In the next two chapters, the committee turns its focus to the interactions between technology and the economy. An overarching theme emerges: economic and societal changes occasioned by technological developments are shaped, not just by the availability of new technologies and their features, but also by ideologies, power structures, and human aspirations and agendas. Technologies are not exogenous forces that roll over societies like tsunamis with predetermined results. Rather, our skills, organizations, institutions, and values shape how we develop technologies and how we deploy them once created, along with their final impact.1
1 For the impact of available skills and markets on the direction of technological changes, see D. Acemoglu, 1998, Why do new technologies complement skills? Directed technical change and wage inequality, Quarterly Journal of Economics 113(4):1055-1090; D. Acemoglu and P. Restrepo, 2016, “The Race Between Machine and Man: Implications of Technology for Growth, Factor Shares and Employment,” No. w22252, National Bureau of Economic Research, Cambridge, Mass.; E. Brynjolfsson and A. McAfee, 2014, The Second Machine Age: Work Progress, and Prosperity in a Time of Brilliant Technologies, W.W. Norton, New York.
For the importance of complementarities in organizations, see E. Brynjolfsson and P. Milgrom, 2013, “Complementarity in organizations” in The Handbook for Organization Economics (R. Gibbons and J. Roberts, eds.), Princeton University Press. On how scarcity might spur innovation, see E. Boserup, 1981, Population and Technological Change: A Study of Long-Term Trends, University of Chicago Press, Chicago, Ill. For the impact of workplace organizations on technology, see E. Brynjolfsson and L. Hitt, 2000, Beyond computation: Information technology, organizational transformation and business performance, Journal of Economic Perspectives 14(4):23-48, and L. Hitt, S. Yang, and E. Brynjolfsson, 2002, Intangible assets:
In this chapter, the committee considers the current state of (1) productivity growth, (2) employment, and (3) income distribution. In each case, the role of technology is considered, recent changes are summarized, and some potential future developments are considered, building on the discussion in Chapter 2 of current and possible future trends in underlying technologies. The committee is keenly aware that making forecasts about social phenomena is perilous. Doing so with respect to the fast-changing and dynamic area of technology is even more challenging. Nevertheless, interpreting societal and economic responses to developments in technology can at least provide a framework for thinking about the future.
In his seminal research on economic growth, Robert Solow found that most of the increases in human living standards have come not from working more hours, and not from using more capital or other resources, but from improved productivity—that is, increases in the efficiency of production as defined by the ratio of output to input. In turn, productivity growth comes from new technologies and new techniques of production and distribution.2 In the mid-1990s, the rate of productivity growth increased significantly in the United States, led by the IT-producing sectors as well as IT-using sectors, a change attributed in part to improvements in the nature and use of IT.3 However, in the past 10 years, U.S. aggregate productivity growth has slowed, according to official statistics from U.S. government agencies. The slowdown preceded the 2008 Great Recession, suggesting that the recession is not the only explanation for
Computers and organizational capital, Brookings Papers on Economic Activity 1:137-199. On the influence of vested interests on blocking of technology, see J. Mokyr, 1990, The Levers of Riches: Technological Creativity and Economic Progress, Oxford University Press, Oxford, U.K. For the influence on macro institutions on technology, see D. Acemoglu and J.A Robinson, 2012, Why Nations Fail: Origins of Power, Prosperity and Poverty, Crown Publishing Group, Chicago, Ill.
2 R.M. Solow, 1959, A contribution to the theory of economic growth, Quarterly Journal of Economics 70(1):65-94, doi: 10.2307/1884513.
3 See D.W. Jorgenson, M.S. Ho, and K.J. Stiroh, 2002, Projecting productivity growth: Lessons from the U.S. growth resurgence, Economic Review Q3:1-13; S.D. Oliner, D.E. Sichel, and K.J. Stiroh, 2007, “Explaining a Productive Decade,” Federal Reserve Board, https://www.federalreserve.gov/pubs/feds/2007/200763/200763pap.pdf; E. Brynjolfsson and L. Hitt, 1995, Computers as a factor of production: The role of differences among firms, Economics of Innovation and New Technology 3:183-199; and L. Hitt, S. Yang, and E. Brynjolfsson, 2002, Intangible assets: Computers and organizational capital, Brookings Papers on Economic Activity 1:137-199; K.J. Stiroh, 2002, “Reassessing the Impact of IT in the Production Function: A Meta-Analysis,” Federal Reserve Bank of New York, http://www.nber.org/criw/papers/stiroh.pdf.
this trend, and has been largely accounted for by slowdowns in the IT-producing and IT-using sectors.4
In some ways, this slowdown in productivity growth is counter to the narrative of increasing advances and adoption of IT. The remainder of this section discusses open issues and questions as well as possible pathways for resolving them.
One hypothesis is that there is an increasing measurement problem in the official statistics on productivity. This has been a longstanding research challenge, recognized at least since Solow5 and Griliches.6 Griliches noted that the economy has been shifting increasingly to sectors where output and output quality are difficult to measure, such as government, health, and finance. Unlike counting bushels of wheat or tons of steel, outputs for medical treatment or bank loans are less standardized.7 Measurement is also very challenging in sectors with rapid technological changes, such as the computer and software industries themselves. Output and productivity measurement require measuring output and input price deflators that reflect changes in quality, which is an enormous challenge. How does one compare a smartphone today with a mainframe from 20 years ago, let alone new apps that have no predecessors? Great progress was made in the 1990s and 2000s on improving price deflators for the hardware parts of IT,8 but the software side has been more challenging. Recent evidence suggests that adoption of cloud computing and other changes are even making it more difficult to assess quality changes for hardware.9
A related, but more fundamental, issue is that productivity is neither a measure of technological progress nor welfare. Productivity is based on gross domestic product (GDP), which is in turn a measure of production or output. However, technological progress can increase welfare without increasing output. For instance, if Wikipedia replaces a paper encyclopedia or a free GPS mapping app replaces a stand-alone GPS device, then consumers can be better off even if output is stagnant or declining.10
4 J.G. Fernald, 2014, Productivity and potential output before, during, and after the Great Recession, NBER Macroeconomics Annual 2014, Volume 29, doi: 10.3386/w20248.
5 R.M. Solow, 1987, “We’d Better Watch Out,” New York Times book review, July 12, p. 36.
6 Z. Griliches, 1994, Productivity, R&D, and the data constraint, American Economic Review 84(1):1-23.
8 See Bureau of Economic Analysis, 2000, “National Income and Wealth Division, Investment Branch” and “Computer Prices in the National Accounts,” April.
9 D.M. Byrne, S.D. Oliner, and D.E. Sichel, 2015, “How Fast are Semiconductor Prices Falling?,” NBER Working Papers 21074, National Bureau of Economic Research, April, doi: 10.3386/w21074.
