Many developed countries have experienced job polarization as occupations that require skills in the middle of the distribution have been replaced by new technologies. This column reports evidence on the relationship between automation and employment in Latin America, analyzing data on five million workers in the six largest economies of the region over two decades. Among the findings is an indication that low-skilled, low-paid women workers who perform routine tasks are particularly vulnerable to automation. Investments in training and education are an essential policy response.
Technological change is one of the main engines of economic growth and social progress. But big changes in technology are also profoundly disruptive, at least in the short run, and may require policy responses to ease the transition.
In recent decades, a new concern has arisen: advances in digital technology and robotics are likely to replace routine tasks that follow well-defined rules, easily automated based on rule-based algorithms. This concern has been examined by the task-based approach to analysis of the workplace: the main idea here is that complementarity or substitutability between technology and labor depends on how susceptible different tasks are to automation.
Most studies of developed countries have found evidence of job polarization: routine tasks are heavily concentrated in the middle of the skills distribution, and hence employment has been increasingly concentrated in high-wage occupations and low-wage occupations, at the expense of traditionally middle-skill jobs.
In a recent study, we explore these issues in the six largest Latin American economies (Argentina, Brazil, Chile, Colombia, Mexico, and Peru, which together represent 79% of the continent’s total population and 86% of its total GDP) over the last two decades. In particular, we document patterns of change in employment by occupation, characterized by different degrees of exposure to routine tasks.
We make use of a rich new dataset – the OECD’s Programme for the International Assessment of Adult Competencies (PIACC) survey – to construct measures of the routine task content of the actual jobs of workers in Latin America. We combine them with microdata from about five million workers in the six largest economies of the region over two decades, from harmonized national household surveys conducted by the Center for Distributive, Labor and Social Studies (CEDLAS) and the World Bank.
We find evidence on six important aspects of the link between automation and employment in Latin America.
Exposure to routine tasks drops with education
Exposure to routinization is heterogeneous across demographic and socio-economic groups and across countries. The most salient asymmetry is among skill groups: our index of routine task content decreases very slowly up to around ten years of education and falls abruptly thereafter (see Figure 1).
On average for the six largest Latin American economies, our index of routine task content is 0.603 for the unskilled, somewhat smaller for the semi-skilled (0.527), and much lower for the skilled workers (0.341).
Figure 1: Index of routinization by years of education and gender
Note: The horizontal axis shows years of formal education. The vertical axis shows the index of routine task content constructed from microdata from PIAAC. The figure reports the unweighted mean value for the six Latin American economies included in the analysis: Argentina, Brazil, Chile, Colombia, Mexico, and Peru.
Automation affects those performing routine tasks
Our results are consistent with previous research indicating that workers performing routine tasks are more likely to have been affected by automation. Over the period, the increase in jobs was significantly decreasing in the degree of routinization (see Figure 2). At least since the mid-2000s, the structure of the labor market in Latin America has moved slowly toward occupations with a lower degree of exposure to routinization.
Figure 2: Growth rate in the number of workers by quintiles of index of routine task content
Note. Occupations sorted by quintiles of routine task content in the horizontal axis.
The size of changes depends on the business cycle
The magnitude of the changes has not been uniform over time. During the expansionary 2000s, employment increased for every group, but especially in those occupations that were less affected by the continuing process of increasing automation (see Figure 2). The pattern was similar though less marked during the more sluggish 2010s.
No evidence of polarization
Given the decreasing pattern of routine task content in education and wages, we do not find evidence of polarization in the labor market.
Some authors suggest possible reasons why, in contrast with advanced economies, polarization does not show up in developing countries data (at least not yet). Different initial occupational distributions, the impact of offshored jobs or the effect of new technologies in fostering sectors that employ middle-skill jobs are some possible explanations.
Employment losses but not wage reductions
In the period under analysis, high-routine occupations experienced reductions in employment share but not in real and relative wages. In fact, the evidence suggests that when the economy was growing, unskilled low-wage workers in high-routine occupations managed to get higher wage raises than the rest. In periods of stagnation, there were few changes in the structure of relative wages.
Unskilled women are particularly vulnerable
We find that low-skilled, low-paid women workers who perform routine tasks are particularly vulnerable to automation. In particular, whereas women’s employment grew substantially more than men’s employment in jobs with low levels of automatability, the gap vanished or even reversed among jobs with high routine task content.
The need for investment in education and training
Our findings suggest that training the most vulnerable individuals, in particular women, and investing in the education of younger generations are key elements to ensure that the future benefits of technological progress are distributed throughout the entire population.
