Worker pay has not kept up with gains in worker productivity in recent decades in the United States, Canada, the United Kingdom, and elsewhere. This raises the question: Does productivity growth actually benefit typical workers by raising their pay—or do the fruits of any increase in productivity growth flow only to investors and people with higher incomes? This question is of great interest to policymakers seeking to reinvigorate productivity growth after the productivity slowdown in recent decades.
In our 2021 paper in the International Productivity Monitor (Greenspon, Stansbury, and Summers 2021), we investigate this question for the United States and Canada.1 The evidence presented there, and summarized in this blog post, suggests that even though productivity has grown much more rapidly than the pay of typical workers in both countries, there is still evidence that higher productivity growth benefits workers—in the sense that pay grows more quickly in periods of higher productivity growth and vice versa. This relationship appears to be stronger in the United States than in Canada.
To determine whether productivity growth raises typical workers' pay, our 2021 paper regresses labor productivity growth on real compensation growth—measured at the average, at the median, and for measures of the compensation of "typical" workers in both the United States and Canada. The possible results fall on a spectrum from complete delinkage (a 1 percentage point higher productivity growth rate is not associated with any change in the compensation growth rate) to complete linkage (a 1 percentage point higher productivity growth rate is associated with a 1 percentage point higher compensation growth rate).2 To focus on the impact of productivity growth over horizons longer than the business cycle, the regressions use three-year moving averages of the growth rate of productivity and compensation and control for unemployment.
The evidence suggests relatively strong linkage between productivity and compensation in the United States, as illustrated in figure 1: In recent decades, in years with a 1 percentage point higher productivity growth rate, the growth rate of average compensation and median compensation was about 0.7 percentage point faster; for the compensation of production and nonsupervisory workers, the growth rate of compensation was about 0.5 percentage point faster.3
For Canada, the evidence on the linkage between productivity and pay is more mixed. For average compensation, regression coefficients suggest moderate linkage: Over the period 1961 –2019, a 1 percentage point higher productivity growth rate is associated with about a 0.4–0.5 percentage point faster rate of compensation growth. For typical workers—whose compensation is proxied using a newly constructed measure of the compensation of hourly paid workers in five industries—the evidence indicates stronger linkage over that period, with a 1 percentage point higher productivity growth rate coming alongside a 1 percentage point higher rate of growth of compensation.4 In more recent periods, estimates for both average and typical compensation are too noisy to rule out strong delinkage (no translation of productivity increases to pay growth), and there is no evidence of linkage for median compensation. The measure used for median hourly compensation, however, likely incorporates significant noise, which may substantially attenuate results estimated from year-to-year changes.
Although the picture is clearer for the United States than Canada, these results suggest that in both countries, marginal increases in productivity growth raise compensation at least somewhat on average, including for typical workers (holding all else equal), despite the divergence between productivity growth and compensation growth in the both countries over the last five decades.
How can there be some linkage between productivity and pay even as they diverge? A productivity growth rate that is consistently faster than the growth rate of median compensation—by a fixed amount, say, a wedge of 1 percentage point—leads to a growing gap between productivity and median pay over time (divergence). But, conditional on this 1 percentage point wedge between the two growth rates existing, it may nonetheless be the case that in years in which productivity growth was faster than average, median pay growth was faster than average too (and vice versa.) This pattern would suggest that incremental increases in productivity growth rate feed into higher pay growth for typical workers but that other, long-term factors have been pulling down pay even as productivity has been acting to raise it, so that on net, pay has grown much less than productivity.
A simple metaphor illustrates the point. Consider a bucket containing some level of water (pay) that is filled from a hose connected to a faucet (productivity growth). The observed divergence between productivity and pay is analogous to the faucet running but the water level not increasing. This outcome could occur because the hose does not actually connect the faucet to the bucket, in which case increasing the water flow will not make any difference to the water level (delinkage: increasing the rate of productivity growth does not affect pay). Or the problem could be that the bucket has a hole. In this case, increased water flow results in a higher water level in the bucket than would have occurred with a slower flow of water but a lower level than would have occurred had the bucket not had a hole (linkage: faster productivity growth increases pay relative to a world with slower productivity growth, even if other factors reduce pay at the same time).
Why does there appear to be weaker linkage between productivity and pay in Canada than in the United States? Canada is a smaller, more internationally open economy than the United States. One possible explanation is therefore that the smaller and more open the economy, the greater the degree to which productivity and real compensation may be determined abroad rather than domestically—and the less one might expect researchers to be able to detect a process in which increases in domestic productivity translate into increases in real compensation. We present some evidence consistent with this hypothesis: There is weaker linkage between average compensation and productivity in regions of the United States—which are similar in GDP and population size to Canada and could therefore be considered small open economies—than for the United States as a whole. The coefficients estimated for US regions are similar in magnitude to those estimated for Canada.
Another interesting point of comparison is that despite much slower labor productivity growth in Canada than in the United States, the growth in real median compensation has been about the same in both countries since 1976, thanks to larger increases in labor income inequality in the United States than in Canada, as illustrated in figure 2.
While the focus of this investigation has been on the relationship between productivity growth and wages, both are of course determined by deeper economic forces, such as the extent of capital accumulation and the nature of technical innovation. It would be valuable in future work to explore the possibility that changes in productivity growth coming from different sources have different impacts on wage growth.
Overall, the findings presented here support the hypothesis that productivity growth still matters for middle-class income growth. Measures to boost productivity growth, such as investment, technological progress, or better functioning markets, are important for raising pay for the average and typical worker. Conversely, a continued productivity slowdown will reduce the likelihood of increased real compensation. This finding does not mean that policymakers should ignore measures to improve redistribution, as productivity growth alone is not enough to substantially raise middle-class incomes, particularly in the face of other downward pressures on pay. But it does mean that policy should not focus only on these issues and ignore productivity growth. To have the greatest impact on middle-class living standards, policymakers should use all measures available—both those that boost productivity growth and those that promote inclusion.
1. This blog builds on the empirical investigation of this question for the United States by Stansbury and Summers (2019).
2. Specifically, we regress the change in the log of annual labor productivity (measured as net output per hour) on the change in the log of hourly real labor compensation, controlling for the level and lagged level of the unemployment rate. In our baseline specifications, all variables are three-year moving averages, standard errors are Newey-West heteroskedasticity and autocorrelation consistent, and compensation is deflated by the personal consumption expenditures (PCE) index in the United States and household final consumption expenditure (HCE) index in Canada.
3. The estimated relationship is somewhat smaller in the past two decades, apparently driven by the fact that the very slow productivity growth after the Great Recession was not matched by as steep a decline in the rate of compensation growth.
4. The five industries in this compensation measure are manufacturing, mining, construction, laundries, and hotels and restaurants. These are the industries that are both most likely to provide employment opportunities for typical workers during times of growth and for which historical time series data are available from 1961 onward. They accounted for 34 percent of total employment in 1961 and 24 percent in 2011.