A. How Routine Are Women’s Jobs?

8. Job routineness as proxy for exposure to automation. Workers’ exposure to automation is determined by the routineness of job tasks, with more routine tasks being susceptible to substitution of labor with capital, particularly ICT capital (Autor, Levy, and Murnane, 2003). This trend has been confirmed for a wide range of countries (Autor and Dorn, 2013; Goos, Manning, and Salomons, 2014; Das and Hilgenstock, 2018).³ The standard measure of job routineness―an index of routine task intensity (RTI)―quantifies the extent of codifiability of tasks performed on the job and serves as a proxy for substitutability of workers and machines. Jobs with a higher share of tasks that can be performed by following a defined set of rules, and are thus easily codifiable, are more susceptible to automation. By contrast, jobs requiring analytical, communicational, and technical skills, are less prone to automation.

9. Methodology and data for assessing job routineness. We develop a new index of job routineness from task composition at work using data from the Program for the International Assessment of Adult Competencies (PIAAC) survey. The PIAAC survey covers adults 16 to 65 years of age and collects detailed information on task composition, task frequency, and extent of ICT use in the workplace for 28 OECD member countries, as well as Cyprus and Singapore.⁴ This data set is uniquely suited for cross-country comparison of the nature of work and its susceptibility to automation.⁵ To construct the routine task intensity (RTI) index, we follow the method outlined in Autor and Dorn (2013) and modified by De La Rica and Gortazar (2016) to match the content of the PIAAC survey. The RTI index evaluates the relative importance of abstract skills, such as reasoning and interpersonal communication, and of non-routine manual skills against the importance of routine tasks, which can be easily automated. Specifically, we calculate RTI for each individual worker i as follows:

in which routine_i, Abstract_i and manual_i are index values of routine, non-routine manual, and abstract skills. The RTI index ranges from zero to one, with higher values indicating that a worker engages in more routine activities.

10. Women conduct more routine tasks than men. The RTI index, on average, is 13 percent higher for female workers across our sample of 30 countries (Figure 1)—a result that is statistically significant. Female workers perform fewer tasks requiring analytical and interpersonal skills or physical labor, and more tasks that are routine, characterized by lack of job flexibility, little learning on the job, and greater repetitiveness. This implies that women are more exposed to automation than men.

View Full Size

Gender Gap in RTI and RTI Components

Citation: Staff Discussion Notes 2018, 007; 10.5089/9781484379769.006.A001

Sources: PIAAC survey; and IMF staff calculations.Note: Routine task intensity (RTI) index is calculated at the individual level using information on routine, abstract, and manual tasks. See Annex I for details. Abstract index describes analytical and interpersonal tasks; manual index describes long hours of physical work (non-routine); routine index describes lack of job flexibility, little learning on the job, and repetitive tasks.*** indicates that gender differences in RTI indices and their components are statistically significant at 1 percent level.

• Download Figure
• Download figure as PowerPoint slide

11. Significant cross-country heterogeneity. Women’s exposure to routine job tasks varies significantly across countries (Figure 2). The exposure is highest in eastern and southern European countries and lowest in Scandinavian and central European countries. For instance, the RTI index level of female workers in Lithuania is 36 percent higher than in Norway. This geographic heterogeneity may be indicative of countries’ different positions along the automation path and selection of women into the labor force. It also reflects differences in the structure of production (e.g., manufacturing vs. services sectors requiring interpersonal communication), technologies adopted, and labor market flexibility, and thus differential distribution of workers across sectors and occupations (IMF 2018; Das and Hilgenstock 2018).⁶

View Full Size

RTI Levels for Women Across Countries

Citation: Staff Discussion Notes 2018, 007; 10.5089/9781484379769.006.A001

Sources: PIAAC survey; and IMF staff calculations.Note: Routine task intensity (RTI) index is calculated at the female level using information on routine, abstract, and manual. See Annex I for details. Country averages are calculated using country-specific sampling weights. Index level varies from 0.45 to 0.61.

