Computerisation and robotics have had a profound effect on labour markets. Using data from Japan, this column finds that female workers are more exposed to risks of computerisation than male workers, and that this tendency is more pronounced in larger cities. The results suggest that supporting additional human capital investment alone is not enough as a risk alleviation strategy against new technology. Policymakers need to address structural labour market issues, such as gender biases in career progression and participation in decision-making positions.

There is growing concern that human jobs are being replaced by the rapid technological progress of artificial intelligence (AI), robotics, and automation (Acemoglu and Restrepo 2017, Brynjolfsson and McAfee 2014, Ford 2015). It is often emphasised that whereas mechanisation has so far replaced blue-collar jobs, recent AI technology, which plays a similar role to the human brain, is mainly replacing white-collar jobs (Dauth et al. 2017, Graetz and Michaels 2018). In fact, pattern recognition based on deep learning plays an important role in companies that collect big data, including image, speech, and texts. AI consulting services are already in use.

In a recent paper (Hamaguchi and Kondo 2018), we explore possible labour market problems that will be caused by computerisation based on AI technology. Our main concern is that the geographical distribution of occupations in a particular country is not uniform. Some occupations are relatively concentrated in urban areas, and others are concentrated in rural areas. This heterogeneity is highly relevant for male versus female workers. When discussing the impacts of AI technology on employment, we should consider these geographical and gender aspects of the occupational distribution.

Fact-finding in the Japanese labour market

To clarify which groups of workers are exposed to the risks of AI technology, we propose a regional employment risk score for computerisation, which is calculated using the regional share of occupations and the probability of computerisation by occupation based on Frey and Osborne (2017).

This risk score takes a value between 0 and 100, with 0 indicating no employment risk of computerisation and 100 indicating full replacement of employment. For example, when all workers are engaged in an occupation with 0 probability of computerisation in a region, the risk score takes the value 0. When all workers are engaged in an occupation with probability of 1 for computerisation in a region, the risk score takes the value 100.

Figure 1 shows the relationship between the employment risk scores of computerisation and city size at the prefectural level. Panel (a) of Figure 1 shows the negative correlation for male workers. By contrast, Panel (b) of Figure 1 shows the positive correlation for female workers. A big question raised by Figure 1 is why gender inequality expands in larger cities.

Figure 1 Employment risk score of computerisation and city size

Image: Reproduced from Figure 3 in Hamaguchi and Kondo (2018).

Our main finding is that female workers tend to have little opportunity to advance in their career and tend to be engaged in occupations that are susceptible to computerisation, such as receptionist, clerical, and sales work.1 Male workers are more likely to get decision-making positions (e.g. managers and supervisors) and professional jobs (e.g. engineers and natural scientists), which are seen as more difficult to replace with AI technology. This tendency becomes more prevalent in larger cities. Consequently, larger cities show a greater gender gap in the employment risk of computerisation.2

Policy implications for AI technology and gender inequality

Currently, policymakers face the twin policy challenges of strengthening the global competitiveness of firms using AI technology and of mitigating the negative employment impact of using AI technology. Importantly, AI technology is essential for firms to survive amidst fierce global competition, but the promotion of AI technology may simultaneously accelerate labour substitution. This study contributes to tackling the latter challenge.

The important policy implication from this study is that supporting additional human capital investment is important, but insufficient, as a means of alleviating the risk of new technology. As shown in Kawaguchi and Toriyabe (2018), there is no significant gender gap in skill levels in Japan. The problem in the Japanese labour market is the gender gap in skill utilisation, which suggests occupational segregation between males and females.3

Corporate managers need to recognise that simple clerical, data-collection, and processing work will be computerised. Human capability within the current vertically centralised decision-making structure, which creates the above-mentioned occupational gender segregation, will no longer be able to deal with the high-speed information flows that such a technology presents. Therefore, decision making needs to be more horizontally decentralised, incorporating the sensibility and perspectives of women. AI technology will amplify the unequal risks of computerisation between males and females if these structural labour market problems remain unresolved.

Finally, we should bear in mind that nobody can exactly predict the future progress of AI and its impact on the labour market. This research field, therefore, should continuously incorporate updated information and contribute to policy debates.