Is There Job Polarization in Developing Economies? A Review and Outlook Downloaded from https://academic.oup.com/wbro/article/39/2/259/7246968 by World Bank user on 19 July 2024 Antonio Martins-Neto , Nanditha Mathew , Pierre Mohnen , and Tania Treibich In this paper we analyze the evidence of job polarization—the relative decline of mid-wage jobs—in developing and emerging economies. We carry out an extensive literature review, revealing that job polarization in these countries is only incipient compared to advanced economies. We then examine the possible moderating aspects explaining this lack of job po- larization. We distinguish three groups of explanations: Limited technology adoption; struc- tural change; and changes in the global value chains. Finally, we suggest new microeconomic data and empirical analyses that should be developed in order to guide evidence-based policy- making addressing those issues in developing and emerging economies. JEL Codes: J24, J63, O33, E24 Keywords: job polarization, technology adoption, tasks, developing countries. Introduction The economic discipline has dedicated a great deal to the possible harmful effects of technological progress on the labor market (Katz and Summers 1989; Katz and Murphy 1992; Levy and Murnane 1992; Card and DiNardo 2002). Throughout recent history, and more famously after the Luddite movement, “technological unem- ployment” has been a persistent debate topic among economists, who have constantly deliberated whether massive waves of unemployment could be around the corner. However, the pessimistic predictions of technological unemployment have yet to be fulfilled. Technical progress did not pave its way through unemployment but rather through changes in the demand and composition of employment. For instance, steam power significantly favored unskilled workers to the detriment of skilled artisans, ac- celerating the transition of low-skilled workers moving out of the farms to better-paid jobs in the cities (Buyst et al. 2018).1 In contrast, subsequent technological waves were skill-using rather than skill-saving. The Digital Revolution in the early 1980s The World Bank Research Observer © The Author(s) 2023. Published by Oxford University Press on behalf of the International Bank for Reconstruction and Development / THE WORLD BANK. This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs licence (http://creativecommons.org/licenses/by-nc-nd/4.0/), which permits non- commercial reproduction and distribution of the work, in any medium, provided the original work is not altered or trans- formed in any way, and that the work is properly cited. For commercial re-use, please contact journals.permissions@oup.com https://doi.org/10.1093/wbro/lkad008 39:259–288 disproportionately and positively impacted the need for skilled workers, increasing the ratio of skilled to unskilled labor in most industries (Katz and Murphy 1992). Not surprisingly, when most developed countries have experienced increasing wage inequality in the past 40 years (Alvaredo et al. 2018), technology-related argu- ments have been at the forefront of explaining these labor market dynamics. The skill- biased technological change (SBTC) hypothesis suggested that technology, precisely the widespread adoption of Information and Communication Technologies (ICT), increased the demand for skilled workers, as they are more capable of using these Downloaded from https://academic.oup.com/wbro/article/39/2/259/7246968 by World Bank user on 19 July 2024 new technologies (see the review by Card and DiNardo 2002), and thereby causing earnings inequality to rise (Goos and Manning 2007; Acemoglu and Autor 2011). For a couple of decades, the SBTC hypothesis worked well in explaining the patterns observed in the data (Machin and Van Reenen 1998). However, it failed to explain another important labor market dynamic: in recent years, the share of high-skill, high-wage, and low-skill, low-wage occupations grew relative to those in the middle of the distribution, resulting in so-called job polarization (Goos et al. 2009). To account for the “hollowing out” of the occupational distribution, a more nuanced analysis focused on the tasks commonly performed by each occupation to explain the so-called job polarization in developed economies. The routine-biased technological change (RBTC) hypothesis argues that computers and robots have diminished the demand for routine, repetitive tasks in production, which are more commonly concentrated among middle-earning workers. On the other hand, tasks performed by unskilled workers, such as waiters or cleaners, and skilled workers, such as managers, are not easily codified and performed by computers (Autor and Dorn 2013; Goos et al. 2014). Evidence of job polarization has been extensively portrayed in developed economies. In the United States, it was first observed by Acemoglu (1999) and later rigorously analyzed by Autor et al. (2003). Beyond this first application, Goos et al. (2009) show a disproportionate increase in high-paid and low-paid employment relative to middle-paid jobs over the period 1993–2006 for 16 European countries, using harmonized data from the European Union Labour Force Survey (ELFS). Moreover, in addition to Michaels et al. (2014) and Goos et al. (2009, 2014), who find evidence of polarization for several OECD and European countries, similar results have also been individually estimated for Germany (Spitz-Oener 2006; Dustmann et al. 2009), the United Kingdom (Salvatori 2018; Montresor 2019), Portugal (Fonseca et al. 2018), and Japan (Ikenaga and Kambayashi 2016). The observed trends in advanced economies indicate that although technological change has not induced a surge in unemployment, it threatens to raise inequality and displace routine workers. However, beyond the context of developed economies, the literature on RBTC and its consequences on labor outcomes remain relatively limited. Understanding the labor market effects of technological change in emerging and developing economies is also important, as inequality and unemployment are already exceptionally high in these contexts. 260 The World Bank Research Observer, vol. 39, no. 2 (2024) The displacement of routine workers would be particularly harmful to less- educated and vulnerable groups who face more difficulties in finding another job and are more likely to transition towards low-stability, low-wage, and high-turnover occupations (Autor and Dorn 2013; Zago 2020). Furthermore, a growing demand for non-routine cognitive tasks would put further pressure on educational systems. In addition to fostering educational attainment, policy-makers in developing and emerging economies would need to respond quickly to the rapid changes in the demand for skills.2 Downloaded from https://academic.oup.com/wbro/article/39/2/259/7246968 by World Bank user on 19 July 2024 This paper attempts to provide a broad survey of job polarization in emerging and developing countries, giving special attention to the theoretical channels that could prevent or slow down job polarization dynamics. Specifically, we stress the roles of technology adoption, structural change, and global value chain (GVCs) participation in explaining differences across countries. Finally, we highlight policy implications that arise throughout the discussion, particularly the need for better data and empirical evidence to support policy design.3 Our review suggests a slower pace of job polarization in most developing and emerging economies, likely related to a significant gap in technology adoption and (or) different paths of structural change. Nevertheless, most of the literature also finds a decline in routine intensity in develop- ing economies (a precondition for job polarization), thus indicating relevant changes in the demand for skills. In addition, we find substantial gaps in the literature, espe- cially micro-level studies, that could significantly improve our understanding of the subject and facilitate the implementation of evidence-based policies. The rest of this paper is organized as follows. Section 2 describes the empirical literature on job polarization in developing economies. Next, Section 3 describes possible factors moderating the effect of automation in developing economies and investigates the interactions between technology adoption in advanced economies and the labor market implications in emerging countries. Section 4 explores the need for more micro-level studies and discusses the policy implications of job polarization in developing countries. The last section concludes. Is There Job Polarization in Developing Economies? Focusing on different regions and countries, as well as various measures of tasks and skills, the literature on job polarization in developing economies is gaining momen- tum (see table 1 for a detailed summary of this literature).4 For instance, Maloney and Molina (2019) use global census data for 67 developing countries and 13 devel- oped economies and, although the results corroborate labor market polarization and labor-displacing automation in developed economies, the authors find little evidence of either effect on developing economies, except for Mexico and China. Das and Hilgenstock (2022) use data on 85 countries since 1990 and observe similar results. In addition, the authors propose a measure of exposure to routinization based on Martins-Neto et al. 261 262 Table 1. Summary of the Existing Literature on Job Polarization in Developing and Emerging Economies Level of analysis Dataset Country Task Result Reference (1) (2) (3) (4) (5) (6) Impact on job polarization Country Global Census Data (IPUMS) 80 developed and developing Occupations (-) Maloney and Molina (2019) countries Country IPUMS, EULFS, household surveys 85 developed and developing O*NET (-) Das and Hilgenstock (2022) countries Country Household surveys Argentina, Brazil, Chile, Colombia, PIAAC (-) Gasparini et al. (2021) Mexico, and Peru Country Household surveys Chile and Mexico STEP (-) Messina et al. (2016) Country Household surveys 10 Central and Eastern European O*NET (-) Hardy et al. (2016) countries Country Household surveys 11 LAC countries O*NET (-) Beylis et al. (2020) Sectors Egyptian Central Agency for Public Egypt Occupations (+) Helmy (2015) Mobilization and Statistics Country Census data China O*NET (-) Ge et al. (2021) Country Household surveys Brazil O*NET (+/-) Firpo et al. (2021) Local Labor Markets CHIP surveys China O*NET (+/-) Fleisher et al. (2018) Sectors National Sample Survey Organization India Occupations (+) Sarkar (2019) Country Household surveys 70 countries O*NET (-) World Bank (2016) Country Household surveys 70 countries O*NET (-) Aedo et al. (2013) Country Household surveys 70 countries O*NET (-) Arias et al. (2014) Country European Union Labour Force Survey Central and Eastern European Occupations (-) Nchor and Rozmahel (2020) countries Impact on task content Country STEP 10 countries STEP (-) Dicarlo et al. (2016) Country Household surveys 86 countries STEP/O*NET (-) Lo Bello et al. (2019) Country STEP and PIAAC 42 countries STEP/PIAAC (-) Lewandowski et al. (2019) Country Household surveys 87 countries STEP (-) Lewandowski et al. (2020) Country STEP and PIAAC 35 countries STEP/PIAAC (-) Caunedo et al. (2021) Country STEP 10 countries STEP (-) Saltiel (2019) Source: Author’s elaboration. Note: The table is separated into groups of papers according to the primary dependent variable in the analyses. Impact on job polarization refers to studies examining the extent of job polarization and, in most cases, without a clear chain of causality between technology adoption and polarization. Impact on task content refers to papers focusing primarily on the differences in the task content of occupations across developed and developing economies. In addition, column 2 refers to the primary occupational dataset, while column 4 describes the measurement of tasks used. Column (5) indicates the sign of the significant relationship tested in each paper, that is, the existence of job polarization (in the impact on job polarization studies), or on the intensity of routine tasks (in the impact on the task content studies). The World Bank Research Observer, vol. 39, no. 2 (2024) Downloaded from https://academic.oup.com/wbro/article/39/2/259/7246968 by World Bank user on 19 July 2024 occupations’ risk of displacement