Rapid wage growth at the bottom has offset rising US inequality

Overall US earnings inequality has plateaued in the last decade, driven by fast earnings growth among low-wage workers, rather than median-wage workers catching up with those at the top

Key Finding

"…Our main finding is that variance in labor earnings has stalled or decreased since around 2012.As shown in Fig. 1 A-C, this finding holds across all of the household surveys and all of the earnings quantities we examine—hourly, weekly, or annual. The variance of log hourly earnings increased in the CPS-ORG from 0.25 in 1982 to about 0.35 in 2012 before stabilizing. The variance of log weekly earnings has actually decreased in recent years; the CPS-ORG and the ASEC both show a slight decline from 2012 to the end of the decade, by about 1% in the case of the CPS-ORG. Inequality in annual earnings has also stabilized in the last decade—although the level where this stabilization occurred depends on the data source, with the ASEC plateauing at a lower level and slightly earlier than the ACS. The similarity in trends across surveys and earnings concepts supports the claim that earnings variance has stopped increasing in recent years…."

Clem Aeppli and Nathan Wilmers, "Rapid wage growth at the bottom has offset rising US inequality,"Proceedings of the National Academy of Sciences, October 3, 2022, https://www.pnas.org/doi/10.1073/pnas.2204305119

"… By contrast, labor markets across the country tightened over the last decade. Workers at the bottom of the national earnings distribution were, therefore, able to benefit from the declining local unemployment rate.The unemployment rate counterfactual (Fig. 4, Lower Left) confirms this description; increasing local unemployment rates to 2018 levels would have boosted earnings at the bottom of the national distribution and helps to explain the earnings compression that we observe…."

Changes In demand

"…Fig. 3 showed that from 2012 to 2018, workplace premiums increased slightly more for lower-paying workplaces than for higher-paying workplaces. However, this pattern is similar to 2002 to 2012, and changes to workplace premiums are relatively small. In contrast, the lowest-paid occupations enjoyed rapid premium increases from 2012 to 2018, following decreases in 2002 to 2012…."

"…The variance of earnings could decrease as all workers’ earnings converge, or it could reflect compression in a single part of the distribution. Fig. 2 shows the trends in the 10th, 50th, and 90th percentiles of hourly (Fig. 2A), weekly (Fig. 2C), and annual (Fig. 2E) earnings. This figure shows that in the 1980s, disparities between the bottom and the middle and between the middle and the top were both increasing. Since the 1990s, however, the gap between workers in the bottom and the middle stabilized, while the gap between workers in the middle and the top of the distribution continued to grow. These patterns are consistent with prior research…In the period since the recovery from the Great Recession, trend differences between top and bottom inequality intensified. Growth in the 10th percentile of earnings accelerated in the last decade—in log weekly earnings from 5.38 in 2012 to about 5.51 in 2018, a 14% boost in real dollars. While earnings growth has been robust during this period, its magnitude has not been uniform across the distribution; median weekly earnings have increased by only 9%, half as fast as the bottom 10th percentile. The 90th percentile has grown in pace with the median—increasing by under 10%. As a result of these different rates of earnings growth, the gap between the 10th and 50th percentiles has shrunk since 2012—shown by the solid lines inFig. 2 B, D, and F. In contrast, the 90 to 50 gap has continued to grow (dashed lines inFig. 2), although more slowly than in the earlier period of stagnant median earnings growth. Earnings in the bottom 10th percentile have advanced toward the median, while earnings at the top remain consistently above the median. InSI Appendix, section B, we directly link these changes in percentiles to the overall variance trend. The change in inequality trends after 2012 was largely due to compression of pay in the bottom two-thirds of the distribution rather than in the top third. The part of the earnings distribution that dominated rising inequality from the 1990s through the 2000s is not the part that explains the recent reversal of the trend since 2012. After decades of rising inequality, overall earnings inequality stopped growing, and possibly declined, since 2012. This fact raises two related questions. How can this unexpected departure from the trend be reconciled with prior research on earnings inequality, which focused on secular drivers of inequality? Additionally, does the recent break in rising inequality mark a new era of labor market conditions, as did the increase in earnings inequality in the early 1980s? We find little evidence that the last decade reversed the main drivers of rising inequality, like shifting worker and occupational composition, union decline, or unequal workplace premiums. Additionally, the top-end inequality that these drivers contribute to has persisted. However, this top-end inequality has been offset by rapid earnings growth for workers at the bottom of the earnings distribution. This earnings growth is not mainly driven by worker composition change, workplace-specific changes, or rising subnational minimum wages. Instead, it is associated with broadly increased premiums in the lowest-paid occupations: occupations that had faced particularly slow wage growth from 2002 to 2012. At the local labor market level, pay increases for low-wage workers were associated with tightening labor markets. These dynamics delivered a period in which earnings growth has been higher for workers at the bottom than at the middle or the top….”

Distributional Changes

"…First, in the dashed line in Fig. 1A, we show the variance of hourly wages based on theOEWS. The OEWS has a top code that increases in nominal terms throughout the period, at around the 98th percentile. However, it shows a decline in inequality similar to, and even sharper than, the household survey sources. Second, the short-dashed line in Fig. 1C shows the variance of annual earnings in the LEHD (from ref. 29 and top coded at the 99.999% level). Unlike the ACS and the ASEC, the LEHD has no information on weeks or hours worked. As a result, the level of inequality is higher in the LEHD, inflated by partial year and marginally labor market-attached workers. However, notwithstanding the level differences, the initially rising trend, subsequent leveling, and small decline are similar to the household sources. Third, in Fig. 1D, we use administrative data without top codes to show another measure of inequality: the share of US labor income going to the top 50%. The SSA shows different levels than PSZ and AS; in 2013, the share is 0.86 in the SSA data and 0.83 in PSZ. However, all three exhibit similar trends; the share of labor income going to the top half increased until about 2012, when it stabilized and even began to drop. This trend was interrupted in the SSA series in 2020, but this is likely due to COVID-19-related unemployment spells, which strongly affect annual earnings inequality measures that do not adjust for partial year employment. Overall, these distinct data, with a distinct inequality concept, also show a stabilization in inequality starting around 2012…"