Over the last five decades, middle-wage jobs diminished in the US as wage inequality increased. This column investigates the relationship between these two phenomena, and finds no evidence that either computerisation or automation (often cited as a source of both trends) produced employment polarisation or increased wage inequality. By examining wages at the individual level (rather than occupation-average wages), the column suggests that the evolution of wages can be better explained by distinct causes—ranging from changing labour market institutions to globalisation—than by observable demographic factors.
Middle-wage jobs in the US are gradually diminishing while wage inequality has been rising. But are the two related? Does the decline in middle-wage jobs represent polarisation of employment, and is the decline a good or a bad thing for workers?
Inequality between top and middle hourly wages has increased steadily for the last 50 years in the US. By contrast, inequality between middle and bottom wages rose sharply in the 1980s; then, after a slight decline, it remained stable for the next 30 years. These gaps are commonly measured as the ratio of the 90th and 50th percentile wages (the wage of the worker earning more than 90% of workers relative to the wage of the worker earning more than 50% of workers) and the ratio of the 50th and 10th percentile wages, as shown in Figure 1.
Economists have investigated many potential explanations for these changes, including the decline of unions (Fortin et al. 2019); the inflation-adjusted minimum wage (Lee 1999); the spread of outsourcing and temporary-agency labour (Feenstra and Hanson 1999); reduced competition and dynamism (Furman and Orszag 2018, Shambaugh et al. 2018); and increased international trade, off-shoring, and technological progress (Blum 2008, Feenstra and Hanson 2003). The first two factors played an important role in the 1980s, and many researchers believe that technology has played an important role throughout.
A complementary analysis focuses on employment shares of low-, middle-, and high-wage workers rather than wage inequality. One of the most striking findings from this work is that the share of middle-wage jobs has declined. This decline can be measured in different ways, one of which is to examine shifts in the occupational composition of employment. Some research foregrounds the role of technology in its examination of these shifts, because in addition to a decline in the employment share of middle-wage occupations and a rise in employment share of high-wage occupations, the share of low-wage occupation employment share has been rising (e.g. Autor 2015a,b and Autor and Dorn 2013 for the US; Goos et al. 2009 and Goos et al. 2014 for other countries).
This pattern of polarisation would be consistent with technology displacing middle-occupation workers to worse jobs if the timing were consistent with the automation of manufacturing and the spread of computers. However, while early research found the low-wage occupation share to be rising beginning in at least the 1980s, our recent research has found it to be rising only from 2002–2012, too late to reflect the introduction of relevant technology (Hunt and Nunn 2019).
Moreover, we argue that because individual wages and average wages in an occupation are only weakly linked, an occupation-based approach is not the best route for seeking links between declining middle-wage jobs and wage inequality, or for examining worker welfare. Over the last 50 years, the share of workers in lowest-wage occupations who are themselves paid low wages has fallen dramatically. This means that low-wage occupations could be growing at the same time that workers are shifting up the wage ladder. Indeed, 86% of the 1973–2018 increase in hourly-wage variance occurred within detailed occupations. Occupations are a useful unit of analysis for other patterns – such as changing organisation of work tasks – but are less useful for examining inequality.
Instead of examining occupation-average wages, we therefore examine wages at the individual level. In this analysis, low, middle, and top wages are defined using cut-offs held constant for inflation. The share of workers earning middle wages has declined for the last 50 years, as shown in Figure 2. But in a departure from employment polarisation, the figure does not show simultaneous increases in the top and bottom group shares. The share of workers in the top bin rises from 0.25 in 1979 (and a similar value in 1973) to 0.35 in 2018, while the share of workers in the bottom wage bin declines from 0.25 in 1979 (and a similar value in 1973) to an all-time low of 0.18 in 2018, after peaking at 0.30 in 1983-1985. This means that shares of workers earning bottom and top wages have generally moved in opposite directions, and do not rise together in the polarising fashion that would provide a link to rising wage inequality.
Whether the decline in the share of middle wage jobs is a positive or negative development for workers depends on what replaces middle-wage jobs, an element seen clearly by examining the decisively different trends for men and women. Women have experienced upward mobility for the last 40 years, with declines in the bottom and middle shares offset by increases at the top. This change is clearly welfare-improving, though women continue to be underrepresented at the top of the distribution. Men, by contrast, fell down the wage ladder during the 1980s, when a falling middle share reflected a decline in welfare, as more men moved from the middle to the bottom than from the top to the middle. Subsequently, the declining middle share reflected increased welfare as men regained ground during the late 1990s and early 2000s. Male workers' fortunes then waned in the Great Recession before waxing during the recovery.
Rising age and education are two of the strongest factors contributing to increased wages for both men and women. For men, declining union membership and shifts from high- to low-wage industries depressed wages, and business cycle conditions provided temporary boosts to or drags on workers. Still, much of the evolution of wages cannot be explained in terms of changes in observable demographic factors.