I24 | Macro Paper Warehouse

Diversifying Society's Leaders? Determinants and Causal Effects of Admission

Mon, 01 Jan 0001 00:00:00 +0000

This paper studies why children from high-income families are more likely to attend Ivy-Plus colleges (Ivy League, Stanford, MIT, Duke, Chicago — 12 colleges total) and whether attending these colleges causally improves post-college outcomes. The authors construct a de-identified panel dataset linking federal income tax records, Department of Education college attendance data, College Board and ACT test scores, and application and admissions records from several Ivy-Plus and flagship public colleges covering approximately 2.4 million students across entering classes from 1998–2015.

The central finding on the input side is that students from families in the top 1% of the income distribution (income above $611,000) are 2.3 times more likely to attend an Ivy-Plus college than middle-class students (defined as the 70th–80th percentiles of the national parental income distribution, approximately $91,000–$114,000) with comparable SAT/ACT scores. Two-thirds of this gap is attributable to higher admissions rates at Ivy-Plus colleges for high-income applicants; conditional on SAT/ACT scores, top-1% applicants are 58% more likely to be admitted than middle-class applicants. The remaining third splits between differences in application rates (roughly 20% of the total attendance gap) and matriculation rates (roughly 12%). In contrast, admissions rates at flagship public colleges are essentially uncorrelated with parental income conditional on test scores.

Three admissions practices drive the high-income admissions advantage at Ivy-Plus colleges. First, legacy preferences: legacy applicants from the top 1% are admitted at more than five times the rate of non-legacy applicants with comparable test scores, demographics, and admissions ratings; children of alumni of a given Ivy-Plus college are not more likely to be admitted to other Ivy-Plus colleges, confirming that legacy status is not merely a proxy for unobservable credentials. Legacy preferences account for 52 of the estimated 168 “extra” top-1% students per average Ivy-Plus class (enrollment ~1,650). Second, non-academic ratings: students from the top 1% have markedly stronger non-academic credentials (extracurricular activities, leadership ratings) partly because they disproportionately attend private high schools whose students receive higher non-academic ratings despite no higher academic ratings; this accounts for 35 additional extra top-1% students. Third, athletic recruitment: the share of recruited athletes rises from 5% among admitted students from the bottom 60% to 13% among those from the top 1%, accounting for 27 additional extra top-1% students.

On the output side, the authors estimate causal effects of attending an Ivy-Plus college using a new research design based on waitlisted applicants. The key identification assumption is that idiosyncratic variation in admissions decisions across waitlisted applicants at one Ivy-Plus college is uncorrelated with admissions decisions at other Ivy-Plus colleges — which the authors verify empirically. Under this assumption, comparisons of admitted vs. rejected waitlisted applicants identify causal effects for marginal students. The marginal student who attends an Ivy-Plus college instead of the average flagship public is approximately 50% more likely to reach the top 1% of the earnings distribution at age 33, nearly twice as likely to attend a highly-ranked graduate school, and 2.5 times as likely to work at a prestigious firm. Attending an Ivy-Plus college increases mean earnings by $101,000 at age 33 relative to a counterfactual mean of $143,000 at state flagships. Effects are concentrated in the upper tail of earnings — the impact on reaching the top quartile is small and statistically insignificant, while impacts on reaching the top 1% far exceed what a constant percentage treatment effect would predict. Effects are larger for students with weaker fallback options (i.e., whose home-state colleges channel fewer students to the top 1%).

Critically, the three credentials driving the high-income admissions advantage — legacy status, athletic recruitment, and high non-academic ratings — are uncorrelated with or negatively correlated with post-college success once the college attended is held constant. Academic credentials (SAT/ACT scores, academic ratings) remain highly predictive of outcomes.

Counterfactual simulations show that eliminating all three high-income admissions preferences and replacing those slots with students having the same test score distribution would increase enrollment from the bottom 95% of the parental income distribution by 8.8 percentage points — comparable in magnitude to the effect of race-based affirmative action on Black and Hispanic enrollment shares. Such a policy would have small effects on monetary leadership outcomes (e.g., Fortune 500 CEO share from bottom-95% families rises by only 0.4 pp, because Ivy-Plus graduates are a small fraction of all top earners) but larger effects on non-monetary leadership positions: the share of senators from the bottom 95% would rise by 1.7 pp and the share of Supreme Court justices by 5.4 pp. With need-affirmative policies (giving low-income students preferences comparable to those currently given to legacy applicants), the share of Supreme Court justices from families in the bottom 60% would rise by 17.5 pp. These predictions assume that the causal share of Ivy-Plus attendance in explaining observational differences in leadership outcomes is the same as that estimated for early-career outcomes, and they ignore general equilibrium effects.

Q: How much more likely are top-1% students to attend an Ivy-Plus college than middle-class students with the same test scores? A: Students from families in the top 1% (income above $611,000) are 2.3 times more likely to attend an Ivy-Plus college than students from the 70th–80th percentile of the parental income distribution (approximately $91,000–$114,000) with comparable SAT/ACT scores. This “missing middle” pattern is stable across entering classes from 1998 to 2018 and persists after controlling for race and ethnicity.

Q: How is the overall attendance gap decomposed into application, admissions, and matriculation? A: Differences in admissions rates explain two-thirds of the gap in Ivy-Plus attendance between top-1% and middle-class students conditional on test scores. Of the estimated 168 “extra” top-1% students per average Ivy-Plus class, 87 come from higher admissions rates for non-recruited athletes, 27 from athletic recruitment, and the remaining slack from application rate differences (accounting for roughly 20% of the overall attendance gap) and matriculation differences (roughly 12%).

Q: How large is the admissions advantage for top-1% applicants at Ivy-Plus colleges? A: Conditional on SAT/ACT scores, applicants from the top 1% are 58% more likely to be admitted to Ivy-Plus colleges than middle-class applicants. Students from the top 0.1% are 2.5 times more likely to be admitted than middle-class applicants with comparable test scores. At flagship public colleges, admissions rates are essentially constant across the income distribution conditional on test scores.

Q: What is the magnitude of legacy preferences and how is it established that legacy is not just a proxy for other credentials? A: Legacy applicants from the top 1% are admitted at more than five times the rate of otherwise comparable non-legacy applicants at the college their parents attended. The paper isolates the legacy effect by showing that children of alumni at a given Ivy-Plus college are only slightly more likely to be admitted at other Ivy-Plus colleges — and the predicted counterfactual admissions rate for legacy students at other colleges closely matches their actual admissions rate — confirming that legacy status is not merely a proxy for other unobservable credentials. Legacy applicants constitute 2.5% of the overall applicant pool but over 9% of top-1% applicants.

Q: How do non-academic credentials differ by parental income, and what drives the difference? A: Top-1% applicants have markedly stronger non-academic ratings (measuring extracurricular participation and leadership traits) compared with other applicants, while the share achieving high academic ratings is essentially constant across the income distribution. Students from the top 1% are much more likely to have attended private high schools, whose applicants receive substantially higher non-academic ratings than students from public high schools with the same SAT/ACT scores. Non-academic ratings account for 35 of the estimated 168 extra top-1% students per Ivy-Plus class.

Q: What is the research design for estimating causal effects, and what is the key identification assumption? A: The authors focus on applicants who are waitlisted at a given Ivy-Plus college and compare those ultimately admitted versus rejected from the waitlist. The key identification assumption is that if different colleges’ admissions committees make correlated assessments of underlying student merit but uncorrelated idiosyncratic admissions errors, then residual variation in admissions outcomes for waitlisted applicants at one college is orthogonal to students’ long-run potential. The authors validate this empirically by showing that waitlist admission at one Ivy-Plus college is uncorrelated with admissions decisions and internal ratings at other Ivy-Plus colleges.

Q: What are the causal effects of attending an Ivy-Plus college on post-college outcomes? A: For the marginal student (one who attends an Ivy-Plus college instead of the average flagship public), attending an Ivy-Plus college increases the probability of reaching the top 1% of the earnings distribution at age 33 by approximately 50%, nearly doubles the probability of attending an elite graduate school, and increases the probability of working at a prestigious firm by approximately 2.5 times. Mean earnings at age 33 increase by $101,000 (relative to a counterfactual mean of $143,000 at state flagships). Effects on reaching the top quartile of earnings are small and statistically insignificant, while effects at the very top tail are disproportionately large.

Q: Why do the findings differ from Dale and Krueger (2002) and related studies finding little effect of selective college attendance on earnings? A: The authors replicate the matriculation design of Dale and Krueger (comparing outcomes conditional on the set of colleges to which students were admitted) and obtain estimates statistically indistinguishable from their waitlist design — the research designs are not the source of disagreement. Instead, the differences arise because (1) the authors have direct college fixed effects rather than relying on average test scores as a proxy for college quality, and (2) the authors focus on upper-tail outcomes (top 1% earnings, elite graduate schools, prestigious firms) rather than log mean earnings, where Ivy-Plus colleges have their largest effects.

