The Technology Trap

by Carl Benedikt Frey

  Housing and Zoning

  Another dilemma is that as skilled cities are becoming more attractive, rising housing prices makes them less affordable. To counteract this, the housing supply must be expanded where new jobs are being created. This will require getting rid of some zoning restrictions, such as minimum lot sizes, height limits, prohibitions on multifamily housing, lengthy permitting processes, and so on, which effectively cap the number of people who can live in thriving places. Because dynamic places like New York and the Bay Area have adopted stringent restrictions on new housing supply, they have effectively limited the number of workers who can participate in the growth created by tech industries. The consequence has been that tech companies find it more difficult to hire due to the rising cost of housing. But more importantly still, an unemployed unskilled worker in Flint who finds a job in Boston cannot afford to live there. As discussed above, the next wave of automation will render many low-skilled jobs redundant, but there are still a variety of in-person service jobs that remain exceedingly hard to automate. Those jobs, it stands to reason, will emerge in skilled cities, where people can afford such services.

  The combined effect of zoning restrictions has been slower economic growth, fewer jobs, lower wages, and higher inequality across the nation. Economists have estimated that in the absence of such restrictions on housing supply, the American economy would be 9 percent larger today, which would mean an additional $6,775 in annual income for the average American worker.50 Abolishing land use restrictions would also have welcome side effects. The breathtaking rise in wealth inequality that has been documented by Thomas Piketty stems almost entirely from housing.51 Inflated house prices due to land-use restrictions are surely part of the reason, and the abolition of those restrictions must therefore be part of the solution.52

  Removing barriers to the expansion and development of skilled cities would help social mobility, too. As Raj Chetty, Nathaniel Hendren, and Lawrence Katz have shown, a person who moves from Oakland to San Francisco at the age of nine will, as an adult, gain more than half of the income differential between the two locations.53 Because zoning restrictions are not distributed randomly but are much more prevalent in high-income cities and neighborhoods, they put people born into less affluent communities at a further disadvantage. Zoning, in other words, has priced lower-income families out of the places with more social capital and better schools.

  Another benefit would be more innovation. Children growing up in places with more inventors, who are thus more exposed to innovation in their early years, are much more likely to become inventors themselves. This, we know, also has an impact on the types of inventions that they are likely to produce. Those growing up in Silicon Valley are more likely to drive innovation in computing, while those who spend their early years in places specializing in medical devices, like Minneapolis, for example, are more likely to invent related technologies.54

  Connectivity

  Transportation infrastructure that linked high-paying labor markets to areas where housing is cheap would also allow more people to tap into strong local economies. Subways or high-speed rail that connected declining places (where jobs have dried up and housing is cheap) with expanding ones (where jobs are abundant and housing is expensive) would serve to level incomes across space. And such connections would boost local service economies in decline, as people spend large parts of their incomes locally. In this light, economists have pointed to the potential benefits of current efforts to connect California’s low-income cities—like Sacramento, Stockton, Modesto, and Fresno—with the Bay Area through high-speed rail.55 Many Californians could remain in Fresno, where housing is cheap, and commute to San Francisco for work.

  In the future, new transportation technologies could also be used to connect places that are much farther apart. Hyperloop technology, which uses a sealed system of tubes to allow people to travel free of air resistance or friction, offers the potential of reaching distant locations at staggering speeds. Hyperloop Transportation Technologies, for example, recently signed agreements with the Illinois Department of Transportation to examine the feasibility of connecting Cleveland and Chicago through a number of different corridors.56 The commute currently takes around 5.5 hours by car or 7.1 hours by public transportation in one direction. The Hyperloop, if successful, is expected to bring this commute down to twenty-eight minutes. All of a sudden, it could become feasible to commute long distances to work.

  Industrial Renewal

  A regrettably less promising approach has been the effort to revitalize cities in decline through place-based policies, which target local industry rather than individuals. It is true that some of these policies have been successful in attracting new jobs, but the costs of doing so have been significant. For example, distressed urban and rural areas that were designated empowerment zones in the 1990s increased local employment through grants, tax credits for business, and other benefits, but each new job is estimated to have cost over $100,000.57 And while large-scale projects to revive communities have generated sustained growth locally, these projects seems to have attracted resources at the expense of other places. The greatest push of this kind in American history was the Tennessee Valley Authority (TVA) Act of 1933, which became law in the midst of the Great Depression. The TVA set out to rapidly modernize the Tennessee Valley’s economy, harnessing promising new technologies like electricity to attract the manufacturing industry. The push included large-scale public infrastructure projects including dams, an extensive road network, and a 650-mile navigation canal. For the valley, the push was surely a good thing: the positive employment contributions were seen as late as the turn of the new millennium, when the area was still growing more rapidly than comparable areas—though the effects have begun to fade. Yet manufacturing jobs created in the valley were offset by losses elsewhere. This is a troubling finding, because federal and local governments are estimated to spend some $95 billion per year on place-based policy programs, which is much more than is spent on unemployment insurance.58

  Big pushes that focus on investment in physical capital are also likely to yield much lower benefits locally today, as manufacturing has become more automated. A more promising way forward for places that are falling behind is to divert resources to investment in human capital. Economists have shown that the presence of a college or university increases the supply of skilled workers not only by educating them, but also by attracting more college-educated people from elsewhere.59 For example, the Land-Grant College Act of 1862 (also known as the Morrill Act) established several land-grant universities and is estimated to have increased labor productivity by 57 percent over a period of eighty years.60 And needless to say, people take their human capital with them if they move. Physical capital stays put.