by information technologies. Using this measure, the authors show that developing economies are significantly less exposed to rou- tinization and that initial exposure to routinization is a strong predictor of long-run exposure. The lack of polarization is further corroborated in Gasparini et al. (2021), who find similar conclusions for Latin America’s six largest economies (Argentina, Brazil, Chile, Colombia, Mexico, and Peru), arguing that although automation has largely impacted workers in routine-intensive occupations, there is no evidence for polariza- Downloaded from https://academic.oup.com/wbro/article/39/2/259/7246968 by World Bank user on 19 July 2024 tion in the labor market. Messina et al. (2016) employ the Skills Toward Employment and Productivity (STEP) Surveys conducted in Bolivia and Colombia as a proxy to measure the task content of jobs in Chile and Mexico. They find few signs of job polarization, except for Chile. In fact, Brazil, Mexico, and Peru present positive growth rates for workers in the middle of the wage distribution. Beylis et al. (2020) study the labor market of 11 Latin American countries (LAC) from 2000 to 2014. Applying the methodology proposed by Autor et al. (2003) and Acemoglu and Autor (2011), the analysis shows substantial changes in the composition of occupations. Although at a different intensity, the demand for routine manual intensive tasks has declined for the entire sample, coupled with a clear and marked increase in the demand for non- routine intensive occupations. Yet, these changes in the labor composition have not resulted in polarized markets. In Central and Eastern European economies, Nchor and Rozmahel (2020) find that despite an increase in the demand for high-skill workers and a decline in middle-skill employment, the rise in low-skill employment is minimal to lead to a U-shape employment distribution which indicates labor polarization. Even among developing and emerging economies, the evidence is not homoge- neous. Hardy et al. (2016) study 10 Central and Eastern European (CEE) countries and point to an increase in non-routine cognitive tasks and a decrease in manual tasks. Nevertheless, contrary to other developed countries and at odds with RBTC, the authors also find that routine cognitive tasks increased in six CEE countries, re- mained stable in two, and declined in the remaining countries. Helmy (2015) studies the Egyptian labor market over the period 2000–2009 and finds suggestive evidence of job polarization, with a decline of 5.9 percent in the share of employment of middle-skilled occupations compared to a growth of 4.5 percent and 1.4 percent for low- and high-skilled occupations. Ge et al. (2021) use census data from China and find that the share of employment in routine manual occupations declined by 25 percentage points from 1990 to 2015. Similarly, Firpo et al. (2021) find evidence of wage polarization in Brazil, but not with respect to employment. In contrast, Fleisher et al. (2018) show that middle-skilled jobs are increasingly transitioning to work in the unskilled and self-employment job categories in China, consistent with the RBTC hypothesis. Similarly, using data from the National Sample Survey Organization from India, Sarkar (2019) also observes increasing job polarization during the 1990s and 2000s. In the period 1984–94, the author finds an up-grading pattern, with a substantial increase in the employment of high-skilled occupations. In contrast, the Martins-Neto et al. 263 following periods show a polarized U-shaped employment growth, with a decline of almost 20 percent for occupations in the 40th percentile of the skill distribution. Table 1 summarizes the main findings of this section, highlighting the context of the studies: the unit of analysis, the data sources, the countries, the task mea- surements, and the impact of technological change on two outcomes of interest, the existence or not of job polarization and the increase of decline in the intensity of routine tasks. Except for the cases of India, the Arab Republic of Egypt, and China, most papers fail to observe job polarization in emerging and developing economies. Downloaded from https://academic.oup.com/wbro/article/39/2/259/7246968 by World Bank user on 19 July 2024 However, as previously discussed, many articles already observe a decline in the routine intensity across low- and middle-income countries—a precondition for job polarization. For the group of papers exploring the impact on task content, all results are negative, suggesting that developing countries are less intensive in non-routine cognitive skills than advanced economies. We will explore these differences in more detail in subsection 4.2 and highlight the need for better measures of tasks across occupations in emerging and developing economies. The Missing Job Polarization The literature has identified three main channels driving job polarization: Technol- ogy adoption; structural change; and participation in global value chains (GVCs). This section discusses how developing economies may differ from advanced ones in each of these aspects and, in turn, how that difference may explain the absence of job polarization in developing and emerging economies. For each channel, we first present the general theoretical mechanisms and then discuss the main differences observed in developing and emerging economies vis-à-vis advanced ones. We begin by examining the role of technology adoption, focusing on why firms may have lower rates of adoption and exploring potential explanations for differences in technology choice. We then move from a micro-level to a macro-level discussion, illustrating the role of structural change and regional differences as key drivers of job polarization. Finally, we open our economy to international trade and discuss how both the micro and macro aspects of a given economy are affected by a country’s participation in GVCs. Although we present each mechanism separately for the sake of simplicity, we emphasize that all of them are interacting forces. For instance, structural change and differences across sectors and regions are to a large extent a combination of firms’ decisions either at the local level or a result of a country’s participation in GVCs. Technology Adoption The “routinization” hypothesis argues that firms combine a continuum of tasks to produce, which can be performed either by capital or labor (Autor et al. 2003; Acemoglu and Autor 2011). Firms will allocate more capital or labor in a given task 264 The World Bank Research Observer, vol. 39, no. 2 (2024) depending on their relative cost and the degree to which tasks can be automated (repetitive and replaceable by code and machines). In past decades, not only did the quality-adjusted ICT and robot prices fall considerably, but these technologies have been particularly successful in carrying out tasks that follow explicit rules (routines) (Michaels et al. 2014; Graetz and Michaels 2018). As a result, firms accelerated the substitution of labor in routine tasks, so workers in routine-intensive occupations were suddenly at high risk of displacement. Traditionally, many routine tasks are concentrated in middle-wage, middle-skill white-collar jobs such as bank clerks, Downloaded from https://academic.oup.com/wbro/article/39/2/259/7246968 by World Bank user on 19 July 2024 or are carried out by blue-collar, less-educated workers, performing, for example, assembly tasks. As firms increase the share of capital in production, the demand for middle-earning jobs should contract, and the labor market should polarize. Yet, while ICT and other automated technologies are expected to be widespread in ad- vanced economies, lower adoption rates can be found in developing and emerging economies. The slow pace of technological adoption in these economies may reflect many aspects, including firms’ capabilities, the extent of informality, and countries’ human capital endowments. Firm Behavior and Capabilities Firms’ ability and willingness to adopt digital technologies are heterogeneous across and within countries. For instance, in the specific cases of Brazil and Vietnam, recent evidence suggests that most firms still rely on pre-digital technologies to perform daily tasks (Cirera et al. 2021a,b). However, rather than a sign of backwardness, firms’ de- cision to not adopt more advanced technologies may be an optimal response to their small scale, local competition, and the relative price of labor and capital. Labor is sub- stantially cheaper in developing economies, and the number of small and informal es- tablishments with a small production scale is larger (we discuss the role of informality below). As Banerjee and Duflo (2005) point out, one reason for the lag in technology adoption could be that the firms are too small to profit from the best technologies. Similarly, when wages are low, the relative price of investment is relatively higher (Hsieh and Klenow 2007) and deters technology adoption. In the context of de- veloped economies, Shim and Yang (2018) show that, in the United States, in high-paying sectors (where therefore, the relative cost of wages compared to capital is higher), there are incentives to replace routine employment. This is confirmed by Lordan and Neumark (2018), who show that minimum wage increases are asso- ciated with a higher probability of replacing routine occupations. In other words, lower wages disincentivize firms in developing countries to adopt more sophisticated technologies. Yet, decisions are not always optimal, and firms may simply not be aware of the available technologies. Due to restricted technological diffusion, advanced technologies have limited diffusion in developing economies—a classic example of Martins-Neto et al. 265 information failure. Acquiring this knowledge can be very costly, and companies may think that adopting new practices would not be profitable (Jensen 1988). Finally, even when managers are aware of best practices, there is a final process of acceptance and implementation. As once stated by Rosenberg (1972, p.191), “in the history of diffusion of many innovations, one cannot help being struck by two characteristics of the diffusion process: its apparent overall slowness on the one hand, and the wide variations in the rates of acceptance of different inventions, on the other.” Technology adoption also depends on firms’ dynamic capabilities (Teece et al. Downloaded from https://academic.oup.com/wbro/article/39/2/259/7246968 by World Bank user on 19 July 2024 1997), that is, their ability to “integrate, build and reconfigure internal and external competencies to address rapidly changing environments.” Therefore, the diffusion of (foreign) new technologies within developing economies also relates to firms’ absorptive capacity (Cohen et al. 1990) and can explain differences in knowledge spillovers and adoption behavior in firms (Fagerberg 1994). Because of institutional and resource constraints in developing economies, firms’ low absorptive capacity could be critical to explaining limited technology adoption. Informal Sector The sizeable informal sector in emerging and developing economies could also impact the patterns of job polarization. The informal sector, which accounts for 90 percent of the economy in developing (low-income) countries and 67 percent in emerging (upper-middle and lower-middle) countries (Bonnet et al. 2019), typically lags in adopting the latest technologies (Cirera et al. 2021), is labor-intensive and has lower productivity compared to the formal sector (La Porta and Shleifer 2014), and most of its workers are engaged in low-skilled services and artisanal production (Falco et al. 2015). Therefore, the potential of technology-driven job displacement is likely less severe in countries with a proportionally large informal economy.5 Moreover, technology adoption in the formal sector displaces workers toward the informal sector and, through this channel, may also affect wages there (Chacaltana Janampa et al. 2018). Using a general equilibrium model, Gomez (2021) finds that an increase in technology adoption in the formal sector results in a larger informal sector and lower wage inequality at the bottom of the skill distribution. Availability of Human Capital Human capital is an essential factor in explaining the adoption of advanced tech- nologies within firms. For