Q: Are the credentials that drive the high-income admissions advantage — legacy, athlete status, high non-academic ratings — predictive of better post-college outcomes? A: No. Recruited athletes, students with higher non-academic ratings, and legacy students have equivalent or lower chances of reaching the upper tail of the income distribution, attending an elite graduate school, or working at a prestigious firm than comparable Ivy-Plus applicants once the college attended is held constant. By contrast, SAT/ACT scores and academic ratings are highly positively predictive of all three post-college outcome measures.

Q: How much could changing admissions practices diversify Ivy-Plus enrollment and subsequently society’s leadership? A: Eliminating legacy preferences, non-academic rating weights, and the differential recruitment of high-income athletes — and filling those slots with students having the same test score distribution as the current class — would increase enrollment from families in the bottom 95% of the parental income distribution by 8.8 percentage points, a magnitude comparable to race-based affirmative action’s effect on Black and Hispanic enrollment shares. For leadership positions, predicted effects are small for monetary outcomes (Fortune 500 CEOs from the bottom 95% would increase by only 0.4 pp) but larger for positions where Ivy-Plus graduates are a larger share: senators from the bottom 95% would increase by 1.7 pp and Supreme Court justices by 5.4 pp. A stronger need-affirmative policy (giving low-income students preferences equivalent to current legacy preferences) would increase the share of Supreme Court justices from the bottom 60% by 17.5 pp.

Q: How are “elite” and “prestigious” employers defined in this study? A: Elite firms are defined as those that disproportionately employ Ivy-Plus graduates relative to flagship public graduates, pulling firms from the top of that ratio ranking until 25% of Ivy-Plus attendee employment is accounted for. Prestigious employers are defined by the residual of that ratio after controlling for the firm’s predicted top-1% income probability — they are firms that disproportionately employ Ivy-Plus graduates conditional on their salaries, capturing high-status jobs that do not necessarily lead to the highest earnings. The paper validates this algorithmic approach against external rankings (Vault.com for law and consulting firms; Scimagoir for hospitals), finding substantial overlap.

Q: How are treatment effect estimates adjusted for heterogeneity in students’ fallback options? A: Causal effects of Ivy-Plus attendance are much larger for students with weaker fallback options — specifically, students whose home-state flagship colleges channel fewer students to the top 1% of earnings. The authors exploit this heterogeneity to estimate the treatment effect for the marginal student who actually switches from a flagship public to an Ivy-Plus college. This heterogeneity also implies that the average causal effect across all admitted students may differ from the effect for the marginal admitted student.

Q: What share of the overrepresentation of top-1% families at Ivy-Plus colleges is attributable to pre-application factors versus admissions practices? A: Of the 245 “extra” top-1% students in an average Ivy-Plus class relative to an unconditionally income-neutral benchmark, 77 (31%) are attributable to the higher test scores of top-1% students (a pre-application factor). The remaining 168 (69%) reflect higher attendance rates conditional on test scores, of which the large majority is attributable to admissions practices (legacy, non-academic ratings, athletic recruitment) rather than application or matriculation rate differences.

Ivy-Plus colleges: The twelve highly selective private colleges comprising the eight Ivy League institutions plus Stanford, MIT, Duke, and the University of Chicago — the focus group of the study, which together account for more than 10% of Fortune 500 CEOs, a quarter of U.S. senators, and three-fourths of Supreme Court justices appointed in the last half century despite enrolling less than 0.5% of Americans.

Missing middle: The pattern by which attendance rates at Ivy-Plus colleges conditional on SAT/ACT scores are lowest for students from the middle class (70th–80th percentile of the parental income distribution, approximately $91,000–$114,000) — lower than both the top 1% and, slightly, the bottom 40% — producing a non-monotone income gradient in attendance.

Legacy preference: An admissions advantage given to applicants whose parent(s) obtained an undergraduate degree from the college to which the student is applying. In the paper’s data, legacy applicants from the top 1% are admitted at more than five times the rate of non-legacy applicants with comparable test scores, demographics, and admissions ratings; the preference is college-specific (children of alumni are only slightly more likely to be admitted at other Ivy-Plus colleges).

Waitlist research design: The paper’s primary identification strategy for causal effects, which exploits idiosyncratic variation in admissions decisions among waitlisted applicants. The design’s validity rests on the empirical finding that waitlist admissions at one Ivy-Plus college are uncorrelated with admissions decisions and internal ratings at other Ivy-Plus colleges, implying that residual variation conditional on being on the waitlist is orthogonal to students’ long-run potential outcomes.

Prestigious employers: Firms defined by the paper’s algorithm as disproportionately employing Ivy-Plus graduates conditional on those firms’ predicted top-1% income probability — capturing high-status employment that does not necessarily lead to the highest earnings (e.g., prominent law firms, consulting firms, elite hospitals). Validated against external rankings (Vault.com, Scimagoir).

Non-academic ratings: Numerical scores assigned by admissions officers measuring aspects of an application outside academic achievement, such as extracurricular activities and leadership traits. In the paper’s data, non-academic ratings differ substantially by parental income — particularly because top-1% applicants disproportionately attend private high schools whose students receive higher non-academic ratings — while academic ratings do not differ across the income distribution.

Surrogate index: A prediction of later earnings outcomes (specifically, probability of reaching the top 1% at age 33 and mean income rank) constructed from individuals’ graduate school attendance and employer fixed effects at ages 22–25, used to extend the outcome window for cohorts observed only early in their careers. The approach follows the terminology and methodology of Athey et al. (2019).

Life-Cycle Wages and Human Capital Investments: Selection and Missing Data

Mon, 01 Jan 0001 00:00:00 +0000

Layer 1 – Overview

Research Question

This paper asks how wage inequalities build up over the life cycle when individual wage trajectories are plagued by interruptions in private-sector participation, and when the standard Missing At Random (MAR) assumption used to handle those gaps may be violated. Specifically, it asks: what is the causal effect of career interruptions on both the level and the dispersion of wages after twenty years of potential experience, and does endogeneity of those interruptions matter for the dispersion result?

Data and Sample

The empirical analysis uses the 2011 DADS Grand Format-EDP panel, a French administrative dataset merging social security records (DADS) and census extracts (EDP). The working sample covers males who entered the private sector between 1985 and 1992, aged 16-30 at entry, and observed through 2011. The authors require at least 15 years of observed private-sector wages, yielding a working sample of 7,004 males and 137,315 person-year observations. Education is grouped into four levels (high-school dropouts, high-school graduates, some college, college graduates). Participation outside the private sector – including public-sector employment, self-employment, unemployment, and non-employment – constitutes the “alternative sector” and generates missing wage observations. On average, cumulative duration outside the private sector is 3.7 years, and the average number of interruptions is 1.44.

Model and Methodology

The paper builds on a structural Ben Porath (1967) human capital model extended to two sectors (private sector and an alternative sector), yielding a reduced-form log-wage equation with five individual-specific coefficients: an intercept (initial human capital), a linear trend in potential experience (growth rate), a curvature term in potential experience (Mincer concavity), the cumulative years of interruptions, and a curvature term in interruptions. Because parameters are individual-specific, the wage equation is a random-coefficient model estimated with a fixed-effects approach.

Selection into the private sector is addressed not by a standard MAR assumption but by a weaker “Missing At Random Conditionally On Factors” (MARCOF) assumption. Sector-preference shocks, human capital prices, and depreciation rates are each decomposed into a common factor (time-varying) and an individual factor loading, plus a residual that is mean-independent of factors and loadings. Conditional on factors and factor loadings, wage residuals and sector choices are independent, making covariates – including the interruption variables – exogenous. The preferred specification includes two unobserved factors, selected by four of six Bai-Ng (2002) information criteria.

Estimation proceeds via an Expectation-Maximization (EM) algorithm adapted from Bai (2009) and Song (2013), with initial values from Moon and Weidner (2018)’s nuclear-norm convex estimator. Because individual parameters converge at rate sqrt(T) and summary statistics of their distributions suffer from incidental-parameter bias, the authors use bias-correction methods from Jochmans and Weidner (2019) for quantiles and inter-decile ranges, and from Arellano and Bonhomme (2012) for variances. Monte Carlo experiments confirm that variances remain poorly corrected even when T > 20, so the paper focuses on inter-decile ranges as the dispersion measure.

Counterfactual “average structural functions” (Blundell and Powell, 2003) are constructed by holding individual parameters fixed and manipulating the history of interruptions. These compare four scenarios: the observed benchmark, the counterfactual with no interruptions (potential wage), the counterfactual with no current-period selection, and both combined.

