The right-hand two-thirds of Figure 4 look like the plots of employment polarization.
From 1979 through 2007, wages rose consistently across the high-skill portion of the figure, which is disproportionately made up of the abstract task-intensive
categories of professional, technical, and managerial occupations. By contrast, wage
growth in the middle-skill, typically routine task-intensive occupations was less rapid
and generally decelerated over time. For the low-education, manual task-intensive
occupations heavily represented on the left-hand side of Figure 4, in the 1980s,
wage growth was a little more rapid than in the middle-skill occupations—and in
the 1990s, it was much more rapid. However, that changed in the 2000s: while
Figure 2 showed that employment growth in these occupations exceeded that in
all other categories between 1999 and 2007, Figure 4 shows wage growth was generally
negative in the low-skill percentiles, lower than in all other categories (Mishel,
Shierholz, and Schmitt 2013). During this time period, my strong hunch is that
the explanation is that declining employment in middle-skill routine task-intensive jobs led middle-skill workers—including new entrants, those displaced from routine
task-intensive jobs, and those who lost jobs during recession—to enter manual
task-intensive occupations instead (Smith 2013; Cortes, Jaimovich, Nekarda, and
Siu 2014; Foote and Ryan 2014).
A final set of facts illustrated by Figure 4 is that overall wage growth was anemic
throughout the 2000s, even prior to the Great Recession. Between 1999 and 2007,
real wage changes were negative below approximately the 15th percentile, and were
below 5 percentage points up to the 70th percentile of the distribution. Indeed,
wage growth was greater at all percentiles during both the 1980s and 1990s than in
the pre-recession 2000s.7 Of course, wage growth was essentially zero at all percentiles
from 2007 to 2012.
Why are the rapidly rising earnings of the top 1 percent (as discussed in
Atkinson, Piketty, and Saez 2011, for example) not strongly evident in Figure 4?
One reason reflects substance; another is an artifact of the data. Substantively, the
plot depicts changes in earnings by occupational percentile rather than wage percentile.
Wage growth by occupational percentile is less concentrated than wage growth
across wage percentiles because the highest earners are found across a variety of
occupations. In addition, the very highest percentiles of earnings are censored in
public use Census and American Community Survey data files, which further masks
earnings gains at extreme quantiles.