What’s more, the opportunity gap is not just bad for the economy. It is bad for democracy, too. College-educated people ages 20–25 are much more likely to engage in activities such as discussing politics, contacting public officials, voluntary work, and so on. More than twice as many people with no more than a high school education are completely detached from all forms of civic life, compared to those who went to college. And in terms of democratic participation, those who went to college are between two to three times more likely to vote in national elections.26 Perhaps more worryingly still, as the political scientists Kay Schlozman, Sidney Verba, and Henry Brady demonstrate, political engagement has also become increasingly intergenerational, so that children tend to inherit their parents’ degree of political participation. In other words, having well-educated or wealthy parents shapes not only children’s job prospects but also their level of engagement in the political sphere.27 This leads to a well-known dilemma. As Robert Dahl noted, “If you are deprived of an equal voice in the government, the chances are quite high that your interests will not be given the same attention as the interests of those who do have a voice. If you have no voice, who will speak up for you?”28 Indeed, the political disenfranchisement of the unskilled, whose interests are no longer represented by mainstream politics, has made the discontents caused by automation harder to address (see chapter 11).
Retraining
How can we help those already in the labor market whose jobs are threatened by AI? Training people out of unemployment is a popular idea and a common response to rapid technological change. In the 1960s, when automation anxiety was at a high, retraining became a national priority. In his 1962 State of the Union address to Congress, President John F. Kennedy urged Congress to enact “the Manpower Training and Development Act [MDTA], to stop the waste of able-bodied men and women who want to work, but whose only skill has been replaced by a machine, or moved with a mill, or shut down with a mine.”29 The MDTA, which was signed into law on March 15, 1962, was the first federal manpower program, designed to train and retrain thousands of workers left behind by automation, though it was soon expanded to train people more broadly. In the period 1963–71, almost two million Americans enrolled in the program. How did they fare thereafter? The economist Orley Ashenfelter set out to evaluate the MDTA in 1978 and found that the answer was not straightforward. The program initially focused on the most easily retrained and was later geared toward more disadvantaged workers, many of whom dropped out. Though Ashenfelter found some evidence that workers earned higher incomes after participating in the program, he concluded that it was difficult to see if the benefits warranted the costs.30
Since the MDTA, federal policy makers have enacted an array of employment and training programs. Many of these are difficult to evaluate, because forgone earnings during training are hard to account for; data on costs for most training programs are sparse; and for the most part, studies trace outcomes only over a few years, meaning that we cannot know to what extent any effects on earnings fade out over time.31 In a recent review of the literature, the economists Burt Barnow and Jeffrey Smith concluded, “Taken together, the recent evidence presents a mixed but somewhat disheartening picture.”32 While the policy conclusion is not that we should dismiss the idea of retraining people later in life, it would be unwise to put too much faith in large-scale training efforts without proof of concept. We must pursue a strategy of trial and error and learn from practical experience what works where. And there are surely some interesting ideas out there that go beyond training programs. Lifelong Learning Accounts, for example, which already exist in Maine and Washington, provide tax incentives for people to invest in their own training and target low-income earners. Eligible citizens may contribute up to $2,500 annually and can receive a refundable tax credit equal to 50 percent of the first $500 contributed and 25 percent of the next $2,000 for any given tax year. Workers can then use these funds at different stages in their career for training purposes, in response to replacement or to advance their careers more generally. Before scaling up such efforts, however, they need to be carefully evaluated.
Major reforms to transform education and training might also be required more broadly. As Harvard University’s Clayton Christensen has forcefully argued, there is no particular reason why people with different learning requirements should have to conform to rigid academic programs that run for a specified period of time. The factory-based education model, which emerged in the aftermath of the Industrial Revolution, gradually expanded across many dimensions with more hours spent in school, more subjects covered, and more years of schooling. And that was a good thing. But if people have to consistently update their skills later in life, more flexible approaches to education will be needed. The learning process could be broken down so that instead of completing a standardized academic program, students could choose from a menu of skills and competencies they wish to acquire. Massive Open Online Courses (MOOCs), for example, can now be used to provide modularized education for people wanting to update their skills. And people can complete courses at their own pace.