  Final Thoughts

  In the mid-nineteenth century, Karl Marx and Friedrich Engels predicted that continued mechanization would mean the continued impoverishment of the working class, just around the time when Britain finally escaped from Engels’s pause. They were right about the past: many Englishmen had been left worse off by the Industrial Revolution. However, they were wrong in thinking that continued progress would lead in the same direction. Like so many others, they were fooled by the mysterious force of technology.

  There have been long periods when things did not work out well for labor. But even those episodes came to an end. The thesis of this book is not that current economic trends must continue indefinitely. On the contrary, there are good reasons to be optimistic about an AI-induced productivity revival, which, besides making us richer on average, would help offset some of the negative effects replacing technologies have on parts of the labor force. But if history is any guide, that could take many years or even decades. And while it is possible that we are at the cusp of a wave of enabling technologies that could reinstate labor in new jobs more broadly, that is unlikely to provide much relief to people in the middle class unless they have the right skills. Even if we assume that AI will spawn gi
gantic new industries, as the automobile did a century ago, Henry Ford’s invention of the assembly line broke complex operations down into simple tasks that could be performed by a person with a fifth-grade education. For more than thirty years now, technological change has created few new jobs that do not require a college degree. In a world that is becoming increasingly technologically sophisticated, new jobs are unlikely to open up for those who would have flocked into the factories before the dawn of automation.

  The economic order that gave rise to a broad middle class has withered, along with the middle-class politics that rested on it. Until the Great Recession the pressures from automation on middle-income households were masked by subsidized credit, which counterbalanced falling wages among non-college-educated workers and left consumption broadly unaffected. The housing boom also meant that abundant construction jobs helped offset some of the job losses in manufacturing—until the burst of the housing bubble.61 In other words, the recession unmasked the steady decline in the wages of the middle class, which helps explain the relatively recent rise of populism.

  Looking forward, the divide between the winners and losers from automation can be expected to grow further. The next wave is not coming just for manufacturing jobs, but also for many unskilled jobs in transportation, retail, logistics, and construction. Thus, while there are good reasons to be optimistic about the long run, such optimism is only possible if we successfully manage the short-term dynamics. People who lose out to automation will quite rationally oppose it, and if they do, the short-term effects cannot be seen in isolation from the long run. In light of the long history of resistance to technology that threatens people’s skills and the recent backlash against globalization, automation cannot be seen as an inexorable fact of life. It is true that unlike the Luddites of the nineteenth century, people now have seen how in the twentieth century technology made everyone richer. As mechanization progressed during the first three-quarters of the twentieth century, wages rose at all levels. But if technology fails to lift all boats in the coming years, broad acceptance of technological change cannot be taken for granted. People have higher expectations than at the time of Engels’s pause. They have the right to vote. And they are already demanding change.

  No single government policy can address the full spectrum of societal challenges brought by automation. Regrettably, proposing apparently easy solutions to a complex set of problems may win elections in the short term, but reality catches up sooner or later. Moderate conservatives and liberals face a tricky balancing act, because exaggerating the effects of automation might prompt fears of mass unemployment and lead to the wrong policy responses, the growth of populist parties, and possibly a backlash against technology itself. At the same time, however, if governments gloss over the social costs of automation, their credibility will diminish. For a long time, governments chose to overlook the costs of globalization and focus on the benefits. Those benefits were indeed significant, but the failure to deal with the individual and societal costs ended up costing mainstream politics its credibility. Governments must avoid making the same mistake with automation. And the stakes could not be higher.

  Some readers might still think that we are entering a new era in which machines take all of the jobs, and of course, there is no way of knowing if that is true. But for now, there is little to suggest that this time is different: our current trajectories look exceedingly similar to those in the classic years of industrialization, and we all know what happened after that. Even assuming that this time is different, however, still means that the challenges ahead lie in the area of political economy, not in technology. In a world where technology creates few jobs and enormous wealth, the challenge is a distributional one. The bottom line is that regardless of what the future of technology holds, it is up to us to shape its economic and societal impact.

  ACKNOWLEDGMENTS

  If this book could be regarded as an invention, it would surely be a recombinant one. It draws upon a vast body of research to which numerous scholars have contributed. I guess my own journey to writing it began in my school years, when my father, Christopher, got back from a business trip with two new books for me. The first was Joel Mokyr’s The Lever of Riches. The second was Clayton Christensen’s The Innovator’s Dilemma. Their work showed me that long-term prosperity derives from technological innovation. But it also made abundantly clear that progress often comes with economic and societal disruption. My lifelong interest in the subject is thanks to my father.