• Download Figure
• Download figure as PowerPoint slide

12. Gender RTI and ICT gaps and female labor force participation. The gender RTI gap―the ratio of the female RTI level to the male RTI level―is correlated with the female labor force participation (FLFP) in a country (Figure 3, left panel). Gender RTI gaps are smaller in countries with higher FLFP, suggesting that more equal representation of women and men in the workplace is associated with more equal division of tasks between men and women (Figure 3, left panel). Turkey, however, is an outlier, exhibiting both the lowest FLFP and the lowest gender RTI gap. To examine if a similar relationship holds for gaps in ICT use in the workplace, we construct an index that measures the level and frequency of computer use at work. Advanced use of computer technologies could indicate higher complementarity of workers and machines.⁷ Gender gaps in ICT use are inversely related to FLFP (Figure 3, right panel). This suggests that in countries with overall low FLFP, women entering the labor force are more likely to work in jobs that are more intensive in ICT use.

View Full Size

Relationship Between Female Labor Force Participation and RTI and ICT Use Indices

Citation: Staff Discussion Notes 2018, 007; 10.5089/9781484379769.006.A001

Sources: International Labour Organization; PIAAC survey, World Bank, World Development Indicators; and IMF staff calculations.Note: Shading of the circle indicates country’s level of GDP per capita. Gender differences in ICT use are not statistically significant in Greece, Italy, Lithuania, and Turkey. ICT = information and communications technology; RTI = routine task intensity.

• Download Figure
• Download figure as PowerPoint slide

13. Explaining gender gaps in RTI: importance of intensive margin. What drives these observed gender differences in the routineness of work? To better understand the determinants, we decompose the gender gap into contributions from individual and job characteristics (e.g., age, numeracy, literacy skills, etc.) worker’s education, and occupational and sectoral choices (see Annex II for details of the decomposition method). The decomposition results show that nearly 13 percent of the unconditional gender RTI gap is explained by occupational choice (Figure 4). These results indicate the type of tasks women perform and their job positions within occupations (the intensive margin) is the main driver of gender disparities in routineness in the work place.

View Full Size

RTI Decomposition: Drivers of RTI Gap

Citation: Staff Discussion Notes 2018, 007; 10.5089/9781484379769.006.A001

Sources: PIAAC survey; and IMF staff calculations.Note: This decomposition is based on the individual-level regression: $\begin{array}{c}RT{I}_i={\beta }_0+{\beta }_{1}Femal{e}_{ic}+{\mathrm{\Sigma }}_k{\beta }_k^{ind}{X}_{ick}^{ind}+{\mathrm{\Sigma }}_m{\beta }_m^{ability}{X}_{icm}^{ability}+{\mathrm{\Sigma }}_n{\beta }_n^{job}{X}_{icn}^{job}+{\alpha }_{ic}+{\sigma }_{ic}+{\tau }_c+{ϵ}_{ic}\end{array}$. See Annex II for further details. Bars indicate the share of unconditional RTI gap explained by a given set of variables. Statistical significance levels: RTI = routine task intensity.*** p<0.01; ** p<0.05; * p<0.1.

• Download Figure
• Download figure as PowerPoint slide

14. Gender RTI gaps across occupations and sectors. Uneven distribution of women and men across occupations is the largest contributor to the gender RTI gaps. Figure 5 (left panel) plots the level of routineness of tasks against corresponding gender gaps by occupation.⁸ We find that gender RTI gaps are persistent across all occupations, increasing in the average routineness level of the occupation. For instance, clerks and elementary occupations are among the most routine occupations. These occupations, however, make up a large share of women’s overall employment, indicating that women are potentially less insulated from automation owing to their occupational choices. The right panel of Figure 5 shows the sectoral distribution of female workers. With the exception of retail trade and services, which employs a significant proportion of the female labor force and has a large gender RTI gap, sectors that employ large proportions of the female labor force (e.g., health and education) also have lower gender RTI gaps. This suggests that greater participation by women on the extensive margin in a sector lowers their relative exposure to job automation.

View Full Size

RTI Levels vs. Gender Gaps by Occupation and Sector

Citation: Staff Discussion Notes 2018, 007; 10.5089/9781484379769.006.A001

Sources: PIAAC survey; and IMF staff calculations.Note: The size of the circle indicates the share of women in that occupation/sector as a fraction of the female workforce. Routine task intensity (RTI) index is calculated at the individual level. Using information on routine, abstract, and manual tasks. See Annex I for details. Gender differences in RTI are statistically significant at 1 percent level for all occupations.