instance, using a large cross-country sample of developed and developing economies, Benhabib and Spiegel (1994) show that human capital affects the speed at which countries absorb technological developments. Comin and Hobijn (2004) examine the diffusion of more than 20 technologies across developed economies and find that countries’ human capital endowment is the most crucial determinant of the pace of technology adoption. As clearly stated by 266 The World Bank Research Observer, vol. 39, no. 2 (2024) Boothby et al. (2010, p. 621), “firms embracing new technology have to obtain new skills and/or to upgrade the skill level of their existing workforce because the attributes of new technology could be significantly different from old technologies.” The literature has largely stressed the lack of managerial capabilities (Bloom and Van Reenen 2010) and workers’ skills in developing economies, which in turn are a crit- ical constraint to innovation and technology adoption (Cirera and Maloney 2017). Educated managers may have a greater understanding of sophisticated technologies and be favorably disposed to adopt them. For instance, using data on digital technol- Downloaded from https://academic.oup.com/wbro/article/39/2/259/7246968 by World Bank user on 19 July 2024 ogy usage, Nicoletti et al. (2020) find empirical evidence that low managerial quality and the lack of ICT skills are negatively associated with technological adoption in 25 European economies. Structural Explanations: Sectors, Regions, and Demographic Change Job polarization is a combination of within-industry and between-industry changes in employment shares, which are, in turn, affected by demographic changes and their effect on the demand for goods and services across firms, sectors and regions. In what follows, we detail how the characteristics of developing and emerging economies in terms of these different dimensions may affect their employment structure and dynamics. Structural Change On the one hand, as technological change replaces routine tasks, a given industry will use less routine employment even while maintaining the same output levels. On the other hand, occupations’ intensity in such routine tasks differs across industries, such that sectoral employment shifts also explain aggregate occupational share changes (Goos et al. 2014). In fact, Foster-McGregor et al. (2021) suggest that the risk of automation (or routine intensity) shows only modest variation within sectors and between countries, but a considerably greater variation between industries within countries. Specifically, manufacturing sectors generally demand relatively larger shares of middle-skilled, routine occupations than agriculture and services. For example, Lee and Shin (2019) find that polarization is faster in manufacturing than in services, and Bárány and Siegel (2018) indicate that job polarization in the United States is directly linked to the decline of manufacturing employment since the early 1950–1960s. Therefore, the level of aggregate routine intensity depends on the sectoral structure of employment—for example, we may expect that the higher the share of manufacturing, the higher the routine intensity for a given country. What do these findings imply regarding employment dynamics in developing and emerging economies? The answer lies in the countries’ trajectories. Often, low-income countries have a significant share of employment in agriculture and Martins-Neto et al. 267 a small percentage of workers engaged in routine tasks in the first place. As coun- tries become more productive in agriculture and start industrializing, they also increase their share of routine occupations. As clearly stated in Das and Hilgenstock (2022, p.100), “the observed increase in the exposure of routinization in developing economies indicates that structural transformation was greater than the offsetting impact from the declining in the price of ICT capital.” Industrialization thus moder- ates the effects of technological change on the demand for routine labor. Overall, Das and Hilgenstock (2022) show that labor markets in low- and middle-income coun- Downloaded from https://academic.oup.com/wbro/article/39/2/259/7246968 by World Bank user on 19 July 2024 tries are significantly less exposed to routinization (lower share of routine-intensive occupations), reflecting the larger share of agriculture in developing economies. In contrast, at more advanced stages of development, countries transition from manufacturing to services and job polarization accelerates.6 Heterogeneity Across Sectors and Regions Much of the literature presented above has relied primarily on aggregate measures, and thus somewhat overlooked job polarization’s regional and sectoral hetero- geneities. It remains unclear if the slow pace of polarization in most developing and emerging economies is a general trend or is confined to a few sectors or regions within countries. Some related evidence can be found for developed economies. Using individual-level data from Statistics Sweden from 2002 to 2012, Henning and Eriksson (2020) find that the decline in manufacturing employment in clusters of previously manufacturing-dominated municipalities drives polarization in the country. In contrast, areas with fast-growing firms in sectors with larger shares of routine workers (extraction industries and lower manufacturing) exhibit the opposite patterns, indicating a greater tendency towards job upgrading. Regional and sectoral differences, and more specifically, the role of extractive industries, could therefore help to explain the modest evidence of job polarization in some emerging economies.7 The commodity boom in the early 2000s led to a significant expansion of the extractive sector in many countries, which is likely to offset the decline in middle-earning jobs across other sectors. Indeed, in many Latin American and African economies, the commodity boom experienced during the 2000s mainly favored low-skilled workers, potentially overshadowing the impacts of ICT adoption (Maloney and Molina 2019). Demography Finally, differences in demographic dynamics across developed and developing economies affect changes in the demand for goods and services as well as the supply of work, therefore resulting in diverging patterns of overall employment. Moreno- Galbis and Sopraseuth (2014) show that population aging in developed economies leads to a rise in the demand for personal services, causing an increase in the 268 The World Bank Research Observer, vol. 39, no. 2 (2024) employment share of low-paid positions. For instance, population aging leads to a rise in the demand for jobs such as cleaners, transportation services in the health indus- try, and housework employees in private homes. In addition, Acemoglu and Restrepo (2021) find that population aging results in a shortage of middle-skilled workers, thus increasing the adoption of automation technologies. However, this pattern contrasts with the demography of most emerging economies. Especially in Africa, countries are experiencing significant growth in the working-age population, resulting in a less in- tense demand for low-paid occupations and an abundance of middle-skilled workers. Downloaded from https://academic.oup.com/wbro/article/39/2/259/7246968 by World Bank user on 19 July 2024 Employment Dynamics in Open Economies Most of the literature on job polarization in developing countries has relied on iso- lated analysis at the country level without considering possible effects stemming from changes in global value chains. The effects of GVCs on job polarization in developing economies are not straightforward. Technological development has drastically re- duced the costs of offshoring jobs to locations with lower labor costs, such that firms in developed economies have off-shored routine-intensive occupations (Acemoglu and Autor 2011; Blinder and Krueger 2013; Goos et al. 2014). In turn, the inflow of routine jobs from advanced countries has likely reduced polarization forces in some host countries (Maloney and Molina 2019). At the same time, new advancements in robotics have reduced the prices of these technologies substantially, resulting in developed economies re-shoring part of their production. The rapid spread of robots in advanced economies could thus have the opposite effect, likely reducing the share of routine workers and accelerating job po- larization in developing economies. Krenz et al. (2021) develop a theoretical model to account for these interactions in which automation in advanced countries increases productivity and reduces the costs of producing in-shore. As a result, part of the production that was previously off-shored to host areas in developing regions may return, although not leading to an improvement in wages for low-skilled workers or the creation of new jobs in the receiving economies. Below, we examine these two contrasting forces affecting job polarization in de- veloping economies. We first highlight the initial findings pointing to the role of offshoring in mitigating job polarization in developing economies. We then point to more recent evidence about the effects of re-shoring and conclude by discussing the specific case of multinationals (MNEs). Global Value Chains and the Routinization of Tasks Early studies on the interactions between global value chains (GVCs) and job polar- ization pointed to different trajectories between developed and developing economies. For instance, Das and Hilgenstock (2022) show that participation in GVCs might Martins-Neto et al. 269 have played a role in the rising number of routine jobs in developing economies, while reducing them in advanced economies. Similarly, Reijnders and de Vries (2018) ex- plore the impacts of both technological change and offshoring on the labor market for several developed and emerging economies. Although the results corroborate an increasing share of non-routine occupations in the labor market of both groups, the authors find that the effect of task reallocation via offshoring reinforces the decline in routine jobs for advanced economies and mitigates it for developing countries. In ad- dition, Lewandowski et al. (2019) test the association between the routine-intensity Downloaded from https://academic.oup.com/wbro/article/39/2/259/7246968 by World Bank user on 19 July 2024 of occupations and technology (computer use), globalization (specialization in global value chains), structural change, and supply of skills in 42 countries at different stages of development. The results generally corroborate the main drivers of job po- larization. On the one hand, technology, structural change, and the supply of skilled workers are positively correlated with routine intensity. On the other hand, glob- alization is positively associated with routine intensity in developing countries and negatively in developed countries, reinforcing the argument that developed countries are offshoring routine occupations to host countries. Finally, Lo Bello et al. (2019) study both supply (e.g., education, age, and age structure) and demand (growth, sector structure, technology, and trade) factors in explaining differences in the skill content of jobs and find that technology adoption is related to de-routinization and trade is an offsetting force in developing economies. New Trends: Reshoring, Robot Adoption, and Job Polarization Recent findings show that automation may be linked to reshoring or decreased offshoring, implying decreasing employment in developing economies.8 Krenz et al. (2021) explore 43 countries and nine manufacturing sectors and provide evidence that robot adoption increases re-shoring activity. Similarly, Kinkel et al. (2015) analyze 3,313 manufacturing firms in seven European countries and find empirical evidence that firms using industrial robots are less likely to off-shore their production outside the region. Without focusing on offshoring per se, a recent strand of the literature also shows that robot adoption in developed economies negatively impacts wages and employ- ment in developing economies. Using data from Mexican local labor markets between 1990 and 2015 and the International Federation of Robotics (IFR), Faber (2020) shows a negative impact of robot adoption on Mexican employment, with a more substantial effect for women and low-educated machine operators in the