10 E. Brynjolfsson, A. McAfee, and M. Spence, 2014,”New World Order: Labor, Capital, and Ideas in the Power Law Economy,” Foreign Affairs, https://www.foreignaffairs.com/articles/united-states/2014-06-04/new-world-order; E. Brynjolfsson and J.H. Oh, 2012, “The
Under this view, gains would show up in broader measures of economic well-being but not in GDP (and in turn not in official productivity statistics). While these measurement issues remain an active area of study, the most recent research suggests that at most only a small fraction of the productivity slowdown can be attributed to measurement problems.11
Another hypothesis is that the reported slowdown in productivity growth in the IT-producing and IT-using sectors is temporary. Brynjolfsson and Hitt found evidence that the productivity benefits of large enterprise systems took up to 7 years to be fully realized, as significant organizational and process changes were typically required to make full use of accompanying software and hardware investments.12 The diffusion and adoption of technologies is time- and resource-intensive and requires much experimentation, with failures and variable time lags along the way. Building on work by Paul David, Syverson discussed the slowdown in productivity growth in the historical context of electrification of the production process at the end of the 19th century.13 He argues that the impact of electrification on productivity came in two distinct waves. The first wave arrived quickly and reflected the adoption of electrification within the existing organization of production. The second wave, delayed by a few decades, reflected new ways of organizing production around this new technology. Achieving the full productivity benefits and impacts of new technology can take decades and may require complementary “co-inventions” of new business practices, infrastructure, and so on, which can dramatically influence the size and distribution of gains from technology and the nature of its societal effects. Similarly, while the first power looms allowed weavers to produce 2.5 times as much cloth per hour, further improvements in the following 80 years in knowledge and skills increased output per hour 80-fold.14 For IT, the value of intan-
Attention Economy: Measuring the Value of Free Digital Services on the Internet,” in Proceedings of the International Conference on Information Systems, December.
11 D.M. Byrne, J.G. Fernald, and M.B. Reinsdorf, 2016, Does the United States have a productivity slowdown or a measurement problem?, Brookings Papers on Economic Activity (forthcoming); C. Syverson, 2016, “Challenges to Mismeasurement Explanations for the U.S. Productivity Slowdown,” mimeo.
12 See E. Brynjolfsson and L.M. Hitt, 2003, Computing productivity: Firm-level evidence, Review of Economics and Statistics 85.4: 93-808. See also T.F. Bresnahan, E. Brynjolfsson, and L. Hitt, 2002, IT, workplace organization and the demand for skilled labor: A firm-level analysis, Quarterly Journal of Economics 117(1): 339-376, doi: 10.1162/003355302753399526.
13 C. Syverson, 2013, “Will history repeat itself? Comments on ‘Is the Information Technology Revolution Over?’ International Productivity Monitor 25:20-36. See also P.A. David, 1990, The dynamo and the computer: An historical perspective on the modern productivity paradox, American Economic Review 80(2):355-61.
14 J. Bessen, 2015, Learning By Doing: The Real Connection Between Innovation, Wages and Wealth, Yale University Press, New Haven, Conn.
gible complements of computer hardware, including changes in business processes and human capital, can have a value 10 times greater than the direct costs of computer hardware, but they also are costly and time-consuming to invent and implement.15
In a related manner, there is evidence that adopting new technologies requires organizational changes and restructuring of business practices that take time.16 For instance, it is easy to imagine significant benefits, eventually, from the widespread digitization of patient data, yet many physicians complain that the adoption process for electronic medical records is slow and cumbersome, with current costs exceeding current benefits.
This perspective may help reconcile the observation of the apparently rapid changes in technology outlined in Chapter 2 with the current sluggish growth in productivity. Yet there are more pessimistic views about the prospects for productivity and economic growth. Some have suggested that recent (post-2000) innovations in information and other advanced technology simply do not have the same high payoff as innovations in earlier periods. The argument is that earlier innovations were in the form of general purpose technologies that had wide application to many industries.17 Alternatively, some have argued that we are in a period of secular stagnation due to weak aggregate demand.18 The suggestion is that persistently weak aggregate demand is acting as an overall drag on economic growth.
Firm-level evidence for the United States and the Organisation for Economic Co-operation and Development shows a widening gap between the most and least productive firms within industries in the post-2000 period.19 Such widening productivity dispersion may reflect increased
15 E. Brynjolfsson and L.M. Hitt, 2000, Beyond computation: Information technology, organizational transformation and business performance, Journal of Economic Perspectives 14(4):23-48; E. Brynjolfsson and L.M. Hitt, 2003, Computing productivity: Firm-level evidence, Review of Economics and Statistics 85(4):793-808.
16 L. Hitt, S. Yang, and E. Brynjolfsson, 2002, Intangible assets: Computers and organizational capital, Brookings Papers on Economic Activity 1:137-199.
17 R. Gordon, 2016, The Rise and Fall of American Growth: The U.S. Standard of Living Since the Civil War, Princeton University Press, Princeton, N.J.
18 L. Summers, 2016, “The Age of Secular Stagnation,” Foreign Affairs, https://www.foreignaffairs.com/articles/united-states/2016-02-15/age-secular-stagnation.
19 D. Andrews, C. Crisculo, and P.N. Gal, 2016, “The global productivity slowdown, technology divergence, and public policy: A firm level perspective,” Brookings, https://www.brookings.edu/research/the-global-productivity-slowdown-technology-divergence/, and R.A. Decker, J. Haltiwanger, R.S. Jarmin, and J. Miranda, 2016, “Declining Business Dynamism: Implications for Productivity?,” Hutchins Center Working Papers presented at Brookings Conference on Slow Growth in Productivity: Causes, Consequences, and Policies, September 2016.
frictions or distortions in the economy (e.g., dampened competition) slowing down the diffusion process or the pace of business dynamism that is critical for moving resources to the more productive firms. The latter has been shown to be an important part of the process of productivity growth, and is discussed further in Chapter 4. From this perspective, the hypothesis is that while the changes in technology outlined in Chapter 2 are indeed occurring, they are slow to show up in economic growth due to slowing diffusion or business dynamism.
All of these hypotheses are active areas of research. The discussion of future research directions in Chapter 6 emphasizes the significance of exploring such critically important questions. It is useful to note that future productivity growth cannot be predicted simply by extrapolating past trends because there is little serial correlation in growth rates from one decade to the next. Instead, future trends will depend on the invention and deployment of new and improved technologies and on the co-inventions by the workforce, organizations, and institutions needed to effectively use them.