• Download Figure
• Download figure as PowerPoint slide

15. Job routineness and gender wage gaps. Growing evidence suggests that rising demand for high-skilled labor has lowered relative wages for more routine occupations (Acemoglu and Restrepo, 2017; Autor, 2015). These are the type of tasks that women typically perform. We empirically assess whether job routineness level matters for earnings for a subset of countries for which wages are available in the PIAAC data set (Annex II contains the description of the empirical specification and decomposition method). Our findings indicate that sectoral differences and other individual and job characteristics (e.g., education, experience, presence of children) explain a large share of existing gender wage gaps (Figure 6), but that job routineness also matters.⁹ Job routineness level accounts for nearly 5 percent of the average unconditional gender wage gap, beyond other factors, such as skill, experience or occupational choice, albeit with significant differences across sectors and countries.¹⁰ Taking the US as an example, a 5 percent wage gap driven by RTI differences would translate into $26,000 in forfeited lifetime income.¹¹ This suggests that gender differences in job task assignment can exacerbate gender earnings gaps, a disadvantage that could be further compounded by the changing landscape of work.

View Full Size

Wage Decomposition

Citation: Staff Discussion Notes 2018, 007; 10.5089/9781484379769.006.A001

Sources: PIAAC survey; and IMF staff calculations.Note: Gender wage gap is estimated at the individual level in the following regression specification: $\begin{array}{c}{wage}_i={\beta }_0+{\beta }_{1}Femal{e}_{ic}+{\beta }_2{TRI}_{ic}+{\mathrm{\Sigma }}_k{\beta }_k^{ind}{X}_{ick}^{ind}+{\mathrm{\Sigma }}_m{\beta }_m^{ability}{X}_{icm}^{ability}+{\mathrm{\Sigma }}_n{\beta }_n^{job}{X}_{icn}^{job}+{\alpha }_{ic}+{\sigma }_{ic}+{\tau }_c+{ϵ}_{ic}\end{array}$ See Annex II. The first bar indicates the unconditional wage gap without including any controls. Subsequent bars indicate the share of unconditional wage gap explained by a given set of variables. Statistical significance levels: *** p<0.01; ** p<0.05; * p<0.1.

• Download Figure
• Download figure as PowerPoint slide

A. Quantifying the Risk of Automation for Women

16. Estimating the risk of automation. Our analysis in the previous section suggests that women perform more routine and less-abstract tasks in the same occupations as their male counterparts, placing them at higher risk of automation. To quantify the potential impact on jobs, we estimate the probability of automation at the level of each individual, accounting for differences in worker characteristics and job tasks (e.g., education level, age, skills, and specific job tasks). As an initial starting point, we use estimates constructed at the occupational level in Frey and Osbourne (2017).¹² These estimates were created by a panel of machine learning experts, determining whether the occupations can be fully codified to be run by computers given current levels of technology. We then project the occupation level estimates onto worker characteristics, including age, education, gender, literacy and numeracy skills, and a broad subset of task characteristics included in the PIAAC data set (see Annex III for details). This allows us to recreate estimates for probability of automation at the level of each individual worker. Our estimates thus incorporate rich information on demographics, skills, and responsibilities in the workplace, as opposed to simply relying on occupational distribution as most existing studies do.¹³

17. Higher probability of automation for women. Women have a higher average probability of automation than men. The average probability of automation among women in our sample is 40 percent, 2 percent higher than the average probability of automation among men―a difference that is statistically significant (Figure 7).¹⁴ Moreover, a larger proportion of the female workforce is at high risk for displacement. We consider all workers with more than 70 percent probability of automation to be at high risk for displacement over the next two decades. Using this definition, we find that 10 percent of all male and female jobs―about 54 million workers in our sample―are at a high risk for automation given the current state of technology. The difference in the probability of automation between men and women is statistically significant with 11 percent of female workers (or 26 million jobs) being at a high risk of automation relative to 9 percent of male workers. However, since there are more men in the labor force than women, this still translates into a slightly larger number of men at high risk of being automated overall in our sample.