manufac- turing sector. Also exploring the Mexican labor market, Artuc et al. (2019) show that an increase of one robot per thousand workers in the United States lowers growth in exports per worker from Mexico to the United States by 6.7 percent. However, the authors did not find evidence of an impact on wage employment or manufacturing wage employment. Kugler et al. (2020) use data from the International Federation 270 The World Bank Research Observer, vol. 39, no. 2 (2024) of Robotics (IFR) to measure automation in the United States and microdata from the Colombian Social Security records to examine the effects of robot adoption in the United States on the Colombian labor market. The results indicate a negative impact on the employment and wages of Colombian workers, especially for women, older and middle-aged workers, and workers employed by SMEs.9 The Role of MNEs The literature has yet to examine the role of MNEs as drivers of job polarization in Downloaded from https://academic.oup.com/wbro/article/39/2/259/7246968 by World Bank user on 19 July 2024 emerging economies. An extensive literature has already provided evidence that MNEs are more productive (Helpman et al. 2004), pay higher wages (Hijzen et al. 2013), and employ a higher share of non-routine jobs (Hakkala et al. 2014). In this context, an increase in foreign direct investment (FDI) could have implications for job polarization in host economies. For instance, Olsson and Tag (2017) exam- ine the impacts of private equity acquisition on the employment composition of recently acquired firms in Sweden and find that workers in less productive firms in routine-intensive occupations are twice as likely to be displaced after buyouts. In the specific case of FDI, Hakkala et al. (2014) rely on Swedish data to study changes in firms’ ownership and find that MNEs demand more non-routine tasks or tasks requiring personal interactions compared to their local counterparts. In addition, Amoroso and Moncada-Paterno-Castello (2018) use data on greenfield FDI for sev- eral European economies to examine the extent to which different types of FDI are related to job polarization. They find that low-skill FDI investments are associated with skill down-grading, while skill-intensive FDI is more commonly associated with skill upgrading. Only investments in ICT are related to job polarization. Yet, as for developing economies, the overall impact on the labor market will de- pend on many factors. In addition to the current economic structure and the target sectors (either low-skill or skill-intensive), the impacts of FDI also rely on foreign firms’ ability to spur technology adoption. Changes in ownership and the increasing share of MNEs in already established sectors could have different impacts. For in- stance, extensive literature has pointed out MNEs’ role in transferring technology and managerial skills (for example, Teece 1977). In this context, if MNEs catalyze tech- nology adoption across local firms, job polarization could emerge as an overall effect of more extensive technology dif fusion. In contrast, a dif ferent strand of the literature stresses that MNEs are more likely to crowd out local firms, use technology that is inappropriate for local circumstances, and limit technology transfer (Oetzel and Doh 2009). As a result, job polarization would be limited to a few MNEs, and the extent of polarization would depend on MNEs’ share in total employment. Tacking Stock Our literature review indicated a significant decline in routine intensity in many developing economies, although with little evidence of job polarization. In addition, Martins-Neto et al. 271 section 3 has explored the reasons for such a lack of polarization in (most) emerging and developing economies and has highlighted some of the main gaps in the litera- ture. We have stressed the need to empirically examine the main drivers of the slow pace of polarization, including countries’ participation in GVCs. A critical argument in our discussion is that structural change and GVC partici- pation can counterbalance the effects of technology adoption on labor demand for routine tasks in emerging and developing countries. Yet, we do not have empirical evidence on this particular process. Also, the observed differences across countries, Downloaded from https://academic.oup.com/wbro/article/39/2/259/7246968 by World Bank user on 19 July 2024 also at a similar level of income or technological knowledge, raise many questions and suggest that further evidence should explore more disaggregated information. For instance, is there within-sector polarization in low- and middle-income coun- tries? Has the process of industrialization curbed the aggregate routine intensity among those economies? Did occupations become less intense in routine tasks over time? Lastly, has the falling demand for routine tasks negatively impacted workers? Answering these questions (and many others) can significantly impact the de- velopment of better-adapted technological, educational, and labor market policies. The following section discusses the opportunities and challenges associated with technology policies in developing and emerging economies and the implications in terms of employment patterns and policies. The Need for More Studies Based on Microdata to Guide Policymaking As mentioned by Case and Deaton (2020, p.261), “[G]lobalization and automation are ultimately beneficial, but they create disruption, especially in the short run, and many less skilled workers lose out.” This conflicting impact of technology poses additional challenges to policy-makers, highlighting the need for complementar- ity in public policies. For instance, while encouraging and facilitating technology adoption, labor market de-routinization calls for robust social protection systems to help workers with low job mobility, especially more disadvantaged groups. For instance, Lewandowski et al. (2017) study the intergenerational disparities in the de-routinization of jobs in 12 European countries and find a significant relation- ship between age groups and shifts in task composition. The decreasing demand for routine occupations also challenges existing education and training systems to re- spond to changing skill demands, especially given the fact that low-educated workers are commonly more affected by the routinization process (Martins-Neto et al. 2022). It is crucial to adequately equip the labor force with the necessary skills to guarantee maximum benefits from recent technological advancements, stimulating the devel- opment of competencies with increasing demand—an excellent example of this is the soft-skills training for employees in the hotels and accommodation industry (for instance, the training from Quality Assurance Agency 2015 in the United Kingdom). 272 The World Bank Research Observer, vol. 39, no. 2 (2024) Ultimately, designing better-fitted policies for skill development, such as programs up-scaling digital skills, vocational training, and better-adapted social protection sys- tems, requires detailed microeconomic studies. Researchers need to move from ag- gregate measurements of polarization into micro-level information to examine differ- ences across firms and workers, including assessing workers’ ability to transition from displacement to re-employment in high-paying jobs in different institutional con- texts. This calls for more systematic and frequent micro-level data collection in devel- oping economies to better understand the task content of occupations specific to each Downloaded from https://academic.oup.com/wbro/article/39/2/259/7246968 by World Bank user on 19 July 2024 country as well as constraints and patterns of technology adoption at the firm level. The remainder of this section presents the main shortcomings that limit a more detailed overview of the effects of technology adoption in low- and middle-income countries. First, we discuss the available measures of technology adoption (sec- tion 4.1) and tasks (section 4.2) and highlight the need for longitudinal and micro- level data. Following this discussion, we point out some of the main gaps in the (empirical) literature, focusing on those that could vastly improve our understanding and facilitate the development of appropriate public policies (section 4.3). Measuring Firm-level Technology Adoption Emerging and developing economies lack information on technology adoption at dis- aggregated levels. Efforts to expand our knowledge in this direction would facilitate a finer understanding of the composition effects of technology adoption and expand our knowledge of the main barriers preventing the adoption of more advanced technologies among those economies. Some recent efforts have provided new evi- dence and data in this direction. For instance, a new survey by the World Bank offers granular information on the adoption (extensive margin) and use (intensive margin) of technologies for both general business functions and sector-specific business func- tions for several emerging and developing economies. Even though there is significant heterogeneity across firms, the results indicate that, on average, firms are adopting manual, pre-digital technologies (Cirera et al. 2021b). In addition, a novel database from UNIDO offers detailed information on the adoption of production technology in developing economies (see, for instance, Delera et al. 2022). The results also point to few firms adopting very advanced technologies and large heterogeneity among firms. However, the continuous evolution of technologies (Dosi 1982) makes it challeng- ing to measure their adoption. Indeed, firms may need to maintain, upgrade or adapt the technologies embedded in their production processes over time—then, which of these decisions should be considered as technology adoption per se? The study of such dynamic systems, i.e., how technologies and their adoption evolve and how firms, workers and their skills co-evolve, requires longitudinal data that tracks firms over time. Further data and research on this would improve our understanding of the Martins-Neto et al. 273 relation between firm characteristics, local availability of skills, and technological paradigms in emerging and developing economies. Measuring the Task Content of Jobs Across Countries Data collection and integration at a decentralized level with a detailed skill mapping system will help local economies to shape policies to foster skill upgrading and place themselves in a better position to respond to the threats and opportunities brought Downloaded from https://academic.oup.com/wbro/article/39/2/259/7246968 by World Bank user on 19 July 2024 by technological change. Measuring Tasks with the O*NET Database The literature on RBTC explicitly explores differences in task composition across occupations to study the labor market consequences of technological development. Within this approach, two main methods have been developed, as also illustrated in column 4 of table A1: the first one using the O*NET database, and the second one building on information about tasks from the PIAAC and/or STEP surveys (see also table A1 for a general comparison of these measures). The first approach focuses on occupational level tasks, which provide information on job characteristics only at the occupational level but not at the worker level. Specifically, authors have used the Dictionary of Occupational Titles (DOT) survey and its updated version, the O*NET. Using the O*NET dataset, Autor et al. (2003) developed a “routine task intensity” index based on the routine, abstract, and manual task content for each occupation.10 The use of the O*NET database allowed for a significant transition in the literature, as we are now able to measure the tasks performed in jobs rather than simply the educational level of workers performing them.11 This measure has also been adopted in the case of studies on developing countries, under the assumption that the task content across occupations is similar across countries.12 However, the assumption that the task content of occupations is similar between countries is obviously a strong one. Differences in technology use are likely to result in different job tasks performed by a machine operator in the United States and those performed by a machine operator in a low-income country. Measuring Tasks with the PIAAC and STEP Surveys In