwards Rebecca, Predicting earnings inequality in 2030: The role of labor and skill supply. Working paper
  4. Goldin Claudia, 2008, Katz
  5. Heckman James J, 2007, LaFontaine
  6. Jorgensen Dale, 1987, Fraumeni
  7. Juhn Chinhui, Workers and their wages: Changing patterns in the United States, ed. Marvin H. Kosters.
  8. Korenman Sanders, Meritocracy and economic inequality, ed. Kenneth Arrow, Samuel Bowles, and Steve Durlauf.
  9. MaCurdy Thomas, Department of Economics
  10. Moore Whitney, National Longitudinal Survey of Youth 1997 (NLSY97): Technical sampling report

Cited by

Sorted by citations

Related

Authors
Altonji, J. G., Bharadwaj, P., & Lange, F.
Affiliations
  • National Bureau of Economic Research (US)
  • University of California San Diego (US)
  • IZA - Institute of Labor Economics (DE)
ORCID
Venue
Journal of Labor Economics
Publisher
University of Chicago Press
Published
2012-10
Volume
30
Issue
4
Pages
783-828
ISSN
  • 0734-306X (ISSN-L)
  • 1537-5307
DOI
10.1086/666536
OpenAlex
W2024319570
Cited by
158
References
39
Citations by year
2026: 5 · 2025: 5 · 2024: 11 · 2023: 9 · 2022: 9 · 2021: 12 · 2020: 15 · 2019: 16 · 2018: 10 · 2017: 8 · 2016: 3 · 2015: 26 · 2014: 11 · 2013: 7 · 2012: 5
FWCI
37.51
Type
journal article
Language
en
Open access
closed
Locations
  • Journal of Labor Economics (publishedVersion)
  • RePEc: Research Papers in Economics (submittedVersion)
Topics
  • Social Sciences › Economics, Econometrics and Finance › Economics and Econometrics › Labor market dynamics and wage inequality
  • Social Sciences › Sociology and Political Science › Intergenerational and Educational Inequality Studies
  • Social Sciences › Political Science and International Relations › Education and Labor Relations
Keywords
youth characteristics, skill distribution, labor market success, adult outcomes, racial and ethnic differences, gender differences, parental education, human capital, technological change, wage inequality, educational attainment trends

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-29.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-29.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review