For all the attention given by politicians in both parties to trade and globalization as a threat to the middle class, by far the most important, past and future, is automation — the replacement of tasks performed by human beings by machines, increasingly enabled by software, especially software that “thinks” or learns on its own, or “artificial intelligence” (AI).
Last year, the Pew Center reported that 72 percent of Americans said they were worried about the impact of automation on jobs – this, despite the unemployment rate at the time being at a twenty-year low (and even lower since). The fears of a jobless dystopia are misplaced. Despite cyclical ups and downs, economies generate new jobs to replace the old ones disrupted by technological change (or by “globalization,” or increased competition from firms in other countries). There are multiple reasons for this, as outlined in my new paper “Meeting the Automation Challenge to The Middle Class and the American Project,” but the simplest is that the savings enabled by automation do not disappear into thin air. They get spent on other goods and services, which creates more jobs elsewhere in the economy.
Worry about wages, not jobs
The real concern about automation is the wages those new jobs pay, or more precisely, on the distribution of those wages. If, as automation proceeds, it continues to favor those especially able to use it at the expense of those who don’t have those skills – in economists’ language continues to be “skills biased” – it will further widen income inequality. More inequality today erodes opportunities tomorrow for the next generation (and the ones thereafter) to realize the American Dream. To use an analogy favored by many labor economists, especially Brookings Senior Fellow Isabel Sawhill, if the rungs on a ladder are driven further apart, it makes it that much more difficult for many to climb the ladder, especially those born at or near the bottom. What, then, should be done? There is wide support for slowing automation down. According to the same Pew survey reporting high levels of anxiety about automation, 87 percent of Americans say want a human in “driverless” cars, just in case. Nearly the same level, 85 percent, favors limiting machines to performing primarily those jobs that are dangerous or unhealthy for humans. Technological advances don’t work in so discriminating a fashion, however. To be sure, robots can defuse bombs at essentially zero risk to human life and are to be vastly preferred for this task – and other dangerous or dirty jobs – than people. There will always be a market for such innovation. It is safe to count on it continuing. But most technological advances will be used to reduce the cost of making goods and delivering services, a process that is and will continue to be invisible to consumers. Elected or regulatory officials will not have an easy time to singling out and slowing down or preventing every such improvement, nor should they. If anything, given the slow pace of productivity growth and thus wage growth since the Great Recession, citizens should want faster automation in the future, because it is key to realizing more rapid economic growth. Here are four ways government can help smooth the transition of workers from jobs that automation will displace to those that automation either will create directly or enable.
Cure #1: Maintain a High-Pressure Economy
Over four decades ago, my first full-time employer and mentor, the late Arthur Okun, authored a pioneering study in the Brookings Papers on Economic Activity showing that workers have the greatest opportunities to realize wage gains at their current jobs and to move to ones with even higher wages and benefits when the economy is “running hot” – that is, at or close to full employment. While the U.S. economy has undergone much change in the intervening years, Okun’s proposition remains as true today as when he outlined it, with two notable differences. First, the long-term or potential growth rate of the U.S. economy since the Great Recession – at roughly 2 percent according to the Congressional Budget Office and even the more optimistic (but less realistic) 3 percent rate projected by the Trump Administration — has fallen well below the near 4 percent pace of the post-war years until 1973, when Okun wrote his article. Slower GDP growth, especially slower productivity growth, also means slower growth in workers’ compensation. Second, whereas in the early 1970s when policy makers thought they had two instruments – fiscal and monetary policy – to keep the economy operating at “high pressure” – today and in the future that may not be the case. With the federal deficit already approaching 4 percent of GDP when the economy is operating very close to full employment, policy makers may be more reluctant to use fiscal stimulus to counter future downturns if, as CBO and virtually every economist outside the Administration projects, the deficit only goes higher. That leaves only the Federal Reserve to keep the economy humming while also trying to keep inflation in check, a difficult balance to maintain. In a recent op-ed in the Wall Street Journal, Harvard’s Martin Feldstein nonetheless argued that the Fed cannot and indeed should not “save jobs from AI and robots,” and indeed urged that attempting to maintain full employment should be dropped from the central bank’s mandate. With all due respect to Feldstein, who is a great economist, this is precisely the wrong advice, unless he can use his powers of persuasion to convince his fellow Republicans and willing Democrats to compromise on a comprehensive deficit reduction plan – involving both entitlement reforms and tax increases – that would restore some freedom for fiscal policy to fight future economic downturns. Failing that outcome, which unfortunately looks unlikely for the foreseeable future, it would be a severe mistake to deprive the Fed of the ability to maintain employment. Otherwise, we will be condemned to suffer the adverse economic, health and political consequences of the jobless dystopia that the automation pessimists claim is in our future. One way for the Fed to prevent this outcome is to target an economy-wide price level, as advocated by Brookings Senior Fellow and former Fed Chairman Ben Bernanke. Such a policy would make it easier for the Fed to keep the economy close to full employment even when the inflation rate may temporarily rise about the central bank’s current informal target of two percent. This is because a price level target would allow the Fed to take account of and make up for low inflation rates in earlier years.