Employment in Recent Years
In the past few years, U.S. employment growth has been fairly robust, with accompanying drops in unemployment. For instance, by early 2016, the unemployment rate fell below 5 percent. However, much of this employment growth can be interpreted as a recovery from the Great Recession, which has been slow despite the fact that it officially ended in 2009. Furthermore, jobs lost in the recession are very different from those that appeared during the recovery.20 While many opportunities continue to be created in fields that do not require a bachelor’s degree, the fastest-growing occupational categories all require a bachelor’s degree or better, and occupations that require a bachelor’s degree are growing at twice the rate of those that do not.21,22
There have also been substantial shifts in employment in various occupational categories. For instance, the employment rate in clerical and
20 A.P. Carnevale, T. Jayasundera, and A. Gulish, 2015, “Good Jobs Are Back: College Graduates Are First in Line,” Georgetown University: Center on Education and the Workforce, https://cew.georgetown.edu/wp-content/uploads/Good-Jobs_Full_Final.pdf.
22 A.P. Carnevale, N. Smith, and J. Strohl, 2010, “Help Wanted, Projections of Jobs and Education Requirements Through 2018,” Georgetown University: Center on Education and the Workforce, https://cew.georgetown.edu/wp-content/uploads/2014/12/HelpWanted.ExecutiveSummary.pdf.
Despite the low unemployment rate, the overall U.S. employment rate (the employment-to-population ratio) remains near a 20-year low. The overall U.S. employment rate exhibited an upward trend through 2000, mostly driven by an increased participation of women in the workforce. It began to decline in the post-2000 period, with a sharp drop during the 2008 Great Recession, from which it has recovered slightly. Some of this trend can be accounted for by the aging of the population. However, declines in the employment rate are especially large for young and less educated individuals. Employment rates of prime age (25-54 year-old) males are still low (84 percent in 2014, near the 50-year low of 81 percent in 2010, as compared to a high of 95 percent in 1967, according to the
23 P. Restrepo, 2015, “Skill Mismatch and Structural Unemployment,” Massachusetts Institute of Technology, http://pascual.scripts.mit.edu/research/01/PR_jmp.pdf.
annual averages from the Bureau of Labor Statistics).24 This highlights that the overall decline cannot be accounted for simply by the aging of the U.S. population.25,26 While both globalization and technology are seen as important factors, many economists view automation as the single most important factor. For instance, Larry Katz, an expert on labor studies and editor of the Quarterly Journal of Economics has said, “Over the long haul, clearly automation’s been much more important—it’s not even close.”27
Among young and less educated workers, the declines have also been especially sharp for certain race/ethnicity groups (especially non-Hispanic blacks). In addition, almost a third of all the unemployed in 2014 were classified as “long term unemployed” (i.e., out of the job market for more than 27 weeks), and many eventually become “discouraged workers,” people that drop out of the market for employment entirely.28 Such patterns are of particular concern in the context of this report, since it is these most vulnerable groups that may be left behind by ongoing changes in technology.
Future Prospects for Technology and Employment
Predictions that new technologies will make workers largely or almost entirely redundant are as old as technological change itself. Although the story might be apocryphal, the famous Roman historian Pliny the Elder recounts how the Roman Emperor Tiberius killed an inventor who had supposedly invented unbreakable glass for fear of what this would do to the glassmaking trade. Queen Elizabeth I similarly refused to grant William Lee a patent for his mechanical knitting machine, arguing, “Consider thou what the invention could do to my poor subjects.”29
It is not only emperors and queens who have feared the implications of new technologies for employment. More famously, British textile workers
24 Organization for Economic Co-operation and Development, “Employment Rate: Aged 25-54: Males for the United States,” retrieved from FRED, Federal Reserve Bank of St. Louis; https://fred.stlouisfed.org/series/LREM25MAUSA156S, March 15, 2017.
25 S.J. Davis and J. Haltiwanger, 2014, “Labor Market Fluidity and Economic Performance,” University of Chicago and NBER, and University of Maryland and NBER, http://faculty.chicagobooth.edu/steven.davis/pdf/LaborFluidityandEconomicPerformance26November2014.pdf.
27 As quoted in Claire Cain Miller, “The Long Term Job-Killer Is Not China: It’s Automation,” The New York Times, December 21, 2016, https://www.nytimes.com/2016/12/21/upshot/the-long-term-jobs-killer-is-not-china-its-automation.html.
28 K. Kosanovich and E. Theodossiou Sherman, 2015, “Trends In Long-term Unemployment,” Bureau of Labor Statistics, http://digitalcommons.ilr.cornell.edu/key_workplace/1399/.
29 M. Finley, 1973, The Ancient Economy, University of California Press, Berkeley, Calif.; D. Acemoglu and J. Robinson, 2011, Why Nations Fail, Crown, New York.
in the early 19th century, known as Luddites, fearing that the new automation coming with power looms, stocking frames, and spinning frames were threatening to replace their expert positions with low-wage laborers, destroyed machines and burned the house of John Kay, the inventor of the “flying shuttle.” In an attempt to halt destructive acts, the British Parliament enacted a law making “machine breaking” a capital offense.30
More recently, the economist John Maynard Keynes predicted that the introduction of new technologies would create considerable wealth but would also generate widespread technological unemployment as machines replaced humans. In 1930, he predicted that the work week would fall to 15 hours and that the “economic problem” of providing for basic needs would be solved.31 Economic historian Robert Heilbroner similarly predicted in 1965: “As machines continue to invade society, duplicating greater and greater numbers of social tasks, it is human labor itself—at least, as we now think of `labor’—that is gradually rendered redundant.”32 Nobel Prize winner Wassily Leontief saw an analogy between human labor and horses of the early 20th century, and in 1952 predicted that humans will follow horses in becoming redundant: “Labor will become less and less important. . . . More and more workers will be replaced by machines. I do not see that new industries can employ everybody who wants a job.”33,34
However, predictions of widespread, technologically induced unemployment have not come to pass, at least so far.35 Technological changes over the last 200 years (and presumably many of those that came before) stimulated demand, created new markets, and fueled wage growth with few adverse consequences for aggregate employment. To be sure, technologies did and will continue to decimate particular occupations. As the Luddites feared, artisans lost their jobs in spinning and then weaving as new technologies automated tasks they had previously performed.
30 K.E. Lommerud, F. Meland, and O.R. Straume, Globalisation and union opposition to technological change, Journal of International Economics 68(1):1-23, 2006.
31 J.M. Keynes, 1933, “Economic Possibilities for Our Grandchildren (1930),” pp. 358-373 in Essays in Persuasion, W.W. Norton & Company, New York.
32 R.L. Heilbroner, 1965, Men and machines in perspective, National Affairs, Fall, pp. 27-36.
33 W. Leontief, 1952, Machines and man, Scientific American 87(3):150-160. For a more recent perspective on this questions, see E. Brynjolfsson and A. McAfee, 2015, “Will Humans Go the Way of Horses?,” Foreign Affairs (July/August), pp. 8-14.