View Full Size

Gender Gap in Probability of Automation

Citation: Staff Discussion Notes 2018, 007; 10.5089/9781484379769.006.A001

Sources: Frey and Osbourne (2017); PIAAC survey; and IMF staff estimates.Note: The probability of automation is estimated using an expectation-maximization (EM) algorithm that relates individual characteristics (age, education, training, among others) and job task characteristics to occupational-level risk of automation. Details on the methodology and variables are included in Annex III. Differences in probability of automation and share of workers with high automatability across gender are statistically significant at 1 percent level. High automatability is defined as having probability of automation >= 0.7.

• Download Figure
• Download figure as PowerPoint slide

18. Age and probability of automation. Among both men and women, younger people are at the highest risk of losing their jobs to automation given current technologies (Figure 8). For instance, 48 percent of women in the 16–19 age group fall into the high risk of automation category. This is potentially driven by the selection of less-educated workers into the labor force, given the high returns from human capital accumulation through education at this age.¹⁵ Moreover, automation can reduce employment, not merely by replacing workers, but by potentially reducing new-job creation in some sectors or creating new high-skilled jobs, which is more likely to affect young entrants as opposed to older incumbents.¹⁶ Women aged 20 to 29 have moderately lower odds of displacement than men in the same age bracket, but these differences are not statistically significant. However, older cohorts of working women (older than 40) are at significantly higher risk for automation than men in the same age cohorts, suggesting greater disadvantage of women among older age groups.¹⁷ This suggests an important role of policies for smoothing transitions for younger workers and ensuring adequate safety nets for older, displaced workers.

View Full Size

High Probability of Automation and Age

Citation: Staff Discussion Notes 2018, 007; 10.5089/9781484379769.006.A001

Sources: Frey and Osbourne (2017); PIAAC survey; and IMF staff estimates.Note: The probability of automation is estimated using an expectation-maximization (EM) algorithm that relates individual characteristics (age, education, training, among other) and job task characteristics to occupational level risk of automation. Details on the methodology and variables are included in Annex III. High automatability is defined as having probability of automation >= 0.7.Statistical significance levels: *** p<0.01; ** p<0.05; * p<0.1.

• Download Figure
• Download figure as PowerPoint slide

19. What underlies the higher probability of automation for women? Our analysis suggests that the likelihood of automation is decreasing in education, numeracy and literacy skills, and in firm size. For instance, the risk of automation is less than 1 percent among workers who have a bachelor’s degree or higher. The most disadvantaged group is women with lower secondary education or less, with nearly 50 percent at high risk for automation, relative to less than 40 percent of men with the same education level.¹⁸ The finding that less well-educated workers could be particularly exposed to automation highlights the importance of increased investment in lifelong learning and retraining. Interestingly, our analysis predicts that workers in small and medium enterprises (SMEs) have a much higher risk of being automated as compared to workers in large enterprises (greater than 1000 workers). This may be a result of SMEs lagging behind large firms in their adoption of digital technologies, resulting in their workforce being more substitutable and less complementary with these technologies.¹⁹ Moreover, while female workers are significantly more likely than male workers to face risk of automation in SMEs, they are significantly less likely to face risk of automation in large firms relative to their male counterparts. This could reflect differences in firm policies regarding hiring or retention, or the selection of preferred workers into large firms.²⁰

20. Cross-country heterogeneity in women’s probability of automation. The proportion of men and women in the workforce who face a high risk of automation is roughly similar in France, the UK, and the US (Figure 9)―countries with large service-dominated economies. In some countries, the differences between men and women in their exposure to high risk of automation are small and statistically insignificant (e.g., Belgium, Denmark, Germany, Slovakia, and Sweden). In Japan, where 4 percent of the male workforce is at risk for automation, 12 percent of the female workforce falls under high risk of being displaced. In some countries (e.g., Austria, Cyprus, Israel, and Korea) women are disproportionately exposed to high risk of automation. In Finland and Poland, however, the female workforce is significantly less exposed to automation than their male counterparts.

