There is increasing evidence that labor markets in developed countries are polarizing or hollowing out. On the one hand, the share of employment in high-skilled, high-paying occupations (managers, professionals and technicians) and low-skilled, low-paying occupations (elementary, service, and sales workers) is growing. On the other hand, the share of employment in middle-skilled, middle-paying occupations (clerks, plant and machine operators) is being squeezed. There is ample evidence of polarization in the United States (see Acemoglu and Autor, 2011; Autor and Dorn, 2013; and Autor (2014) for a less technical discussion), and also in Western Europe (Goos, Manning, and Salomons, 2014). Harrigan, Reshef and Toubal (2016), more recently, document the same phenomenon in France, using firm-level data.
In the World Development Report 2016 “Digital Dividends”, we show, for the first time, that the labor market is also hollowing out in developing countries (Figure 1). Since 1995, the average decline in the share of routine employment has been 0.39 percentage points a year or 7.8 percentage points for the period. This is a slower pace of labor market polarization than what is observed in developed countries, where, on average, the share of routine labor in employment has fallen by about 0.59 percentage points a year or almost 12 percentage points since 1995.
Figure 1: The labor market is becoming polarized in both developed and developing countries
Annual average change in employment share (percentage points), circa 1995-circa 2012, in high income countries (Panel A) and developing countries (Panel B)
Labor market polarization in developing countries, however, is not uniform. China, for example, is an exception, since the mechanization of agriculture increased the share of routine employment. Ethiopia, with a large share of employment in manual occupations, is also not polarizing; neither is Mongolia or certain Latin American countries where other factors—such as a commodity-driven boom benefiting low-skilled workers—could play a larger role in shaping labor markets.
Thinking of becoming a clerk or machine operator? Think again…
What explains these trends? Technological change is one factor. Greater computer power and internet connectivity reduce the need for workers in tasks and jobs that are codifiable and routine, and thus automatable. Machine operators and clerical support workers, for example, perform routine tasks that are easily automated. Frey and Osborne (2013) estimate that in the next two decades, from a purely technological standpoint, these jobs have more than an 80 percent probability of being automated. If you are curious, for economists, the probability of automation is close to 50 percent.
Fewer than half of today’s school children in China, Croatia, or Thailand can expect to find a job in an occupation that exists today. Even the jobs that do not disappear will be transformed. If instructions can be clearly written out, chances are an algorithm can do that job. This is partly what is behind the fall in the share of employment in middle-skilled jobs; it is in this category where most routine jobs fall. The remaining tasks and jobs require complex skills that complement technology, such as creativity, critical thinking, problem-solving, or teamwork, that is, non-routine cognitive and socio-emotional skills. Or, they require physical dexterity and human contact, features that machines lack.
The impact of digital technologies on jobs depends, therefore, on the tasks at hand and how technology either complements or substitutes workers in those tasks. In some cases, technology augments labor by complementing workers. Both the researcher and the hairdresser do tasks that are non-routine and not easily programmed into a computer, but technology makes the researcher (who uses more advanced skills at work) much more productive while barely affecting the hairdresser. This means that technology is skill-biased. In other cases, workers are in jobs that are routine, whether mostly manual or mostly cognitive, and are susceptible to automation and to seeing their jobs profoundly transformed or vanishing. In these cases, technology is labor-saving.
In addition to technological change, other factors are also likely to drive labor market polarization. Globalization, urbanization, and structural transformation fundamentally shape labor markets and in some cases are likely to be even more important than technology. In some Latin American countries, commodity-driven economic booms have boosted low-wage earnings and show no polarization at the aggregate level. Yet, technological change seems to be an important part of the explanation. Globalization does not fully explain the observed hollowing out.
Given the weight of China in manufacturing supply chains, the decline in routine labor in the rest of the world could simply be the result of a shift of routine labor to China’s manufacturing sector. Between 2000 and 2010, the share of employment in routine occupations in China rose from 19 to 27 percent. Yet manufacturing employment in China is also polarizing, for the increase in routine labor is explained by the mechanization of agriculture.
Moreover, across most countries with relevant data, employment is polarizing even within services, suggesting an additional effect in skill demand over and above what can be explained by trade or the structural transformation of developing economies. This evidence is also consistent with the evidence for OECD countries linking changes in skills requirements to technological changes, even within occupations.
While these changes are good for overall productivity, they can be deeply disruptive for people and could increase inequality. High-skill workers are the biggest winners when paired with digital technologies. Digital technologies complement their skills and allow them to be more productive and to find more rewarding opportunities and higher wages. Globally, returns to education remain high at 10 percent per year and are even higher for those using technology at work. The poor, with no access to technology and lacking skills, see few of the direct downsides of technological adoption and only partial benefits.
The main concern is that the ladder to the middle class is pulled away as middle-skilled jobs disappear or are fundamentally transformed. Such medium-skilled jobs, critical to the growth of the middle class, are giving way either to high-skilled jobs that only a small share of workers qualify for—or to low-skilled jobs that face increasing competition and most likely declining wages. As we document in the World Development Report 2016, the jobs where workers are likely to lose out are disproportionally held by the least educated and the bottom 40 percent of the income distribution. As a result, the biggest risk from the digital revolution is not massive unemployment, but widening income inequality.
The most important public policy response is to help everyone to gain the skills that technology complements, and not those that technology replaces. These skills remain difficult for technology to emulate, but also for education and training systems to provide, leaving many workers unprepared for the modern world of work. In Albania, Indonesia, Jordan, Malaysia, and Peru, more than half of youth are de facto functionally illiterate according to the latest results from the Programme for International Student Assessment (PISA), that is, they can read and write but can hardly understand a text or think critically. In the era of Google, when most information is at our fingertips, this is unlikely to be good enough.