The debate over technological unemployment — do digital machines kill jobs? — continues to heat up. At Slate, Matthew Yglesias hacks together a chart, from Federal Reserve data, that he claims shows that technological unemployment doesn’t really exist. He plots the country’s economic output (red line) against the hours worked by non-management personnel (blue line) for the last ten years:
They seem to match up pretty well, with the growth in hours worked lagging a bit behind the growth in output — an indication of a healthy increase in productivity. Comments Iglesias, “The big rise and fall and rise again in output is caused by a big rise and fall and rise again in the amount of time people put on the job. Or alternatively, the big rise and fall and rise again in working time is caused by a big rise and fall and rise again in the amount of demand for goods and services.” It’s the economy, stupid. Employers’ recent investments in fancy new technology don’t seem to be altering the old economic laws. There’s no such thing as technological unemployment, at least not when you look at the big picture.
Some commenters suggested that the “covariance” of the two lines isn’t really as tight as Iglesias makes it out to be, that the blue line may be flattening out. Wrote one: “The problem is not that the red line and blue line are growing apart, the problem is that, while the red line has more than recovered since the recession, the blue line is still below recession levels.” Wrote another: “Matt’s right that there is a lot of covariance between the two curves, but that’s hardly surprising, and even people who believe that technology is displacing workers would agree that large-scale macroeconomic trends will strongly affect labor demand — it would be difficult to believe data that didn’t show a strong relation between the two. But visual inspection also shows that the overall linear trend for GDP (i.e. the regression line) has a clearly larger slope than the trend for aggregate hours.”
Now, Andrew McAfee has responded with some charts of his own. Here’s one that looks particularly at the manufacturing sector, comparing output growth (blue line) and job growth (red line) over a longer period:
Not much recent covariance there. Writes McAfee: “That really looks like technological unemployment to me, especially when manufacturing employment is also on the decline in Germany and Japan, in China, and around the world. When this is the case, it means that employment changes are not due to jobs moving around the world in search of cheap human labor; they’re due to machine labor becoming at least as capable as and cheaper than humans.”
And here’s a longer term look at the overall economy, comparing output (red line), jobs (green line), and hours worked (blue line).
Certainly, the economic cycle still matters, but it doesn’t seem to be the whole story. McAfee again: “We are still adding jobs and working more hours in non-recession years, but not as quickly as we used to. Since the end of the 2001 recession real GDP has increased by just about 20%. The number of hours worked, however, has increased by only 2.8% over that same time, and the total number of jobs by 1.9%.”
The economy is a complex beast, and none of these charts shows the actual effect of new technology on employment. But the specter of technological unemployment certainly hovers over McAfee’s charts. Dismissing it as a myth seems at best premature.
SPECIAL BONUS CHART: And just to prove I can create a line chart too, here’s one, made with Google’s Ngram Viewer, showing mentions of the phrase “technological unemployment” in books over the past ninety years or so. (Click to enlarge it.) Ngram doesn’t go beyond 2008, but I bet when you make the chart in the future you’ll see the start of a new spike in 2013.
Library of Congress photo.