View Full Size

Gender Gap in High Risk of Automation Across Countries

Citation: Staff Discussion Notes 2018, 007; 10.5089/9781484379769.006.A001

Sources: Frey and Osbourne (2017); PIAAC survey; and IMF staff estimates.Note: The probability of automation is estimated using an expectation-maximization (EM) algorithm that relates individual characteristics (age, education, training, among other) and job task characteristics to occupational level risk of automation. Details on the methodology and variables are included in Annex III. High automatability is defined as having probability of automation >= 0.7. Difference in automatability = (Share of females with high automatability)/(Share of males with high automatability).

• Download Figure
• Download figure as PowerPoint slide

21. Sectoral variation in likelihood of automation. Overall, our analysis suggests the accommodation and food services, retail trade, and transportation sectors have the largest exposure to risk of automation (Figure 10, left panel). While retail trade and accommodation and food services sector employ roughly similar proportions of the male and female workforce, men are disproportionately represented in the transport sector, and therefore more exposed to risk of automation owing to their sectoral choice.²¹ Women are, however, vastly overrepresented in education, health, and social services, which are at a low risk of being automated, with 34 percent of the female workforce working in these sectors relative to only 11 percent of the male workforce. However, gender gaps in the risk of automation varies within sectors (Figure 10, right panel). For instance, despite the large share of female workers in education, women face a higher likelihood of being automated than men. This suggests that the extent to which female workforce is at risk of being automated depends not just on sectoral choices but also the job composition within a sector.

View Full Size

Automation Across Sectors

Citation: Staff Discussion Notes 2018, 007; 10.5089/9781484379769.006.A001

Sources: Frey and Osbourne (2017); PIAAC survey; and IMF staff estimates.Note: The probability of automation is estimated using an expectation-maximization (EM) algorithm that relates individual characteristics (age, education, training, among other) and job task characteristics to occupational level risk of automation. Details on the methodology and variables are included in Annex III. High automatability is defined as having probability of automation >= 0.7. Difference in automatability = (Share of females with high automatability)/(Share of males with high automatability). The size of the circle indicates the share of women in that sector as a fraction of the female workforce.

• Download Figure
• Download figure as PowerPoint slide

22. Importance of occupation and job characteristics. Differences in the choice of occupation as well as differences in job characteristics within occupations drive gender gaps in the threat of automation. Elementary occupations are most exposed to risk for automation, with nearly 40 percent of workers at high risk of being displaced (Figure 11). Roughly similar proportions of the male and female workforce are employed in elementary occupations, with no statistically significant difference in the numbers of male and female workers at risk of automation. Women are overrepresented among service workers, however, and female service workers face a higher risk of automation than their male counterparts. ²² Legislators, managers, and professionals are well insulated from the threat of displacement by automation, with less than 1 percent of the workforce in these occupational categories being at high risk of automation. Among professionals, however, while the threat from automation is overall low, women are twice as likely as their male counterparts to be at high risk for automation. In retail trade, where the overall risk of automation is very high, female workers are significantly less likely than male workers to perform abstract tasks—reflecting the fact that there are fewer women in managerial positions.

View Full Size

Gender Gap in High Risk of Automation Across Occupations

Citation: Staff Discussion Notes 2018, 007; 10.5089/9781484379769.006.A001

Sources: Frey and Osbourne (2017); PIAAC survey; and IMF staff estimates.Note: The probability of automation is estimated using an expectation-maximization (EM) algorithm that relates individual characteristics (age, education, training, among other) and job task characteristics to occupational level risk of automation. Details on the methodology and variables are included in Annex III. High automatability is defined as having probability of automation >= 0.7. Difference in automatability = (Share of females with high automatability)/(Share of males with high automatability). The size of the circle indicates the share of women in that occupation as a fraction of the female workforce.

• Download Figure
• Download figure as PowerPoint slide

23. Some positive trends. The distributional shift toward technical and professional occupations has accelerated for women in the last two decades. Looking at the types of jobs gained and lost between 1994 and 2016 for a sample of OECD countries, we find that most job growth has been on the high-skill end, and that women have benefited from this more than men.²³ Figure 12 indicates an overall shift away from clerical occupations towards service and retail workers, technicians, and professionals―a trend that is more pronounced for women. Women appear to be increasingly opting into occupations that are relatively more insulated from the risk of automation. The increase in women in managerial and professional roles is particularly encouraging in this regard, but more needs to be done. These findings are also consistent with growing educational participation of women. According to the Pew Research Center, in 1994, slightly more than 60 percent of male and female graduates in the United States were enrolled in university. By 2014, the figure for women had jumped to 71 percent, while that for men was broadly unchanged. More women than men are also completing college now than in the past, which has implications for their employment patterns and labor market prospects.