Main Findings

Downward bias from omitting interruptions and factors. Omitting interruption variables and unobserved factors strongly downward biases estimated returns to experience after 20 years. Most of this bias is attributable to interruptions rather than to the interactive factor effects: selectivity is mainly captured through the interruption channel, not through residual factor structure.
Effect on mean wages. Potential experience increases log wages by approximately 65% over 20 years, consistent with cross-country evidence from homogeneous Mincer equations. The average cost of interruptions after 20 years is approximately 10% of log wages. Reassigning interruptions to the beginning of the working life has a persistent negative effect on mean log wages that never fully recovers over 20 years, while reassigning them to the end increases mean wages above the no-interruption benchmark at every experience level.
Effect on wage dispersion – a new stylized fact. Interruptions decrease, not increase, the inter-decile range of log wages after 20 years. After 20 years, with an average interruption duration of 2.47 years, interruptions decrease the inter-decile range by 0.52 log points (approximately 38%). This compression operates differentially: the 90th percentile falls by 0.34 and the 10th percentile rises by 0.18.
Endogeneity explains the dispersion compression. When years of interruption are randomly reassigned across time (holding total interruption years fixed), the inter-decile range diverges upward from the observed benchmark after about 5 years. This shows that the dispersion-reducing effect of actual interruptions is due to the endogenous timing of those interruptions – specifically to the negative correlation between the timing of interruptions and potential log wages – rather than to the correlation between the structural coefficients on interruptions and potential wages (which is also negative, with a Spearman rank correlation of -0.32 between eta_i1 and eta_i3). Endogenously chosen interruptions smooth inequality over time.
Current-period selection is negligible. Current-period selection into private-sector employment has no statistically significant effect on median, mean, variance, or inter-decile range of wages at any experience level, as confirmed by the small inter-decile range of the interactive factor component.

Scope Conditions

Results pertain to cohorts of French males entering the private sector between 1985 and 1992, restricted to those with at least 15 observed private-sector years. The French context is distinctive: wage inequality in the working population was stable over 1985-2011, driven in part by minimum wage policy and payroll tax exemptions for lower-skilled workers, in contrast to rising inequality in the United States and Germany. Results on timing of interruptions (eta_i3 and eta_i4) are identified only for individuals with at least two interruptions followed by re-entry (roughly those with K_T >= 2). The paper does not analyze female wages.

Layer 2 – Q&A

Q1: What is the structural model and how does it generate a reduced-form wage equation?

The model is a Ben Porath (1967) two-sector human capital model in which individuals divide time between investing in human capital and earning wages in either the private sector (e) or an alternative sector (n). Human capital accumulation in each sector has a sector-specific return rate (rho^s) and depreciation (lambda^s_t). Period utility is log income minus a quadratic investment cost, plus a sector preference shock. Solving the dynamic program backwards (because of log-linearity) yields closed-form optimal investments that are linear in the individual-specific terminal value of human capital (kappa). The resulting log-wage equation (Proposition 5) is a function of five terms: an intercept (eta_i0), a linear trend in potential experience t (eta_i1), a geometric curvature term beta^{-t} (eta_i2), cumulative years of interruptions x^(3)_it (eta_i3), and a curvature in interruptions x^(4)_it (eta_i4), all with individual-specific coefficients. This provides a tractable random-coefficient structure.

Q2: What is the MARCOF assumption and why is it weaker than MAR?

MARCOF – Missing At Random Conditionally On Factors – posits that sector-preference shocks, human capital prices, and depreciation rates each follow factor structures: a common time-varying factor (phi_t) multiplied by an individual loading (theta_i) plus an i.i.d. residual. The residuals are assumed mean-independent of factors and loadings, and independent over time. Under standard MAR, missingness is assumed independent of outcomes conditional on observables alone. Under MARCOF, residuals in the wage equation and the sector choice equation are independent conditional on (unobserved) factors and factor loadings. This is weaker than MAR because it allows the unobservable determinants of wages and participation to share common factors, accommodating the high persistence observed in human capital stocks (20-year lag correlation of 0.28, far above the geometric decay benchmark of 0.024).

Q3: How are the individual-specific parameters identified?

Under exogenous selection (or, under MARCOF, conditional on factors), identification of eta_i0, eta_i1, and eta_i2 requires variation in potential experience within the individual’s time series. Identification of eta_i3 and eta_i4 separately requires individuals to experience at least two spells out of the private sector each followed by re-entry (at least four transitions, so K_T >= 2). An individual with only one interruption spell generates proportional variation in x^(3) and x^(4), so only a linear combination of eta_i3 and eta_i4 is identified. The “flat spot” approach – using the observed fact that individuals aged 50-55 have stopped investing in human capital – separately identifies time, cohort, and age effects and provides the restriction that factors are orthogonal to the level, trend, and curvature in potential experience.

Q4: What do the distributions of estimated individual-specific coefficients look like?

Focusing on the main (two-factor) specification with bias correction: the median of the growth parameter eta_i1 is positive (consistent with rising wages with experience) and the median of the curvature parameter eta_i2 is negative (consistent with concavity). However, heterogeneity is substantial: the 90th percentile of eta_i1 is 6.2 times the median, and the first quartile of eta_i1 is negative (implying declining potential wages for a non-negligible share). For the interruption coefficients eta_i3 (year of interruptions) and eta_i4 (curvature), bias-corrected medians are close to zero in the sub-sample with >=2 interruptions, but dispersion is large and symmetric around zero. Bias correction reduces the 90th percentile of eta_i1 by approximately 20% and reduces the absolute 10th percentile of eta_i3 by approximately 27%.

Q5: How important are interruptions relative to potential experience and factors in explaining wage variation?

A wage decomposition using inter-decile ranges (preferred over variance due to bias) shows that the potential experience component is the largest contributor to wage dispersion, followed by the interruption component (described as “sizable”), while factors play a minor role. Crucially, the potential experience and interruption components are highly negatively rank-correlated: the Spearman rank correlation between the growth coefficient eta_i1 and the interruption coefficient eta_i3 is -0.32. This negative correlation is central to understanding why interruptions compress dispersion rather than expanding it.

Q6: What is the finding on the effect of interruptions on mean wages, and what does the timing experiment show?

After 20 years, the average cost of interruptions (relative to a counterfactual of no interruptions) is approximately 10% of log wages. The timing of interruptions matters: reassigning interruptions to the beginning of the working life causes a persistent loss in mean log wages that does not fully recover over the 20-year horizon, while reassigning them to the end raises mean log wages above the no-interruption level at every experience level. For median wages, the early-interruption loss is eventually recovered (median log wages do catch up), but the mean does not catch up. These asymmetries are consistent with early interruptions having a larger negative effect on human capital accumulation due to the geometric structure of investment returns.

Q7: What is the key finding on wage dispersion and what explains it?

Interruptions compress the inter-decile range of log wages by 0.52 log points (approximately 38%) after 20 years, with average interruption duration of 2.47 years. This compression is asymmetric: the 90th percentile of wages falls by 0.34 and the 10th percentile rises by 0.18. The dispersion-reducing effect is established by comparing the benchmark (observed interruptions) to the counterfactual of no interruptions. When interruptions are instead randomly reassigned across time (holding total interruption duration fixed), the inter-decile range diverges upward from the benchmark starting around 5 years of experience. This demonstrates that the compression is due to the endogenous timing of interruptions – individuals who have high potential wages tend to time their interruptions in ways that reduce the measured spread of actual wages – rather than to the negative structural coefficient (eta_i3 < 0 for high-wage workers on average).

Q8: How does the paper handle the incidental parameter problem for distributional statistics?

Because individual parameters are estimated at rate sqrt(T) and the panel is unbalanced (some individuals observed for as few as 15 years while the model has up to 7 individual parameters), standard distributional statistics like the variance suffer from substantial incidental parameter bias. Monte Carlo experiments show that bias-corrected variance estimates remain strongly biased even at T > 20. Inter-decile ranges are better behaved and the Jochmans and Weidner (2019) bias-correction procedure reduces their bias satisfactorily. This is why the paper reports inter-decile ranges as its primary dispersion measure rather than variances. The bias in corrected inter-decile ranges is at most approximately 10% of the uncorrected estimate.

Q9: What does the paper show about the MAR assumption in the context of this data?

The results directly challenge the MAR assumption that is standard in the life-cycle earnings literature. Under MAR, interruptions would be treated as random conditional on observables, and their endogeneity would be ignored. The paper shows that treating interruptions as endogenous (through the MARCOF + structural model approach) substantially changes estimated returns to experience (there is a strong downward bias when interruptions and factors are omitted) and reverses the sign of the effect of interruptions on dispersion (under exogenous interruptions, randomly reassigned, dispersion would be higher than observed; the actual compression is an artifact of endogenous timing). The conclusion is that MAR assumptions produce systematically misleading pictures of life-cycle wage inequality dynamics.

Q10: What are the robustness and external validity considerations?

The working sample excludes individuals observed fewer than 15 years. A robustness exercise compares the subsample observed 10-14 years to a censored version of the 20+ subsample with matched marginal distributions of observation counts. Median profiles for the uncensored and censored 20+ samples are similar, and inter-decile ranges are slightly more dispersed in the censored sample only for potential experience greater than 7. However, the 10-14 year sample shows substantially different patterns – larger median gaps between benchmark and no-interruption cases, and a larger inter-decile range – consistent with lower private-sector returns to human capital for that group. The authors conclude that selection into the 15+ working sample matters, and results are explicitly restricted to that working sample. The French context (stable aggregate wage inequality, minimum wage policy) limits direct comparability to countries with rising inequality.

Key Concepts

MARCOF (Missing At Random Conditionally On Factors): The paper’s central identifying assumption, weaker than standard MAR. It posits that sector-preference shocks, human capital prices, and depreciation rates follow factor structures (common time-varying factor x individual loading + i.i.d. residual), and that residuals are mean-independent of factors, loadings, and their own histories. Conditional on factors and loadings, wage residuals and sector-choice residuals are independent, making selection exogenous.