response to this caveat, a second approach has used worker-level information pro- vided by new household surveys such as the Program for International Assessment of Adult Competencies (PIAAC) by the Organisation for Economic Co-operation and Development (OECD) and Skills Toward Employment and Productivity (STEP) by the World Bank. Both surveys attempt to measure tasks and skills across the developing world.13 274 The World Bank Research Observer, vol. 39, no. 2 (2024) Dicarlo et al. (2016) construct a measure of the skill content of occupations for 10 low and middle-income countries using the STEP skill measurement surveys and compare it with that of the United States. A number of exciting facts result from this comparison: (a) first, along the skill dimension, occupations are ranked similarly across countries; (b) second, workers in higher-income countries use analytical and interpersonal skills more frequently; (c) lastly, there are significant differences in the skill content across countries, so that assuming that the US skill content is a good proxy for developing countries is wrong and likely to impact the estimates. Messina Downloaded from https://academic.oup.com/wbro/article/39/2/259/7246968 by World Bank user on 19 July 2024 et al. (2016) also employ the STEP Surveys conducted in Bolivia and Colombia as a proxy for the routine/abstract/manual content of jobs in Latin America. They show that Latin American occupations exhibit a higher manual content than similar oc- cupations in the United States. Similar results are discussed in Lo Bello et al. (2019), who apply the STEP survey for a more significant number of developing countries. The authors argue that indexes based on US data do not provide a fair approximation of routine cognitive and non-routine manual skill content of jobs in developing countries. Lo Bello et al. (2019) also point out two caveats in using the STEP Surveys. First, as estimates are based on workers’ responses, it is assumed that workers do not differ in their view of tasks performed at work. However, this assumption may not hold as most questions are subjective. Second, the survey focuses on urban areas, thus under-representing the agricultural sector. Lewandowski et al. (2019) combine the STEP and PIAAC surveys and develop a harmonized measure of the task content of occupations based on Acemoglu and Autor (2011).14 The authors find that workers in developed economies perform mostly non-routine cognitive analytical and non-routine cognitive interpersonal tasks. In contrast, workers in developing economies perform routine tasks more in- tensively. Moreover, Lewandowski et al. (2020) explore the PIAAC survey for several countries and develop a regression-based methodology to predict the country-specific routine task intensity of occupations, thus overcoming the lack of survey data for several large developing economies, such as Brazil and India. Besides corroborating that occupations in developing countries are more routine intensive, the authors also find that from 2000 to 2017, the gap in average routine-task intensity with respect to developed countries has increased. In contrast, Gasparini et al. (2021) use harmonized national household surveys for Latin America’s six largest economies combined with task content based on information from the PIAAC surveys conducted in Chile, Mexico, Peru, and Ecuador. Applying the mean results derived from these four economies, the authors find a strong linear correlation between their measure of routine intensity and the routine task index developed by Autor and Dorn (2013). Finally, Caunedo et al. (2021) construct a measure of occupational task content using the PIAAC and STEP surveys from 2006 to 2015 and find that developed countries use non-routine analytical and interpersonal tasks more intensively than developing countries. In contrast, developing countries use routine cognitive and Martins-Neto et al. 275 routine-manual tasks more intensively. In addition, the authors show that countries are converging to similar task intensities over this period. Within-occupations Variance Another important aspect besides differences in task intensity across occupations is the extent of within-occupations variance in tasks. As discussed above, both DOT and O*NET provide information only at the level of occupations, not workers. Therefore, the implementation of worker-level surveys, including the PIAAC and STEP surveys Downloaded from https://academic.oup.com/wbro/article/39/2/259/7246968 by World Bank user on 19 July 2024 discussed above, allow us to study within-occupation differences. For example, Autor and Handel (2013) explore data from the Princeton Data Improvement Initiative (PDII) survey (formerly STAMP) and document that tasks vary substantially within occupations in the United States. Stinebrickner et al. (2019) take advantage of data from the Berea Panel Study and explore the contribution of task content to wage growth, finding that high-skilled tasks pay substantially more than low-skilled tasks. In the context of developing economies and to the best of our knowledge, Saltiel’s (2019) is the only paper to consider the returns to worker-level task measures. The author explores work-level data from the STEP survey for 10 low- and middle-income countries, finding substantial variance in task intensity within occupations and sug- gesting that non-routine analytic and interpersonal tasks are associated with sizable wage premiums. In addition, the empirical findings suggest that more educated workers sort into occupations with higher non-routine task content. Evolution of Tasks Over Time Despite the recent developments in task measurement across the developing world, the literature still lacks information on the evolution of tasks. Not only do occu- pations differ across countries, but they also evolve over time. For instance, using data from job ads from The Boston Globe, The New York Times, and The Wall Street Journal, Atalay et al. (2020) demonstrate that words related to routine tasks have declined in frequency over the period from 1950 to 2000 in the United States. Fur- thermore, Garcia-Couto (2020) harmonizes data from three different rounds of the Dictionary of Occupation Title (DOT) and the Occupational Information Network (O*NET) and finds that the cognitive intensity of occupations has increased during recent decades and that a significant share of changes in wages is due to increases in the return and the intensity of cognitive tasks. Similar trends are also observed by Cassidy (2017) and Spitz-Oener (2006), who use the German Qualification and Career Survey conducted by Federal Institute for Vocational Education and Training (BIBB) and the Institute for Employment (IAB). As for developing economies, it remains unclear whether (and to what ex- tent) changes in tasks within occupations are similar to what we observe in ad- vanced economies. Most analyses still rely on occupational and sector composition 276 The World Bank Research Observer, vol. 39, no. 2 (2024) information to determine the extent of polarization without a clear understanding of changes in task requirements over time. An obvious reason for this gap is the lack of longitudinal data sources, which subsequent rounds of the STEP and PIAAC surveys could overcome. Thus, in addition to expanding the number of countries covered in the study, especially emerging economies, it is also critical to gather information on worker-level tasks within countries over time. Another way forward would be to use job ads from job platforms to study the demand for digital skills and non-routine tasks in developing countries. Following the methodology proposed by Atalay et al. Downloaded from https://academic.oup.com/wbro/article/39/2/259/7246968 by World Bank user on 19 July 2024 (2020), researchers could explore other platforms to study the evolution of tasks de- manded in some emerging economies. Yet, researchers should also be aware of some issues in using job ads data, particularly that they under-represent certain sectors and occupations, for instance, the construction sector and occupations related to the production and transportation of goods. In addition, these job ads might not capture jobs from the informal sector, which represents a significant share of the workforce in developing and emerging economies. Future Research Directions As discussed in section 3, there is little evidence of the underlying mechanisms explaining the slow polarization pace in low- and middle-income countries. Geo- graphical, sectoral, and firm heterogeneities have largely been overlooked, as most studies have focused on aggregate measures. In many cases, the lack of research stems from inappropriate information. In this context, firm-level details on the adop- tion of more advanced technologies and longitudinal measures of tasks as described above will enable a significant leap in the literature. In addition, tracking workers’ transitions across occupations and in and out of unemployment could improve public policies and help design or improve a safety net minimizing the harms of technological change. For instance, the literature has not explored the extent to which the declining demand for routine occupations takes place within worker categories or through changes in the composition of workers. If workers can easily transition between routine and non-routine occupations, tech- nological unemployment becomes less of an issue. Public policies can play a crucial role if job polarization occurs through workers’ composition changes. For instance, Cortes et al. (2020) show that most of the decline in routine occupations in the United States is linked to the inflow rates to routine employment (from unemployment and non-participation, i.e., less workers starting a routine job) rather than the outflow rates (more routine workers losing their job). Moreover, Maczulskij (2019) explores Finnish data and shows that most of the relative increase in non-routine occupations compared to mid-level routine occupations is a within-worker phenomenon in the decomposition analysis. In contrast, the share of low-skilled non-routine manual tasks has increased mainly through entry dynamics. Martins-Neto et al. 277 Additionally, we need a more detailed analysis of the effects of labor-displacing automation on workers’ labor prospects, especially in the context of increasing digitalization. One crucial empirical question concerns which types of workers have a more pronounced decline in wages and increase in unemployment duration fol- lowing the event of displacement. Despite the long-term drop in demand for routine tasks, little is known about the short-term impacts of technological change at the individual level, and less so in the context of developing countries. Although most empirical results point to a lack of polarization among those economies, it is still Downloaded from https://academic.oup.com/wbro/article/39/2/259/7246968 by World Bank user on 19 July 2024 unclear whether workers previously employed in routine-intensive occupations are already facing the negative implications of automation. In the context of advanced economies, a number of studies show that automation increases the probability of incumbent workers separating from their employers (Bessen et al. 2019), and that displaced workers in routine-intensive occupations are more likely to face long-term unemployment and a decrease in wages and number of days worked (Bessen et al. 2019; Blien et al. 2021; Goos et al. 2021). However, the literature on developing economies is much thinner. Except for Martins-Neto et al. (2022), who find that displaced individuals in routine-intensive occupations face longer unemployment rates in Brazil, no other study has sought to investigate the implications of routinization at the individual level in the context of developing coun- tries. A detailed account of the effects of displacement on different kinds of workers could help in assessing the differential impacts on employment and income distribu- tion. This in turn will help to categorize more disadvantaged workers, thereby for- mulating specific policies for various categories (including unemployment benefits). Therefore, while the literature on job polarization in developing countries is rel- atively new, the research agenda should concentrate on understanding the factors behind