View Full Size

Changes in Occupational Shares by Gender (1994–2016)

Citation: Staff Discussion Notes 2018, 007; 10.5089/9781484379769.006.A001

Sources: International Adult Literacy Survey; PIAAC survey; and IMF staff calculations.Note: Shift in occupation is calculated as the difference between share of female (male) workers in occupation X and country Y in 1994, and the share of female (male) workers in the same occupation and country in 2016. Top and bottom values of the intervals represent country-specific maximum and minimum values for occupational share differences.

• Download Figure
• Download figure as PowerPoint slide

24. Global challenge: projecting occupational and sectoral risks for the world. To gauge the impact of automation on global employment, we use International Labor Organization (ILO) estimates of employment by gender and apply our sector-wise probability of being at high risk of automation to the global employment projections by sector.²⁴ We estimate that 180 million female jobs are at high risk of being automated comprising more than 14 percent of the global female workforce. At the global level, much of this automation will be driven by agriculture and retail trade, with more than 50 million female jobs being potentially displaced in each, followed by accommodation and food services. In extrapolating our results to other countries, we are assuming that technology is immediately transferrable between different economies at different stages of development and with differing relative costs of labor and capital. The extent of technological diffusion is, and may continue to be, heterogeneous. Hence, caution should be exercised in interpreting our results beyond the sample of countries for which we have micro-level data. These estimates, however, still serve as a useful guide for the potential for automation in economies that may be at earlier stages of technological progress than those in our sample.

25. Caveats and rise of “gig economy.” Our estimates for probability of automation assume current levels of technology and prevailing bottlenecks in the use of computer-controlled equipment. As such, our analysis presents a lower bound for the potential impact of automation. Given the speed of technological advancement in recent years, further improvements in the state of computing could result in more tasks being automated than predicted by the current level of technology. Our estimates are also based on the technological feasibility of automation as opposed to the economic feasibility: tasks could be automated given the current state of technology, but the costs of automation may be prohibitive relative to the prevailing cost of labor.²⁵ Moreover, the impact could differ for men and women. For instance, Hawksworth, Berriman, and Goel (2018) find that female workers could be more affected by automation over the next decade, but male jobs could be at higher risk in the longer term due to the nature of technological change. Our results of a significant gender gap in the risk of automation, however, are robust to using alternate measures for probability of automation at the occupation level by Brynjolfsson, Mitchell, and Rock (2018).²⁶ Finally, we do not account for the rising contribution of the “gig” economy on labor supply and demand. On the one hand, the gig provides additional avenues to earn income by supplying services—either digitally or physically—on an on-demand or short-term basis.²⁷ Platforms facilitating delivery of these services, however, may reallocate employment across tasks and sectors, with ambiguous implications for women in the workforce.²⁸ An analysis of the gig economy is thus crucial for a comprehensive analysis of the attendant opportunities and challenges for the future of work.

B. Looking Ahead: Opportunities and Challenges

26. Deep dive of expanding sectors. In this subsection, we focus our analysis on specific sectors that are relevant from the perspective of female employment and the future of work. For instance, the ICT sector is expected to expand rapidly with improvements in the availability and efficiency of communication technology. Similarly, health and social services, which is the largest employer of women, will experience a significant increase in demand owing to rapid population aging and rising longevity in advanced economies.

