When the Factory Closes, the Town Keeps Paying the Bill

Neil Irwin's June 2018 piece for The New York Times opens with a simple conditional: if artificial intelligence and robotics make a large share of today's jobs obsolete, what should Washington do? The "if" matters. At the time of writing, the unemployment rate sat at its lowest level in 18 years, and the economy's actual complaint was that productivity was too weak, not too strong.

So the debate is about a future that may never arrive. We think that is a reason to take it seriously rather than to shrug. Policy ideas take years to design and pass, and a labor market shock does not wait for Congress to finish arguing.

The article closes with a lesson from the recent past. Globalization and automation hollowed out American manufacturing from the 1980s through the early 2000s, and millions of factory workers lost their jobs. The damage to those communities is still visible, and it feeds many of the social and economic problems people argue about today.

Now suppose a similar wave reaches service workers: clerks, processors, analysts, drivers. The service sector employs far more Americans than factories ever did, which makes the stakes larger. Whether the wave comes or not, the question of who absorbs the cost of a technological transition is one investors should care about, because markets, tax rules and corporate margins all depend on the answer.

Five Policy Ideas That Sit Between Doing Nothing and Universal Basic Income

Universal basic income is the headline-grabbing answer, a monthly payment to every citizen large enough to cover basic needs. It draws support from free-market libertarians and socialists alike, which is unusual in itself. It is also expensive and politically hard.

Irwin points to a quieter set of proposals from economist Mark Paul at the Roosevelt Institute, a liberal think tank. His paper, "Don't Fear the Robots," argues that several modest steps, none very radical alone, could spread the gains from technology widely. First is a firmer commitment by the Federal Reserve to a legally mandated maximum employment goal, even if that means tolerating a little more inflation risk.

Second is a rewrite of intellectual property law so that patent and trademark holders enjoy shorter monopolies. Over time, that would shift more of the payoff from innovation toward labor and away from capital. Third is work-sharing, modeled on German programs that held unemployment down through downturns.

The arithmetic of work-sharing is easy to follow. If new technology means a company needs 20 percent less labor, it can lay off one worker in five, or it can trim everyone's hours by 20 percent. Society keeps its employment, workers keep their attachments to the firm, and the company keeps its trained staff.

Fourth and fifth are publicly funded higher education and training, and portable benefits. The assumption behind the whole bundle is that technology will change what work the economy needs rather than eliminate the need for work. That echoes the earlier shifts from farm to factory and from factory to office.

A Rare Overlap Between the Left, Consultants and Conservatives

What surprised us most in the article is how much these ideas overlap across ideological lines. The McKinsey Global Institute has published extensive analysis suggesting that information technology and robotics endanger millions of jobs in the coming decade. Its answer leans heavily on subsidized education and training.

Susan Lund, a McKinsey partner, floats the idea of lifelong learning accounts, funded by government or by employers. The goal would be to let a worker take a two-month leave to study and keep pace with change. McKinsey also recommends making health insurance and retirement benefits portable, so that contractors and frequent job changers do not lose stability each time they move.

Conservatives are warier of an activist government, but Michael Strain of the American Enterprise Institute is open to flexibility. His worry is a bifurcated labor market: workers with advanced skills earn more, while those without them watch technology erode demand for their services and their pay. If market wages for large groups fall well below what people would be willing to work for, he argues, much larger earnings subsidies may be justified.

An expanded earned-income tax credit is the obvious vehicle. It raises the effective pay of low-wage jobs so that taking them beats staying home. Strain also pushes back on the left, noting that a higher minimum wage could backfire if technology is already pushing market wages down for some workers.

The agreement is partial, and each camp frames its goals differently. Still, education, portability and wage subsidies show up on lists from very different authors, which suggests they are the ideas most likely to survive a real legislative fight.

Why Policy Choices Move Corporate Profits

Each of these proposals has a price tag and a payer. Training accounts funded by employers raise labor costs. Shorter patent terms cut the returns on innovation. A Fed that tolerates more inflation to protect jobs changes the path of interest rates. None of that is neutral for company earnings or for asset prices.

Take the patent idea. If monopoly periods shrink, the profits that technology leaders earn on a breakthrough fade sooner, and more of the gain goes to wages or to competitors. Investors who assume that a dominant technology position earns outsized returns for decades would have to shorten their time horizon.

Take the Fed idea. A stronger tilt toward maximum employment, with more tolerance for inflation, would likely mean policy rates staying lower for longer than a strict price-stability approach would suggest. That helps borrowers and equity valuations, and it hurts holders of long fixed-rate bonds whose real returns get eroded.