34 Acemoglu and Restropo, 2016, “The Race Between Machine and Man.”
35 We do note, however, that the length of the workweek has dropped significantly since the 1800s, when a workweek of longer than 70 hours was not uncommon. For more discussion of this point, see Economic History Association, “Hours of Work in U.S. History,” EH.net, https://eh.net/encyclopedia/hours-of-work-in-u-s-history/, and R.J. Gordon, 2016, The Rise and Fall of American Growth: The US Standard of Living Since the Civil War, Princeton University Press, Princeton, N.J.
Similarly, the replacement of horses by automobiles eliminated the need for blacksmiths. But as these jobs disappeared, new ones sprang up to operate, manage, and service the new technologies. For instance, in the late 1800s, the replacement of the stagecoach by the railroad went hand in hand with the creation of new work for managers, engineers, machinists, repairmen, and conductors. Simultaneously, there was a boom in a range of new service occupations, from teaching to entertainment to sales.36
Nonetheless, simultaneous automation of a broader range of tasks could create unemployment or perhaps reduce aggregate levels of employment for an extended period of time. As noted in the previous section, over the past 20 years, the share of people working—the employment-to-population ratio—has declined. While there are many factors at work, it is possible that technological substitution for certain types of labor is part of the explanation. As compensation falls for tasks that can increasingly be done by machines, some people may choose to work less or not at all, finding other alternatives, including increased leisure or family time, applying for disability benefits, or investing in education, to become relatively more attractive. Over the longer term, there may be a continuation of the long-term decline in the share of hours worked as society as a whole becomes wealthier and leisure becomes relatively more attractive.
What happens depends, in part, on whether new technologies automate and replace workers in existing tasks more rapidly than they create new demands for labor. Which will be the case is difficult to answer, because it is easier to see how new technologies coming down the line will automate existing tasks than it is to imagine tasks that do not yet exist and how new technologies may stimulate greater consumer demand. Further, the future of employment is not only a question of the availability or necessity of tasks to be performed, but how they are organized, compensated, and more generally valued by society. These are matters of business strategy, social organization, and political choices and not simply driven by technologies themselves.
Consider self-driving cars. In principle, driving and delivery occupations could be automated with the use of such technologies over the next several decades. Visions of a future with fully automated vehicles have captured many people’s imagination. However, there are numerous social and cultural as well as technological roadblocks to such an outcome. These include such factors as consumer trust; the fact that there will be a long period of mixed-use road use, with both autonomous driving and manual driving cars sharing the roads; and the infrastructure require-
36 Acemoglu and Restrepo also provide evidence that there is a large contribution of new occupations to employment growth in the last three decades: Acemoglu and Restrepo, 2016, “The Race Between Machine and Man.”
ments needed to expand usage and performance of self-driving cars. It is possible that ongoing development of these technologies, including infrastructure, will create more jobs than are lost in the wake of self-driving vehicles, but it is likely that the skills required for such jobs will be quite different from those currently possessed by today’s truckers and taxi drivers. The new jobs are likely to rely more heavily on analytic, cognitive, and technical skills. Indeed, even in the near term, as self-driving technologies are being developed, the occupation of trucking37 is likely to be transformed. For example, additional IT-based capabilities for driver simulation training can help improve the skill sets of more drivers than would be possible otherwise. In the longer term, increased automation will reduce the need for additional drivers and ultimately reduce overall demand for truck drivers. “Platooning,” where a single lead vehicle driven by a driver is followed by a string of other self-driving vehicles, is already emerging as a viable technology for highway driving and can affect employment numbers. As transportation costs drop due to partial automation, it is possible that lower per-unit costs will lead to increased demand (e.g., for more delivery services), resulting in a partially counteracting force in the opposite direction toward increased demand for drivers. Self-driving cars also offer a good illustration of the variable and mixed impact of technology on employment, as well as the long and often uneven march of technology development, which complicates the ability to make accurate long-term projections.
In addition to eliminating some jobs while creating others, technological developments can create new occupations without reducing employment in older occupations. New medical imaging technologies are a case in point. Prior to the development of computer-controlled imaging modalities such as ultrasound, computed tomography scanners, and magnetic resonance imaging, most technicians who worked in radiology departments operated standard X-ray machines and fluoroscopes. The jobs associated with these technologies were not significantly altered by the arrival of digital imaging. Instead, new technicians’ occupations arose: the sonographer, the computed tomography technologist, and the MRI technologist as well as the technicians’ occupations who service such machines. Thus, in the case of medical imaging, the overall number
37 Truck driving remains a significant source of employment and middle-class jobs in the United States. In fact, according to a recent analysis by NPR, in 2014, “truck, delivery, and tractor drivers” were the most common occupational category in 29 states, see Q. Bui, 2015, “Map: The Most Common Job in Every State,” NPR, http://www.npr.org/sections/money/2015/02/05/382664837/map-the-most-common-job-in-every-state. See also T. Reddy, 2007, “Fleets Eye Safety Gains to Cut Insurance Costs,” Transport Topics, http://www.ttnews.com/articles/printopt.aspx?storyid=17563, for figures on truck driver hiring and claims of inability to fill positions.
of occupations, and hence people employed as technicians, expanded. Furthermore, with the march of technology, both in terms of advances in imaging and with developments at the intersection of imaging and other areas of biomedical engineering, radiologists began to specialize in particular imaging modalities and in whole new radiology subdisciplines such as “interventional radiology,” extending the range of opportunities for careers within radiology and increasing the need for radiologists.38
Nonetheless, technologies can also have an impact on how tasks are allocated and how job categories and tasks associated with particular organizational forms and structures are designed. For instance, they can shift the allocation of tasks across occupations such that some occupations contract as the work they once performed is shifted onto members of other occupations.39 An instructive example is the advent of word processing technologies and online databases in the context of a typical academic department in the university. As recently as the 1980s, administrative assistants answered phones, interacted with students, kept paper records of accounts, filed documents, and typed letters, memos, and manuscripts for faculty (who often wrote first drafts by hand). Today, administrative assistants continue to answer phones and interact with students, but few type documents for faculty. Professors now use a computer to create and revise their own documents. Some faculty also enter their own data on travel expenses and other activities directly into databases, tasks previously performed by administrative assistants. These and other changes removed certain tasks from administrative assistants and transferred them to faculty, which can be viewed as an instance of “disintermediation.”40 Because of the increased efficiency of producing and storing documents and because faculty have assumed the task of producing documents, universities now employ fewer administrative assistants, and some of those who remain have acquired new skills and tasks, such as the maintenance
38 S.R Barley, 1986, Technology as an occasion for structuring: Evidence From observations of CT scanners and the social order of radiology departments, Administrative Science Quarterly 31:78-108; S.R Barley, 1990, The alignment of technology and structure through roles and networks, Administrative Science Quarterly 35:61-103.
39 J. Bessen, 2015, “How Computer Automation Affects Occupations: Technology, Jobs and Skills,” Law and Economics Working Paper No. 15-49, Boston University School of Law, Boston, Mass.