27. Women underrepresented in ICT jobs and STEM. In the growing ICT sector, the share of women at high-risk of displacement is also significantly greater than the share of men (Figure 13 top panel). This is not surprising as a larger proportion of men are in managerial and professional positions (87 percent as opposed to 72 percent for women)—positions more likely to entail performing abstract tasks (e.g., applying human judgment and supervising AI-systems). Moreover, there are nearly four times as many women employed as clerks or service workers (26 percent as opposed to 7 percent for men). The types of tasks performed with the help of technology by such workers are typically more routine (e.g., working with spreadsheets and conducting basic transactions online). Moreover, while there may be near parity in the number of men and women graduating with degrees in social sciences and mathematics (at least in the US) this is not true of computer science. Women make up less than 20 percent of tertiary school graduates in computer science in OECD countries (OECD, 2018).²⁹ Persistent gender differences in field of study may mean that women will benefit less from the new job opportunities in STEM-related occupations.

View Full Size

Occupational and Task Differences in ICT and Health & Social Services

Citation: Staff Discussion Notes 2018, 007; 10.5089/9781484379769.006.A001

Sources: PIAAC survey; and IMF staff calculations.Note: The abstract index describes analytical and interpersonal tasks; the manual index describes long hours of physical work; the routine index describes lack of job flexibility, little learning on the job, and repetitive tasks. See Annex I for details. Gap = Male index level/Female index level. Gap >1 indicates that male workers score higher on this index.

• Download Figure
• Download figure as PowerPoint slide

28. Building on existing advantages. The lower likelihood of automation in healthcare and social services is likely driven by the smaller gender gap in managerial and professional positions, and the smaller gaps in abstract and manual tasks performed (Figure 13, bottom panel). Indeed, the gap in the manual index is actually reversed, with women performing more non-routine manual work, such as nursing. With populations aging across most advanced economies, demand for health and social care services will only continue to increase over the coming decades.³⁰ Studies show that women tend to agglomerate in industries where there is higher gender parity, which suggests that trends in labor supply by women in health and education services can be expected to continue. Yet new technologies are already changing caregiving jobs (West, 2015). Coping with aging populations will require both more human workers and greater use of artificial intelligence, robotics, and other advanced technologies to complement and boost productivity of workers in healthcare services. Workers in these sectors will need to acquire the necessary digital and softer skills that will be demanded.

Empowering women in the workplace.

31. Endowing women with the requisite skills. There is growing consensus that investments in human capital are likely to have the biggest payoffs for both men and women in the face of technological change (Grigoli, Koczan, and Topalova, 2018; IMF 2018; World Bank 2018). Indeed, skills will provide the most important safeguard against displacement from technology and allow women to benefit from the new work opportunities that are created. The demand for narrow job-specific skills and competencies, however, is waning, and demand for advanced cognitive skills (critical thinking, problem solving, and learning agility) and socio-emotional skills (creativity, curiosity, and adaptability) is on the rise. This suggests an urgent need to adapt and reform education systems and workforce training to reduce skill-mismatches for a changed workplace and remove barriers to lifelong learning. Not all the work on education needs can be done by governments alone. Employers are best placed to identify skills gaps in a more technology-enabled workforce. Businesses could thus play a more active role in education and training, including divulging information on future skills demand and providing better learning opportunities themselves.

32. Building skills early. Raising the quality and quantity of the human capital of new labor market entrants will be important. In this regard, ensuring universal and high quality basic and secondary education is critical to help individuals adapt to new technologies (Autor, 2013), but the right balance should be found between endowing women with the necessary tools in specialized STEM fields, where they are at a relative disadvantage, and a general curriculum that emphasizes critical thinking, decision-making, problem-solving skills, and empathy (Deming, 2015). Early investment in women in STEM fields with a digital and quantitative focus and peer mentoring can help breakdown gender stereotypes and increase retention in these fields.³⁴ Our findings suggests that the risk of automation is less than 1 percent among female workers who have a bachelor degree or higher, compared with 50 percent among women with lower secondary degrees, highlighting the importance of higher education. Across most advanced economies, more women than men are now tertiary graduates, which is good news, but ensuring skill acquisition and long-term skill matching in tertiary education will be important.