Cure #2: Insure Wages
Spanning more than three decades, I and multiple co-authors have advocated wage insurance for displaced workers, initially just those victims of import competition but later for all displaced workers for any reason. The idea is to compensate displaced workers for some portion of their wage loss in accepting a new lower-paying job. Most wage insurance proposals set the compensation fraction at 50 percent, and trigger payments only when workers take a new job and stop at two years from when they lost their job. The ticking clock nature of when the compensation kicks in would give workers incentives not to remain unemployed looking for the “ideal” job that may never materialize. The compensation should be limited in time, say two years, and be capped. For years I thought the annual cap should be set at $10,000 but given the severity of job and income loss during and since the Great Recession, a higher cap, such as $20,000 might be in order. In 2016, CBO reportedly estimated the annual cost of wage insurance at about $3 billion annually, an amount that could easily be funded, if not part of a larger adjustment package, in various ways, such as through an increase in the federal unemployment insurance tax. In his last State of the Union address, President Obama briefly voiced his support for a universal wage insurance program – one available to workers displaced for any reason – but given President Trump’s continuing efforts to reverse anything that Obama implemented or endorsed, there seems little chance now that he would support the idea. That shouldn’t kill it. Those Republicans and Democrats wanting to preserve freer trade and avoid a “techlash” against automation-driven productivity improvement should be interested in addressing workers’ legitimate concerns about displacement. Wage insurance had bipartisan support in 2000 in the recommendations of Commission on Trade Deficit. The case for wage insurance is just as strong now — maybe stronger given current fears of automation — than it was then.
#3 Finance Lifetime Learning
Wage insurance is not a silver bullet to middle class anxieties. It is meant primarily to cushion the short-term pain of job loss and to ease transition to other jobs. In this respect, wage insurance is analogous to government assistance for apprenticeships, in principle a good idea and one for which additional government support also is appropriate, but a program likely to be limited to helping younger workers get their first, good job. Neither wage insurance nor apprenticeships are designed to facilitate the efforts of workers already in the labor force to move to new, potentially higher paying and ideally more secure careers. For that, more than on-the-job training is required: in community or even four-year colleges, or for-profit institutes offering certificates for various kinds of jobs. At Brookings, Darrell West has written thoughtfully about how government can provide financial assistance for this type of continuous skills upgrading, in the form of new tax-deferred accounts (like IRAs) for retraining, with dollar-for-dollar government matches, up to $1000 per year. I believe the government can and should do more, since many individuals may not be able or have the foresight to save up enough in their “Lifetime Training Accounts” (LTAs) to pay for training when they may most need it – when they’re out of a job. As I have laid out elsewhere with Council on Foreign Relations Senior Fellow Ted Alden, government can solve this problem by extending “income contingent” loans, which so far have been made available for those attending college, to a wide range of training opportunities that individuals may want, or need, throughout their working lives. Income contingent loans, as their label implies, are not repaid by regular repayments of some principal, with interest, but rather through repayments calculated on a percentage of the borrowers’ income (more money borrowed, the higher that percentage, up to a ceiling). To prevent borrowers from being cheated by fly-by-night training centers, loan funds should be available for appropriately certified providers and/or for institutions that provide current information about their placement records post-training (with strong penalties for false reporting). To keep loan repayments manageable, the loan balances in these LTAs would have a lifetime ceiling (such as $50,000-100,000, inflation adjusted). The budgetary cost for these lifetime training loan accounts should be limited. That is because only any subsidies related to the loans are counted as federal expenditures, not the loan funds themselves. To minimize the subsidy expenditure, the lifetime loan program can be structured so that income-contingent repayments from highly successful borrowers, who will repay more than they borrowed (up to a cap) largely, if not entirely, offset the costs for those borrowers, who for any number of reasons, are unable to fully repay their loans.
Cure #4: Target Distressed Places
Wage insurance and lifetime training accounts, as needed as they are, cannot help all workers who want to help themselves but are committed for family and other reasons to staying in parts of the country that are being or threaten to be left behind in our modern economy. The 2017 tax bill implicitly recognized this by offering generous capital gains tax incentives for investments in a limited number of high poverty “Opportunity Zones” designated by the states. But as Brookings Senior Fellow Adam Looney notes, the 2017 bill may simply promote real estate investments that increase land values in and thus “gentrify” poorer areas, doing little or nothing for the lifetime earnings prospects of poor people who live in them. An alternative, and potentially more cost-effective approach, is to subsidize employment or jobs in lagging areas, as Benjamin Austin, Edward Glaeser and Lawrence Summers suggested in a paper presented earlier this year at Brookings, to appear in the next issue of the Brookings Papers on Economic Activity. Although the authors do not offer a detailed plan for doing this, a tax credit for new hires (so as not to provide windfall benefits on existing jobs) could be aimed at areas with abnormally high unemployment or poverty. Subsidizing jobs in such locations has the advantage of inducing investment along with it, which is not necessarily the case with targeted investment incentives only.
Worries about the job displacement effect of future automation nonetheless are so great that, according to the Pew survey already cited, healthy majorities (in the 60 percent range) support having the government guarantee a universal basic income (UBI) or establish a national service program to be a last resort source of employment. Apart from being unnecessary for reasons noted at the outset of this blog, neither idea is fiscally realistic, given an already large and growing federal deficit, but also due to pressing demands for added federal funding for other programs — for infrastructure and fixing Obamacare, to name just two obvious priorities – that would ease middle class anxieties. Assisting those who help themselves, through programs like wage insurance and financial assistance for lifetime training, will be far less costly, and once Americans realize the full fiscal implications, far more politically acceptable, than paying people who can work but choose not to or running a potentially huge federal jobs program. The middle class justifiably is worried about automation. Policymakers need to offer some practical solutions in response.  Those co-authors include, in alphabetical order: Martin Baily, Lael Brainard, Gary Burtless, Lori Kletzer, Robert Lawrence, and Robert Shapiro.