40 Disintermediation refers to the elimination tasks or people in a supply chain or work flow because the tasks are now done by someone positioned earlier in the work flow (R. Benjamin and R. Wigand, 1995, Electronic markets and virtual value chains on the information superhighway, Sloan Management Review 36:62-67; A.M. Chircu and R.J. Kauffman, 1999, Strategies for Internet middlemen in the intermediation/disintermediation/reintermediation cycle, Electronic Markets 9(1-2):109-117; U. Schultze and W.J. Orlikowski, 2004, A practice perspective on technology-mediated network relations: The use of Internet-based self-serve technologies, Information Systems Research 5(1):87-106.
The committee notes that the effects of technologies on employment can be shaped by interests and social dynamics beyond merely the technological dimension. For example, computer-mediated communications, especially those facilitated by the Web, such as e-mail, computer teleconferencing, and the ability to easily and almost instantaneously transfer documents of all kinds across space (and hence time zones), were initially thought of as simply more efficient ways to communicate. But because these technologies did not require co-location, companies began using such technologies to both outsource and offshore a variety of tasks and even jobs, ranging from clerical to engineering work. There is nothing about computer-mediated communication technologies that preordained such developments. Instead, they are the result of choices (strategic or otherwise) by decision makers in organizations about how the technologies would be deployed and what they would be used to achieve, along with market forces encouraging the adoption of cost-efficient processes. Choices regarding the development of technologies can also be influenced by the same interests and social dynamics. For example, it has been suggested that the decision to develop technologies that automate rather than augment the human role in the machine tool industry was driven by the combined interests of the U.S. Air Force and the Massachusetts Institute of Technology servomechanisms laboratory.42
Human Skills Versus Automation
Consideration of whether technology can replace human workers has prompted discussion about the subtle complexity of human skills. A recent paper43 by economist David Autor invoked the philosopher Michael Polanyi: “We can know more than we can tell. . . . The skill of a driver cannot be replaced by a thorough schooling in the theory of the motorcar; the knowledge I have of my own body differs altogether from
41 Bill Gates predicted in 2016 that for “pure labor substitution for jobs that are largely physical and visual manipulation—driving, security guard, warehouse work, waiter, maid, that threshold—I don’t think you’d get much disagreement that over the next 15 years the robotic equivalents in terms of cost, in terms of reliability, will become a substitute to those activities” (E. Klein, 2016, “Bill Gates: The Energy Breakthrough That Will ‘Save Our Planet’ Is Less Than 15 Years Away,” Vox, last updated February, http://www.vox.com/2016/2/24/11100702/bill-gates-energy).
42 D.F. Noble, 1984, Forces of Production, Transaction Publishers Piscataway, N.J.
43 D. Autor, 2014, “Polanyi’s Paradox and the Shape of Employment Growth,” Working Paper 20485, National Bureau of Economic Research, Cambridge, Mass.
the knowledge of its physiology.”44 This phenomenon, that tacit knowledge often is greater than explicit cognition, is referred to as Polanyi’s Paradox. Currently, computational systems are far from being able to use creativity, intuition, persuasion, and imaginative problem solving, or to coordinate and lead teams. Autor and others have argued that many highly valued and important human capabilities may never be automated.45 As technology becomes more sophisticated and encapsulated, managing human interfaces may become the dominant component of more and more jobs; there is already evidence that social skills are in increasing demand and valued in the labor market.46 For instance, while drivers for a ride-sharing service do not need to be experts on the internal combustion engine or smartphone to do their jobs, they do need reasonably good interpersonal skills to be successful in this era of online ratings. Educational programs, even those in vocational disciplines like business and engineering, may need to add interpersonal and creative skills to their mix of hard analytical skills.
The resistance to automating multiple dimensions of human intellect and of the “presence” and leadership of people suggests that there will likely be enduring market value for traits and factors that are uniquely human, as only humans will be able to perform certain types of work for the foreseeable future, if not forever. To what extent are these human attributes, including creativity, empathy, interpersonal skills, leadership, mentoring, and physical presence currently valued in the U.S. labor force, and how will these uniquely human capabilities be valued in the U.S. labor force in the future? A number of these uniquely human attributes include cognitive and “noncognitive” skills. On the other hand, recent improvements in machine learning have enabled significant technical advances in, for example, the field of self-driving cars, suggesting that even Polanyi’s archetypal example of driving a motorcar is not immune to automation. The rapidly growing attendance at research conferences on artificial intelligence (AI), like the annual Neural Information Processing Systems and Association for the Advancement of Artificial Intelligence conferences, demonstrates that an increasing number of researchers are attempting to address these challenges, and most of them now focus on approaches that enable machines to learn how to do tasks, from recognizing and labeling objects to understanding speech, improving dexterity and mobility, and mastering increasingly complex games and puzzles.
44 M. Polanyi and A. Sen, 2009, The Tacit Dimension, University of Chicago Press, Chicago, Ill.
45 D.H. Autor, 2015, Why are there still so many jobs? The history and future of workplace automation, Journal of Economic Perspectives 29(3):3-30.
46 D.J. Deming, 2015, “The Growing Importance of Social Skills in the Labor Market,” National Bureau of Economic Research, doi: 10.3386/w21473.
Income and Wealth Distribution in Recent Years
It is generally understood that, by increasing productivity, IT will tend to increase overall income—although without a guarantee that these gains will be evenly distributed. Furthermore, while it is common to focus on average levels of income and income growth, the distribution of those gains can also have an effect on well-being. This is true not only because absolute levels directly affect the quality of life of particular groups, but also because broad perceptions of unfairness can have a negative psychological impact, and inequality can contribute to sociopolitical tensions.
Since the mid-1970s, the United States has experienced significant growth in inequality in both income and wealth. This is the subject of a large amount of literature and has been documented in great detail by Acemoglu, Autor, Katz, Piketty, Saez, and many others.47 One aspect of this is the growing dispersion between productivity growth and median worker compensation, as most of the income growth went to the top of the income distribution. Over the past several decades, IT and automation have been a significant driver of this increase in inequality, although there are also other forces at work.
Much popular attention has been focused on the rising share of income of the top 1 percent of each of these distributions. While this increase has been substantial, with the share of income accruing to the top 1 percent of households increasing from about 10 percent to over 20 percent between 1980 and 2012, there have also been increases in earnings inequality within the other 99 percent, accounted for largely by the increasing skills premium associated with a 4-year college degree. For example, the absolute median earnings gap between those with a high school and a college degree approximately doubled from 1980 to 2012, as the real wages of college graduates rose and those of less educated workers fell through about 2000.48
A related phenomenon is the falling share of GDP paid to labor relative to owners of capital (illustrated in Figure 3.5).49 This trend affects not only income, but also wealth (to which an individual’s income contributes
47 For reviews, see, for example, D.H. Autor, L.F. Katz, and M.S. Kearney, 2008, Trends in US wage inequality: Revising the revisionists, Review of Economics and Statistics 90.2:300-323;T. Piketty, 2014, Capital in the Twenty-First Century, Harvard University, Cambridge, Mass.; and T. Piketty and E. Saez, 2013, Top incomes and the Great Recession: Recent evolutions and policy implications, IMF Economic Review 51 (3):456-478, doi: 10.1057/imfer.2013.14.
48 D.H. Autor, 2014, Skills, education, and the rise of earnings inequality among the ‘other 99 percent,’ Science 344(6186):843-851.
49 L. Karabarbounis and B. Neiman, 2013, The global decline of the labor share, Quarterly Journal of Economics 128(1):61-103, doi: 10.3386/w19136.
over time). It suggests that trends in income are increasingly favoring those who have already accrued wealth. This decline in the labor share of GDP, if sustained, will affect the distribution of wealth as well as that of income, expanding the share of total income flowing to wealth holders.50
Many factors are likely at work in this landscape of inequality; technological change, social biases, increased globalization and trade, the decline in labor union density and power,51 declines in the real minimum wage, changing norms regarding executive compensation, growing economic deregulation, changes in tax rates, and growing oligopoly—or in some cases, simple monopoly52—are among the hypothesized causes of increased inequality of income and wealth over the past 40 years.53 However, for the purposes of this study, the committee focuses on the role of technology in income and wealth distributions.
As with employment, the case that technological advances have contributed to wage inequality is strong. For most of the 20th century, real median incomes—incomes of people at the 50th percentile—grew at least as fast as overall real GDP per person, suggesting that the benefits of improved technological progress were widely shared. But since the late 1970s, productivity and GDP per person have continued to grow, while median incomes have stagnated (illustrated in Figure 3.2), reflecting growing income inequality over a period of significant technological change.
There is a debate in the research literature, and indeed, among committee members, about how much of the increase in inequality should be attributed to technology. There are three prominent narratives implicating technological change as a force toward greater inequality over the last several decades. First, many new technologies have replaced labor-intensive, routine, and physical tasks and expanded demand for labor in jobs that require social skills, numeracy, abstract thinking, and flexibility.54 This shift is often said to be responsible for higher returns to a college education and the widening income gap between skilled and less
50 For this and other statistics on wealth inequality, see E.N. Wolff, 2012, The Asset Price Meltdown and the Wealth of the Middle Class, New York University, New York; A.B. Atkinson, T. Piketty, and E Saez, 2011, Top incomes in the long run of history, Journal of Economic Literature 49.1:3-71; and T. Piketty, 2014, Capital in the Twenty-First Century, Harvard University, Cambridge, Mass.
51 See, for example, B. Western and J. Rosenfeld, 2011, Unions, norms, and the rise in US wage inequality, American Sociological Review 76(4):513-537, doi: 10.1177/0003122411414817.
52 J. Furman and P. Orszag, 2015, “A firm level perspective on the role of rents in the rise of inequality,” White House, https://www.whitehouse.gov/sites/default/files/page/files/20151016_firm_level_perspective_on_role_of_rents_in_inequality.pdf.
53 See T. Piketty, Capital in the Twenty-First Century, Harvard University, Cambridge, Mass, p. 70.
54 See, for example, D.J. Deming, 2015, “The Growing Importance of Social Skills in the Labor Market,” National Bureau of Economic Research, doi: 10.3386/w21473.
skilled workers. Second, as labor-intensive tasks are automated, the share of income going to capital relative to labor can increase, which may also help to explain the falling share of labor in overall GDP (as illustrated in Figure 3.5) both in the United States and abroad. Third, improvements in communication technologies have contributed to what has been termed the “superstar phenomenon” whereby the most successful performers in any occupation can now command a larger share of the global market. This in part reflects their improved ability to sell to not only customers in local markets, but also with greater ease to those in regional, national, and even global markets as improved communications technologies reduce the costs of reaching a broader audience.55 Geography and consumer ignorance have become less important as barriers to entry, making it
55 D. Coyle, 1997, “Rich man, poor man, superstar,” Independent, http://www.independent.co.uk/news/business/rich-man-poor-man-superstar-1271342.html; S. Rosen, The economics of superstars, The American Economic Review 71(5):845-858, 1981.
easier for sellers with a superior product to gain a dominant market share.56 This phenomenon is also sometimes said to explain the growing proportion of the national income garnered by the top 1 percent of the wage distribution.57
Changes in IT also seem to be playing a role in the changing demand for skills and the earnings inequality for the other 99 percent. Technology can be a complement for highly skilled workers, as well as a substitute for low- or medium-skill workers. This is often called the skill-biased technological change hypothesis.58,59,60 When new technologies have different skill requirements than older ones, they tend to favor the hiring of workers possessing these skills. Unless supply changes sufficiently, this will shift wages in favor of the more skilled group.61 In recent decades, the impact of IT has been uneven across the skill distribution.62 Since the 1970s, males with graduate or college education have seen their wages grow, while those with a high school education or less have seen falling wages63 (see Figure 3.3). It has been suggested that this divergence is exacerbated by an increasing reliance on technology in the workplace, as the skills required to work with these technologies are more readily
56 E. Brynjolfsson, Y.J. Hu, and M.S. Rahman, Competing in the age of omnichannel retailing, MIT Sloan Management Review 54(4):23, 2013.
57 See, for example, E. Brynjolfsson and A. McAfee, 2014, The Second Machine Age, W.W. Norton & Company, New York.
58 D. Card and J.E. DiNardo, 2002, Skill-biased change and rising wage inequality: Some problems and puzzles, Journal of Labor Economics 20:4, doi: 10.3386/w8769.
59 D.H. Autor, L.F. Katz, and M.S. Kearney, 2008, Trends in U.S. wage inequality: Revising the revisionists, Review of Economics and Statistics 90(2):300-323.
60 C. Goldin and L.F. Katz, 2007, “The Race between Education and Technology: The Evolution of U.S. Educational Wage Differentials, 1890 to 2005,” National Bureau of Economics, doi: 10.3386/w12984.
61 This perspective is different than the common claim that new technologies always create inequality. In fact, many new technologies of the 19th century automated previously skilled occupations and expanded unskilled assembly work which paid lower wages than the prior forms of work. For instance, the Luddites may have been misguided in their tactics of smashing mechanical spinning and weaving machines, but they were right that the way these machines were used was bad news for them: relatively highly paid textile workers were gradually replaced by machines and less well paid machine tenders. In general, if there is a mismatch between the skills of the workforce and the skill requirements of new technologies, changes in the structure of pay will tend to follow.
