Task Trade and the Wage Effects of Import Competition by Abigail Cooke University at Buffalo, SUNY Thomas Kemeny University of Southampton David L. Rigby University of California, Los Angeles CES 16-03 January, 2016 The research program of the Center for Economic Studies (CES) produces a wide range of economic analyses to improve the statistical programs of the U.S. Census Bureau. Many of these analyses take the form of CES research papers. The papers have not undergone the review accorded Census Bureau publications and no endorsement should be inferred. Any opinions and conclusions expressed herein are those of the author(s) and do not necessarily represent the views of the U.S. Census Bureau. All results have been reviewed to ensure that no confidential information is disclosed. Republication in whole or part must be cleared with the authors. To obtain information about the series, see www.census.gov/ces or contact Fariha Kamal, Editor, Discussion Papers, U.S. Census Bureau, Center for Economic Studies 2K132B, 4600 Silver Hill Road, Washington, DC 20233, [email protected]. To subscribe to the series, please click here. Abstract Do job characteristics modulate the relationship between import competition and the wages of workers who perform those jobs? This paper tests the claim that workers in occupations featuring highly routine tasks will be more vulnerable to low-wage country import competition. Using data from the US Census Bureau, we construct a pooled cross-section (1990, 2000, and 2007) of more than 1.6 million individuals linked to the establishment in which they work. Occupational measures of vulnerability to trade competition – routineness, analytic complexity, and interpersonal interaction on the job – are constructed using O*NET data. The linked employer-employee data allow us to model the effect of low-wage import competition on the wages of workers with different occupational characteristics. Our results show that low-wage country import competition is associated with lower wages for US workers holding jobs that are highly routine and less complex. For workers holding nonroutine and highly complex jobs, increased import competition is associated with higher wages. Finally, workers in occupations with the highest and lowest levels of interpersonal interaction see higher wages, while workers with medium-low levels of interpersonal interaction suffer lower wages with increased low-wage import competition. These findings demonstrate the importance of accounting for occupational characteristics to more fully understand the relationship between trade and wages, and suggest ways in which task trade vulnerable occupations can disadvantage workers even when their jobs remain onshore. Keyword: Task trade; import competition; wages; globalization; occupational characteristics; trade * * This research uses confidential data from the U.S. Census Bureau. Any opinions and conclusions expressed herein are those of the authors and do not represent the views of the U.S. Census Bureau. All results have been reviewed to ensure that no confidential information has been disclosed. Kemeny and Rigby acknowledge support for this project from the National Science Foundation under grant number BCS-0961735. Help by Arnie Reznek and Frank Limehouse at the U.S. Census Bureau on disclosure avoidance review is gratefully acknowledged. 1: Introduction Over the past half century, dramatic improvements in transportation technologies have significantly reduced the cost of moving goods (e.g., Levinson, 2006). Advances in communication technologies have simultaneously reduced the cost of information exchange. These changes have fostered great expansion in trade in services, and have enabled firms to better develop and control value chains distributed across the globe (Bonacich & Wilson, 2008; Coe & Yeung, 2015; Gereffi & Korzeniewicz, 1994; Kleibert, 2015). The rise in the complexity and length of global production networks, related-party trade, and foreign direct investment flows indicate that growing numbers of businesses are exploiting newfound capacities to fragment production across developed and developing economies alike. Unlike traditional spatial divisions of labor based on finished goods such as those Ricardo described, or on monopolistic competition between varieties of goods as analyzed by Krugman (1979), the contemporary reworking of the production and trade landscape involves a much finer-grained division of production possibilities that Baldwin (2006), Blinder (2006) and Grossman and Rossi-Hansberg (2006) tie not to industries but to production tasks. Tasks represent all the incremental steps of the production process necessary to design, test, construct, assemble, sell, and deliver intermediate goods and, eventually, final products and services. More specifically, Grossman and Rossi-Hansberg define tasks as “the finest possible addition to the value added of