The Impacts of New Digital Technologies on Gender Equality in Developing Countries

This paper investigates the impact of new digital technologies, including advances in artificial intelligence, on men’s and women’s jobs across sectors in developing and transition economies. On the one hand, new digital technologies may have destructive effects on jobs when they replace human workers. On the other hand, they may have transformative effects by changing occupations without necessarily substituting human workers. This paper uses two measures, the computerization probabilities estimated by Frey and Osborne (2017) and past advances achieved in artificial intelligence provided by Felten et al. (2018) to better account for different aspects of digitalization. The empirical analysis is based on the large representative STEP Skill Measurement Surveys of individuals residing in urban areas of selected developing and transition countries. The results suggest that there are strong gender differences concerning skill endowments, which represent the bottlenecks to computerization. Women in developing and transition economies are significantly less likely than men to have skills that protect them from the destructive digitalization, namely analytical, non-routine manual, interpersonal, advanced ICT and socio-emotional skills. This result is robust across sectors, but gender differences are more pronounced in manufacturing than in services. Moreover, the results reveal that women on average face a higher computerization risk (destructive digitalization) of their jobs than men. However, women are less likely to benefit from advances in AI (transformative digitalization). For both measures of digitalization, the results are more pronounced for the manufacturing sector than for services. In addition, a higher level of formal education decreases the impact of destructive digitalization; however, highly educated individuals are more strongly affected by transformative digitalization. Implications of the results for policymakers are discussed.