Inequality is Central to the Productivity-Pay Gap

Matt Yglesias is an insightful writer, but his recent article, “Hillary Clinton’s favorite chart is pretty misleading” is itself very misleading. Since the Clinton campaign’s “favorite chart” is an EPI chart, which Jared Bernstein and I originally came up with twenty years ago, I think it’s important to set the record straight. The main problem is that Yglesias does not actually engage with the chart he says he’s criticizing.

The chart compares the growth of average productivity since 1948 with the growth of the hourly compensation (all wages and benefits) of production/nonsupervisory workers, a group comprising 82 percent of payroll employment (blue collar workers in manufacturing and non-managers in services).

The point is to show that the pay of a typical worker has not grown along with productivity in recent decades, even though it did just that in the early post-war period. That is, it shows a substantial disconnect between workers’ pay and overall productivity—a disconnect that has not always existed. We use data on production/nonsupervisory workers because there is no other data series on the pay of a typical worker that goes back to the early post-war period. The point of the chart is to show not only the current divergence but also that it was not always present—also, these data tend to move with the economy-wide median wage.

Yglesias argues that the major reason for the divergence is the different methods that must be used to adjust each line for inflation. This is flat wrong—the vast majority of the divergence is caused by rising inequality (both inequality among wage earners and a shift from labor to capital incomes). Yglesias does not explain why he wipes out the inequality story by shifting the discussion to a comparison of average compensation (the pay of all workers, including the top one percent) and productivity, rather than focusing on the actual lines in the Clinton/EPI graph. A key characteristic of inequality is that the average for all workers diverges from the median, as the former is pulled up by very large values at the top of the scale while the latter is not.

By focusing on the average compensation Yglesias has omitted from his analysis the biggest factor that generates the divergence between the pay of a typical worker and productivity growth portrayed by the Clinton campaign chart: rising inequality of compensation. That’s a pretty big omission. I would say it is very misleading to start with a chart that demonstrates the impact of inequality and then claim there is no big divergence when you make inequality disappear by comparing average pay instead of that of the typical worker.

I quantified the factors behind the divergence of median hourly compensation and productivity for the period between 1973 and 2011 in a 2012 paper, “The wedges between productivity and median compensation growth.” An older version of the Clinton campaign chart is Figure A in that paper. (We are actually in the process updating this analysis, so stay tuned.)

The three wedges that are responsible for the productivity-pay gap are:

  1. Changes in labor’s share: an overall shift in how much income in the economy is received as compensation by workers and how much is received by owners of capital;
  2. Compensation inequality: growinggaps in wages, benefits, and compensations between the top 1 percent, and high, middle-, and lowwage workers;
  3. “Terms of trade”: the faster growth of the prices of what workers buy relative to the prices of what they produce.

My analysis of these three wedges and their importance in particular sub-periods is the table below. The first two items are dimensions of rising inequality, while the third item is the one highlighted by Yglesias as the “big problem.”

Table 1

Reconciling growth in median hourly compensation and productivity, 1973–2011

197379 197995 199500 200011 197311
A. Basic trends (annual growth)
Median hourly wage -0.26 -0.15 1.50 0.05 0.10
Median hourly compensation 0.56 -0.17 1.13 0.35 0.27
Average hourly compensation 0.59 0.55 2.10 0.95 0.87
Productivity 1.08 1.29 2.33 1.88 1.56
Productivity-median compensation gap 0.52 1.46 1.21 1.53 1.30
B. Explanatory factors (percentage-point contribution to gap)
Inequality of compensation 0.02 0.72 0.97 0.59 0.61
Shifts in labor’s share of income 0.03 0.23 -0.40 0.69 0.25
Divergence of consumer and output prices 0.46 0.51 0.64 0.24 0.44
Total 0.52 1.46 1.22 1.52 1.29
C. Explanatory factors (percent contribution to gap)
Inequality of compensation 4.8% 49.6% 80.0% 38.9% 46.9%
Shifts in labor’s share of income 5.5% 15.4% -32.5% 45.3% 19.0%
Divergence of consumer and output prices 89.7% 35.0% 52.5% 15.8% 34.0%
Total 100.0% 100.0% 100.0% 100.0% 100.0%

Note: Totals for panels A and B do not exactly match due to rounding

Source: Analysis of Mishel and Gee (2012) Table 1

Copy the code below to embed this chart on your website.

The biggest wedge over the 1973-2011 period was rising compensation inequality, which was responsible for nearly half (46.9 percent) of the total divergence. Changes in labor’s share of income was the least important factor (19 percent), and the “terms of trade” (the factor Yglesias focuses on) was the second most important (34 percent). The results for recent years—the most relevant ones for current policymakers—are different. From 2000 to 2011, the erosion of labor’s share of income was the most important factor (45.3 percent), inequality next most important, and the “terms of trade” the least (15.8 percent). Inequality of compensation contributed 38.9 percent of the divergence between 2000 and 2011. So, the two inequality items (falling labor’s share and rising compensation inequality) explained 84 percent of the divergence in the 2000s, when the pay and productivity divergence was larger than in any other period.

Having set aside the whole issue of rising inequality of compensation, Yglesias identifies the “terms of trade” as the really big deal. This is what he means when he says, “the bigger problem is that both lines are indexed to inflation — using different inflation indexes.” He is referring to the fact that the prices used to adjust productivity (a combination of the prices of consumption, investment, exports, imports and government) differ from the inflation adjustment for compensation, which relies solely on consumption prices. Note that my analysis shows his “bigger problem” was primarily a problem from the 1970s and late 1990s, and has had very little effect since 2000.

Yglesias correctly notes that real wages “really have risen much too slowly over the past 40 years,” but he adds that “Clinton’s version of the chart makes it look like rising productivity isn’t part of the solution.” He’s wrong to suggest that the chart implies that productivity growth is unimportant. Productivity growth is surely important but it is also surely an insufficient way to grow real wages and compensation for the typical worker. There must also be a policy agenda to reconnect productivity and pay. That is why EPI has offered our Agenda to Raise America’s Pay and why our former colleague, Jared Bernstein, has a book on “The Reconnection Agenda.” And that is why the tweet from Hillary Clinton that went along with our chart—“That bargain has eroded. It is our job is to make it strong again”—is right on target. Reconnecting those lines should be the central focus of economic policy and of economic debate during the upcoming campaign.