Between 1964 and 2003 the cohort-based instrumental-variables (causal, IV) estimate of the return to a year of schooling in the United States rose from about 4.8 to about 8.4 percent, and cohort-based instrumental-variables estimates indicate that the acceleration of the measured (conventional, OLS) education premium after 1980 is explained almost entirely by a rise in ability bias — the growing wedge between conventional and causal estimates that, in the paper's own reading, reflects either a rising market price of unobserved skill or ability levels that have become more dispersed, or more strongly sorted, relative to educational choices. This problem record states what that analysis left open: which economic mechanism or mechanisms drove the rise in ability bias; whether the rise persisted, reversed, or proved temporary beyond the 2003 data endpoint; and how far the phenomenon extends beyond the design's scope, which identifies returns for the middle-ability subpopulation — roughly the middle 80 percent of the ability distribution — in the United States, and whose extension to other countries the paper itself proposes as future work. The follow-up literature documents adjacent facts — falling prices of measured cognitive skill, growing ability sorting across education groups, and a post-2000 reversal in the demand for skill — but has not identified the mechanism or updated the ability-bias series, so the problem remains open.

Background

Labor economists measure the return to education as the percentage wage gain attributable to an additional year of schooling, conventionally estimated by regressing log wages on years of schooling in a Mincer earnings regression. Since Griliches (1977), it has been understood that this ordinary least squares (OLS) estimate is contaminated by ability bias: workers who obtain more schooling also tend to have higher pre-existing ability, which raises their wages regardless of schooling, so the regression conflates the productivity of education with the labor-market value of traits formed much earlier in life. Instrumental variables (IV) address this by using a variable that shifts schooling but is plausibly unrelated to individual ability; Angrist and Krueger (1991) used quarter of birth interacting with compulsory-schooling laws. Kaymak (2009) instead exploited differences in educational attainment across birth cohorts: if average ability is similar among nearby cohorts, then cohort-to-cohort differences in attainment translate into earnings differences only to the extent that education is productive. Applying this to U.S. March Current Population Survey data for men aged 25–60 over 1964–2003 (born 1914–1968), with a Decennial Census replication, he found that the causal (IV) return to a year of schooling rose from about 4.8 to about 8.4 percent, while the ability bias — measured as the gap between the OLS and IV estimates, interpretable as the market price of unobserved ability times its correlation with schooling — rose from about 1.8 to about 4.7 percent, so that the post-1980 acceleration of the measured education premium is explained almost entirely by the rise in ability bias. The design has three features that matter for what remains open. First, the instrument identifies the return for the subpopulation whose attainment responds to cohort variation — roughly the middle 80 percent of the ability distribution, as evidenced by the well-ordered (monotone) cumulative distribution functions of attainment across adjacent cohorts — leaving returns for the least- and most-able workers outside the identified range. Second, the decomposition of market-wide time effects from cohort effects in the return assumes cohort effects are smooth, modeled as a quartic polynomial in birth year; sharp unobserved cohort-specific shocks correlated with attainment (beyond the cohort-size and labor-market-entry conditions the paper tests) would violate this. Third, the estimates end in 2003 and cover only the United States. Prior to this paper the evidence on the role of ability was conflicting: Taber (2001) concluded that rising returns to unobserved ability accounted for essentially all of the 1980s rise in the college premium, while other proxy-based and selection-model approaches were inconclusive or found no role for ability (surveyed in Card 2001). Whether the rise in the education premium reflects genuine returns to schooling or a rising price of ability matters directly for policy — education investment versus early-childhood skill formation — and for how the field interprets one of its central stylized facts.

Problem Statement

Explain the post-1980 rise in ability bias — the widening gap between the OLS and the cohort-IV estimates of the return to schooling — that accounts almost entirely for the measured acceleration of the U.S. education premium, and establish the persistence and scope of the phenomenon. The paper reads the widening gap disjunctively: it “suggests that the price of unobserved skill has also gone up or, alternatively, that the ability levels have become more dispersed relative to educational choices,” and determining which force operated, and through what mechanism, is part of the open question. Concretely, the problem left open by the cohort-based analysis of the 1964–2003 U.S. series is: (1) identify and quantify the economic mechanisms — candidate explanations include skill-biased technical change such as computerization, shifts in industry composition toward sectors that reward cognitive skill, increases in the dispersion of ability within cohorts or changes in how strongly ability and educational attainment are sorted, institutional changes, or other forces — that produced the rise in the gap between conventional (OLS) and causal (instrumental) estimates of the return to schooling, in a way consistent with the documented series and with the partially conflicting dynamics of prices of measured cognitive skill; (2) determine whether the rise persisted, reversed, or proved temporary, using data beyond 2003 (one hypothesis the paper raises is that a temporary rise would subside once the skills demanded by new technology could be taught in formal education); and (3) establish the scope of the phenomenon beyond the design: the benchmark estimates are identified for roughly the middle 80 percent of the ability distribution in the United States, and a uniform estimate of ability bias across countries is not readily available. These three sub-questions are jointly posed: sub-question (1) is the core mechanism objective named in the title, while (2) and (3) establish whether the rise persisted and how far the phenomenon extends. Each is independently attackable: an answer to (2) or (3) can be settled without a full mechanism decomposition, and (1) can be assessed against the documented series even where (2) or (3) remain open. In (3), "scope" means both the tail-ability ranges outside the middle 80 percent — the least- and the most-able workers — and the cross-country extension that the paper proposes. The paper itself links the mechanism question to persistence: “Alternative explanations of rising ability bias differ in their predictions for the persistence of the elevated role of ability.” Any answer inherits the design's identification assumptions: the ability-bias series is measured under the requirements that cohort effects in earnings and in the return to education are smooth functions of birth year and that short-run cohort variation in average ability is orthogonal to average attainment; the paper explicitly controls for the documented cohort-size and labor-market-entry scenarios and finds them quantitatively unimportant, and notes that unmodeled year-to-year cohort-specific shifts in the return correlated with attainment would bias estimated levels upward while their effect on the estimated changes over time “would be ambiguous.”

