Abstract
IMF Research Perspectives Spring-Summer 2020
Mariya Brussevich
Era Dabla-Norris
Salma Khalid
An imaginary conversation between a worker and a machine
WORKER: Will you replace my job?
MACHINE: Perhaps.
WORKER: How likely?
MACHINE: Well, it depends on your gender, age, and education.
Automation is hardly a novel phenomenon; traditional sectors such as agriculture and manufacturing have experienced large substitutions of labor with machine capital in the past. With acceleration in computerization of manual labor in recent years, concerns are rising that a new wave of displacement by automation is upon us, with female workers rather than their male counterparts on the front line.
Why would female workers be more vulnerable? Because women perform more routine tasks than men-tasks that are more prone to automation. Moreover, women perform fewer tasks requiring analytical input or abstract thinking, where technological change can complement workers’ skills and improve their productivity.
Recent IMF research has zoomed in on how the threat of automation varies by workers’ gender, age, and education in 30 advanced and emerging market economies. The worker-level microdata from the Organisation for Economic Co-operation and Development’s
Programme for the International Assessment of Adult Competencies (PIAAC) permits analysis of exposure to automation at the individual level.
The Gender Routineness tGAP
The index of routine task intensity (RTI) evaluates the relative importance of abstract skills, such as reasoning and interpersonal communication, and nonroutine manual skills against routine tasks that are repetitive and hence can be easily automated: RTI, = Routine,- Abstract, – Manual,,. The RTI gap is the ratio of female-to-male RTI.
Women, on average, perform more routine and less abstract tasks than men across all countries, sectors, and occupations. Differences in occupational distribution across genders explain most of the RTI gap between women and men. For instance, women are overrepresented in clerical occupations, where RTI gaps are high, and underrepresented among managers, legislators, and senior officials, where RTI gaps are low.
Risk of Automation
The risk of automation for individual workers can be quantifed in two steps. The first step uses estimates for the risk of automation at the occupational level, which depends on whether workers in these occupations can be replaced by state-of-the-art computer-controlled equipment. The second step relates these automation risks to individual workers based on worker characteristics-such as gender, age, education, literacy, and numeracy skills—and the characteristics of tasks they perform at work-such as the use of technology, the use of interpersonal skills, and the flexibility of work processes.
Using this approach, the authors find a large gender difference in the risk of automation. Eleven percent of the female workforce is at a high risk of automation (defined as having at least a 70 percent estimated likelihood of automation), relative to 9 percent of the male workforce, with 26 million female jobs at stake in 30 countries. While younger cohorts of male and female workers have similar levels of exposure to automation, women older than 40 are at significantly higher risk than men. Gender gaps in automation are also highest for less educated workers and for clerical and sales workers (Figure 1).
Gender Gaps in High Risk of Automation by Generation and Educational Level
(Difference in automatability between females and males)
Citation: IMF Research Perspectives 2020, 001; 10.5089/9781513551180.053.A004
Sources: Organisation for Economic Co-operation and Development, Programme for the International Assessment of Adult Competencies; and IMF staff estimates.Note: The probability of automation is estimated using an expectation-maximization algorithm that relates individual characteristics (age, education, and training, among others) and job task characteristics to occupational-level risk of automation. Bars represent the gender difference in automatability = (share of females at high risk for automation) / (share of males at high risk for automation). High automatability is defined as probability of automation ≥ 0.7. Individuals between ages 20 and 39 are defined as younger generation. Individuals between ages 40 and 65 are defined as older generation. Statistical significance levels on bars reflect t-tests of the differences between proportions of male and female workers at high risk of replacement by technology. Statistical significance levels: *** p < 0.01; ** p < 0.05; * p < 0.1.Gender Gaps in High Risk of Automation by Generation and Educational Level
(Difference in automatability between females and males)
Citation: IMF Research Perspectives 2020, 001; 10.5089/9781513551180.053.A004
Sources: Organisation for Economic Co-operation and Development, Programme for the International Assessment of Adult Competencies; and IMF staff estimates.Note: The probability of automation is estimated using an expectation-maximization algorithm that relates individual characteristics (age, education, and training, among others) and job task characteristics to occupational-level risk of automation. Bars represent the gender difference in automatability = (share of females at high risk for automation) / (share of males at high risk for automation). High automatability is defined as probability of automation ≥ 0.7. Individuals between ages 20 and 39 are defined as younger generation. Individuals between ages 40 and 65 are defined as older generation. Statistical significance levels on bars reflect t-tests of the differences between proportions of male and female workers at high risk of replacement by technology. Statistical significance levels: *** p < 0.01; ** p < 0.05; * p < 0.1.Gender Gaps in High Risk of Automation by Generation and Educational Level
(Difference in automatability between females and males)
Citation: IMF Research Perspectives 2020, 001; 10.5089/9781513551180.053.A004
Sources: Organisation for Economic Co-operation and Development, Programme for the International Assessment of Adult Competencies; and IMF staff estimates.Note: The probability of automation is estimated using an expectation-maximization algorithm that relates individual characteristics (age, education, and training, among others) and job task characteristics to occupational-level risk of automation. Bars represent the gender difference in automatability = (share of females at high risk for automation) / (share of males at high risk for automation). High automatability is defined as probability of automation ≥ 0.7. Individuals between ages 20 and 39 are defined as younger generation. Individuals between ages 40 and 65 are defined as older generation. Statistical significance levels on bars reflect t-tests of the differences between proportions of male and female workers at high risk of replacement by technology. Statistical significance levels: *** p < 0.01; ** p < 0.05; * p < 0.1.At the sectoral level, more women in accommodation and food services, retail trade, and transportation face a high risk of automation (Figure 2). Women are also overrepresented in sectors relatively less exposed to automation, such as education and health care. However, even within less-automation-prone sectors, women face a higher risk than men. These differences suggest that not only selection of men and women across sectors but also variation in task composition performed within these sectors determine the likelihood of automation.