Interactive effects / factor structure for selection: An approach in which unobserved confounders are modeled as a bilinear product of time-varying common factors (phi_t) and individual factor loadings (theta_i). This allows flexible correlation between wage processes and participation choices without requiring exclusion restrictions or instrumental variables. The paper’s preferred specification uses two unobserved factors identified by Bai-Ng information criteria.

Average structural functions: Objects defined by Blundell and Powell (2003) that integrate counterfactual outcomes (wages evaluated at a manipulated interruption history) over the distribution of individual-specific parameters. They allow estimation of the causal impact of a change in interruption timing or presence while holding individual structural parameters fixed, under identification conditions analogous to those of Chernozhukov et al. (2013).

Individual-specific coefficients (random coefficients): The five parameters (eta_i0, eta_i1, eta_i2, eta_i3, eta_i4) governing each individual’s wage equation, with structural interpretations: initial log human capital, return to potential experience, curvature (Mincer concavity), effect of cumulative interruption years, and curvature in interruptions. Their individual-specificity is the source of the incidental parameter problem for distributional statistics.

Flat spot approach: An identification device (from Heckman, Lochner, and Taber, 1998; Bowlus and Robinson, 2012) that uses median wages of workers aged 50-55 – who are assumed to have stopped investing in human capital – as consistent estimates of human capital prices by education group and year. This separates the volume of human capital from its price, and provides the restriction identifying the level, trend, and curvature factors from the time-varying unobserved factors phi_t.

Interruption variables x^(3) and x^(4): Reduced-form variables derived from the structural model summarizing the history of private-sector participation gaps. x^(3)_it is the cumulative number of periods spent in the alternative sector prior to date t; x^(4)_it is a geometric-weighted version of those interruptions that reflects the timing (early vs. late) through the discount factor beta. They enter the wage equation with individual-specific coefficients that are identified only for workers with at least two complete interruption spells.

Mincer dip: A U-shaped profile in wage variance (or inter-decile range) over potential experience, predicted by the Ben Porath model because high-return workers invest more at the start of their careers (reducing current wages), causing their wage profile to cross below then above low-return workers. Estimated in this paper at approximately 5 years of potential experience under the main specification.

Incidental parameter bias in distributional statistics: The bias that arises when estimating moments or quantiles of the distribution of individual-specific parameters that converge at rate sqrt(T) rather than sqrt(N). The paper shows through Monte Carlo experiments that variance estimates remain substantially biased even after Arellano-Bonhomme (2012) correction when T >= 20, while inter-decile ranges corrected by Jochmans-Weidner (2019) are more reliable.

Marginal Returns to Public Universities

Mon, 01 Jan 0001 00:00:00 +0000

This paper asks whether enrolling in an American public university generates positive net returns for marginal students — those who barely qualify for admission — and whether those returns justify public expenditures. The question is policy-relevant because marginal students have weak academic preparation, face high dropout risk, and the net returns to expanding admission margins are theoretically ambiguous.

The author assembles administrative records spanning all 35 public universities in Texas, covering the universe of Texas public high school graduates from 2004–2014 (approximately 2.7 million students). Texas public universities collectively enroll over 10 percent of all American public university students. The data link high school records (test scores, demographics, coursework, attendance, disciplinary infractions) to college application and admission records, postsecondary enrollment and degree completion records, financial aid packages, institutional expenditure data from IPEDS, and quarterly earnings records from the Texas Workforce Commission unemployment insurance system.

The identification strategy exploits hundreds of decentralized SAT/ACT score cutoffs in university admissions — varying across schools and application years — that generate sharp discontinuities in admission probability. A fuzzy regression discontinuity design compares applicants just above versus just below each cutoff. On average, crossing a cutoff raises the probability of admission by 27 percentage points and the probability of enrolling at the target university by 15 percentage points. Density tests and pre-college covariate balance validate the smoothness assumptions. The typical cutoff complier is more disadvantaged than the average college applicant but comparable to the average Texas high school graduate.

Roughly half of cutoff compliers would fall back to another, typically less selective, four-year institution if rejected; 43 percent would fall back to a two-year community college; and only about 6 percent would forgo higher education entirely. The pooled estimates therefore blend intensive-margin effects (more selective versus less selective four-year college) with extensive-margin effects (four-year college versus community college or no college).

Main causal findings for enrollment compliers: the typical marginally admitted student completes approximately one additional year of credits in the four-year sector and becomes 12 percentage points more likely to ever earn a bachelor’s degree from any institution. About half of the additional four-year credits are offset by 15 fewer credits in the two-year sector, and associate degree or certificate completion falls by 7 percentage points. All bachelor’s degree gains are in non-STEM fields; STEM degree completion shows no detectable increase. Compliers become about 3 percentage points more likely to hold a graduate degree by 10 years out.

On earnings, admitted compliers earn less than rejected counterparts in the first five years due to continued enrollment. Year six is the crossover point; by years 8–12, compliers earn a stable 8.6 percent earnings premium in log terms (8.2 percent in dollar ratio terms, representing a LATE of $3,339 against an untreated complier mean of $40,829), with earnings ranks rising approximately 4 percentiles from a base near the 50th percentile.

Marginally admitted students pay no additional net tuition on average: $4,600 in additional gross tuition is nearly fully offset by grant aid, though they take on $5,300 more in student loans. Society incurs approximately $10,000 in additional educational expenditures per complier. Internal rates of return are 26 percent for students, 16 percent for society, and 7 percent for the government budget. At a 3 percent discount rate, the lifetime net present value of enrolling the typical marginal applicant is approximately $80,000 — $70,000 accruing to the student and $10,000 to taxpayers.

Earnings gains are similar across institutions of varying selectivity, but significantly smaller for low-income compliers, who spend more time enrolled, complete fewer degrees, and major in less lucrative fields. A bounding method shows that extensive-margin compliers (those who would otherwise not attend any four-year college) experience larger effects than intensive-margin compliers.

Q: What is the core research question and why is credible evidence scarce? A: The paper asks whether enrolling marginal students in American public universities generates positive net returns — private, social, and fiscal — and what drives heterogeneity in those returns. Credible evidence is scarce because most existing work is correlational and fails to account for selection bias: individuals with more college education may have had pre-existing advantages, confounding college’s causal effect with systematic sorting into it. Even if average returns are positive, the policy-relevant question is whether the marginal student — who has weak preparation and high dropout risk — represents a good investment.

Q: What is the regression discontinuity design, and what does the first stage look like? A: The author infers hundreds of decentralized SAT/ACT score cutoffs across approximately 700 application cells (combinations of university, year, GPA quartile, and test type) by searching for the score value with the largest discontinuity in admission and enrollment within each cell. This procedure delivers a superconsistent estimator of each cell’s true cutoff. Pooled across all cells, crossing a cutoff raises the probability of admission by 27 percentage points and the probability of enrollment at the target university by a precisely estimated 15 percentage points. The density of applicants and a rich set of pre-college characteristics run smoothly through the cutoffs, supporting the exclusion restriction.

Q: Who are the cutoff compliers, and are they representative of any broader population? A: Compliers — applicants who enroll in the target university if and only if they barely cross its cutoff — comprise approximately 15 percent of marginal applicants. In observable characteristics, compliers are roughly representative of the broader population of marginal applicants at the cutoff. They are significantly more disadvantaged than the average public university applicant, but broadly comparable to the average Texas public high school graduate in terms of academic preparation and family income.

Q: What are the next-best alternatives for marginal applicants who are rejected? A: Approximately 47 percent of compliers would fall back to another Texas four-year college (mostly public), 43 percent to a two-year community college, and approximately 9 percent would not enroll in any Texas institution. National Student Clearinghouse data for the 2008–2014 cohorts confirm that only 4 percent of untreated compliers attend a college outside the THECB universe, meaning approximately 6 percent of all compliers truly forgo higher education altogether if rejected. The empirically relevant extensive margin is therefore between the four-year sector and the two-year sector, not between college and no college.

Q: How does cutoff crossing change the institutional characteristics a complier experiences? A: Compliers are propelled into substantially better-resourced environments: the average math test score of college peers rises by half a standard deviation; peers are 12 percentage points less likely to have been low-income; gross tuition rises by $2,400 (a 42 percent increase over the untreated complier mean of $5,700); educational spending per student rises by $3,200 (43 percent over the untreated mean); peers’ 10-year BA completion rate rises by 28 percentage points; and peer mean earnings 8–12 years after college entry are $6,700 higher.

Q: What are the educational attainment effects? A: Cutoff crossing causes compliers to complete approximately 28 additional credits at any four-year institution (roughly one full year of a four-year program) and increases the probability of ever earning a bachelor’s degree by 12 percentage points, raising the completion rate from approximately 40 percent to just above 50 percent. About 15 fewer two-year sector credits are offset against the four-year gains, and associate degree or certificate completion falls by 7 percentage points. All bachelor’s degree gains are in non-STEM fields; there is no detectable increase in STEM degrees. Graduate degree completion rises by approximately 3 percentage points by 10 years out.