References

  1. Kaymak, Barış (2009). "Ability Bias and the Rising Education Premium in the United States: A Cohort-Based Analysis." Journal of Human Capital 3(3): 224–267. DOI: 10.1086/651167. https://doi.org/10.1086/651167
  2. Griliches, Zvi (1977). "Estimating the Returns to Schooling: Some Econometric Problems." Econometrica 45(1): 1–22. DOI: 10.2307/1913285. https://doi.org/10.2307/1913285
  3. Angrist, Joshua D., and Alan B. Krueger (1991). "Does Compulsory School Attendance Affect Schooling and Earnings?" The Quarterly Journal of Economics 106(4): 979–1014. DOI: 10.2307/2937954. https://doi.org/10.2307/2937954
  4. Taber, Christopher R. (2001). "The Rising College Premium in the Eighties: Return to College or Return to Unobserved Ability?" Review of Economic Studies 68(3): 665–691. DOI: 10.1111/1467-937X.00185. https://doi.org/10.1111/1467-937X.00185
  5. Card, David (2001). "Estimating the Return to Schooling: Progress on Some Persistent Econometric Problems." Econometrica 69(5): 1127–1160. DOI: 10.1111/1468-0262.00237. https://doi.org/10.1111/1468-0262.00237
  6. Castex, Gonzalo, and Evgenia Kogan Dechter (2014). "The Changing Roles of Education and Ability in Wage Determination." Journal of Labor Economics 32(4): 685–710. DOI: 10.1086/676018. https://doi.org/10.1086/676018
  7. Hendricks, Lutz, and Todd Schoellman (2014). "Student Abilities During the Expansion of US Education." Journal of Monetary Economics 63: 19–36. DOI: 10.1016/j.jmoneco.2014.01.008. https://doi.org/10.1016/j.jmoneco.2014.01.008
  8. Açıkgöz, Ömer Tuğrul, and Barış Kaymak (2014). "The Rising Skill Premium and Deunionization." Journal of Monetary Economics 63: 37–50. DOI: 10.1016/j.jmoneco.2014.01.002. https://doi.org/10.1016/j.jmoneco.2014.01.002
  9. Naylor, Robin, and Jeremy Smith (2009). "Ability Bias, Skewness and the College Wage Premium." Working paper, AgEcon Search record 271293. DOI: 10.22004/ag.econ.271293. https://doi.org/10.22004/ag.econ.271293
  10. Beaudry, Paul, David A. Green, and Benjamin M. Sand (2016). "The Great Reversal in the Demand for Skill and Cognitive Tasks." Journal of Labor Economics 34(S1): S199–S247. DOI: 10.1086/682347. https://doi.org/10.1086/682347
  11. Valletta, Robert (2016). "Recent Flattening in the Higher Education Wage Premium: Polarization, Skill Downgrading, or Both?" NBER Working Paper No. 22935. DOI: 10.3386/w22935. https://doi.org/10.3386/w22935
  12. Donovan, Kevin, and Christopher Herrington (2019). "Factors Affecting College Attainment and Student Ability in the U.S. Since 1900." Review of Economic Dynamics 31: 224–244. DOI: 10.1016/j.red.2018.07.003. https://doi.org/10.1016/j.red.2018.07.003
  13. Nikolov, Plamen, Nusrat Jimi, and Jerray Chang (2020). "The Importance of Cognitive Domains and the Returns to Schooling in South Africa: Evidence from Two Labor Surveys." Labour Economics 65: 101849. DOI: 10.1016/j.labeco.2020.101849. https://doi.org/10.1016/j.labeco.2020.101849
  14. Caponi, Vincenzo, and Miana Plesca (2021). "Post-Secondary Education in Canada: Can Ability Bias Explain the Earnings Gap Between College and University Graduates?" Posted content, Ryerson University Library and Archives. DOI: 10.32920/ryerson.14638566. https://doi.org/10.32920/ryerson.14638566
  15. Kaymak, Barış (2025). "Quantifying the Signaling Role of Education." Federal Reserve Bank of Cleveland Working Paper No. 25-02. DOI: 10.26509/frbc-wp-202502. https://doi.org/10.26509/frbc-wp-202502