Tip of the hat to Tyler Cowen at the Upshot for flagging Chad Syverson’s excellent paper on the productivity slowdown. The slowdown in productivity growth over the past decade in the US and across the industrialized world is by now widely recognized. Many skeptics—especially on the West Coast—have speculated that the measured slowdown is merely the effect of mismeasurement of value-added in the new digital economy. That is, since the Internet is largely free, GDP does not capture the enormous gains in welfare from recent advances such as smartphones, social media, and so on. Syverson calls it the Mismeasurement Hypothesis (MH).
Syverson sets up a series of tests for MH. First, he investigates whether there is a systematic difference between advanced economies. If MH holds, then the measured slowdown should be more pronounced in countries with greater broadband access or where the tech sector looms relatively larger. He shows that there is no evidence that this is the case.
Second, he shows that even the most expansive estimates of consumer surplus (the difference between what a consumer is willing to pay and actually pays for anything) arising from digital technologies is orders of magnitude smaller than the missing output. While the former is in tens or hundreds of billions of dollars, the missing output is of the order of three trillion dollars. Finally he shows that if MH holds, Tech would’ve registered far faster growth rates in measured output that it has since 2004 (otherwise the surplus would have to rise faster and faster compared to measured output).
Syverson’s clever analysis supports what the Policy Tensor has long suspected: The slowdown in productivity is very real. This is, of course, very counter-intuitive since everyday experience suggests that there has never been a more innovative time, even outside Tech. Innovations in hydraulic fracturing and horizontal drilling have revolutionized the economics and geopolitics of oil and gas. Despite the collapse in the price of hydrocarbons, dramatic declines in the cost of producing renewable energy have finally brought a green future within our grasp. Vaping has already helped millions quit smoking by delinking harmless nicotine from lethal tar. Robots are revolutionizing manufacturing and war. Alas, rapid advances can be found everywhere but in the productivity numbers.
One criticism of Syverson’s analysis is that he uses a measure labor productivity, output per hour. The problem is that it is entirely possible for output per hour to rise because of capital deepening and vice versa. In other words, the slowdown may reflect slower growth in capital deployed per worker than a deceleration in real productivity gains. In order to check that this does not affect the results, I compared a measure of multi-factor productivity from the OECD to output per hour from the BLS.
Multi-factor productivity or Total Factor Productivity (TFP) is calculated by taking out the contributions of capital deepening and labor input from real output growth. For instance, if real output is growing at 3 per cent while both capital deployed and labor deployed are growing at 1 per cent, then TFP growth would be 1 per cent. This is the real source of rising living standards. And it is only this component of growth that truly deserves to be called productivity growth.
In the chart above, the bars represent annual TFP growth (left scale), the curved line is output per hour (right scale), and the black horizontal lines display average rates of TFP growth in selected periods. We see that that TFP and output per hour are highly correlated and that TFP growth accounts for about half the gains in output per hour. We can be assured that Syverson’s analysis is robust to alternate measures of productivity. In particular, using TFP we should be able to recover his ballpark of three trillion dollars in lost output that would need to be accounted for by surplus arising from innovations in Tech.
Calculated annual TFP is cyclical which in theory it should not be since it is supposed to capture long term developments. This is because the computation assumes that the economy is in equilibrium (i.e., no business cycle). Because both labor and capital are sticky, the residual (TFP) picks up some of the cyclical movement. Therefore, specific annual numbers may not be a reliable guide. But averaged over the business cycle they are quite reliable.
We should therefore pay attention to TFP growth numbers averaged over longish periods. I calculate that TFP growth averaged 0.7 per cent in 1985-1995, 1.6 per cent in 1996-2005, and 0.5 per cent in 2006-2014. The big question is whether we should expect the recent tepid growth in productivity to continue in the medium term or whether the current stagnation is transitory.
To see why this is important just note that, had productivity grown at the same pace over the last ten years as it did over the previous ten, the median American household would be making about twenty thousand dollars more every year.