  Over the past four years of writing, I have accumulated many debts. This book could not have been written without generous financial support from Citigroup. I’m especially indebted to Andrew Pitt and Robert Garlick at Citi, whose genuine intellectual curiosity made this project possible. Special thanks also go to Sarah Caro, my editor at Princeton University Press, for her guidance and many thoughtful comments. Chinchih Chen has done a fabulous job of providing diligent research assistance. And my long-standing friend Thor Berger has read many different versions of this manuscript, for which I’m enormously thankful. I’m also grateful to Ian Goldin, Logan Graham, Jane Humphries, Frank Levy, Jonas Ljungberg, Joel Mokyr, Michael Osborne, and Anil Prashar for reading all or part of this manuscript and providing invaluable comments.

  Above all, my family has long been gracefully supportive of my many professional preoccupations, including this one. They are the ones who have kept me sane.

  APPENDIX

  FIGURE 5

  Constructed following R. C. Allen, 2009b, “Engels’ Pause: Technical Change, Capital Accumulation, and Inequality in the British Industrial Revolution,” Explorations in Economic History 46 (4): 418–35, appendix I, using the sources below:

  Gross domestic product (GDP) factor cost estimate from C. H. Feinstein, 1998, “Pessimism Perpetuated: Real Wages and the Standard of Living in Britain during and after the Industrial Revolution,” Journal of Economic History 58 (3): 625–58; B. Mitchell, 1988, British Historical Statistics (Cambridge: Cambridge University Press), 837, for 1830–1900.

  Real output per capita from N. F. Crafts, 1987, “British Economic Growth, 1700–1850: Some Difficulties of Interpretation,” Explorations in Economic History 20 (4): 245–68.

  Average full-employment weekly earnings for the United Kingdom for 1770–1882 from Feinstein 1998, appendix table 1, 652–53; average full-employment weekly earnings for the United Kingdom for 1883–1900 from Feinstein, 1990, “New Estimates of Average Earnings in the United Kingdom,” Economic History Review 43 (4): 592–633.

  Cost of living index for 1770–1869 from R. C. Allen, 2007, “Pessimism Preserved: Real Wages in the British Industrial Revolution” (Working Paper 314, Department of Economics, Oxford University), appendix 1.

  Great Britain/United Kingdom cost of living index for 1870–1900 from C. H. Feinstein, 1991, “A New Look at the Cost of Living,” in New Perspectives on the Late Victorian Economy, edited by J. Foreman-Peck (Cambridge: Cambridge University Press), 151–79.

  I have converted the wage index for 1882 onward from Feinstein 1990, based on 1880–81, the benchmark year in C. H. Feinstein, 1998, “Pessimism Perpetuated: Real Wages and the Standard of Living in Britain during and after the Industrial Revolution,” Journal of Economic History 58 (3): 625–58. The nominal wage for 1770–1881 is derived from Feinstein 1998 and for 1882–1900, it is derived from Feinstein 1990.

  Following Allen 2009b, I used the growth rate of real output per capita from N. F. Crafts 1987, table 1, to extrapolate backward to 1770.

  All GDP, wage, and population data have been collected from R. Thomas and N. Dimsdale, 2016, “Three Centuries of Data–Version 3.0” (London: Bank of England), https://www.bankofengland.co.uk/statistics/research-datasets.

  FIGURE 9

  Constructed following R. J. Gordon, 2016, The Rise and Fall of American Growth: The U.S. Standard of Living since the Civil War (Princeton, NJ: Princeton University Press), figure 8-7, using the sources below:

  1929–2016 U.S. real
GDP data, 1870–2016 production workers’ hourly compensation (nominal dollars), and the 1870–1928 GDP deflator are collected from L. Johnston and S. H. Williamson, 2018, “What Was the U.S. GDP Then?,” http://www.measuringworth.org/usgdp//.

  1870–1929 nominal gross national product (GNP) is collected from N. S. Balke and R. J. Gordon, 1989, “The Estimation of Prewar Gross National Product: Methodology and New Evidence,” Journal of Political Economy 97 (1): 38–92, table 10.

  Total civilian man-hours for 1870–1947 are from J. W. Kendrick, 1961, Productivity Trends in the United States (Princeton, NJ: Princeton University Press), table A-X.

  Total civilian man-hours for 1948–66 are from J. W. Kendrick, 1973, Postwar Productivity Trends in the United States, 1948–1969 (Cambridge, MA: National Bureau of Economic Research [NBER] Books), table A-10.

  Total private average weekly hours of production and nonsupervisory employees’ data for the years 1967–75 are from Bureau of Labor Statistics, 2015. “Employment, Hours, and Earnings from the Current Employment Statistics Survey” (Washington, DC: U.S. Department of Labor).

  Average weekly hours at work in all industries and in nonagricultural industries for the years 1976–2016 are from Bureau of Labor Statistics, 2015, “Labor Force Statistics from the Current Population Survey” (Washington, DC: U.S. Department of Labor).

  FIGURE 14

  Constructed following B. Milanovic, 2016b, Global Inequality: A New Approach for the Age of Globalization (Cambridge, MA: Harvard University Press), figure 2-1, using the sources below: