Robert Gordon has a new paper up at NBER, “Revisiting U.S. Productivity Growth Over the Past Century with a View of the Future” [subscriber-link[. It is awesome. After offering a wide-ranging survey, he tries to tease out the lessons of the last twenty years for the next twenty years:
The 20‐year period 1987‐2007 combines the inexplicably slow productivity growth of 1987‐95, the temporarily ebullient period 1995‐2000, and the interesting 2000‐07 period that in some dimensions looks like more normal behavior. The seven years between 2000:Q4 and 2007:Q4 were neatly divided in half, with extremely rapid productivity growth between 2000:Q4 and 2004:Q2 (2.68 percent), and much slower growth from 2004:Q2 to 2007:Q4 (1.36 percent), averaging out to 2.02 percent for the seven‐year interval. As argued above the productivity growth “explosion” of 2001‐04 rested on a combination of savage corporate cost cutting and delayed learning from the internet revolution. Once profits had recovered the pressure for cost cutting disappeared, and eventually the delayed learning subsided as well.
Unfortunately, the relevance of the 2000‐07 period is tainted by the bizarre behavior of the labor‐market variables. Given that the unemployment rates were similar at the beginning and end, there was an unprecedented drop in hours per capita, with aggregate hours (H) growing at only 0.36 percent per year while the working‐age population grew at 1.24 percent per year. Hours per capita fell by 0.88 percent per year, or a cumulative exponential 6.4 percent decline between 2000 and 2007.
The sharp decline in hours per capita in 2007 to some extent reduced output growth but it also raised productivity growth, in that some of the reduction of labor hours represented a permanent transition to a more aggressive stance of management against labor (Gordon, 2010b). To this extent that 2.02 percent growth in total‐economy output per hour in 2000‐07 may be misleading as an indicator for future forecasts. Since forecasts of future growth of aggregate hours are bunched together at 0.7 percent (see for instance Jorgensen et al. 2008, Table 2), we anticipate that at least half of the return from 0.38 in 2000‐07 to 0.70 on average after 2007 will be offset by slower productivity growth.
This argument that the 2000‐07 growth rate of total economy labor productivity was unsustainably high, due to the role of cost‐cutting in boosting productivity at as offset to a sharp decline in labor hours, suggests that the long‐run growth rate of productivity in the total economy is likely to be closer to the 1.79 percent of 1987‐2007 than the 2.02 percent of 2000‐07. By the same reasoning, the growth rate of NFPB sector productivity, shown in Table 10 to be 2.48 percent in 2000‐07, is more likely to be closer to the 2.23 percent of 1987‐2007. [Emphasis added]
Gordon also references a phenomenon that Tino Sanandaji and I discussed in a recent article for National Review, i.e., the role of labor quality:
Jorgenson, Ho, and Stiroh (2006), show that educational attainment has reached a plateau in the US, in their estimates implying that the improvement in labor quality (the BLS labor composition effect) will gradually decline toward zero over the next 10 to 15 years. Similarly, Goldin and Katz (2008) both lament and explain the plateau in U. S. educational attainment. They point out (2008, Figure 9.1, p. 327) that unlike most European nations, where a catching‐up process has made the 25‐34 age group much better educated than the 55‐64 age group, in the U. S. the educational attainment of both age groups is the same, the very definition of a plateau.
Since the continuation of the labor quality adjustment depends largely on increasing educational attainment, yet this increase seems to have disappeared, Table 10 cuts in half the labor quality adjustment for 2007‐2027 as compared to its average value of 1987‐2007. The difficulties of further increasing U. S. educational attainment are well documented by Goldin and Katz, including a rapidly increased relative price of higher education and a paucity of opportunities for less privileged students to gain fellowship support.
Tino and I frame the labor quality issue in terms of the “achievement gap” between white and Asian students and black and Latino students:
Let us extrapolate the historical average of 2 percent annual per capita income growth until 2050, and take aging and demographic change into account. Aging will shrink the working-age population, bringing income growth down to about 1.75 percent. Also, whites and Asians will go from two-thirds to a little over half of the working-age population — and the effect of that shift will depend on changes in the achievement gap.
The worst-case scenario is that the gaps between whites and non-whites in education and earnings will not change. In this scenario, the skills and earnings of the American work force decline, and per capita income growth falls to 1.49 percent per year. However, if we assume that policy reform or assimilation will close half of the educational-achievement gap by 2050, and that this in turn will close the earning gap by half, then the average growth rate per capita will be 1.85 percent per year. In this second scenario, the growing Hispanic population not only doesn’t reduce income growth, but actually mitigates some of the effects of population aging.
Comparing the two scenarios vividly illustrates the economic value of closing the achievement gap. The alternative futures available to us are an economy producing $38 trillion per year and an economy producing $44 trillion per year.
Basically, Tino and I conclude that the school reform agenda is of central importance to America’s economic future. This isn’t exactly a groundbreaking conclusion, and the methods we use are obviously plagued by limitations. But this is one reason why I tend to think that collective bargaining reform at the state and local level is actually a more important than, say, the top marginal tax rate.