Long-Term Effects of Artificial Intelligence on Employment and Wages in India

<title>Abstract</title> <p>This paper examines the long-term employment and wage effects of artificial intelligence (AI) diffusion in the Indian labour market, using a dual-track analytical strategy. Track A applies a full inferential statistics pipeline descriptive statistics, normality diagnostics, independent-samples t-tests, one-way ANOVA with Tukey post-hoc comparisons, Pearson/Spearman correlation, chi-square tests of independence, multiple linear regression, difference-in-differences (DiD) and triple-difference estimation, and logistic regression to a calibrated, literature-anchored simulated microdata panel built to reproduce the marginal distributions (occupation, education, gender, rural/urban location, wage levels, and informal-employment shares) reported in India's Periodic Labour Force Survey (PLFS), because raw PLFS/EPFO/ESIC microdata could not be retrieved directly within this study's data-access environment. Track B analyses the actual, published official aggregates PLFS unemployment, worker-population, and labour-force participation rate series (2017-18 to 2023-24), and Employees' Provident Fund Organisation (EPFO) payroll releases to triangulate the simulated findings against real, cited government statistics. The combined evidence shows a robust skill-polarisation pattern consistent with the international literature: occupations with high generative-AI exposure display a statistically significant wage premium for graduates in the post-GenAI period (triple-interaction coefficient = 0.907, p &lt; .001) but a significant wage penalty and higher probability of informalisation for workers below graduate level (DiD coefficient = -0.126, p &lt; .001). The chi-square test confirms a significant association between AI exposure and informal employment (χ²(2) = 234.90, p &lt; .001, Cramér's V = 0.217). Track B shows that India's real aggregate unemployment rate fell steadily from 6.0% (2017-18) to 3.2% (2023-24, slope = -0.51 pp/year, p &lt; .001), but this decline predates large-scale generative-AI diffusion and is therefore not, on its own, attributable to AI. The paper concludes that AI's long-term labour-market effect in India is best characterised not as aggregate job destruction but as an intensifying skill and formality divide, with policy implications for reskilling, social protection for informal workers, and caste- and gender-aware labour-market monitoring.</p>