This study uses data collected from school students in Mumbai to investigate how they perceive subjective expected returns for different levels of education in an environment that includes labour market discrimination. We are particularly keen to observe subjective returns to education for different social identity groups, such as gender, religion, and caste.

Despite lower actual returns to education in the labour market, students from Other Backward Castes and Scheduled Castes and Tribes do not have significantly different beliefs relative to their advantaged peers.

To estimate the differences in subjective returns to education, we use a novel statistical technique: distributional regression. This method can be a complement to the mean estimates, enabling the assessment of heterogeneity in expected earnings for different social groups in India.

Using the distributional regression, we observe that females and Muslims have left-skewed earnings distributions relative to our reference group (male, upper caste Hindus).