Q: What is the earnings trajectory, and when does the premium materialize? A: Admitted compliers earn less than rejected counterparts in the first five years after application because they remain enrolled longer. Year six is the crossover point. By years 8–12, the earnings premium stabilizes at approximately 8.6 percent in log terms and 8.2 percent in dollar ratio terms (a LATE of $3,339 against an untreated complier mean of $40,829). Earnings rank rises by approximately 4 percentiles from a base near the 50th percentile. These results are robust across sandwich earnings, all-quarters-with-earnings, and zero-imputed specifications.

Q: What does the cost-benefit analysis show? A: Marginally admitted students pay no additional net tuition on average: $4,600 in additional gross tuition is nearly fully offset by additional grant aid. They do borrow $5,300 more in student loans, likely financing higher room, board, and consumption costs at four-year colleges. From society’s perspective, compliers generate approximately $10,000 in additional educational expenditures. Cumulative undiscounted earnings benefits surpass costs after 8 years for students, 11 years for society, and 19 years for taxpayers. At a 3 percent discount rate, the lifetime net present value is approximately $80,000 total — $70,000 accruing to the student and $10,000 to taxpayers — with internal rates of return of 26 percent for students, 16 percent for society, and 7 percent for the government budget.

Q: Does selectivity of the admitting institution predict larger earnings returns? A: No. Compliers at more selective institutions experience substantially larger increases in peer quality than those at less selective institutions, but they are also less likely to be on the extensive margin of four-year enrollment and experience smaller BA attainment gains. These factors roughly offset, producing no systematic difference in earnings gains across institutions of varying selectivity. More selective institutions also impose no additional cumulative cost on society, while compliers actually pay slightly less in additional net tuition at more selective schools.

Q: How does the commonly used measure of college value-added (mean peer earnings) compare to actual complier returns? A: Mean peer earnings overpredicts actual value-added for marginal students by a factor of two: compliers attend an institution with $6,700 higher average peer earnings as a result of admission but gain only $3,300 themselves. The measure also overpredicts the earnings return to selectivity by a factor of three: a 100-SAT-point increase in target school selectivity predicts $3,000 higher peer earnings but only a statistically insignificant $900 higher gain in the complier’s own earnings.

Q: How do earnings returns differ by family income? A: Compliers from low-income families experience significantly smaller earnings gains compared to higher-income compliers. The gap is not explained by differential changes in college quality induced by admission. Instead, low-income compliers gain fewer degrees despite spending more time in college and major in less lucrative fields, consistent with related findings in the literature on family income gaps in degree completion and major choice.

Q: How do earnings returns differ by gender and by race? A: Female and male compliers eventually earn similar log earnings and earnings rank gains, but women reach their gains more quickly — likely because men take longer to finish college. White and Asian compliers experience similar earnings gains and BA completion improvements as Black and Hispanic compliers, despite white and Asian students experiencing larger increases in college selectivity and spending per student as a result of admission.

Q: What is the method for separating intensive- and extensive-margin effects? A: The two complier types are not directly distinguishable in the data. The author first uses an endogenous but strong stratification variable — having at least one other Texas public university admission offer — to identify some mean potential outcomes for each type. He then imposes an empirically-informed rank assumption to bound the remaining unknown mean potential outcomes, delivering tightly informative upper and lower bounds on each margin’s effects without requiring full nonparametric identification. The results show that pooled effects are driven by larger returns for extensive-margin compliers who would not have attended any four-year college, with smaller contributions from intensive-margin compliers shifting between four-year institutions.

Q: How do this paper’s earnings estimates compare to prior studies, and what explains the differences? A: This paper’s 8 percent earnings gain is smaller than the 17–26 percent reported in prior studies (Zimmerman 2014: 22%; Kozakowski 2023: 26%; Smith, Goodman, and Hurwitz 2025: 17%; Bleemer 2024: 21%; Hoekstra 2009: 20%). The differences are likely explained by the much larger educational attainment and institutional quality gains induced by those studies’ natural experiments: in Zimmerman (2014), enrollment compliers gain roughly three additional years of four-year education versus one year in this paper; in Bleemer (2024), compliers experience roughly $30,000 more in institutional spending per student versus approximately $3,000 in this paper.

Q: What are the scope conditions for these results? A: The results pertain to marginal applicants to Texas public universities (excluding UT-Austin, which uses holistic admission with no detectable SAT/ACT cutoffs) from the 2004–2014 high school graduation cohorts. The identified effects are local average treatment effects for compliers — applicants who would enroll in the target university if and only if they barely crossed its admission cutoff — and do not represent effects for always-takers or infra-marginal students. Earnings are measured only for Texas-based workers covered by the state unemployment insurance system, which captures an estimated 90 percent of the civilian labor force.

Cutoff complier: An applicant who enrolls in their target university if and only if their SAT/ACT score barely exceeds that university’s admission cutoff. Compliers are the population whose behavior — and thus whose treatment effects — are identified by the fuzzy RD design. They comprise approximately 15 percent of marginal applicants and are more disadvantaged than the average public university applicant but broadly comparable to the average high school graduate.

Extensive versus intensive margin: The extensive margin refers to the contrast between attending any four-year college versus falling back to a two-year community college or no college. The intensive margin refers to the contrast between attending a more selective versus a less selective four-year institution. Approximately half of cutoff compliers are on each margin; the paper treats them as economically distinct parameters requiring separate identification.

Fuzzy regression discontinuity (RD) design: An identification strategy that uses the discontinuous jump in admission probability at a test score cutoff as an instrument for enrollment, recovering the LATE for compliers via the ratio of the reduced-form discontinuity in outcomes to the first-stage discontinuity in enrollment. “Fuzzy” refers to the fact that crossing the cutoff changes admission and enrollment probabilities with a discrete jump rather than with certainty.

Internal rate of return (IRR): The discount rate at which the net present value of an investment equals zero — here, the discount rate equating the discounted stream of earnings benefits to the discounted stream of costs. The paper estimates IRRs separately for students (26 percent), society (16 percent), and the government budget (7 percent), reflecting different cost and benefit definitions from each perspective.

Rank assumption (bounding method): An empirically-informed assumption about the ordering of mean potential outcomes across latent complier types (extensive vs. intensive margin) that, combined with partial identification from a strong endogenous stratification variable, yields tight upper and lower bounds on each margin’s causal effects without requiring full nonparametric identification.

Net tuition: Gross tuition charges minus grant aid. For the typical marginal complier, gross tuition rises by $4,600 but is nearly fully offset by additional grant aid, yielding approximately zero additional net tuition cost — meaning the private financial cost of attending a public university for marginal students is effectively zero on net, though they take on $5,300 more in student loans to finance room, board, and consumption.

Sandwich earnings measure: A procedure applied to quarterly state earnings data that retains only quarters with positive earnings sandwiched between other quarters with positive earnings, discarding high-variance transition quarters between employment spells. Annualized by multiplying the quarterly average by four; used to reduce noise from entry and exit transitions in administrative earnings records.

The Effect of Education Policy on Crime: An Intergenerational Perspective

Mon, 01 Jan 0001 00:00:00 +0000

This paper studies the intergenerational effects of education policy on crime, asking whether a compulsory schooling reform that reduced crime among those directly exposed also reduced crime among their children. The authors exploit the staggered municipal rollout of Sweden’s comprehensive school reform, implemented gradually between 1949 and 1962 across more than 1,000 municipalities, which increased compulsory schooling by one to two years, abolished tracking into academic and vocational streams after 6th grade, and introduced a uniform national curriculum. The parent generation consists of all individuals born in Sweden between 1945 and 1955 (approximately 447,000 men and 450,000 women), and their children form the child generation (426,721 sons observed from age 15 to 29). Crime is measured by administrative conviction records from the Swedish National Council for Crime Prevention covering 1973–2010.

The empirical strategy is difference-in-differences, comparing changes in conviction rates across cohorts in municipalities that implemented the reform at different times, with treatment assigned based on the parent’s birth municipality to avoid endogenous sorting bias. Standard errors are clustered at the municipality level. Parallel trends validity is supported by three tests: results are unchanged when municipality-specific linear trends are included, placebo tests using incorrect reform dates yield effects indistinguishable from zero, and residuals from crime regressions show no correlation with municipality-specific trends.

The main finding is a significant 0.79 percentage point (pp) decline in conviction rates among sons of fathers exposed to the reform (p-value < 0.002), representing a 3.4 percent reduction relative to baseline. The decline spans multiple crime types: violent crime fell by 0.27 pp, traffic-related crime by 0.45 pp, fraud by 0.22 pp, and other offenses by 0.41 pp — percentage reductions of three to six percent across categories. Multiple convictions fell by 0.43 pp (5.8 percent). These second-generation effects are driven entirely by paternal exposure: the impact of maternal reform exposure is an order of magnitude smaller and statistically insignificant, and the difference between paternal and maternal effects is itself significant (p-value 0.048 for any conviction, 0.009 for multiple convictions). Effects on daughters in the child generation are much smaller, with only the residual “other crime” category showing a significant 0.129 pp (15.5 percent) decline.