Gender Gap in High Risk of Automation across Sectors
Citation: IMF Research Perspectives 2020, 001; 10.5089/9781513551180.053.A004
Sources: Frey and Osbourne (2017); Organisation for Economic Co-operation and Development, Programme for the International Assessment of Adult Competencies; and IMF staff estimates.Note: The probability of automation is estimated using an expectation-maximization algorithm that relates individual characteristics (age, education, and training, among others) and job task characteristics to occupational-level risk of automation. Bars represent, respectively, the proportion of male and female workers at high risk of replacement by technology. High automatability is defined as probability of automation ≥ 0.7. Dots (and associated percentages) reflect the proportion of the male and female labor force employed in each sector. Statistical significance levels on the sectors reflect the t-test of the differences between proportions of male and female workers at high risk of replacement by technology. Statistical significance levels: *** p < 0.01; ** p < 0.05; * p < 0.1.Gender Gap in High Risk of Automation across Sectors
Citation: IMF Research Perspectives 2020, 001; 10.5089/9781513551180.053.A004
Sources: Frey and Osbourne (2017); Organisation for Economic Co-operation and Development, Programme for the International Assessment of Adult Competencies; and IMF staff estimates.Note: The probability of automation is estimated using an expectation-maximization algorithm that relates individual characteristics (age, education, and training, among others) and job task characteristics to occupational-level risk of automation. Bars represent, respectively, the proportion of male and female workers at high risk of replacement by technology. High automatability is defined as probability of automation ≥ 0.7. Dots (and associated percentages) reflect the proportion of the male and female labor force employed in each sector. Statistical significance levels on the sectors reflect the t-test of the differences between proportions of male and female workers at high risk of replacement by technology. Statistical significance levels: *** p < 0.01; ** p < 0.05; * p < 0.1.Gender Gap in High Risk of Automation across Sectors
Citation: IMF Research Perspectives 2020, 001; 10.5089/9781513551180.053.A004
Sources: Frey and Osbourne (2017); Organisation for Economic Co-operation and Development, Programme for the International Assessment of Adult Competencies; and IMF staff estimates.Note: The probability of automation is estimated using an expectation-maximization algorithm that relates individual characteristics (age, education, and training, among others) and job task characteristics to occupational-level risk of automation. Bars represent, respectively, the proportion of male and female workers at high risk of replacement by technology. High automatability is defined as probability of automation ≥ 0.7. Dots (and associated percentages) reflect the proportion of the male and female labor force employed in each sector. Statistical significance levels on the sectors reflect the t-test of the differences between proportions of male and female workers at high risk of replacement by technology. Statistical significance levels: *** p < 0.01; ** p < 0.05; * p < 0.1.Reassuringly, since the 1990s, more women than men have shifted from elementary and clerical occupations toward professional jobs, providing more insulation from displacement by technology. Moreover, job growth in aging economies is likely to be stronger in traditionally female-dominated sectors such as health and social services, where jobs require cognitive and interpersonal skills and thus are less prone to automation.
The overall positive trends displayed by the intergenerational analysis suggest that there is room for optimism about the future of work for women. An important caveat, however, is in order. The data set does not allow tracking of workers over time, so these results could also reflect widening gender gaps in routinization and thus risk of automation over the women’s life cycle. For instance, a large part of gender inequality in the labor market can be explained by a “childbearing penalty.” Thus, there may be an important role for policies that smooth transitions of younger female workers and ensure adequate safety nets for older, displaced workers (Brussevich and others 2018).
