Statistical Discrimination Revisited: Explaining the Early Gender Wage Gap with Graduate Data

This paper revisits the statistical discrimination model of Phelps (1972) to explain why a gender wage gap emerges immediately at labour-market entry, despite women's superior academic performance. We focus on graduates and extend the framework by adding a productivity-relevant attribute - willingness to work abroad or IT skills - that is correlated with gender and differs across fields of study. Employers observe noisy individual signals and coarse group-level statistics by gender and field, and optimally combine them when setting wages. Within this setting, gender differences in the distribution of these attributes can generate an entry wage premium for men even when women have higher average human capital. We test this mechanism using AlmaLaurea microdata on master's graduates from the University of Bologna (2015-2022). We calibrate the model for the full sample and separately for Economics & Management and Engineering. Human capital alone cannot reproduce the observed wage differences, while augmenting the model with willingness to work abroad or IT skills brings predicted and actual gaps into close alignment. Complementary wage regressions show that mobility intentions explain a substantial share of the raw gender wage gap across fields, whereas IT skills matter primarily in Engineering and only marginally in the aggregate. The combined evidence from the model calibration and the empirical analysis supports an extended statistical discrimination channel operating through gendered distributions of mobility and IT-related attributes.