The labor-cost point matters most for companies that employ large numbers of people in routine, automatable tasks. If lifelong learning accounts or portable benefits become mandatory, the bill lands on the employer. Firms that have already automated heavily face a smaller increase; firms with big, hands-on payrolls face a larger one.

All of this is speculative today, which is Irwin's own caveat. We would treat it as a scenario to watch, not a forecast to trade on.

What It Could Mean for Citigroup, Goldman Sachs and Morgan Stanley

The article does not mention any company, but the assets we are tracking here, Citigroup (C), Goldman Sachs (GS) and Morgan Stanley (MS), are a useful test case. Big banks employ enormous numbers of people in work that automation could reach: operations, compliance checks, trade processing, back-office reconciliation. They are also heavily exposed to interest rate policy.

Start with the risk side. A bank with a large operations workforce could see two pressures at once: technology that lets it cut headcount, and policy that makes cutting costlier through work-sharing expectations, retraining mandates or benefit rules. Citigroup, with its global retail and operations footprint, has the most people doing process-heavy work of the three. That makes it the most sensitive to rules about how displaced workers are treated.

Now the opportunity side. Goldman Sachs and Morgan Stanley earn more of their revenue from trading, advice and wealth management, where the output per employee is high and the share of payroll in routine tasks is smaller. If automation lowers their processing costs, the savings fall to margins faster. Morgan Stanley's wealth management business could also benefit if workers with portable retirement accounts and training funds become a bigger pool of clients.

Here is a rough scenario with numbers. Suppose a bank spends 55 percent of revenue on compensation and that a retraining mandate adds 1 percent to payroll. That is about 0.55 percent of revenue, which on a business earning a 25 percent pre-tax margin trims profit by roughly 2 percent. Painful but manageable. If the same bank had already automated 10 percent of its routine work, the extra cost would be partly offset.

Our point is not that one bank wins and another loses. It is that policy responses to automation change cost structures, and a bank's mix of routine and high-skill work tells you which direction the effect leans.

Putting the Idea to Work in a Portfolio

Because nothing here has been legislated, the practical step is to build a watch list, not to rebalance. We would track three developments: proposals for lifelong learning accounts or portable benefits in Congress, any change in how the Fed describes its employment mandate, and debate over patent term reform. Any of them gaining traction would change the cost picture for large employers.

For timing, the better signal is when proposals attach to a funding mechanism. Ideas that float in think tank papers move nothing. A bill with a payroll tax or an employer mandate attached would move estimates of labor costs, and that is when we would revisit assumptions about the banks above.

Implementation has its own difficulties. Headcount data for banks is reported unevenly, and a firm's automation progress is rarely disclosed in a way you can compare across peers. We would lean on annual reports for compensation as a share of revenue, and on management commentary about efficiency programs, and treat the comparison as approximate.

A second difficulty is concentration. Citigroup, Goldman Sachs and Morgan Stanley move together with the broader financial sector and with interest rates. Owning all three does not diversify away a rate or regulation shock, so a position built on this theme should stay small relative to the whole portfolio.

Finally, remember the opposite possibility. If productivity stays weak, as it was in 2018, the robots may take years to matter and the policy debate may fade. Betting heavily on a disruption that does not arrive costs real returns.

What We Would Do With This

The core of the article is that a tech-driven labor shock, if it comes, will probably be met with a mix of moderate measures rather than one grand scheme. Training, portable benefits, work-sharing, wage subsidies and a firmer employment mandate each have supporters on more than one side of the aisle. That makes them more realistic than universal basic income.

For investors, the consequence is a gradual change in who pays for transitions. Employers with large routine workforces may carry more of the cost, and companies with long patent monopolies may see them shrink. Interest rates may stay lower if the Fed leans toward employment.

Here are the steps we would take. First, look up compensation as a share of revenue for the banks you own and compare it across Citigroup, Goldman Sachs and Morgan Stanley. Second, note which of them describe automation programs in their filings and how much they say they have saved.

Third, set a calendar reminder to check for legislation that attaches funding to training or portable benefit ideas, since that is the point at which the cost becomes concrete. Fourth, keep any position tied to this theme small and avoid assuming the three banks diversify each other.

Last, hold on to the lesson in Irwin's final paragraph. The factory closures of the 1980s through the early 2000s showed how long communities can suffer when a transition is left to run its course. Paying attention early, as citizens and as investors, costs little.