Wage Insurance
It must also be acknowledged that retraining is not going to be the answer for everyone. People who experience dislocation late in life and see their skills rendered redundant might find it easier to take on a low-skilled job, even if it pays less. As noted above, displaced worker studies consistently show that many end up in jobs that pay less well than the jobs they previously held, and this is especially true of older people. Retraining and unemployment insurance do little to help displaced workers whose new jobs mean a significant pay cut. However, wage insurance—which compensates workers if they are forced to move to a job with a lower salary—would help ensure that fewer people are left worse off by automation. And it would make unskilled work pay more, relative to joblessness, which would likely reduce nonworking rates among the unskilled (see chapter 9). In America, wage insurance currently exists as a component of Trade Adjustment Assistance, a federal program designed to reduce the negative effects of imports felt by workers in some sectors. But the program is restricted to workers over fifty who do not earn over $50,000 a year. At the very least, it should be expanded to cover other sources of dislocation, like automation, which can lead to a permanent drop in people’s income. In the words of the economist Robert LaLonde, “Whereas private markets offer insurance for storms and fire, no such insurance is available when a middle-aged worker loses a job and suffers a permanent drop in wages. There is a market failure here, and government should correct it.”33
Tax Credits
In the popular press, universal basic income (UBI) has become a widely discussed way of limiting individual losses resulting from automation and deindustrialization. Of course, there are arguments in favor of UBI that have nothing to do with technological change, but this is not the place to dwell on them. The question here is whether it provides a good way of addressing the discontents brought about by the rise of the robots. In essence, UBI—which is closely tied to Milton Friedman’s old idea of a negative income tax—would give people a minimum income regardless of whether they worked or not. They could then earn additional income if they decided to work. The way it was originally conceived, UBI would replace other existing welfare programs. If introduced this way, the downside is that it would increase inequality, unless people were willing to accept a significant increase in tax rates. Because existing welfare programs are designed to help those in need, at the lower end of the income distribution, UBI (which, as the name suggests, would be received by everyone) would effectively redistribute income back to those at the higher end. But more fundamentally, the welfare state emerged the way it did because most citizens felt uncomfortable transferring resources to those not in need.34 UBI, in other words, would require a significant shift in attitudes and politics, which is highly unlikely given the sharpening economic and political divide in the past few decades. Growing economic segregation has meant that people rarely have firsthand
knowledge of the realities other people face, which has led to diminishing cross-class loyalty (see chapter 11).
A change in attitude might come if people are faced with a serious threat of AI-driven mass unemployment. But for the time being, there is little to suggest that widespread joblessness is imminent. As discussed above, AI is a long way from being able to replace workers in all domains, and the new technologies on the horizon will not all arrive at the same time, nor will they be adopted overnight. What’s more, as the historical record makes abundantly clear, fears that work will disappear have always turned out to be false alarm. If we think that this time is different, we should at least be able to explain why. Yet when we look at previous episodes of automation anxiety, like those of the 1830s, 1930s, 1960s, and 2010s, it is striking how much technology has advanced, but how little the debate has progressed. When I was researching this book, I struggled to find a single argument for why this time should be different that had not been made in earlier debates about automation.
Another false claim is that UBI is preferable to the welfare state because people do not like work. Back in 1970, for example, Walter Reuther, a vocal proponent of UBI, looked forward to the day when workers could spend less time on the job and devote themselves to music, painting, and scientific research. As we grew richer, he argued, we would work less and spend more time doing more self-fulfilling things. Yet most people find fulfillment and meaning in their work, whereas time-use studies show that the unskilled, who have seen their prospects in the labor market deteriorate, spend much of their time in front of the television, despite many studies showing that there is a negative correlation between television consumption and individual well-being.35 Contrary to the anthropologist David Graeber’s witty essay on “bullshit jobs,” in which he claims that most people spend their working lives doing work they perceive to be meaningless, large-scale survey evidence shows the exact opposite.36 And a wide range of studies across many countries and periods of time has consistently shown that people who work are happier than those who do not.37 As Ian Goldin puts it, “Individuals gain not only income, but meaning, status, skills, networks and friendships through work. Delinking income and work, while rewarding people for staying at home, is what lies behind social decay.”38
Thus, rather than subsidizing everyone, whether they work or not and regardless of their income, which is what UBI would do, it makes more sense to specifically target low-income groups in the labor market who have seen their earnings capacity fall. In contrast to UBI, which for the reasons outlined above remains controversial, such policies have broad support. In a recent op-ed in the Washington Post, for example, the economist Glenn Hubbard, who served as chairman of the Council of Economic Advisers in the George W. Bush administration, pointed out that “the economic-growth-lifts-all-boats camp needs to confront the question of what happens when growth alone fails to generate inclusion.”39 Hubbard argued for the introduction of a set of vouchers targeting low-income individuals, to be used for their training and their children’s education. He also argued for expanding the eligibility for the Earned Income Tax Credit (EITC).
The EITC is a negative income tax that is available only to working low-income individuals and already has a good track record. Scholars have found that those receiving it have seen dramatic increases in take-home income. They have found that its expansion helped put single parents back into the workforce. And the children of people receiving the credit have benefited enormously, in terms of both well-being and educational attainment, as reflected in higher math and reading scores as well as higher college enrollment rates.40 Thus, unsurprisingly, American states with more generous EITC policies also have higher rates of intergenerational mobility.41 As the sociologist Lane Kenworthy summarizes the research, “Government cash transfers of just a few thousand dollars could give a significant lifelong boost to the children who need it most.”42
For these reasons, the EITC should be expanded. First, making it more generous for low-income households with children would help level the playing field in terms of increasing the chance that children from disadvantaged backgrounds can move up in the ranks. Second, there is a good case for extending it to citizens without qualifying children, for whom the subsidy is currently minimal. As noted, the divide between college-educated people and others is likely to continue to grow. Thus, governments must make low-paying jobs pay more to improve incentives to work and reduce inequality, and the EITC or an equivalent provides a credible way of doing so. And if the past is any guidance, some of its cost will be offset by rising labor force participation among the unskilled.