The asymmetry between paternal and maternal transmission is explained by the first-generation effects of the reform. For men, the reform increased schooling by 0.32 years, earnings by approximately 1 percent, the probability of white-collar employment by 1.2 percent, cognitive skills by 0.14 standard deviations, noncognitive skills by 0.17 standard deviations, spousal earnings by 1,022 SEK per year, and overall household income by approximately 1 percent. For women, the reform increased education by 0.21 years but did not raise earnings, household income, or white-collar employment, and did not reduce their already low crime rates. Only 13 percent of women in the 1945–55 cohorts were at or below the compulsory schooling threshold, versus 20 percent of men, substantially limiting the reform’s bite for women.

A mediation analysis decomposes the intergenerational transmission through three channels: fathers’ education accounts for 64.8 percent of the indirect effect, the decline in paternal crime accounts for 18.5 percent, and the increase in household disposable income accounts for 16.7 percent. The direct effect (unexplained by these mediators) accounts for 48 percent of the total effect. The paper also documents that children of treated fathers attended schools with lower peer crime rates and lived in neighborhoods with lower youth crime rates, supporting a neighborhood and peer effects channel alongside human capital and role-model channels.

Scope conditions: the study covers male children observed to age 29 in Sweden; results apply to a context of near-universal administrative records, a specific postwar schooling reform, and cohorts born 1945–1955 in a Nordic welfare state.

Q: What is the magnitude of the intergenerational crime reduction caused by the reform?

A: Sons of fathers exposed to the reform experienced a 0.79 pp decline in conviction rates (p-value < 0.002), corresponding to a 3.4 percent reduction relative to the baseline conviction rate of approximately 24 percent for the child generation by age 29. Multiple convictions fell by 0.43 pp, a 5.8 percent reduction. These magnitudes are similar in percentage terms to the direct crime reduction the reform caused among fathers themselves.

Q: Does the reform’s intergenerational effect on crime differ by the sex of the treated parent?

A: Yes. The intergenerational effect is driven entirely by paternal exposure to the reform: the effect of maternal exposure is an order of magnitude smaller and insignificant at any conventional significance level. The difference between paternal and maternal effects is statistically significant, with p-values of 0.048 for any conviction and 0.009 for multiple convictions. The paper attributes this asymmetry to the much weaker first-generation effects of the reform on women’s earnings, household income, crime rates, and neighborhood sorting.

Q: Which crime types declined significantly among sons of treated fathers?

A: Significant declines were found in violent crime (−0.27 pp, Romano-Wolf p-value 0.09), traffic-related crime (−0.45 pp, RW p-value 0.057), fraud (−0.22 pp, RW p-value 0.09), and other offenses (−0.41 pp, RW p-value 0.047), each representing a three-to-six percent reduction relative to the mean incidence of that crime type. Property crime and drug-related crime did not show significant declines.

Q: What were the direct effects of the reform on the parent generation’s human capital?

A: For men, the reform increased schooling by 0.32 years, earnings by approximately 1 percent, the probability of white-collar employment by 1.2 percent, cognitive skills by 0.14 standard deviations, and noncognitive skills by 0.17 standard deviations, all measured at military enlistment. Spousal earnings increased by 1,022 SEK per year and overall household income rose by approximately 1 percent. For women, education increased by 0.21 years and marriage market matches improved, but earnings, household income, and white-collar employment probability did not increase significantly.

Q: Why did the reform have stronger first-generation effects on men than on women?

A: The average share of individuals at or below the compulsory schooling threshold — the margin at which the reform was binding — was 20 percent for men but only 13 percent for women in the 1945–55 cohorts. Because fewer women were constrained by the old compulsory schooling limit, the reform increased their education by less and produced smaller downstream effects on earnings and labor market outcomes.

Q: What are the three channels through which the reform reduces child crime, and what is the relative contribution of each?

A: The paper identifies three channels: (1) the human capital channel, whereby increased parental education raises household income and child human capital; (2) the role model channel, whereby reduced paternal crime participation directly reduces son’s crime; and (3) the neighborhood and peer effects channel, whereby higher income enables sorting into lower-crime neighborhoods and better schools. The mediation analysis attributes 64.8 percent of the indirect effect to fathers’ increased education, 18.5 percent to the decline in paternal crime, and 16.7 percent to the increase in household disposable income. The direct effect unexplained by these three mediators accounts for 48 percent of the total effect.

Q: What is the role model effect, and how strong is it in the parent generation?

A: The role model channel operates through the strong intergenerational persistence in crime participation: sons are 2.06 times more likely to participate in crime if their fathers have been convicted (Hjalmarsson and Lindquist, 2012). The reform reduced the incidence of any conviction among treated men by 1.5 pp and repeat convictions by 1.5 pp — the latter representing an approximately 8 percent decline from a lower base. For women, the reform produced no reduction in crime, providing no analogous role model improvement through the maternal channel.

Q: How does neighborhood and school peer quality change for children of treated fathers versus treated mothers?

A: Sons of fathers exposed to the reform moved to neighborhoods with lower youth crime rates (−0.087 pp) and attended schools with lower peer crime rates (−0.077 pp). In contrast, sons of mothers exposed to the reform experienced higher neighborhood crime rates (p-value 0.06) and higher school peer crime rates (p-value 0.01), the opposite direction. This asymmetry helps explain why only paternal treatment generates significant second-generation crime reductions.

Q: What happens to other outcomes for children of treated fathers beyond crime?

A: Sons experienced a 1.2 percentile increase in school GPA (RW p-value 0.05), a 2.3 pp increase in employment (RW p-value 0.04), a matching 2.3 pp decline in unemployment benefit receipt, a reduction in hospitalization of 2.4 days (17 percent, RW p-value 0.02), and a decline in prescribed drugs of 31 doses (2.8 percent, RW p-value 0.09). The decline in prescribed drugs for sons is driven by nervous system drugs and painkillers, pointing to improved mental health. Daughters of treated fathers show a significant reduction in welfare dependency but no other significant improvements.

Q: How does the paper validate the parallel trends assumption?

A: Three tests are reported. First, including municipality-specific linear trends leaves the main coefficient unchanged (p-value 0.85 for the trend terms themselves). Second, placebo contrasts using incorrect reform implementation dates produce effects indistinguishable from zero for all tested dates. Third, graphical inspection of regression residuals shows no correlation with municipality-specific trends. Together these provide strong support for the identifying assumption.

Q: Are the results sensitive to using a linear probability model instead of a nonlinear model?

A: A Monte Carlo experiment was conducted replicating observed crime rates across municipalities and imposing the estimated average treatment effect. Assuming the true data-generating process is a probit model, the linear probability model biases the estimated average effect upward by only 5 percent — a difference that is statistically indistinguishable from zero in the actual data — validating the OLS approach.

Q: What is the broader policy implication of the findings?

A: The results show that well-designed education policies can reduce crime not only among the directly treated generation but also among their children, amplifying the social benefits of reform across generations. The authors interpret this as consistent with the theoretical framework of Becker and Tomes (1979) on intergenerational transmission of human capital, and suggest that education policy evaluations that focus only on the treated generation substantially understate total social returns.

Intergenerational transmission of education reform effects: the phenomenon whereby an education policy that raises parental human capital produces improvements in children’s outcomes — including crime — through multiple channels including resource increases, parental role modeling, and neighborhood sorting, beyond any direct policy exposure of the child generation.

Comprehensive school reform (Sweden, 1949–1962): a nationally mandated restructuring of compulsory schooling that extended required attendance by one to two years, abolished selection into academic and vocational tracks after 6th grade, and introduced a uniform national curriculum, rolled out staggered across 1,055 Swedish municipalities.

Human capital channel: the mechanism by which increased parental education raises earnings and household income, enabling greater investments in children’s development and exploiting complementarity between parental and child human capital in the skill production function, thereby raising children’s opportunity cost of crime.

Role model channel: the mechanism by which reduced parental crime participation directly reduces children’s crime, operating through the transmission of norms and information across generations; identified empirically by the strong intergenerational correlation in convictions (sons with convicted fathers are 2.06 times more likely to be convicted themselves).

Neighborhood and peer effects channel: the mechanism by which increased parental income from the reform enables sorting into residential neighborhoods and schools with lower youth crime rates, exposing children to peers less involved in illegal activities and thereby reducing their own crime participation.

Mediation analysis: a decomposition method following Heckman, Pinto, and Savelyev (2013) that quantifies the share of a total treatment effect accounted for by specific intermediate variables (here: fathers’ education, fathers’ crime participation, and household disposable income) versus the direct unexplained effect.

Conviction rate: the proportion of individuals in a given generation and observation window who received at least one criminal conviction in Swedish administrative records; used as the primary outcome measure because it captures offenses that led to a court appearance, excluding minor infractions resolved by direct fine.

The Impact of EITC on Education, Labour Market Trajectories, and Inequalities

Mon, 01 Jan 0001 00:00:00 +0000

This paper studies the effect of the Earned Income Tax Credit (EITC) on educational attainment and labor market trajectories through two complementary approaches. Using policy discontinuities at U.S. state borders—exploiting variation in state EITC generosity set as a percentage of the federal EITC—the paper finds that an increase in the state EITC leads to a statistically significant increase in the high school dropout rate. The mechanism is that a tax credit targeted at low-wage (low-skilled) workers increases the value of low-skilled employment and reduces the relative return to schooling, generating a powerful disincentive to pursue long-term studies. A structural life-cycle matching model with directed search and endogenous educational choices, search intensities, hirings, hours worked, and separations is developed to quantify the long-run general equilibrium effects: in the long run, EITC reduces the proportion of high-skilled workers, with ambiguous effects on income inequality that depend on the competing channels through which EITC affects both the supply and demand sides of the labor market.