a good or service done by a particular factor of production” (2012, p. 595), with the production of intermediate goods comprised of “bundles” of tasks. The proliferation of task trade has relied on the ability to integrate and control distributed production networks, involving the efficient and cost-effective coordination of the movement of goods, services, and information across the globe. Task trade is remapping patterns of global trade and 2 specialization. But it may also demand that we reconsider the links between trade and welfare. For instance, trade in tasks may involve shifts in trade distributional impacts. Specifically, education is becoming a less reliable indicator of whether that job is vulnerable to trade (e.g., Baldwin, 2006; Blinder, 2006; Blinder & Krueger, 2013). This article responds to the need for more nuanced ways of capturing how trade influences worker well being. It builds on prior work that considers workers to be distinctive because of the kinds of tasks they perform. This work argues that relatively routine tasks, meaning those that can be readily codified, are increasingly vulnerable to being offshored (Baldwin 2006; Blinder 2006; Leamer and Storper 2001), or replaced by computers, robots and other technology (e.g. Autor, Levy, and Murnane 2003). In contrast, tasks that involve complex judgment-based decisions, and those requiring interpersonal interaction, are more costly to offshore and automate. This paper explores whether the job characteristics of US workers influence how import competition from low-wage countries shapes their wages. We expect that workers will be more vulnerable to low-wage country import competition if the tasks they perform are (a) highly routine, (b) require little analytic complexity, and (c) involve scant interpersonal interaction. To test these hypotheses, we build standard measures of low-wage import competition using annual US Trade data, and relate these to a host of individual and establishment characteristics, derived from the US Census Bureau. We estimate pooled cross-sectional models predicting a worker’s annual wages for 1.6 million workers across the years1990, 2000 and 2007. A worker’s occupation is the basis for their task profile, which we construct using information from the US Department of Labor’s O*NET database. Overall, this dataset allows us to model the effect of low-wage import competition on the wages of workers with different occupational characteristics, net of the effects of education, demographics, and establishment characteristics. In relation to 3 recent work in the same area, we add value in a few ways. Unlike Ebenstein et al (2014), we look beyond MNC offshoring to include arms-length sourcing captured in import records. Unlike Baumgarten et al. (2013), the present paper includes establishment characteristics whose omission may bias the attempt to link outcomes to trade. And we complement work by Hummels et al. (2014) by offering results from a different country context: the US, rather than Denmark. To preview the findings, we show that task intensity mediates the effect of low-wage import competition on workers’ wages. Import competition from low-wage countries is associated with lower wages for workers with highly routine manual jobs and workers with jobs that have low analytic complexity. At the same time, workers in jobs with low routine manual tasks and high analytic complexity earn higher wages when there is greater low-wage import competition. Together, these effects on high- and low- task intensity workers have a polarizing influence on wages along a task-intensity continuum, rewarding workers at one end of the spectrum and penalizing workers at the other end. The trade and wage relationship mediated by interpersonal task intensity is less straightforward. Workers in occupations with high levels of interpersonal interaction have higher wages when there is greater import competition from low- wage countries. Interestingly, the same is true for workers with the lowest levels of interpersonal interaction in their jobs, though the size of the coefficient is much smaller. Only workers with medium-low levels of interpersonal interaction in their occupations suffer lower wages with increased low-wage import competition. Interaction effects show that the mediating relationship of task intensity is non-linear. These results demonstrate the importance of accounting for occupational characteristics to more fully understand the relationship between trade and wages and suggest ways in which task-trade-vulnerable occupations disadvantage workers even when their jobs remain onshore. 4 The remainder of this article proceeds as follows. Section 2 provides a review of the literature on task trade and labor market effects. Section 3 outlines an empirical model to capture the importance of task characteristics in describing the relationship between trade and wages. Data sources, variable construction, and a series of empirical concerns are discussed. Section 4 presents the results from estimating a series of related statistical models. Section 5 concludes, summarizing the key findings. 