– Good Grades for Hard Work? A Lab-in-the-Field Study of Effort and Educational Inequality

Gil-Hernández, C.J., Palacios-Abad, A., and Radl, J.

DiSIA Working Paper 2025/09

https://labdisia.disia.unifi.it/wp_disia/2025/wp_disia_2025_09.pdf

Despite its importance for status attainment and meritocracy, measuring effort remains elusive, often relying on indirect proxies or unreliable self-reports. This study examines how objective (cognitive effort, CogEff) and subjective (teacher-perceived effort, TpEff) measures of student effort contribute to educational inequality. We examine the predictive capacity of effort for educational performance and test the mediating and moderating roles of effort in the relationship between parental socioeconomic status (SES) and school grades. Drawing on original, representative “lab-in-the-field” data from 1,270 fifth-graders in Spain and Germany, who performed three different incentivized real-effort tasks engaging various executive functions, four key findings emerge. First, both CogEff and TpEff predict grade point average (GPA), with TpEff having a powerful effect, more predictive even than IQ or parental SES. Second, effort—especially TpEff—is unequally distributed by parental SES and explains a substantial share of the SES-based GPA gap, on par with IQ. Third, roughly half of the GPA gap by social origin remains unexplained even after accounting for academic merit (IQ + effort). Fourth, while grading returns to CogEff are independent of SES, high-SES students are significantly less penalized for low TpEff than low-SES peers. Overall, effort predicts academic success and shapes educational (in)equality. High-SES students show higher average effort and can afford to be perceived as lazy, while hardworking low-SES students can overcome disadvantage through greater returns to teacher-perceived effort. We discuss the findings’ implications for student agency, educational inequality, and fair evaluations.

– Unfair Educational Inequality Worldwide: A Theory-Driven Machine Learning Approach

Brunori, P., Gil-Hernández, C.J., and Triventi, M. (In progress)

Social stratification research extensively scrutinised educational inequality patterns, yet three limitations persist: 1) unsystematic theoretical and empirical conceptualization of inequality of educational opportunity (IoP); 2) failure to account for individual circumstances’ intersectionality; 3) a focus on Western countries. This study addresses these gaps with a theoretically-informed analysis of unfair educational inequalities by student’s ascribed circumstances (Roemer 1998), examining PISA data (2003-2022) across 390 country-years. We adopt a data-driven machine learning approach reducing estimation bias and modelling arbitrariness: IoP captures outcome distribution inequalities between complex socio-demographic types—by sex, migrant background, home language, geographical area, and parental education/occupation. The share of unfair IoP ranges from 10% to 30% over total math achievement variation, with a trendless pattern, considerable cross-national disparities, and father/mother occupation as the most important circumstance. There is no efficiency-equality trade-off since mean achievement and IoP negatively correlate. We discuss the implications of our findings for IoP formalization in social stratification research.

– Within-Couple Child Penalty by Assortative Mating: Household Specialization or Doing Gender?

Gil-Hernández, C.J., Brini, E., Guetto, R., Maitino, M, Ravagli, L., and Vignoli, D. (In progress)

A large literature documents gendered earnings responses to childbirth, yet most studies adopt an individual perspective and overlook that child penalties are generated within couples. We develop a dyadic life-course approach to examine how the first childbirth reshapes partners’ relative income trajectories and how pre-birth resource asymmetries moderate these penalties in the traditional Southern European context. Using Italian administrative tax records, we reconstruct a retrospective panel of nearly 85,000 married couples in Tuscany from 2003 to 2022. We extend the standard event-study framework by incorporating couple-fixed effects, FE-individual-slope specifications that absorb unit-specific linear trends, and dynamic difference-in-differences estimators to address staggered childbirth timing and heterogeneous treatment effects. The findings reveal a large and persistent child penalty, reaching nearly 58 percentage points by ten years after birth on the extensive margin and about 22 percentage points on the intensive margin. The penalty is smallest among hypogamous couples, where women’s educational advantage appears to protect against post-birth losses. By contrast, it is the largest among couples in which women were the primary earners before childbirth, suggesting that couples whose economic arrangements deviate most strongly from conventional gender norms re-traditionalize most sharply upon parenthood. We discuss implications for competing accounts of household specialization, bargaining, and gender norms, and argue that education and relative earnings capture distinct dimensions of intra-couple status.

– The Gender Grading Gap in Italy: Classroom Heterogeneity and Trajectories from Student Population Data

Abbiati, G., Gil-Hernández, C.J., and Lievore, I. (In progress)

Girls, historically disadvantaged, outperform boys in education across most OECD countries, despite persistent gender gaps in STEM and labour markets. While boys outperform girls in standardised math assessments, girls consistently receive higher grades across subjects. This paradox has sparked growing interest in the Gender Grading Gap (GGG), defined as the advantage girls receive in teacher-assigned grades compared to boys with similar performance on external assessments. The GGG has been mainly attributed to gender differences in classroom behaviour and teachers’ implicit stereotypes. This study investigates three dimensions of the GGG not fully addressed in the literature: (1) variation across school subjects, (2) the role of school/classroom context and peer composition, and (3) temporal dynamics over the school career. First, we provide a systematic, theory-driven analysis of the GGG in literacy (Italian) and numeracy (Math). Second, we examine contextual moderators—such as classroom gender balance, socioeconomic composition, and migrant background—highlighting how peer effects may influence behaviour, performance norms, and grading criteria in gendered ways. We also test whether grading practices reflect compensatory/reinforcing preferences by classroom-level gender gaps in standardised tests. Third, we assess trends in the GGG across grades and cohorts. This study utilises INVALSI population data, encompassing over 7 million students (grades 2, 5, and 8) across 371,423 classrooms from 2012 to 2024. Estimation is performed separately by subject and grade at the classroom level, allowing analysis of grading heterogeneity and moderation by class gender, ESCS, and migration composition. The study finds that the GGG is larger in Italian and lower-secondary schools and shaped by classroom context.

– Trends and Mechanisms of Unfair Educational Inequality in Spain: A Normative Approach with Machine Learning

Gil-Hernández, C.J., and Salas-Rojo, P. (In progress)

This article formalizes John Roemer’s normative theory of (in)equality of opportunity (IoP) through machine learning techniques to analyze unfair inequalities in academic achievement, as well as their evolution and underlying mechanisms. Using PISA data for Spain (2003–2022), we examine how eight circumstances beyond students’ control (related to geographic area, sex, socioeconomic status, and migrant background) define distinct ascribed opportunity types and contribute to explaining variation in mathematics scores. The results show that IoP accounts for up to 20% of total variance, follows a stable inverted U-shaped trajectory over time, and that mother's occupation is the most relevant circumstance. The article discusses the implications of these findings for theories of social justice and meritocracy.