Ekonomi & Bisnis editorial
Open AccessOA2023
The impact of artificial intelligence on labor markets in developing countries: a new method with an illustration for Lao PDR and urban Viet Nam
This paper proposes a new methodology to translate measures of AI impact from the US to developing countries by comparing textual descriptions of work activities. Using SBERT, it assesses the suitability of occupations for machine learning in Lao PDR and urban Viet Nam, revealing that Vietnamese workers are more concentrated in occupations affected by AI.
Francesco Carbonero; Jeremy Davies; Ekkehard C. Ernst; F. Fossen; Daniel K. Samaan; Alina Sorgnerยท Journal of evolutionary economicsยท 2023ยท DOI 10.1007/s00191-023-00809-7
The core problem
Artificial intelligence (AI) is transforming labor markets globally, but research has predominantly focused on advanced economies, neglecting developing countries. The impact of AI on labor markets varies across countries due to differences in occupational structures and the composition of tasks within occupations. This paper addresses this gap by proposing a new methodology to translate existing measures of AI impacts, developed for the US, to countries at various levels of economic development. The method assesses semantic similarities between textual descriptions of work activities in the US and workers' skills elicited in surveys for other countries. The authors implement this approach using the measure of suitability of work activities for machine learning provided by Brynjolfsson et al. (2018) for the US and the World Bank's STEP survey for Lao PDR and Viet Nam. The approach allows characterizing the extent to which workers and occupations in a given country are subject to destructive digitalization (putting workers at risk of displacement) versus transformative digitalization (benefiting workers). The authors find that workers in urban Viet Nam, compared to Lao PDR, are more concentrated in occupations affected by AI, requiring adaptation or risking partial displacement. The method based on semantic textual similarities using SBERT is advantageous compared to approaches transferring AI impact scores across countries using crosswalks of occupational codes.
Innovation
The authors propose a novel method to assess the impact of AI on labor markets in developing countries by leveraging semantic textual similarity. The core idea is to compare textual descriptions of work activities from the US, for which measures of AI impact exist, with descriptions of skills and tasks from surveys in other countries. Specifically, they use the suitability of work activities for machine learning scores from Brynjolfsson et al. (2018) for the US, which are based on the O*NET database. For the developing countries, they use the World Bank's STEP (Skills Toward Employment and Productivity) survey, which provides detailed information on workers' skills and tasks. The method employs Sentence-BERT (SBERT), a state-of-the-art natural language processing model, to compute semantic similarities between US work activities and developing country survey responses. This allows mapping the AI impact scores to occupations in the developing countries without relying on crosswalks of occupational codes, which can be imprecise due to differences in occupational classifications and task compositions across countries. The approach distinguishes between destructive digitalization, where workers are at risk of displacement, and transformative digitalization, where workers benefit from AI. The method is illustrated using data from Lao PDR and urban Viet Nam.
Introduction
Artificial intelligence (AI) is transforming labor markets globally, but research has predominantly focused on advanced economies, neglecting developing countries. The impact of AI on labor markets varies across countries due to differences in occupational structures and the composition of tasks within occupations. This paper addresses this gap by proposing a new methodology to translate existing measures of AI impacts, developed for the US, to countries at various levels of economic development. The method assesses semantic similarities between textual descriptions of work activities in the US and workers' skills elicited in surveys for other countries. The authors implement this approach using the measure of suitability of work activities for machine learning provided by Brynjolfsson et al. (2018) for the US and the World Bank's STEP survey for Lao PDR and Viet Nam. The approach allows characterizing the extent to which workers and occupations in a given country are subject to destructive digitalization (putting workers at risk of displacement) versus transformative digitalization (benefiting workers). The authors find that workers in urban Viet Nam, compared to Lao PDR, are more concentrated in occupations affected by AI, requiring adaptation or risking partial displacement. The method based on semantic textual similarities using SBERT is advantageous compared to approaches transferring AI impact scores across countries using crosswalks of occupational codes.
Why it matters
The paper's main contribution is the development of a method that overcomes the limitations of transferring AI impact measures across countries using occupational code crosswalks. By directly comparing the textual descriptions of work activities, the method captures the heterogeneity in task composition within occupations across countries, which is crucial for accurately assessing AI's impact. The findings highlight that developing countries are not homogeneous in their exposure to AI; urban Viet Nam, with its more advanced economy and occupational structure, faces greater exposure to AI than Lao PDR. This has important policy implications: workers in Viet Nam may need retraining and adaptation strategies to cope with the transformative and destructive effects of AI, while Lao PDR may have more time to prepare. The authors note that the method can be extended to other countries with available survey data. However, they also acknowledge limitations, such as the reliance on the US-based AI impact measure and the quality of survey data. Overall, the study provides a valuable tool for policymakers and researchers to understand and mitigate the labor market impacts of AI in developing economies.
Who should read this
Pembaca Research Digest dan praktisi rumpun ini.
Opening member contentโฆ