2: Task trade and labor market effects: A brief review of the literature Trade in intermediate goods has been recognized for some time (for example, Baldwin, 2006; Coe, Hess, Yeung, Dicken, & Henderson, 2004; Dixit & Grossman, 1982; Robert C. Feenstra & Gordon H. Hanson, 1996; Gereffi & Korzeniewicz, 1994; Grossman & Rossi- Hansberg, 2006, 2008; Helpman, 1984). However, the concept of trade in tasks rather than intermediates is not merely old wine in new bottles. Task trade demands substantial retheorization, especially in regards to trade’s welfare impacts. Attention to tasks suggests that who will be affected by trade is more “unpredictable” than previous theories of trade would lead us to believe (e.g., Baldwin, 2006; Baldwin & Robert-Nicoud, 2006). It is no longer low-skill workers or production workers who are vulnerable to international competition, and the jobs that are vulnerable can shift quickly as technological changes enable better and cheaper coordination across long distances. Moreover, the fine grained level of competition in task trade makes clear that the effects will be differentially located within firms and within groups of workers previously thought to share the same fate (Baldwin & Robert-Nicoud, 2006). Thus, recent work suggests that production tasks in high-wage economies requiring substantial formal education or 5 skill development are no longer “safe” from offshoring (Baldwin, 2006; Blinder, 2006), complicating both modeling efforts as well as policy responses. This prompts reconsideration of what characteristics might be more definitive in understanding the contours of trade impacts, some of which rest on what makes something tradable, and thus possible to offshore, or not. Characteristics of tasks that make them vulnerable to offshoring An evolving body of work identifies key characteristics of different tasks that influence the ease with which they might be produced in different locations. Many important issues are being developed in the literature (e.g., see Blinder & Krueger, 2013), including: the precise nature of these characteristics; which are most important; how to measure them; and the number of jobs in high-wage economies that could be affected. Building from extant theoretical and empirical contributions we focus herewith on three key task characteristics: manual routineness, analytical complexity, and interpersonal interaction. Autor, Levy, and Murnane define routine tasks as those requiring “a limited and well- defined set of cognitive and manual activities, those that can be accomplished by following explicit rules” (2003, p. 1280). This task profile is in decline in high-wage economies like the US, UK and Germany (e.g., Autor, Levy, & Murnane, 2003; Goos & Manning, 2007; Spitz-Oener, 2006), and research has suggested it responds negatively to investments in technology, as computers and other forms can efficiently replicate such tasks. At the same time it captures intuitions about shifting patterns of trade, such as the relocation of most garment and footwear production – as an example of routine labor – from higher-wage to low-wage countries. This 6 intuition has been formalized in economic models, such as Grossman & Rossi-Hansberg (2006), and Baldwin and Robert-Nicoud (2014).1 Leamer and Storper (2001) also argue that routineness allows some tasks to be performed far away from the headquarters or management. They argue that routineness is not necessarily a characteristic of individual tasks but rather of the coordination between them. For example, can the tasks be coordinated with codifiable information or do they require more tacit information necessitating trust and understanding between the parties? Newly fractured production processes may technically be performed in any number of places, but coordinating all the parts can be costly enough to keep the task fragments located together. Later, when the new process and the coordination of the set of tasks is routinized and codified, those tasks are more likely to move abroad to cheaper locations. The newness of fragmentation in a particular production process (Leamer and Storper, 2001) is not possible to observe directly in any of the data available to us. However certain aspects of the coordination factor are possible to capture in two further characteristics: analytical complexity (also called nonroutiness in the literature) and interpersonal interaction. Autor et al. define ‘nonroutiness’ in opposition to their characterization of routine tasks as involving “problem-solving and complex communication activities” (2003, p. 1280). However, we prefer Oldenski’s (2011) related concept of ‘complexity’ as involving creativity, problem- solving, and decision-making, because it indicates that there are aspects of this concept that do not follow perfectly along a continuum from routine to nonroutine, but involve additional characteristics. Conceptually helpful, it also corresponds to the possibilities of constructing measures for routineness and complexity that are negatively correlated, but not perfectly so. 1 The model Grossman and Rossi-Hansberg (2006) develop is agnostic as to what actually makes a task vulnerable to offshoring or not (e.g., p. 13), but their discussion of tasks tends towards the routine/nonroutine division. (e.g., p. 10-11). 