Summary based on a working paper version, AI-assisted and human-reviewed. See the linked published article for the authoritative version.

In depth

Q1. What is the empirical strategy for identifying the effect of EITC on education?

The paper identifies the causal effect of state EITC on education by exploiting policy discontinuities at U.S. state borders, comparing contiguous PUMA pairs on opposite sides of state borders that differ in state EITC generosity. State EITC rates are set as a percentage of the federal EITC and have varied considerably since the mid-1980s. Borrowing from the minimum wage literature (Dube et al., 2010; Hagedorn et al., 2015), the border-discontinuity design controls for local labor market conditions that vary continuously across state borders while isolating the effect of the discrete EITC policy difference.

Q2. What is the labor market mechanism linking EITC to education?

EITC raises the value of low-skilled employment by directly increasing the earnings of low-wage workers, which in turn reduces the relative return to investing in education, generating a powerful disincentive to pursue long-term studies. When directed search is present—as supported by recent empirical studies—educational decisions affect both job-finding probabilities and labor incomes over the life cycle. EITC’s subsidization of low-skilled work contracts the education premium in this framework, making the forgone earnings cost of staying in school larger relative to the low-skilled employment option supported by the EITC.

Q3. What does the life-cycle matching model contribute?

The structural life-cycle matching model with directed search and endogenous educational choices, search intensities, hirings, hours worked, and separations quantifies the general equilibrium and long-run effects of EITC that purely reduced-form studies cannot capture—including the feedback of an expanded low-skilled labor force on equilibrium wages and job creation. The model endogenizes labor demand, capturing both household responses (education, hours, search intensity) and firms’ responses (job creation and destruction). It is solved and estimated to replicate the life-cycle profile of labor market variables.

Q4. What are the long-run implications for inequality?

In the long run, EITC reduces the proportion of high-skilled workers in the economy, with ambiguous effects on income inequality because of offsetting channels: EITC directly increases earnings of low-skilled workers, but by expanding the supply of low-skilled labor it may also depress low-skilled wages; additional channels through unemployed workers’ search effort and employed workers’ hours further complicate the net effect. The model is used to determine the optimal design of the EITC that balances the income-support objective against these unintended long-run effects.

Key concepts

state EITC : a supplement to the federal Earned Income Tax Credit set as a fixed percentage of the federal credit; varies across states; used in this paper as the identification source for the effect of EITC generosity on education via border discontinuities. directed search : a labor market framework in which workers and firms direct their search to specific submarkets with posted wages; in this setting, educational choice affects both job-finding probabilities and wages over the life cycle, amplifying the disincentive effects of EITC on education relative to random-search models. education-EITC disincentive : the mechanism by which EITC targeted at low-wage workers raises the relative value of low-skilled employment and reduces the return to schooling, generating an increase in high school dropout rates as a side effect of the anti-poverty policy.

What Works and for Whom? Effectiveness and Efficiency of School Capital Investments Across the U.S.

Mon, 01 Jan 0001 00:00:00 +0000

What Works and for Whom? Effectiveness and Efficiency of School Capital Investments Across the U.S.

Research Question

This paper investigates which types of school facility investments benefit students (as measured by test scores) and are valued by homeowners (as measured by house prices), and for which student populations these investments are most effective. Prior state-level studies had reached conflicting conclusions about the returns to school capital spending, and no nationwide evidence had distinguished impacts across spending categories or student backgrounds.

Data and Methodology

The authors assemble a novel panel dataset covering approximately 14,000 school bond referenda in 29 U.S. states and 10,146 districts enrolling 71% of all U.S. students, for the period 1990–2017. The dataset combines: (1) ballot-level bond election records including vote shares, proposed amounts, and ballot text; (2) district-level test scores from the Stanford Education Data Archive (SEDA) extended backward to 2003 for all states and as early as 1995 for some, normalized to a national scale via NAEP; (3) a Census-tract-level house price index (Contat and Larson, 2022) aggregated to school districts; and (4) NCES district finance and demographic data.

Bond ballot texts are classified into eight spending categories using text-analysis: classroom construction/renovation; HVAC; other infrastructure (plumbing, roofs, furnaces); safety and health (pollutant removal, building safety); STEM equipment and labs; athletic facilities; land purchases; and transportation vehicles.

The identification strategy exploits quasi-random variation from close bond elections, building on the dynamic regression discontinuity (DRD) framework of Cellini et al. (2010). A key methodological contribution is a stacked DRD design that addresses heterogeneous treatment effects correlated with timing: each treatment cohort (districts that narrowly authorize a bond in year c) is matched against “clean controls” — districts that also proposed a bond in the same cohort but narrowly failed to authorize it and did not authorize any bond in the following ten years. Cohorts are stacked, and a dynamic RD model is estimated controlling for cohort fixed effects and a district’s bond proposal history.

Main Findings with Quantitative Magnitudes

Average effects. Bond authorization raises capital spending by approximately $1,650 per pupil cumulatively over five years. Test scores increase gradually, reaching 0.079 standard deviations (sd) higher five to eight years after authorization, and 0.073 sd higher nine to twelve years after. 2SLS estimates, amortizing spending over a 30-year project life at a 9% depreciation rate, imply that a $1,000 increase in the flow value of capital spending raises test scores by 0.048 sd. House prices rise by approximately 9% eight to nine years after authorization. When house price effects are estimated against only locally-financed capital spending (not state aid), the 2SLS estimate is 0.8% per $1,000 — roughly consistent with efficiency — suggesting that the larger reduced-form house price response is driven primarily by state aid that supplements local funds rather than by an inefficiently low ex ante spending level.

Heterogeneity by spending category. Category-specific estimates reveal that only certain project types raise test scores: HVAC (+0.20 sd, largest effect), safety and health (+0.15 sd), other infrastructure/plumbing/roofs (+0.15 sd), STEM equipment (+0.15 sd implied), and classroom space (+0.10 sd), all measured three to six years post-election. By contrast, bonds for athletic facilities, land purchases, and transportation produce no detectable effects on test scores. The pattern for house prices is the inverse: athletic facilities generate a 17% house price increase; classroom space generates 14%; STEM generates 11% — while HVAC and safety/health bonds produce no significant effect on house prices. The correlation between category-level test score and house price estimates is −0.07, indicating these are largely orthogonal outcomes.

Heterogeneity by student socioeconomic status. Effects are concentrated in districts serving socioeconomically disadvantaged students (top tercile of the share of students eligible for free or reduced-price meals, denoted low-SES). In low-SES districts, bond authorization raises test scores by 0.13 sd after seven years and house prices by 15%; in high-SES districts, neither outcome shows a significant effect. 2SLS estimates confirm that a $1,000 increase in cumulative spending raises test scores by 0.08 sd in low-SES districts but produces no detectable change in high-SES districts. The SES gradient persists after conditioning on spending amounts, spending categories, and baseline capital stock, indicating that students in disadvantaged districts have higher marginal returns to capital improvements independent of these channels. High-minority districts (top tercile of Black and Hispanic share) similarly see a 0.12 sd test score gain and 15% house price gain after seven years, versus 0.04 sd and 3% in low-minority districts.

Role of baseline capital stock. Among districts with below-median capital stock, test score effects are 0.20 sd in low-SES districts seven years post-election. Even among above-median-stock districts, low-SES districts see house price effects exceeding 10% while high-SES districts see no effect. Differences by SES persist after conditioning on capital stock.

Policy simulation. Closing the spending gap between high- and low-SES districts (approximately $1,000 over 10 years) without changing the composition of spending would raise low-SES test scores by roughly 0.08 sd, closing about 8% of the roughly 1 sd achievement gap. Targeting that same additional spending toward HVAC and safety/health (the highest-impact categories) would generate test score increases approximately three times as large, potentially closing up to 25% of the observed achievement gap.

Reconciling prior literature. Replicating state-level estimates, the authors show that Ohio’s positive effects are explained by a high share of bonds in low-SES districts funding infrastructure, while Texas’s near-zero effects reflect a high share of bonds in higher-SES districts funding classrooms and athletic facilities.

Q&A: Analytical Steps, Mechanisms, and Robustness

Q1: What is the first-stage effect of bond authorization on capital spending, and does it contaminate other spending categories?

A1: Bond authorization raises per-pupil capital spending by approximately $700 per year at two years post-election and $590 at three years, with cumulative spending $1,650 higher over five years in treated districts relative to districts that narrowly failed to authorize a bond. Bond revenues are legally restricted to capital uses, and the paper confirms that non-capital (current) spending and instructional spending are not affected following authorization. This establishes a clean first stage: bond authorization raises only capital outlays.

Q2: Why does the standard DRD estimator of Cellini et al. (2010) require refinement, and what problem does the stacked DRD design solve?

