Skip to Content

U.S. Productivity Growth: Miserable Or Just Mismeasured

Theories on why the GDP still hasn't lived up to expectations

Person working in an automobile factory

Rather than a productivity slowdown, we may be in an early phase of significant efficiencies in industry, spurred by modern innovations in robotics and artificial intelligence. However, these efficiencies may not yet be reflected in the numbers because agencies that compile the data use methods more suited to an earlier era.

After the global financial crisis of 2008, many economists projected that the United States economy would steadily recover, eventually returning to pre-crisis strength, or at least close to it, within a matter of years. Yet, almost a decade later, the current state of the economy continues to defy expectations, falling short of a full recovery by a long shot. Indeed, real gross domestic product (GDP), a standard measure of economic strength, may've improved significantly from -2.8% in 2009, yet it was still just 1.6% in 2016 and 1.4% in the first quarter of 2017, well short of the 2.5%-3.0% believed to reflect a healthy economy.1

Why has the recovery failed to fulfill expectations? One popular explanation is that productivity, an engine of the economy, has itself exhibited only sputtering growth for more than a decade (see Exhibit 1). But has it? Two main theories have emerged to explain the shortfall in productivity numbers — and their rationales couldn't be more different.

Graph representing productivity change in business and manufacturing sectors

The first theory proposes that the computer age, whose innovations helped fuel the last great period of productivity growth in the U.S., from roughly the late 1980s to the early 2000s, has been running on fumes for some time now, and that we are actually in the midst of an extended slowdown with almost no end in sight. The second explanation maintains that U.S. productivity, far from stagnant, is actually improving, but official productivity numbers do not reflect this progress because they fail to incorporate a range of efficiencies stemming from recent technological innovations. In short, it's either a case of anemic productivity growth or a measurement problem. So which is it?

Are We Less Productive?

According to one school of thought whose members include economist Robert Gordon, a professor at Northwestern University, we are in a long-term productivity slump. In their view, we’ve already seen most of the economic stimulus afforded by the widespread integration of the personal computer (PC). Essentially, we are still using the same software and computers that we were 10 years ago, their thinking goes, so it’s hardly a surprise that productivity growth has been depressed. We may have to wait for the next great innovation to see significant growth again, they say, but there is little on the horizon that might do the trick. Gordon and others reckon that we saw something similar around 1970. The benefits of older technological advances like electricity and the internal combustion engine, and more recent ones such as air conditioning and the U.S. interstate system, had by then all run their course, they say, and productivity entered a slump that essentially lasted until the late 1980s, when PCs became widespread.2 In short, low productivity growth may be the new normal.

Are our Measurements Off?

Rather than a slowdown, economists such as Joel Mokyr, also a professor at Northwestern University, think we are in the early phases of significant industry efficiencies, spurred by modern innovations in robotics, artificial intelligence (AI), natural gas and oil extraction, and other developing fields. However, these efficiencies are not fully reflected in the numbers, Mokyr and others believe, because the Bureau of Labor Statistics and other agencies that compile the data use methods more suited to an earlier era.

We tend to share such views, believing that while productivity is a big part of economic growth, in the last 10 years we have seen very little growth reflected in the actual statistics; and rather than a long-term slowdown in growth, we think there is a sizable problem with how growth is measured. Computers and software helped improve productivity starting about 25 years ago. But, unlike Gordon and his cohorts, we believe that kind of technological advantage is ongoing in areas that have developed only recently, and may provide production efficiency for years to come — although they may be overlooked in the data.


One area that may be susceptible to undercounting is robotics. Programmable machines appeared on automotive assembly lines in the 1970s and now can be found in almost every industry, from healthcare and agriculture to electronics and construction. What’s fairly new is a robot’s ability to basically learn on the go, through AI. It’s a feature that has the potential to exponentially increase robots’ efficiency, accuracy and safety, even as they become more affordable. AI could bring further efficiencies to a variety of areas of industry, including electrical grids, railroads, online search engines, shopping websites and investment practices. It’s also likely to be an integral part of autonomous cars and trucks, helping to create significant potential savings for consumers and businesses through fewer collisions and injuries, lower fuel consumption and less time wasted looking for parking.

Image representing pros and cons of robot workforce

A widely used definition of productivity is GDP per hours worked — usually considered to be human hours. Yet, while overall unemployment numbers have dropped significantly, from 9.8% in January 2010 to 4.8% in January 2017, according to the Bureau of Labor Statistics, humans have steadily been losing jobs to machines for some time now. Ball State University estimated in 2015 that of all U.S. manufacturing jobs eliminated between 2000 and 2010, 87% were lost due to automation. Meanwhile, PricewaterhouseCoopers LLC calculated in 2017 that the U.S. could see 38% of all jobs lost to technology by the early 2030s. We believe that in this transition from human labor to robot labor lies a key area where a more accurate way of measuring productivity growth may be needed.

Consumer Surplus

Another reason we think there is a sizable measurement problem is that while many technological advancements have helped industries, they have also benefitted consumers. Consumers may have to pay a fee for Internet access, for example, but they then have direct access to countless digital services at little or no additional cost. Consider communication: There was a time when calling a relative living across the world meant you could soon expect a large bill from your service provider. Nowadays, you can chat for hours and pay next to nothing.

This evolution in communication and other services such as online shopping are part of what’s sometimes called consumer surplus — in other words, when they’re spending less for vital services or products, consumers may have more money to spend elsewhere. We think that it’s an element of efficiency that does not yet appear to be included in productivity data — but perhaps it should be.

The Future

There are major concerns as the wave of robotics, AI, and automation gains speed. The adoption rate of these new technologies may be quicker than in prior periods of change, which may add to societal stresses, especially if the existing workforce is unable to adapt their skills to keep up with the current pace of innovation. Indeed, a paradox may result, by which the considerable displacement of workers, due to the efforts by businesses to increase efficiency and maximize profits, may also depress growth of aggregate demand — after all, robots have no use for restaurants, clothing, or entertainment. Finally, these trends may act as a sustained tailwind for the recent rise of populist movements and contemporary approaches to reduce inequality through schemes such as universal basic income or so-called social dividends.

At the same time, with the current U.S. unemployment rate at historically low levels, an inflection point for productivity may be at hand. With unemployment so low, businesses that are unable to find workers, or are having to pay a lot more to entice them, could finally make the productivity-enhancing investments that enable them to produce more for each employee in their payroll.3

Ultimately, we think there is room for optimism. Even as people lose jobs, technology could open up a whole range of new ones. We might not know yet what they are, but that’s not necessarily a bad thing. Something similar to this happened at the start of the Industrial Revolution. While some groups in Britain — the so-called Luddites, for instance — opposed the use of machines in the workplace because they put people out of work, the revolution began a cultural transformation that, in a sense, we are still enjoying. In the long run, it created a once-inconceivable range of jobs, as well as improvements in our lifestyle, health and longevity, all of which have long been a normal part of modern life. It’s that kind of potential that is almost certainly not reflected in the official numbers. Assuming that one day that potential is in fact reflected, productivity data will, we believe, present a more accurate picture of growth.

Related Insights

Back to Top