7 Oldenski offers the likelihood of problems arising that management must solve as determining which activities are likely to be actually offshored. She thus ties in some of the insights from Leamer and Storper that the coordination of the tasks is as important as the tasks themselves in determining what parts of the production process are not only technically footloose but likely to leave. The interpersonal interaction characteristic speaks partly to the Leamer and Storper (2001) tacit coordination factor, but also draws on the personal/impersonal division Blinder (2006) develops with regard to the services sector. Blinder offers this distinction as the key factor in whether a task can be offshored, with impersonal services being those “that can be delivered electronically over long distances with little or no degradation in quality” (p. 114). This concept is also applicable to business-service oriented occupations within the manufacturing sector, such as management jobs. This concept (whether a task is “interactive” or not) is also used by Becker et al. (2013) and Baumgarten et al (2013). Theoretical models Beyond the models commonly used to understand impacts of trade on workers and firms (e.g., Bernard, Jensen, & Schott, 2006; Ethier, 2005; Robert C. Feenstra & Gordon H. Hanson, 1996; Robert C. Feenstra & Gordon H. Hanson, 1996; Feenstra & Hanson, 2001), there are a number of explicit models of task trade.2 The present research draws most heavily upon Grossman and Rossi-Hansberg (2008), who develop a model of task trade in which what is 2 Other models of task trade effects on labor markets include: an extension of the Grossman and Rossi-Hansberg model that investigates the job destruction and creation effects of offshoring by relaxing full-employment conditions (Kohler & Wrona, 2011) and finds that jobs are destroyed as offshoring occurs, but the productivity effect can compensate for the job destruction effect in the long term under certain conditions; an update of the basic Heckscher-Ohlin framework that conceptualizes offshoring as ‘shadow migration’ of endowments, finding that Stolper-Samuelson predictions hold for the home country, implying that in countries like the U.S., inequality in the wages paid to skilled- and unskilled labor should rise with increased offshoring; and offshoring within a monopolistic competition framework (e.g., (Robert-Nicoud, 2008); Grossman and Rossi-Hansberg 2012). 8 traded, or offshored, is determined by weighing the costs of monitoring and controlling workers in another country against the potential savings from lower labor costs in that other country. The costs of coordinating workers from a distance are assumed to be lower for more routine tasks than for nonroutine tasks, and routine tasks are more likely to be performed by low-wage workers and nonroutine tasks by high-wage workers (e.g., Autor et al., 2003). Reductions in trade costs, particularly communications costs, lead to increased offshoring of trade-vulnerable tasks. In this model, the increase in offshoring reduces costs and affects wages in the high- wage (onshore) country in three ways: through terms of trade effects (reducing the price of the imported goods since they are likely made by workers with lower wages); labor supply effects (with demand decreasing for workers with the task trade vulnerable characteristics), and; productivity effects (where the onshore workers refocus on higher-productivity tasks). The aggregate effect of these three wage effects is not clear from the model itself. The first two effects suggest that (real) wages for workers in the home country will fall, but the third effect suggests that average wages could rise. It is likely that these three effects of offshoring impact workers differentially. Those workers least able to respond to the new challenges of higher productivity tasks could still benefit from rising wages tied to average productivity increases, but are less likely to directly benefit from these shifts in general. These distributional effects are confirmed in Rojas-Romagoas (2010) who runs numerical simulations of the Grossman and Rossi-Hansberg model and finds that with nearly all combinations of endowments, robust to a wide a range of parameters, the model leads to increased inequality in the onshore, high-wage country. 9