the slow pace of job polarization and examining the heterogeneities of this process, especially those related to firm-level differences in technology adoption and the adverse impacts at the worker level. As discussed in this section, researchers could expand our understanding of the many heterogeneities surrounding labor market trends in emerging economies while exploring matched employer-employee databases.15 Two other dimensions that require further research are the roles of the type of technology and the way technology adoption affects firms’ internal organization. First, employment dynamics depend on the nature of technologies, i.e., which skills they complement or substitute. This has been the focus of several recent works highlighting the new patterns linking digital technologies and demand for skills in advanced economies (Frey and Osborne 2017; Acemoglu and Restrepo 2020). How- ever, such a relationship is also mediated by firms’ own organizational routines and adaptations, which affect how technologies remodel production and workers’ tasks within firms (Dosi and Nelson 2010; Dosi and Virgillito 2019; Ciarli et al. 2021). Firms intentionally invest in organizational arrangements, practices, and routines to 278 The World Bank Research Observer, vol. 39, no. 2 (2024) create new business models in response to the changing and increasingly complex technological landscape (Colfer and Baldwin 2016). The employment effects of technology in developing and emerging economies could therefore be significantly related to the complex interplay between technologies, innovation, and skills driven by organizational restructuring, highlighting the need for urgent attention and more research in this area. Conclusions Downloaded from https://academic.oup.com/wbro/article/39/2/259/7246968 by World Bank user on 19 July 2024 While studying the impact of technological change on jobs and how it affects economies and societies, one must recognize the existing differences among coun- tries that emerge from different socioeconomic systems, levels and distributions of income, institutional contexts, and industrial structures. The nature and long-term impact of technologies created and adopted in different economies very much relate to existing institutional and political contexts. In this review, we have highlighted the impacts of technology adoption on the labor market, focusing on the extent of job polarization in developing and emerging economies. The evidence synthesis suggests that, in advanced economies, the rapid spread of ICTs and robots has resulted in increasing inequality and the “hollowing out” of the occupational distribution, with a significant decline in the demand for routine occupations (Spitz-Oener 2006; Acemoglu and Autor 2011). Yet, in economies at lower levels of income per capita, the pace is considerably slower, with little evidence of labor market polarization or labor-displacing automation (Maloney and Molina 2019; Firpo et al. 2021; Gasparini et al. 2021). In section 3, we explored the possible mechanisms slowing job polarization in developing economies, suggesting the critical role of firms’ and workers’ capabilities in slowing technology adoption and the off-shoring of routine-intensive jobs from advanced economies to some host developing countries. Other moderating aspects in- clude lower wages and different economic structures in emerging economies. We also highlighted the need for more research on the moderating sources, especially those associated with differences in the relative cost of inputs (lower wages in developing countries) and the role of MNEs in slowing or accelerating job polarization. Finally, in section 4, we have stressed the need for micro-level studies and explo- ration of the different mechanisms preventing job polarization in those economies. These studies would enhance our understanding of the main barriers to technology adoption and the adverse effects at the worker level, thus allowing for the devel- opment and implementation of better-adapted policies fitted to developing and emerging economies’ specific contexts. Appendix A Martins-Neto et al. 279 280 Table A1. Comparing the Different Measures of Tasks O*NET STEP PIAAC Countries United States Albania, Armenia, Azerbaijan, Bolivia, Bosnia & Australia, Austria, Belgium (Flanders), Canada, Chile, Herzegovina, Colombia, Georgia, Ghana, Kenya, Czech Republic, Denmark, Ecuador, Estonia, Kosovo, Lao PDR, Macedonia, Serbia, Sri Lanka, Finland, France, Germany, Greece, Hungary, Ukraine, Vietnam, and the Yunnan Province in Indonesia, Ireland, Israel, Italy, Japan, Kazakhstan, China. The third wave of the China Urban Labor Korea, Latvia, Lithuania, Mexico, Netherlands, Survey (CULS) includes a section based on the New Zealand, Norway, Peru, Poland, Russian STEP survey. It includes information on Federation, Singapore, Slovak Republic, Slovenia, Guangzhou, Shanghai, and Fuzhou on the coast, Spain, Sweden, Turkey, United Kingdom (England Shenyang in the northeast, Xian in the northwest, and Northern Ireland), United States and Wuhan in central China Measure Composite measures of O*NET work activities and Workers are asked about specific tasks. STEP Workers are asked about specific tasks. Often, the work context importance scales. For each questions typically refer to whether workers PIAAC questions refer to the frequency of occupation, experts assign a score—between 1 perform a specific task as part of their job or not performing a task (five categories ranging from and 5 to the 44 existing tasks “never” to “every day”) Caveats *Assumption that the task content of occupations *Only covers urban areas *Does not cover large developing economies, is similar across countries and constant over * Does not cover large developing economies, including, for instance, Argentina, Brazil, time including, for instance, Argentina, Brazil, Bangladesh, India, Nigeria, and South Africa * Includes “numerous potential task scales, and it Bangladesh, India, Nigeria, and South Africa *The mapping between tasks and skills is not trivial is rarely obvious which measure (if any) best *The mapping between tasks and skills is not trivial * Subject bias in workers’ response, especially given represents a given task construct” (Acemoglu *Subject bias in workers’ response, especially given that most questions are subjective and Autor 2011, p.1078) that most questions are subjective * Sample size is not large enough to develop * No variation in the task scores within *Sample size is not large enough to develop disaggregated classifications at the country level occupations disaggregated classifications at the country level Advantages * Offers task content of occupations at * Variation in the task scores within occupations * Variation in the task scores within occupations disaggregated levels and with easily-available * Estimation for a number of developing countries, * Estimation for a number of developing countries crosswalks to most classifications including low-income economies Source: Authors’ elaboration. Note: STEP and PIAAC also present differences in the way the data is collected and in the way the proficiency of respondents is estimated (see Keslair and Paccagnella 2020). The World Bank Research Observer, vol. 39, no. 2 (2024) Downloaded from https://academic.oup.com/wbro/article/39/2/259/7246968 by World Bank user on 19 July 2024 Notes Antonio Martins-Neto (corresponding author): UNU-MERIT and Maastricht University, The Netherlands, martinsneto@merit.unu.edu; Nanditha Mathew: UNU-MERIT, The Netherlands, mathew@merit.unu.edu; Pierre Mohnen: UNU-MERIT and Maastricht University, The Nether- lands, p.mohnen@maastrichtuniversity.nl; Tania Treibich: Maastricht University, The Nether- lands, OFCE, France and Institute of Economics and EMbeDS, Scuola Superiore Sant’Anna, Italy, t.treibich@maastrichtuniversity.nl. Downloaded from https://academic.oup.com/wbro/article/39/2/259/7246968 by World Bank user on 19 July 2024 This research is funded by the Comprehensive Innovation for Sustainable Development (CI4SD) Re- search Programme of the United Nations University—Maastricht Economic and Social Research In- stitute on Innovation and Technology (UNU-MERIT), Maastricht, the Netherlands. Antonio Martins- Neto, Pierre Mohnen and Tania Treibich acknowledge the support by the European Union’s Horizon 2020 Research and Innovation Program under grant agreement No. 822781—GROWINPRO. Nanditha Mathew and Tania Treibich acknowledge the support by the European Union’s Horizon 2020 Research and Innovation Program under grant agreement No. 101004703—PILLARS. 1. Chin et al. (2006) show that, in addition to skill-replacing dynamics, steam power also had some elements that were skill-biased, causing a rise in the demand for engineers. Nevertheless, as pointed out by O’Rourke et al. (2013), novel technologies were, on average, skill-saving in the early nineteenth century. 2. Job polarization and the decline in the middle-class could also have important political implica- tions. For instance, Birdsall (2010) suggests that the middle class is an “indispensable” force to achieve more sensible economic policy, robust and responsive political institutions, and thus more sustained growth. 3. We restrict our analysis to the impacts of digital technologies and automation (robots) on the labor market. Automation refers to computer-assisted machines, robotics, and artificial intelligence, such that robots are a subset of automation. Recent developments in artificial intelligence (AI) make it likely that they will replace more tasks in production, with estimations suggesting that high-paying, non-routine occupations are at particular risk of displacement (Webb 2019). Yet, due to the short evaluation time, we do not discuss the possible implications of the more recent and advanced technologies such as AI and the internet of things (IoT). 4. Despite the growing discussion around job polarization in developing economies, one of our re- search’s main challenges was the initial search for articles on the topic. When searching on the Web of Science and on Scopus using different keywords related to job polarization and developing economies, we identified only a few articles, among which only some were actually about developing and emerging economies. To overcome this challenge, we have extensively relied on citations and Google Scholar to find additional working papers, articles, and reports, which has resulted in the identification of about 20 articles focusing on job polarization in developing and emerging economies. 5. The considerable presence of informal firms in low-income countries relates to countries’ capa- bilities and is due to inadequate access to education but also corruption, regulation, and the lack of proactive policies to embrace the informal economy (Etim and Daramola 2020). 6. A similar explanation relates to the wage structure. The decline in the demand for routine- intensive occupations only leads to job polarization if these occupations are in the middle of the wage distribution and if the wage distribution reflects the skills structure. Nevertheless, routine occupations in emerging economies could be ranked differently, given the sizeable informal sector and wage-setting institutions. For example, using data from 10 OECD countries, Haslberger (2021) documents that RBTC can lead to occupational upgrading rather than polarization, as countries differ in terms of the occupa- tional routine-wage hierarchies. In other words, given that in many developing countries, the number of workers engaged in codified tasks is small and, in some cases, concentrated in low-wage occupations, routinization could lead to occupational upgrading. Martins-Neto et al. 281 7. Besides differences across sectors, firms of the same industry also present considerable hetero- geneity in their employment and wage structures (see Helpman et al. 2017 for Brazil and Harrigan et al. 2021; Domini et al. 2022 in the case of France). In the context of developing countries, it could be the case that there is a polarization process within firms, but it is compensated by the fact that large and growing firms are more intensive in middle-earning occupations. Therefore, reallocation dynamics (i.e., changes in the market shares of firms within sectors) among firms with different occupational structures may explain why occupational shares at the aggregate level do not change. 