62 World Economic Forum, Global Agenda Council on Employment, 2014, “Matching Skills and Labour Market Needs: Building Social Partnerships for Better Skills and Better Jobs,” http://www3.weforum.org/docs/GAC/2014/WEF_GAC_Employment_MatchingSkillsLabourMarket_Report_2014.pdf.
63 D.H. Autor, L.F. Katz, and M.S. Kearney, 2008, Trends in U.S. wage inequality: Revising the revisionists, Review of Economics and Statistics 90(2):300-323; D.H. Autor, 2014, Skills, education, and the rise of earnings inequality among the ‘other 99 percent,’ Science 344(6186):843-851.
acquired through higher education.64 Related to this challenge, the skills gap continues to widen such that more than half of employers report that they have had difficulty finding qualified job applicants to fill certain jobs, which they believe to be in part due to education gaps.65,66
In the 1980s and 1990s, these changes in technology—along with complementary factors such as globalization, deregulation, and deunionization—have likely contributed to the reduction in demand for middle-level skills, and this has been reflected in both the quantity of jobs and in wages for middle-skill workers (see Figure 3.4).
In particular, workers doing routine tasks (such as production tasks in manufacturing or clerical tasks) have seen their demand decline due to multiple factors, including changing technology. This is reflected in a decline in manufacturing employment even as output has grown to an all-time high. Globalization has further eroded the demand for such skills in advanced economies like the United States. In contrast, there have been expanding job opportunities in both high-skill, high-wage occupations like professional, technical, and managerial occupations and low-skill,
64 R. Valletta, 2015, “Higher Education, Wages, and Polarization,” Federal Reserve Bank of San Francisco, http://www.frbsf.org/economic-research/publications/economicletter/2015/january/wages-education-college-labor-earnings-income/.
65 CareerBuilder, “The Shocking Truth about the Skills Gap,” 2014, http://www.careerbuildercommunications.com/pdf/skills-gap-2014.pdf, accessed August 16, 2016.
66 A recent survey by Manpower group found this to be a global trend, with 38% of surveyed employers reporting a talent shortage in 2015, a 7-year high; see Manpower Group, 2015, 2015 Talent Shortage Survey, http://www.manpowergroup.com/wps/wcm/connect/db23c56008b6-485f-9bf6-f5f38a43c76a/2015_Talent_Shortage_Survey_US-lo_res.pdf?MOD=AJPERES.
But changes in skill demand are far from the only factors at work. In addition, technology may be helping to drive a decline in the labor force participation rate and a broader shift in the labor-capital relationship, as advanced technology is embodied in capital equipment that replaces many workers.69,70 This shift has happened not only in the United States but in many nations around the world. Figure 3.5 illustrates the global decline of the labor share of income.
The reductions in real compensation for labor include the decline of
67 D. Acemoglu, 2009, Changes in unemployment and wage inequality: An alternative theory and some evidence, American Economic Review 89(5):1259-1278.
68 D. Acemoglu, 2001, Good jobs versus bad jobs, Journal of Labor Economics 19(1):1-21.
69 B. Neiman and L. Karabarbounis, 2013, The global decline of the labor share, Quarterly Journal of Economics 129(1):61-103.
70 E. Brynjolfsson and A. McAfee, 2014, The Second Machine Age: Work, Progress, and Prosperity In a Time of Brilliant Technologies, W.W. Norton & Company, New York.
defined-benefit pensions and cuts in employer-provided health insurance.71 In addition to technological change, the shift in the labor-capital ratio likely reflects increased monopoly power and rents by firms in many industries and the growing share of residential housing in the capital stock.72,73
Another factor that is important in this context is that advances in IT often involve significant supply-side economies of scale: the cost of making the first copy of a new enterprise system, video game, or mobile app is often significant, but the cost of making an additional copy is very low. Thus companies whose software has more users will tend to have lower average costs per user. Likewise, there are often demand-side economies of scale as well, which are more commonly called “network effects.” Network effects occur when users get more benefits from a product or platform when it is adopted by more users.74,75 For instance, the value of Facebook or LinkedIn is greater if other people, especially friends, family, or colleagues, also use the same application. Similar logic holds for many business-to-business platforms and, to some extent, even productivity software like word processors, spreadsheets, and presentation software, because they are compatible between users and enable file sharing and collaboration.
Economies of scale on the supply side, the demand side, or both tend to a greater prevalence of “winner take all” outcomes where the leading firm in a particular platform does exceptionally well. Even if the gains are shared across all the workers within the winning firms, this still leads to concentration of the gains to a relatively small share of workers.76
That said, gains have not been evenly distributed, with the top employees in firms seeing the biggest gains on average.77 Founders and
71 J.A. Cobb, 2015, Risky business: The decline of defined benefit pensions and firms’ shifting of risk, Organization Science 26(5):1332-1350.
72 J. Furman and P. Orszag, 2015, “A Firm-Level Perspective on the Role of Rents in the Rise of Inequality,” https://www.whitehouse.gov/sites/default/files/page/files/20151016_firm_level_perspective_on_role_of_rents_in_inequality.pdf.
73 M. Rognlie, 2015, “Deciphering the fall and rise in the net capital share,” Brookings Paper on Economic Activity, March 19-20, http://www.brookings.edu/~/media/projects/bpea/spring-2015/2015a_rognlie.pdf.
74 J. Farrell and G. Saloner, 1987, Competition, compatibility and standards: The economics of horses, penguins, and lemmings, in Product Standardization and Competitive Strategy (L. Gabel, ed.), North Holland.
75 G.G. Parker, M.W. Van Alstyne, and S.P. Choudar, 2016, Platform Revolution: How Networked Markets Are Transforming the Economy—And How to Make Them Work For You, W.W. Norton & Company, New York.
owners of the most successful IT firms and entrepreneurs who have introduced new products, business models, and platforms are among the biggest winners. This is consistent with the rise of the “innovator class,” which suggests that neither labor nor capital owners will necessarily benefit the most from technological advances. Instead, the rapid proliferation of digital technology may enable a third class—people who can create and distribute new products, services, and business models—to prosper immensely.78 In addition, there is evidence that pay for top executives, like CEOs, has grown fastest in industries that use IT the most intensively, which is consistent with them being able to gather more detailed information and relay their decisions more effectively across more people and assets. This can amplify their power, importance, and relative pay.79
While the gender wage gap has narrowed since the 1980s, it is important to note that significant wage gaps by both race and gender persist in the United States. According to a recent analysis by the Pew Research Center,80 2015 median hourly earnings for women were just 83 percent of white men’s, and black workers’ only 75 percent of white men’s. Much but not all of these gaps have been explained by varying education and experience, or the fact that different groups participate at different rates in different industries and occupational fields. Even when controlling for education, white men out-earned all groups except for Asian men; white and Asian women out-earned Hispanic and black men and women. How these disparities will shift in the future as a result of technological change is an open question; gender and occupational fields are discussed further in Chapter 4.