Regulation
A different set of policies is needed to make it easier to move between jobs. Regulatory barriers to job switching are bad for productivity, wages, and equality. Of course, there are good reasons to insist on people like doctors and nurses to be licensed to practice. Yet the U.S. government practice of allowing only licensed practitioners to receive pay in a growing number of professions is worrying. To become a hair shampooer in Tennessee, for example, one must complete seventy days of training and pass two exams. Across America, the share of workers requiring a license to perform their jobs legally expanded from 10 percent in 1970 to almost 30 percent in 2008.43 Because obtaining a license often requires considerable investments in human capital and licensing fees, Americans who lose their jobs to machines are less likely to switch into licensed occupations, and workers in those jobs are less likely to switch out. Licensing requirements also often vary considerably across states—and even more so across countries—meaning that those in licensed jobs frequently have to make additional investments in obtaining a license when they move. Hence, unsurprisingly, economists have found that places with more people in licensed occupations also have higher rates of joblessness.44
In addition, the use of noncompete clauses—in which the employee agrees not to take a similar job at a competing firm for a prespecified period of time after leaving their current firm—has increased in many American states, providing another hurdle for engineers, scientists, and professionals seeking to move to expanding firms. The ability to change jobs easily is frequently mentioned as a prime reason for the success of Silicon Valley. Indeed, the departure of Gordon Moore and Robert Noyce from Fairchild Semiconductor to found Intel in 1968 was a critical moment in the area’s history. As is well known, worker mobility is much higher in the computer industry in California in general, and in Silicon Valley in particular, than elsewhere in America.45 One widely accepted explanation credits the California Civil Code of 1872, which outlawed all covenants in employment contracts—thus ensuring that Moore and Noyce could set up Intel after leaving Fairchild.46
The economist Steven Klepper has found that the same sort of dynamism underpinned the success of Detroit’s automobile industry in its heyday. In this regard, the rise of Detroit had much in common with the rise of Silicon Valley.47 Noncompete clauses were long prohibited in Michigan, too. But the Michigan Antitrust Reform Act of 1985, repealed Michigan’s—and thus Detroit’s—ban on the enforcement of such clauses. My research with Thor Berger shows that the technological dynamism of Michigan declined in response, as fewer workers shifted into new computer-related jobs relative to those in other similar states that did not see any change in legislation.48 In other words, the repeal exacerbated the decline of Detroit as an innovation hub. Thus, while it is unclear how much the rolling back of excessive occupational licensing practices and the abandonment of noncompete clauses would help, it is certainly worth finding out.
Relocation
The computer revolution has been a double-edged sword for American cities (see chapter 10). Cities with skilled populations were better able to take advantage of skill-intensive computer technologies and have subsequently prospered. In contrast, many of America’s problems are concentrated geographically, where middle-class jobs have been replaced by robots. Looking forward, even if new and improved substitutes for face-to-face interactions are developed, they ca
nnot substitute for spontaneous encounters that require physical proximity. Any digital communication must always be planned on at least one end, which means that the type of random interactions that occur in a workplace cannot happen at distance. Rather, the value of proximity will probably increase as AI makes production more skill intensive. Thus, the curse of geography is likely to intensify.
Historically, migration was the mechanism by which cities adjusted to trade and technology shocks. Workers moved to areas where new industries, spawned by the Second Industrial Revolution, created an abundance of well-paying, semiskilled manufacturing jobs. In the Great Migration, millions of Americans left the South for flourishing smokestack cities like Chicago and Buffalo. And agricultural workers left the farms for booming cities like Pittsburgh and Detroit. More and more people moving to higher-productivity areas served to equalize incomes across regions. But migration is no longer the equalizer it once was. While symbolic analysts are still highly mobile, the unskilled have become less likely to migrate since the dawn of the computer revolution (see chapter 10). One reason might be financial. Even if skilled cities provide better employment opportunities, moving is an investment that requires liquidity up front. Thus, as Enrico Moretti has convincingly argued, there is a case for subsidizing relocation.49 Mobility vouchers could pay for themselves by shifting the unemployed into paid employment elsewhere, while serving to equalize incomes across space. Some will argue that mobility vouchers might serve to accelerate the exodus from communities in decline, leaving parts of America in an even more dire state, but even those who stayed put would likely benefit in terms of having a better chance of finding a job.
The Technology Trap Page 40