8. Although some evidence suggests that automation in advanced economies is yet to impact FDI flows (Hallward-Driemeier and Nayyar 2019). 9. In addition to changes in world trade, the COVID-19 pandemic may also have had an impact on Downloaded from https://academic.oup.com/wbro/article/39/2/259/7246968 by World Bank user on 19 July 2024 the pace of digital adoption in developing economies. While initial evidence suggests that the pandemic has accelerated the digital transformation of businesses, it also indicates widening the digital divide (Avalos Almanza et al. 2023). 10. The O*NET database covers nearly 1,000 occupations in the United States and provides occu- pational level task indexes estimated by experts, who rank occupations based on workers’ interviews. Autor et al. (2003) selected a number of relevant variables for each of the five conceptual categories: non-routine analytic tasks; non-routine interactive tasks; routine cognitive tasks; routine manual tasks; and non-routine manual tasks. For instance, in measuring routine manual activity, the authors use the variable FINGDEX, an abbreviation of Finger Dextery. 11. The literature on developed economies has also explored the survey of employees carried out by the German Federal Institute for Vocational Training (Bundesinstitut f ür Berufsbildung; BIBB) and the Research Institute of the Federal Employment Service (Institut f ür Arbeitsmarkt und Berufsforschung; IAB) (see, for instance, Spitz-Oener 2006, for additional details). However, the database only includes bi- nary information on whether the worker either performs a specific task or not, and aggregate measures are based on the share of each category of tasks (abstract, routine and manual). In our review in sec- tion 2, authors have opted for using the O*NET database when studying job polarization in developing economies. 12. For example, World Bank (2016) and Maloney and Molina (2019) follow Autor and Dorn’s (2013) classification and define nine groups of occupations coded according to the major categories in the International Standard Classification of Occupations (ISCO) to study job polarization (see also Aedo et al. 2013 and Arias et al. 2014). 13. The use of direct worker-level information on the specific tasks performed on the job was pio- neered by Handel (2008), who developed the STAMP survey. 14. Lewandowski et al. (2020) also present different task measures based on STEP and PIAAC data from other authors. 15. Such data is available in several developing economies, including Brazil, Mexico, South Africa, Morocco, and Tunisia. References Acemoglu, D. 1999. “Changes in Unemployment and Wage Inequality: An Alternative Theory and Some Evidence.” American Economic Review 89 (5): 1259–78. Acemoglu, D., and D. Autor. 2011. “Skills, Tasks and Technologies: Implications for Employment and Earnings.” In Handbook of Labor Economics, Volume 4 of Handbook of Labor Economics, edited by O. Ashenfelter and D. Card, Chapter 12, 1043–171. Elsevier, Amsterdam, The Netherlands. Acemoglu, D., and P. Restrepo. 2020. “The Wrong Kind of Ai? Artificial Intelligence and the Future of Labour Demand.” Cambridge Journal of Regions, Economy and Society 13 (1): 25–35. https://academic. oup.com/cjres/article-abstract/13/1/25/5680462?redirectedFrom=fulltext. ———. 2021. “Demographics and Automation.” Review of Economic Studies 89 (1): 1–44. 282 The World Bank Research Observer, vol. 39, no. 2 (2024) Aedo, C., J. Hentschel, J. Luque, and M. Moreno. 2013. “From Occupations to Embedded Skills: A Cross- country Comparison.” Policy Research Working Paper No. 6560, World Bank , Washington, DC. Alvaredo, F., L. Chancel, T. Piketty, E. Saez, and G. Zucman. 2018. The World Inequality Report. Harvard University Press, Cambridge, USA. Amoroso, S., and P. Moncada-Paterno-Castello. 2018. “Inward greenfield FDI and Patterns of Job Polar- ization.” Sustainability 10 (4): 1219. Arias, O. S., C. Sanchez-Paramo, M. E. Davalos, I. Santos, E. R. Tiongson, C. Gruen, N. de Andrade Falcao, G. Saiovici, and C. A. Cancho. 2014. Back to Work: Growing with Jobs in Europe and Central Asia. The World Bank. Downloaded from https://academic.oup.com/wbro/article/39/2/259/7246968 by World Bank user on 19 July 2024 Artuc, E., L. Christiaensen, and H. J. Winkler. 2019. “Does Automation in Rich Countries Hurt Devel- oping Ones?: Evidence from the U.S. and Mexico.” Policy Research Working Paper No 8741, World Bank, Washington, DC. Atalay, E., P. Phongthiengtham, S. Sotelo, and D. Tannenbaum. 2020. “The Evolution of Work in the United States.” American Economic Journal: Applied Economics 12 (2): 1–34. Autor, D., and M. Handel. 2013. “Putting Tasks to the Test: Human Capital, Job Tasks, and Wages.” Journal of Labor Economics 31 (S1): S59–96. Autor, D. H., and D. Dorn. 2013. “The Growth of Low-Skill Service Jobs and the Polarization of the US Labor Market.” American Economic Review 103 (5): 1553–97. Autor, D. H., F. Levy, and R. J. Murnane. 2003. “The Skill Content of Recent Technological Change: An Empirical Exploration.” Quarterly Journal of Economics 118 (4): 1279–333. Avalos Almanza, E. I., X. Cirera, M. J. Vargas Da Cruz, L. Iacovone, D. Medvedev, G. Nayyar, and S. Reyes Ortega. 2023. “Firms’ Digitalization during the COVID-19 Pandemic: A Tale of Two Stories.” Policy Research Working Paper No. 10284, World Bank, Washington, DC. Banerjee, A. V., and E. Duflo. 2005. Growth Theory through the Lens of Development Economics. Volume 1 of Handbook of Economic Growth. Chapter 7, 473–552. Elsevier, Amsterdam, The Netherlands. Barany, Z. L., and C. Siegel. 2018. “Job Polarization and Structural Change.” American Economic Journal: Macroeconomics 10 (1): 57–89. Benhabib, J., and M. M. Spiegel. 1994. “The Role of Human Capital in Economic Development Evidence from Aggregate Cross-Country Data.” Journal of Monetary Economics 34 (2): 143–73. Bessen, J. E., M. Goos, A. Salomons, and W. Van den Berge. 2019. “Automatic Reaction: What Happens to Workers at Firms That Automate?” Law and Economics Research Paper, Boston University School of Law. Beylis, G., R. Fattal-Jaef , M. Morris, A. R. Sebastian, and R. Sinha. 2020. “Going Viral: COVID-19 and the Accelerated Transformation of Jobs in Latin America and the Caribbean.” World Bank Latin American and Caribbean Studies. World Bank, Washington, DC. doi:10.1596/978-1-4648-1448-8. Birdsall, N. 2010. “The (Indispensable) Middle Class in Developing Countries; or, The Rich and the Rest, Not the Poor and the Rest.” Working Papers 207, Center for Global Development. Blien, U., W. Dauth, and D. H. Roth. 2021. “Occupational Routine Intensity and the Costs of Job Loss: Evidence from Mass Layoffs.” Labour Economics 68, 101953. Blinder, A. S., and A. B. Krueger. 2013. “Alternative Measures of Offshorability: A Survey Approach.” Journal of Labor Economics 31 (S1): 97–128. Bloom, N., and J. Van Reenen. 2010. “Why Do Management Practices Differ across Firms and Coun- tries?” Journal of Economic Perspectives 24 (1): 203–24. Bonnet, F., J. Vanek, and M. Chen. 2019. “Women and Men in the Informal Economy: A Statistical Brief .” International Labour Office, Geneva. http://www.wiego. org/sites/default/files/publications/files/ Women% 20and% 20Men% 20in% 20the% 20Informal 20. Martins-Neto et al. 283 Boothby, D., A. Dufour, and J. Tang. 2010. “Technology Adoption, Training and Productivity Perfor- mance.” Research Policy 39 (5): 650–61. Buyst, E., M. Goos, and A. Salomons. 2018. “Job Polarization: An Historical Perspective.” Oxford Review of Economic Policy 34 (3): 461–74. Card, D., and J. E. DiNardo. 2002. “Skill-biased Technological Change and Rising Wage Inequality: Some Problems and Puzzles.” Journal of Labor Economics 20 (4): 733–83. Case, A., and A. Deaton. 2020. Deaths of Despair and the Future of Capitalism. Princeton University Press, Princeton, New Jersey, USA. Cassidy, H. 2017. “Task Variation within Occupations.” Industrial Relations: A Journal of Economy and Downloaded from https://academic.oup.com/wbro/article/39/2/259/7246968 by World Bank user on 19 July 2024 Society 56 (3): 393–410. Caunedo, J., E. Keller, and Y. Shin. 2021. “Technology and the Task Content of Jobs across the Development Spectrum.” NBER Working Paper No. 28681, National Bureau of Economic Research, Cambridge, MA. Chacaltana Janampa, J., V. Leung, and M. Lee. 2018. “New Technologies and the Transition to Formality the Trend towards E–formality.” ILO Working Paper No. 994998792502676, International Labour Organization, Geneva, Switzerland. Chin, A., C. Juhn, and P. Thompson. 2006. “Technical Change and the Demand for Skills during the Sec- ond Industrial Revolution: Evidence from the Merchant Marine, 1891–1912.” Review of Economics and Statistics 88 (3): 572–8. Ciarli, T., M. Kenney, S. Massini, and L. Piscitello. 2021. “Digital Technologies, Innovation, and Skills: Emerging Trajectories and Challenges.” Research Policy 50 (7): 104289. Cirera, X., D. A. Comin, M. J. Vargas Da Cruz, and K. Lee. 2021. “Firm-Level Adoption of Technologies in Senegal.” Policy Research Working Paper No. 9657, World Bank, Washington, DC. Cirera, X., D. A. Comin, M. J. Vargas Da Cruz, K. Lee, and A. Soares Martins Neto. 2021a. “Firm-Level Technology Adoption in the State of Ceara in Brazil.” Policy Research Working Paper No. 9568, World Bank, Washington, DC, USA. ———. 2021b. “Firm-Level Technology Adoption in Vietnam.” Policy Research Working Paper No. 9567, World Bank, Washington, DC. Cirera, X., and W. F. Maloney. 2017. The Innovation Paradox: Developing-country Capabilities and the Unre- alized Promise of Technological Catch-up. World Bank Publications. Cohen, W. M., and D. A. Levinthal. 1990. “Absorptive Capacity: A New Perspective on Learning and Innovation.” Administrative Science Quarterly 35 (1): 128–52. Colfer, L. J., and C. Y. Baldwin. 2016. “The Mirroring Hypothesis: Theory, Evidence, and Exceptions.” Industrial and Corporate Change 25 (5): 709–38. Comin, D., and B. Hobijn. 2004. “Cross-country Technology Adoption: Making the Theories Face the Facts.” Journal of Monetary Economics 51 (1): 39–83. Cortes, G. M., N. Jaimovich, C. J. Nekarda, and H. E. Siu. 2020. “The Dynamics of Disappearing Routine Jobs: A Flows Approach.” Labour Economics 65, 101823. Das, M., and B. Hilgenstock. 2022. “The Exposure to Routinization: Labor Market Implications for De- veloped and Developing Economies.” Structural Change and Economic Dynamics 60: 99–113. Delera, M., C. Pietrobelli, E. Calza, and A. Lavopa. 2022. “Does Value Chain Participation Facilitate the Adoption of Industry 4.0 Technologies in Developing Countries? World Development 152, 105788. Dicarlo, E., S. Lo Bello, S. Monroy-Taborda, A. M. Oviedo, M. L. Sanchez Puerta, and I. Santos. 2016. “The Skill Content of Occupations across Low and Middle Income Countries: Evidence from Harmonized Data.” IZA Discussion Paper No. 10224, Institute of Labor Economics (IZA). Domini, G., M. Grazzi, D. Moschella, and T. Treibich. 2022. “For Whom the Bell Tolls: The Firm-level Effects of Automation on Wage and Gender Inequality.” Research Policy 51 (7): 104533. 284 The World Bank Research Observer, vol. 39, no. 2 (2024) Dosi, G. 1982. “Technological Paradigms and Technological Trajectories: A Suggested Interpretation of the Determinants and Directions of Technical Change.” Research Policy 11 (3): 147–62. Dosi, G., and R. R. Nelson. 2010. “Technical Change and Industrial Dynamics as Evolutionary Pro- cesses.” Handbook of the Economics of Innovation 1: 51–127. Dosi, G., and M. E. Virgillito. 2019. “Whither the Evolution of the Contemporary Social Fabric? New Technologies and Old Socio-economic Trends.” International Labour Review 158 (4): 593–625. Dustmann, C., J. Ludsteck, and U. Schonberg. 2009. “Revisiting the German Wage Structure.” Quarterly Journal of Economics 124 (2): 843–81. Etim, E., and O. Daramola. 2020. “The Informal Sector and Economic Growth of South Africa and Nige- Downloaded from https://academic.oup.com/wbro/article/39/2/259/7246968 by World Bank user on 19 July 2024 ria: A Comparative Systematic Review.” Journal of Open Innovation: Technology, Market, and Complexity 6 (4): 134. Faber, M. 2020. “Robots and Reshoring: Evidence from Mexican Labor Markets.” Journal of International Economics 127, 103384. Fagerberg, J. 1994. “Technology and International Differences in Growth Rates.” Journal of Economic Literature 32 (3): 1147–75. Falco, P., W. Maloney, B. Rijkers, and M. Sarrias. 