A common theme in business is that automation will continue to sap U.S. employment in manufacturing even as productivity rises.

This week the New York Times‘ Louis Uchitelle surmised that “manufacturing is unlikely to be capable of producing a great deal of additional employment,” making the familiar argument:

Modern assembly-line machinery continues to eliminate jobs. Production has been increasing, but factories are doing this with fewer people.

From a post-World War II peak of 19.6 million workers in 1979, employment in manufacturing has declined to 12.5 million, according to the Bureau of Labor Statistics. Put another way, manufacturing has dropped from 23 percent of the total work force to nearly 8 percent. That happened even as factory output rose in total value to $2.154 trillion in 2018. America is making more goods and materials with fewer people.

But to make his case, Uchitelle shorts the argument. Manufacturing bottomed out in January of 2010 at 11.3 million jobs, two years after President Obama took office, but has since then recovered, adding 1.5 million to a total today of 12.8 million manufacturing jobs.

How, then, to square conventional wisdom — that American manufacturing is losing jobs to automation — with the positive trend the past six years?

Let’s back up. Clearly, Obama walked in to a buzzsaw: Manufacturing employment was in free-fall from 17.3 million jobs when President Bush took office in January 2000 to the low-mark in 2010. In all, the economy shed 6 million manufacturing jobs during the Bush presidency.

But from 2010 to 2016, Obama added 900,000 or so jobs, and President Trump’s on a faster pace, adding 480,000 or so jobs in 26 months of his presidency. I’ve pointed to growth in new manufacturing industries as a source for new employment, and clearly red-hot market sectors like food and beverage are piling up the numbers. Manufacturers are generally bullish on Trump’s tax policies and his efforts to reset trade policy with manufacturing powers like China. It’s also more economical to make things in the U.S. than in decades, and fabricators in growth industries like aerospace are finding that innovation is enabling more domestic production. The numbers would look even more impressive if companies were able to fill the half-million or so manufacturing job openings that bedevil recruiters.

It’s possible we’ve reached a peak and employment will head down again. Or, perhaps we’ve found a new equilibrium, where huge jobs losses to offshoring are a thing of the past, and steady growth in new industries will sustain moderate growth.

If the Trump administration is able to maintain it’s hot start, economists and pundits might even be forced to reassess the premise that American manufacturing employment is in irreversible decline. Today President Trump’s cheerleaders boast of his gains compared to Obama, which seems the wrong argument to make. If a Trump manufacturing surge continues at its current pace, to add a million or so jobs in four years, it could reshape the narrative of manufacturing in the U.S.

How likely is it?

The numbers are daunting. The economy added 2.9 million manufacturing jobs 40 years ago, from June 1975 to June 1979, a staggering number by today’s standards. There’s no period in the past four decades that comes close to this surge in manufacturing employment. A Trump recovery of a million manufacturing jobs would be equally impressive, given technology’s influence.

Mining more jobs from legacy industries to add over half a million net new jobs in the next two years seems implausible. Or does it? The president may do well to add carrots to a manufacturing strategy that today relies on sticks.

His manufacturing legacy may depend on it.

Bart Taylor is publisher of CompanyWeek. Contact him at btaylor@companyweek.com.