AI AND SERVICES-LED GROWTH: EVIDENCE FROM INDIAN JOB ADVERTS

How does the deployment of Artificial Intelligence (AI) in Indian firms affect labour demand and wages? Rapid advances in machine learning have spurred an intense debate about the labour market consequences of AI. Online job adverts show that demand for AI-related skills has grown almost exponentially and concurrently in several countries around the world since 2015. Yet detailed empirical evidence on the extent of AI deployment and its distributional impacts remains scarce, particularly beyond a handful of advanced economies. For low- and middle-income countries, the use cases and impacts of AI need not be the same as for advanced economies. AI could have important consequences for their development trajectory, especially for countries promoting services-led growth. In India, for example, many of the services industries that have driven structural change, productivity growth and job creation, such as Business Process Outsourcing (BPO), are highly exposed to machine learning-based automation, raising questions over the future viability of a services-led development model and its promise of promoting widespread prosperity. This project aims to fill this knowledge gap by shedding light on the labour market impacts of AI in India – the archetypical pioneer of a services-led development model. 

Using the job adverts data, the researchers first document several patterns in AI-related hiring in India. There was a rapid take-off in demand for AI-related skills after 2016, particularly in the IT, finance and professional services industries, closely mirroring patterns found for advanced economies. AI demand increased from 0.37% of all job vacancies in 2015 to 1.03% in 2019, coinciding with an increase in demand for specific ‘deep learning’ skills, along with ‘natural language processing’ to a lesser extent.

The results show that growth in AI demand has a significant negative impact on the growth in non-AI postings and average wage offers by establishments. These negative effects on vacancy growth are most pronounced for higher-skilled professional and managerial occupations, notably engineering professionals and general and corporate managers. Using the classification of Acemoglu & Autor (2011), the researchers find that AI lowers demand for occupations that are typically non-routine task intensive, both overall and within the affected managerial and professional occupation groupings. This stands in sharp contrast to findings for computerisation and robotics, which have been shown to lower demand for routine tasks. 

Taken together, the results show sizable impacts of AI on high-skilled, non-routine, analytical work within establishments in India’s predominantly urban, white collar service sector. These effects of AI in its early years of adoption contrast with those for the computer and robot revolutions, where routine work lost out. AI jobs pay a substantial wage premium, but these opportunities are highly concentrated in certain industries, cities and large firms, providing benefits for a small group with AI skills at the expense of demand for other types of high-skilled workers. However, these displacement effects are driven by older ‘incumbent’ firms that are not AI ‘producers’.

When looking for wider effects at the district level, the researchers find little evidence of displacement. This could suggest that negative within-establishment effects are offset by growth in new startups, including those focused on AI production, or that the aggregate impacts of AI are not yet economically meaningful. Tracing the aggregate impacts of future advances in, and deployment of, AI on innovation and employment will be an important task for future research.