A2: The original CFR estimator assumes treatment effects are uncorrelated with the timing of treatment — an assumption potentially violated when, for example, bonds financing HVAC (high-impact) versus athletic facilities (amenity-focused) have different propensities to be proposed at different points in time. The stacked DRD design avoids “forbidden comparisons” by comparing each treatment cohort only against clean controls that propose but fail to authorize a bond in the same year and do not authorize any bond in the subsequent ten years. This ensures consistency even when treatment effects are heterogeneous across cohorts and correlated with timing.

Q3: How do the authors validate the quasi-random assignment assumption of the regression discontinuity design?

A3: Three tests are performed. First, a McCrary (2008) density test on the vote margin distribution shows no discontinuity at the cutoff in the pooled or stacked data (p-values of 0.59 and 0.24, respectively), though discontinuities are found in Arkansas, Missouri, and Oklahoma — those three states are excluded. Second, pre-election district covariates (income, education, SES shares, enrollment, revenues, expenditures) are smooth around the cutoff in both datasets. Third, pre-election trends in test scores and house prices are flat and parallel between marginally approved and marginally rejected districts.

Q4: How are the eight spending categories constructed, and how many bonds are successfully classified?

A4: Categories are drawn from the SchoolBondFinder.com classification produced by The Amos Group, then refined by splitting capital improvements into HVAC versus other infrastructure, splitting construction/renovation into classroom versus athletic facility projects, and adding land purchases as a separate category. Keyword-based text analysis of ballot language successfully assigns 75% of the approximately 14,000 bonds to at least one of the eight categories. More than two-thirds of classified bonds receive multiple category designations, with a mean of 2.9 categories per proposed bond and 3.2 per authorized bond.

Q5: Why do HVAC bonds raise test scores but not house prices, while athletic facility bonds raise house prices but not test scores?

A5: The authors interpret this divergence as reflecting what different types of improvements offer to different stakeholders. HVAC improvements reduce excessive heat and air pollution exposure in classrooms, directly improving students’ learning experiences — consistent with Park et al. (2020) on heat and Gilraine and Zheng (2022) on air pollution. These improvements are not visibly salient to homeowners without school-age children and carry no amenity value for the broader community. Athletic facilities, by contrast, are highly visible and provide a community amenity valued in the housing market regardless of their impact on academic instruction. The near-zero correlation (−0.07) between category-level test score and house price estimates confirms that the two outcomes respond to largely distinct features of capital investments.

Q6: What are the three candidate explanations for the larger effects of bond authorization in low-SES districts, and which explanations survive empirical scrutiny?

A6: The three candidates are: (1) larger spending increases after authorization in low-SES districts; (2) a different composition of spending categories (more toward high-impact HVAC and safety); and (3) higher marginal returns per dollar for disadvantaged students, holding spending size and composition fixed. The data confirm all three operate, but the third is the residual: 2SLS estimates show a $1,000 increase raises test scores by 0.08 sd in low-SES districts versus a statistically zero effect in high-SES districts, and within-category estimates show HVAC bonds raise scores by 0.27 sd in low-SES districts but have no detectable effect in high-SES districts. Differences by SES also persist after conditioning on the estimated baseline capital stock, though low capital stock accounts for part of the gap.

Q7: How does the role of state aid alter the interpretation of the house price effect for spending efficiency?

A7: A 9% house price increase after bond authorization, if taken at face value under Brueckner’s (1979) efficiency test, would suggest the ex ante level of school capital spending was inefficiently low. However, state grants that partly match local bond revenues raise actual spending without raising local property taxes proportionally. When the 2SLS house price effect is estimated against only locally financed capital spending (using proposed bond size as the relevant measure), the implied house price increase is just 0.8% per $1,000 — consistent with rough efficiency on average across the full sample. The authors conclude that the large reduced-form house price response is driven primarily by the capitalization of state aid, not by an undersupply of capital investments at the aggregate level.

Q8: Does household sorting account for the observed test score and house price gains following bond authorization?

A8: Bond authorization produces small but detectable compositional changes: the share of high-SES students is approximately 3 percentage points higher seven years after an election (a roughly 4% increase relative to an average share of 0.73), while enrollment and the share of white students are largely unaffected. However, controlling for district-by-year shares of each sociodemographic group only slightly attenuates the test score and house price estimates, indicating that sorting accounts for a small share of the observed gains.

Q9: Are the findings robust to alternative research designs?

A9: The results are robust to five alternative estimation approaches: (1) the original one-step TOT estimator of Cellini et al. (2010); (2) a version of the stacked DRD where clean controls are districts that do not approve any bonds in the full [c−5, c+10] window; (3) a version that matches treated and control districts in each cohort based on bond history; (4) a version not controlling for future bond history; and (5) the extended two-way fixed effects (ETWFE) estimator of Wooldridge (2021). Results are also robust to linear polynomials with different slopes and quadratic polynomials of the vote margin.

Q10: How does the capital stock measure illuminate mechanism, and what are its limitations?

A10: The authors construct a district-level capital stock as the 30-year depreciated sum of capital spending from Census of Governments data (1967–2017) at a 5% depreciation rate. This stock is negatively correlated with the share of low-SES students, confirming that more disadvantaged students attend schools in worse structural condition. Conditioning on this proxy, the SES gradient in bond impacts is reduced but remains. Among districts with below-median capital stock, low-SES districts see test score gains of 0.20 sd after seven years, while among above-median-stock districts the gap narrows to approximately 0.10 vs. 0.05 sd. A key limitation is that detailed school-condition data are unavailable nationally, so the capital stock is a proxy only.

Q11: What is the quantitative policy implication of the targeting exercise?

A11: On average, low-SES districts receive about $97 per pupil per year less in capital spending than high-SES districts, so closing this gap over ten years implies approximately $970 in additional cumulative spending. Without changing spending composition, this would raise test scores by roughly 0.08 sd in low-SES districts, closing about 8% of the approximately 1 sd achievement gap between high- and low-SES districts. Redirecting that same additional spending toward the highest-impact categories (HVAC and safety/health) would generate test score gains roughly three times larger, potentially closing up to 25% of the observed achievement gap.

Q12: How do the cross-state differences documented in prior literature map onto the paper’s heterogeneity findings?

A12: The authors replicate earlier state-level estimates and show that Ohio’s relatively large positive effects — found by Conlin and Thompson (2017) — are explained by a high concentration of bonds in low-SES districts funding infrastructure, while Texas’s near-zero effects — found by Martorell et al. (2016) — reflect a high share of bonds in higher-SES districts funding classrooms and athletic facilities. Wisconsin and Michigan, which showed null effects in earlier studies, similarly have bond compositions and student demographics that predict small impacts under the paper’s heterogeneity framework.

Key Concepts

Stacked Dynamic Regression Discontinuity (Stacked DRD). The paper’s primary estimation strategy, which combines the dynamic RD framework of Cellini et al. (2010) with a stacked-cohort design adapted from the staggered difference-in-differences literature. For each treatment cohort (year in which a bond barely passes), “clean controls” are defined as districts that also proposed a bond in the same year but narrowly failed to authorize it and did not authorize any subsequent bond within ten years. Cohort-specific datasets are stacked and estimated jointly with cohort fixed effects, ensuring that estimates are robust to treatment effect heterogeneity correlated with timing.

Clean Controls. Districts used as the counterfactual for treated districts in a given cohort: those that propose a bond in the same year as the treated cohort, barely fail to authorize it, and remain untreated for ten subsequent years. Their “clean” status is quasi-random because their future non-authorization results from narrow electoral loss rather than any endogenous district choice.

Bond Spending Categories. Eight mutually-non-exclusive classifications of bond spending derived from ballot text using keyword analysis: classroom space; HVAC; other infrastructure (plumbing, roofs, furnaces); safety and health (pollutant removal, compliance upgrades); STEM equipment and labs; athletic facilities; land purchases; and transportation. These categories are defined in the paper not by administrative accounting codes but by the stated intended use of funds in ballot language.

Treatment-on-the-Treated (TOT) Estimator. The CFR estimator that captures the effect of bond authorization against the counterfactual of never authorizing a bond in the foreseeable future, achieved by including leads and lags of a district’s bond proposal history as controls. This addresses the problem that multiple elections over time make simple treated-vs-control comparisons confounded by past and future bond activity.

Capital Stock (District-Level Proxy). A measure of each district’s accumulated school facility capital at a given point in time, constructed as the depreciated 30-year running sum of capital expenditures from the Census of Governments, using a 5% annual depreciation rate. Used as a proxy for facility conditions in the absence of nationally available building-quality data, and confirmed to be negatively correlated with district share of low-SES students.

Brueckner Efficiency Test. An application of the theoretical framework linking public good provision levels to house price responses. If a spending increase raises house prices, the initial spending level was below the efficient level; if it lowers house prices, spending was too high. In this paper, the test is refined to use only locally-financed capital spending as the explanatory variable, to strip out the capitalization of state aid and isolate the efficiency assessment for locally-determined spending.

Socio-Economic Status (SES) Terciles. Districts are ranked by the share of students eligible for free or reduced-price school meals as of 1995. “Low-SES districts” refers to those in the top tercile of this share (most disadvantaged); “high-SES districts” refers to those in the bottom tercile (least disadvantaged). Effects are estimated separately for these subsamples throughout.