2015. “Heterogeneity in Subjective Well-being: An Application to Occupational Allocation in Africa.” Journal of Economic Behavior & Organization 111 (C): 137–53. Firpo, S., A. Portella, F. Riva, and G. Úbida. 2021. “The Changing Nature of Work and Inequality in Brazil (2003-19): A Descriptive Analysis WIDER Working Paper Series Wp-2021-162.” World Institute for Development Economic Research (UNU-WIDER), Helsinki, Finland. Fleisher, B., W. McGuire, Y. Su, and M. Q. Zhao. 2018. “Innovation, Wages, and Polarization in China.” MPRA Paper No. 87105, University Library of Munich, Germany. Fonseca, T., F. Lima, and S. C. Pereira. 2018. “Job Polarization, Technological Change and Routinization: Evidence for Portugal.” Labour Economics 51: 317–39. Foster-McGregor, N., O. Nomaler, and B. Verspagen. 2021. “Job Automation Risk, Economic Structure and Trade: A European Perspective.” Research Policy 50 (7): 104269. Frey, C. B., and M. A. Osborne. 2017. “The Future of Employment: How Susceptible Are Jobs to Com- puterisation?” Technological Forecasting and Social Change 114: 254–80. Garcia-Couto, S. 2020. “Harmonizing Task Intensities across Time.” Unpublished. https://www. santiagogarciacouto.com/research Gasparini, L., I. Brambilla, G. Falcone, C. Lombardo, and A. César. 2021. “Routinization and Employment: Evidence for Latin America.” CEDLAS Working Paper No. 0276, CEDLAS, Universidad Nacional de La Plata. Ge, P., W. Sun, and Z. Zhao. 2021. “Employment Structures in China from 1990 to 2015.” Journal of Economic Behavior & Organization 185: 168–90. Gomez, M. 2021. “Job Polarization and the Informal Labor Market.” Documentos de trabajo – Alianza EFI 019418, Alianza EFI. Goos, M., and A. Manning. 2007. “Lousy and Lovely Jobs: The Rising Polarization of Work in Britain.” Review of Economics and Statistics 89 (1): 118–33. Goos, M., A. Manning, and A. Salomons. 2009. “Job Polarization in Europe.” American Economic Review 99 (2): 58–63. ———. 2014. “Explaining Job Polarization: Routine-biased Technological Change and Offshoring.” American Economic Review 104 (8): 2509–26. Goos, M., E. Rademakers, and R. Röttgen. 2021. “Routine-biased Technical Change: Individual-level Evidence from a Plant Closure.” Research Policy 50 (7): 104002. Martins-Neto et al. 285 Graetz, G., and G. Michaels. 2018. “Robots at Work.” Review of Economics and Statistics 100 (5): 753–68. Hakkala, K. N., F. Heyman, and F. Sjöholm. 2014. “Multinational Firms, Acquisitions and Job Tasks.” European Economic Review 66 (C): 248–65. Hallward-Driemeier, M. C., and G. Nayyar. 2019. “Have Robots Grounded the Flying Geese?: Evidence from Greenfield FDI in Manufacturing.” Policy Research Working Paper No. 9097, World Bank, Washington, DC. Handel, M. J. 2008. Measuring Job Content: Skills, Technology, and Management. University of Wisconsin- Madison, Institute for Research on Poverty. Downloaded from https://academic.oup.com/wbro/article/39/2/259/7246968 by World Bank user on 19 July 2024 Hardy, W., R. Keister, and P. Lewandowski. 2016. “Technology or Upskilling? Trends in the Task Com- position of Jobs in Central and Eastern Europe.” HKUST IEMS Working Paper No. 2016-40, HKUST Institute for Emerging Market Studies. Harrigan, J., A. Reshef , and F. Toubal. 2021. “The March of the Techies: Job Polarization within and between Firms.” Research Policy 50 (7): 104008. Haslberger, M. 2021. “Routine-biased Technological Change Does Not Always Lead to Polarisation: Evidence from 10 OECD Countries, 1995–2013.” Research in Social Stratification and Mobility 74: 100623. Helmy, O. 2015. “Skill Demand Polarization in Egypt.” Middle East Development Journal 7 (1): 26–48. Helpman, E., O. Itskhoki, M.-A. Muendler, and S. J. Redding. 2017. “Trade and Inequality: From Theory to Estimation.” Review of Economic Studies 84 (1): 357–405. Helpman, E., M. J. Melitz, and S. R. Yeaple. 2004. “Export versus FDI with Heterogeneous Firms.” Amer- ican Economic Review 94 (1): 300–16. Henning, M., and R. H. Eriksson. 2020. “Labour Market Polarisation as a Localised Process: Evidence from Sweden.” Cambridge Journal of Regions, Economy and Society 14 (1): 69–91. Hijzen, A., P. Martins, T. Schank, and R. Upward. 2013. “Foreign-owned Firms around the World: A Comparative Analysis of Wages and Employment at the Micro-level.” European Economic Review 60 (C): 170–88. Hsieh, C.-T., and P. J. Klenow. 2007. “Relative Prices and Relative Prosperity.” American Economic Review 97 (3): 562–85. Ikenaga, T., and R. Kambayashi. 2016. “Task Polarization in the Japanese Labor Market: Evidence of a Long-term Trend.” Industrial Relations: A Journal of Economy and Society 55 (2): 267–93. Jensen, R. 1988. “Information Cost and Innovation Adoption Policies.” Management Science 34 (2): 230– 9. Katz, L. F., and K. M. Murphy. 1992. “Changes in Relative Wages, 1963–1987: Supply and Demand Factors.” Quarterly Journal of Economics 107 (1): 35–78. Katz, L. F., and L. H. Summers. 1989. “Industry Rents: Evidence and Implications.” Brookings Papers on Economic Activity 20: 209–90. Keslair, F., and M. Paccagnella. 2020. “Assessing Adults’ Skills on a Global Scale: A Joint Analysis of Results from PIAAC and STEP.” OECD Education Working Paper No. 230, OECD Publishing. Kinkel, S., A. Jager, and C. Zanker. 2015. “The Effects of Robot Use in European Manufacturing Compa- nies on Production Offshoring outside the EU.” International Annual EurOMA Conference, Neuchâtel, Switzerland. Krenz, A., K. Prettner, and H. Strulik. 2021. “Robots, Reshoring, and the Lot of Low-skilled Workers.” European Economic Review 136, 103744. Kugler, A. D., M. Kugler, L. Ripani, and R. Rodrigo. 2020. “U.S. Robots and Their Impacts in the Trop- ics: Evidence from Colombian Labor Markets.” NBER Working Paper No. 28034, National Bureau of Economic Research, Cambridge, MA. 286 The World Bank Research Observer, vol. 39, no. 2 (2024) La Porta, R., and A. Shleifer. 2014. “Informality and Development.” Journal of Economic Perspectives 28 (3): 109–26. Lee, S. Y. T., and Y. Shin. 2019. “Horizontal and Vertical Polarization: Task-Specific Technological Change in a Multi-Sector Economy.” Working Paper No. 888, Queen Mary University of London, School of Economics and Finance. Levy, F., and R. J. Murnane. 1992. “U.S. Earnings Levels and Earnings Inequality: A Review of Recent Trends and Proposed Explanations.” Journal of Economic Literature 30 (3): 1333–81. Lewandowski, P., R. Keister, W. Hardy, and S. Gorka. 2017. “Routine and Ageing? The Intergenerational Divide in the Deroutinisation of Jobs in Europe. IZA Discussion Paper No. 10732, IZA Institute of Downloaded from https://academic.oup.com/wbro/article/39/2/259/7246968 by World Bank user on 19 July 2024 Labor Economics. Lewandowski, P., A. Park, W. Hardy, and D. Yang. 2019. “Technology, Skills, and Globalization: Explain- ing International Differences in Routine and Nonroutine Work Using Survey Data.” IBS Working Pa- per No. 04/2019, Instytut Badan Strukturalnych. Lewandowski, P., A. Park, and S. Schotte. 2020. “The Global Distribution of Routine and Non-routine Work.” IBS Working Paper No. 06/2020, Instytut Badan Strukturalnych. Lo Bello, S., M. L. Sanchez Puerta, and H. J. Winkler. 2019. “From Ghana to America: The Skill Con- tent of Jobs and Economic Development.” Policy Research Working Paper No. 8758, World Bank, Washington, DC. Lordan, G., and D. Neumark. 2018. “People versus Machines: The Impact of Minimum Wages on Au- tomatable Jobs.” Labour Economics 52: 40–53. Machin, S., and J. Van Reenen. 1998. “Technology and Changes in Skill Structure: Evidence from Seven OECD Countries.” Quarterly Journal of Economics 113 (4): 1215–44. Maczulskij, T. 2019. “Occupational Mobility of Routine Workers.” Working Paper No. 327, Palkansaajien tutkimuslaitos, Labour Institute for Economic Research. Maloney, W. F., and C. Molina. 2019. “Is Automation Labor-Displacing in the Developing Countries, Too? Robots, Polarization, and Jobs.” World Bank, Washington, DC, USA. Martins-Neto, A., X. Cirera, and A. Coad. 2022. “Routine-biased Technological Change and Employee Outcomes after Mass Layoffs: Evidence from Brazil.” MERIT Working Paper No. 2022-014, United Nations University – Maastricht Economic and Social Research Institute on Innovation and Technol- ogy (MERIT). Messina, J., G. Pica, and A. M. Oviedo. 2016. “Job Polarization in Latin America.” Inter-American De- velopment Bank, Washington, DC. Unpublished. Available at http://www.jsmessina.com Michaels, G., A. Natraj, and J. V. Reenen. 2014. “Has ICT Polarized Skill Demand? Evidence from Eleven Countries over Twenty-Five Years.” Review of Economics and Statistics 96 (1): 60–77. Montresor, G. 2019. “Job Polarization and Labour Supply Changes in the UK.” Labour Economics 58: 187–203. Moreno-Galbis, E., and T. Sopraseuth. 2014. “Job Polarization in Aging Economies.” Labour Economics 27: 44–55. Nchor, D., and P. Rozmahel. 2020. “Job Polarization in Europe: Evidence from Central and Eastern Eu- ropean Countries.” DANUBE 11 (1): 52–74. Nicoletti, G., C. von Rueden, and D. Andrews. 2020. “Digital Technology Diffusion: A Matter of Capabil- ities, Incentives or Both?” European Economic Review 128, 103513. Oetzel, J., and J. P. Doh. 2009. “MNEs and Development: A Review and Reconceptualization.” Journal of World Business 44 (2): 108–20. Olsson, M., and J. Tåg. 2017. “Private Equity, Layoffs, and Job Polarization.” Journal of Labor Economics 35 (3): 697–754. Martins-Neto et al. 287 O’Rourke, K., A. Rahman, and A. Taylor. 2013. “Luddites, the Industrial Revolution, and the Demo- graphic Transition.” Journal of Economic Growth 18 (4): 373–409. Quality Assurance Agency. 2015. “Hospitality, Leisure, Sport and Tourism.” Technical report, Subject Benchmark Statement, Gloucester: QAA. Reijnders, L. S., and G. J. de Vries. 2018. “Technology, Offshoring and the Rise of Non-routine Jobs.” Journal of Development Economics 135: 412–32. Rosenberg, N. 1972. “Factors Affecting the Diffusion of Technology.” Explorations in Economic History 10 (1): 3–33. Saltiel, F. 2019. “Comparative Evidence on the Returns of Tasks in Developing Countries.” Technical Downloaded from https://academic.oup.com/wbro/article/39/2/259/7246968 by World Bank user on 19 July 2024 report, Mimeo. https://fersaltiel.github.io/tasks_mostrecent.pdf. Salvatori, A. 2018. “The Anatomy of Job Polarisation in the UK.” Journal for Labour Market Research 52 (1): 1–15. Sarkar, S. 2019. “Employment Change in Occupations in Urban India: Implications for Wage Inequal- ity.” Development and Change 50 (5): 1398–429. Shim, M., and H.-S. Yang. 2018. “Interindustry Wage Differentials, Technology Adoption, and Job Po- larization.” Journal of Economic Behavior & Organization 146: 141–60. Spitz-Oener, A. 2006. “Technical Change, Job Tasks, and Rising Educational Demands: Looking outside the Wage Structure.” Journal of Labor Economics 24 (2): 235–70. Stinebrickner, R., T. Stinebrickner, and P. Sullivan. 2019. “Job Tasks, Time Allocation, and Wages.” Journal of Labor Economics 37 (2): 399–433. Teece, D. 1977. “Technology Transfer by Multinational Firms: The Resource Cost of Transferring Tech- nological Know-how.” Economic Journal 87 (346): 242–61. Teece, D. J., G. Pisano, and A. Shuen. 1997. “Dynamic Capabilities and Strategic Management.” Strategic Management Journal 18 (7): 509–33. Webb, M. 2019. “The Impact of Artificial Intelligence on the Labor Market.” http://dx.doi.org/10.2139/ ssrn.3482150 World Bank. 2016. World Development Report 2016: Digital Dividends. Washington, DC: World Bank. Zago, R. 2020. “Job Polarization, Skill Mismatch and the Great Recession.” Working Paper No. 755, Banque de France. The World Bank Research Observer © The Author(s) 2023. Published by Oxford University Press on behalf of the International Bank for Reconstruction and Development / THE WORLD BANK. This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs licence (http://creativecommons.org/licenses/by-nc-nd/4.0/), which permits non- commercial reproduction and distribution of the work, in any medium, provided the original work is not altered or trans- formed in any way, and that the work is properly cited. For commercial re-use, please contact journals.permissions@oup.com https://doi.org/10.1093/wbro/lkad008 39:259–288 288 The World Bank Research Observer, vol. 39, no. 2 (2024)