Occupational Choice and Labour Market Dynamics: An Empirical Study Using Plfs 2023–24

India’s labour market has changed significantly since the economic liberalisation of the 1990s. While the economy has grown tremendously since 1990s’, the supply of jobs has not kept pace, leading to jobless growth and increasing disparities in the quality of employment. The labour market is still segmented with several inequalities in skill, gender, caste, education, and region, and this study aims to analyse how education, gender, caste, and region as determinants impact occupational choice in India. This paper studies occupational choice by combining three employment types (self-employed, regular salaried, casual) with four skill categories to form twelve occupational groups. A central empirical challenge is that occupations are observed only for employed individuals so that naïve occupational-choice estimates may be biased by non-random selection into employment. Using PLFS 2023-24 microdata (with 2018-19 to 2023-24 rounds used for descriptive context) and survey weights, the study applies a two-step generalised selection correction for multinomial choice. First, a Probit employment-participation model is estimated, and the inverse Mills ratio (IMR) is constructed. Second, occupational choice conditional on employment is modelled using a multinomial logit augmented with the IMR. Model validity is assessed using Hausman-style IIA diagnostics and a nested logit robustness check that allows within-employment-type correlation in unobserved utilities.Results show a steep education gradient: higher education substantially increases the probability of high-skill regular employment and high-skill self-employment and reduces low-skill outcomes. Vocational training has heterogeneous effects across segments. Persistent gender, caste, and rural-urban disparities remain after controls, and significant IMR terms indicate selection bias is empirically relevant. Robustness checks provide no evidence that baseline findings are driven by IIA misspecification.