Future Prospects for Income Distribution
An important variable for understanding and shaping future developments is the way education and vocational training institutions might ameliorate the mismatch between new technologies and existing skills. The early 20th century was also a period of rapid technological change in the U.S. economy. However, existing evidence suggests that wage inequality declined during this era, in part because the U.S. education system substantially increased the availability of workers with primary and
78 E. Brynjolfsson, A. McAfee, and M.I. Spence, 2014, “New World Order: Labor, Capital, and Ideas in the Power Law Economy,” Foreign Affairs, https://www.foreignaffairs.com/articles/united-states/2014-06-04/new-world-order.
79 E. Brynjolfsson, H. Kim and G. Saint-Jacques, 2015, CEO pay and information technology, in Firms and the Distribution of Income: The Roles of Productivity and Luck, National Bureau of Economic Research, Cambridge, Mass.
80 E. Patten, 2016, “Racial, Gender Wage Gaps Persist in US Despite Some Progress,” Pew Research Center. http://www.pewresearch.org/fact-tank/2016/07/01/racial-gender-wage-gaps-persist-in-us-despite-someprogress.
secondary schooling. That does not mean that simply investing more in the same sorts of education is the best way to reduce inequality, however. More out-of-the-box thinking may be necessary for prescribing what the educational system needs to deliver. The existing evidence so far suggests that greater numeracy skills and ability to think abstractly might help. A cocktail of other skills are also likely to be affected, including cognitive, physical, interpersonal, networking, and problem-solving skills. But an emphasis on any fixed set of skills might be too backward-looking. As technology continues to evolve, the future workforce will also likely need to evolve. This suggests that flexibility will likely continue to be a valued skill—one that may be particularly underprovided by the current schooling system—in particular, to enable lifelong learning and adaptability to a changing labor market.
This discussion suggests that for society to make the best use of new technologies without increasing income inequality, adjustments are necessary. Although the heaviest burden of adjustment is likely to fall on the skills, competencies, and flexibility of workers, the perspective that technology is not a force of nature and can be shaped and adapted by societal decisions suggests that technology can have positive societal impacts if it is designed with certain values in mind. Can society continue to harvest the benefits of new technologies while at the same time modifying their implementation so that they create more work for those at the margins of society, make better use of existing workers, and yield deeper satisfaction for workers?
The history of numerically controlled, or programmed, machine tools is instructive on this point. In the early days of numerical control, the task of writing programs for machine tools was allocated to programmers, not machinists. The programs were encoded on paper tapes or punch cards and fed to the machine tools. Machinists simply monitored the tools. This led some to argue that machinists would be deskilled because management desired to separate cognition from execution. But with the rapid improvements in the microchip, computers became small enough to embed in the machine tools, and machinists now had access to the computers and the programs. They began to learn to alter and eventually revise the code that controlled their tools. In short, programming became part of the machinist’s task.81 More recent technologies like the Baxter and Sawyer robots from Rethink Robotics take this a step further, enabling line workers to “program” the machines simply by physically guiding them through the steps, rather than by writing computer code.82
It bears repeating that this discussion of technology and inequality
81 D.F. Noble, 1984, Forces of Production, Transaction Publishers.
should not be read to imply that all aspects of inequality and its recent increase in much of the Western world can simply be attributed to technological change: one must also consider the role played by the demise of unions, concessions on wages and job structures, the offshoring of work to places with lower wage rates, tax policy, and the increasing contribution played by financial investments in the accumulation of wealth. Indeed, increasing inequality in the United States has a political dimension, a phenomenon that is, of course, not new. Politically powerful individuals and groups have long deployed their power in order to increase their economic rents. This is seen today in the form of some business groups campaigning for special treatment of their line of business or for the continuation of tax loopholes. Nevertheless, there is a technological element to such rent-seeking, and to inequality as well. Some of the most spectacular salaries on Wall Street are paid by hedge funds that deploy more and more sophisticated computers and algorithms to execute trades or arbitrage information, before their competitors, in a matter of milliseconds. This is just one facet of the potential use of technology in ways that do not necessarily advance social welfare, but create large benefits for those who deploy and control it.
In sum, new technologies can participate in eliminating occupations, creating new occupations, shifting the distribution of tasks among occupations, and altering the geographical division of labor.83 The dynamics of employment and wage inequality are driven by forces that are not exclusively or even primarily technological in nature: institutional patterns, legal structures, tax structures, and managerial ideologies that emphasize the accumulation of wealth also play a role. Most importantly, these forces interact with technology, and vice versa, to shape ultimate outcomes. Consequently, the ways that any particular technology, or a range of new and rapidly changing technologies, will affect employment and income are not predetermined, but rather a function of choices.
The committee ends by again cautioning the reader against believing that all the effects of a technology on employment and inequality are inherent in the nature of the technology itself. How technologies affect work and employment hinge not only on the constraints and affordances
83 There are some interesting parallels to the effects of import competition. For instance, Autor, Dorn, and Hanson (2013) found that rising Chinese imports caused higher unemployment, lower labor force participation and reduced wages in the local markets that had manufacturing plants with competing products (D.H. Autor, D. Dorn, and G.H. Hanson, 2013, The China syndrome: Local labor market effects of import competition in the United States, American Economic Review 103.6:2121-2168).
of the technology, but also on complex interactions among technologies, organizations, skills, institutions, markets, culture, and public policies. Nevertheless, on the basis of previous research, the committee believes several generalizations are possible.
- Productivity is the key driver of increased living standards. In turn, innovation, diffusion, or adoption of technology is the key driver of improvements in productivity. The most important technology of this era is IT. However, the existence of technology alone is not enough to enhance productivity. Effective use of technology typically requires a shift toward complementary skills profiles in the workforce and adaptation of business processes, organization of work, and institutional processes. These changes can be costly and take decades to play out.
- While productivity is still growing, it does not appear to be growing as rapidly as it did in the late 1990s and early 2000s. This may reflect a combination of factors, including mismeasurement of some of the benefits, reduced dynamism, the inherent lags often associated with the implementation of new technologies, or secular stagnation.
- Technology has transformed employment by automating some tasks and creating the need for new ones, a trend that is likely to continue. As a result, some entire occupations may become obsolete, and new occupations will come into being. Employment will shift from one occupation to another, while some occupations will simply experience changes in their required skill sets. While the share of people working has declined over the past 20 years, shifts within and across occupations will likely be much more economically significant than changes in the overall level of employment.
- New computerized technologies do appear to have contributed to increased income inequality and are likely to continue to do so as long as they replace skills and tasks historically associated with low-wage or middle-wage occupations. The jobs that remain tend to require more abstract, cognitive skills, or they provide personal services that are not currently economically valued. These differences will tend to be mainly reflected in wages and incomes, although they may also show up to some extent in hours worked and overall employment as well.