33. Encouraging lifelong learning. Adapting and upgrading the skills and competencies of those already in the workforce will be equally important. The adult education systems currently in place in many countries tend to reinforce existing gender and economic disparities, with greater frequency of reskilling and upskilling by more educated, high-income workers with digital literacy skills and access to the internet. Women, who are most exposed to the risk of automation, typically work in occupations with potentially lower on-the-job learning opportunities and are also the least likely to participate in training. Fiscal policy instruments, such as publicly subsidized vouchers to ease access to formal education, portable individual learning accounts (e.g., in France), preferential loans, and tax deductions (e.g., in Netherlands) could incentivize both men and women to invest in lifelong learning. Incentivizing the private sector to encourage human capital investments through tax benefits and other incentives (e.g., payroll taxes dedicated to subsidizing training opportunities, public grants for subsidizing training) could also be considered. There is also merit in targeting retraining tax incentives to particular sectors where women continue to face large skill gaps relative to men.

34. Fostering gender parity in management positions. Failure to retain female talent up the job ladder and create a level playing field has important consequences for their risk of automation. Although women are, on average, more educated than men in most advanced economies and now participate more fully in professional and technical occupations than two decades ago, their chances to rise to positions of leadership are only 28 percent of those of men (WEF, 2017b). Greater involvement of women in senior management has been found to strengthen economic performance (Cuberes, Newiak, and Teignier, 2017; Sahay, Čihák, and others, 2018).³⁵ Importantly, the payoffs from gender diversity are significantly higher in high-tech and knowledge-intensive sectors—sectors which are likely to dominate in the future. Family-friendly policies discussed above can play an important role in boosting women’s retention and career progression, but setting relevant recruitment and retention targets for organizations, as well as mentorship and training programs to promote female talent into managerial positions can also help close gender gaps. Mandatory gender quotas on hiring and promotion of women in the public and private sectors could be considered. In Norway, for instance, following the adoption of quota legislation for corporate boards in 2003, women representation on all company boards increased to about 40 percent (OECD, 2016).

35. Bridging digital gender divide. The impact of technology on the risk of automation and the ability to benefit from the new job opportunities created depends on access. Digital technologies, and the flexible form of working that they may enable, could boost women’s employment rates, even in the presence of social, religious, and cultural barriers. This trend could also help foster more gender-balanced career paths and reduce earnings inequalities. Large gender gaps persist, however, in access to and use of ICT: 60 percent of the global population, most of them women in emerging and developing economies, still have no access to the internet; 250 million fewer women are online than men; and 200 million fewer women than men own a mobile phone across developing countries (ITU, 2017). The lack of digital infrastructure is thus particularly detrimental for women’s labor market outcomes. Ensuring equal access to finance is important as many women face affordability barriers. Investment through public, private, or public-private- partnerships, will also be essential to support technological adoption and close digital gender gaps. National connectivity policies should also apply a gender lens to ensure equal access for all. Finland, for instance, has defined access to the internet at broadband speeds as a legal right and has pursued a universal access policy.

Easing transitions for workers

36. Smoothing adjustment in the face of technological change. Given that older female workers are at disproportionately higher risk of displacement than their male counterparts, ensuring gender equality in support for displaced workers will be essential.³⁶ In the short term, direct support for both male and female workers can be provided through selected active labor market policies. These can either include direct support for displaced individuals—such as job search assistance, retraining, and income or geographical mobility support—or employment incentives—such as hiring and wage subsidies (IMF, 2018). Policies to cope with labor market adjustment may be inadvertently biased against women if such schemes do not provide support to sectors where women dominate (OECD, 2017). Ensuring that training and benefits are linked to individuals rather than jobs can help improve reemployment prospects for both men and women. Where needed and desired, the tax/benefit system also has an important role to play in redistributing income after market outcomes for both men and women.

37. Adapting social protection to new forms of work. The rise of flexible, nonstandard employment will put pressure on traditional forms of social protection. The eligibility for social insurance—which includes pension and unemployment benefits—has traditionally been conditional on accumulated contribution records. Nonstandard workers, and thus many women, are likely to be at a disadvantage relative to those on standard work contracts. Tax and benefit systems will need revamping to close coverage gaps, which are particularly wide for women, and allow for portability of benefits to prevent the loss of social benefit entitlements when workers move between jobs. Several countries are testing basic income guarantees linked to technology and other types of noncontributory schemes (e.g., expanding social pensions and earned income tax credit) to address greater income decline and uncertainty generated by the impact of automation on jobs, which could be particularly beneficial for older women who are at high risk of being displaced.