A prevalent story explains wage inequality as a simple consequence of growing employer demand for skills and education—often thought to be driven by advances in technology. According to this explanation, because there is a shortage of college-educated workers, the wage gap between those with and without college degrees is widening. The expected boost to workers’ pay from a four-year college degree is known as the “college premium.”

Despite its great popularity and intuitive appeal, this story about recent wage trends driven more and more by a race between education and technology does not fit the facts well, especially since the mid-1990s. The growing inequality of note is that between the top (or very top) and everyone else. The pulling away of the very top cannot be explained by education differences, but rather the escalation of executive and financial sector pay.

Even when looking at the relative changes in the 95^th percentile of wage earners compared to the 50^th percentile of wage earners, and comparing that gap with the college wage premium from 2000 to 2018, it is clear that gains in the college wage premium have been very modest and far less than the continued steady growth of the 95/50 wage gap. Therefore, it is highly implausible that the growth of unmet employer needs for college graduates has driven wage inequality.

The evidence suggests the demand for college graduates has grown far less in the period since the mid-1990s than it did before then. This is difficult to square with contentions that automation or changes in the types of skills employers require have been more rapid in the 2000s than in earlier decades. Rather, automation has been slower in the recent period than in earlier decades as seen in the pace of productivity, capital, information equipment, and software investment—and in the speed of changes in occupational employment patterns.

The first figure below compares the change in the college wage premium over 1979–2000 and 2000–2018 with the change in the log 95/50 wage gap. The college wage premium is the percent by which average hourly wages of four-year college graduates exceed those of otherwise equivalent high school graduates, controlling for gender, race and ethnicity, age, and geographic division. The 95/50 wage ratio is a representation of the level of inequality within the hourly wage distribution, comparing how much more the 95th percentile worker is paid relative to the 50th percentile worker. Both wage gaps are measured in log changes and shown as annual changes.

Figure A

The figure shows that the regression-adjusted college wage premium grew rather quickly between 1979 and 2000 then rose at a much slower rate in the 2000s, about as tenth as much. It had already slowed considerably by the mid-1990s, as noted above. In contrast, the 95/50 wage gap grew somewhat faster in the more recent period. When we compare the relative size of the changes in each gap from 2000 to 2018, it is clear the very modest gains in the college wage premium in recent years have not been large enough to plausibly drive the continued steady growth of the 95/50 wage gap. In fact, the log 95/50 wage gap grew 10 times faster than the college premium over this period.

Between 1979 and 2000, the log 95/50 wage ratio and the regression adjusted college wage premium grew at roughly the same pace. Increased employer demand for education as a prime driver of inequality appeared to be a more plausible story then. As we see in the latter period, the growth in the college wage premium cannot explain the growth in inequality and therefore the correspondence in the earlier period also shouldn’t be over-interpreted as education driving the 95/50 wage gap. The truth is there were other factors in the economy driving both slower median wage growth as well as the college premium such as the decline in unionization, allowing for excessive unemployment, globalization, and the fact that the overtime threshold was allowed to wither.

The more salient story between 2000 and 2018 is not one of a growing differential of wages between college and high school graduates, but one of growing wage inequality between the top (and the tippy top) relative to the vast majority of workers. Wage inequality is driven by changes within education groups (among people with the same education) and not between education groups. From 2000 to 2018, the overall 95th-percentile wage grew over three times faster than wages at the median (25.1 percent versus 7.0 percent). Among college graduates, there has been a significant pulling away at the very top of the wage distribution. The figure below shows the change in college wages from 2000 to 2018 for various deciles of the college wage distribution.

Figure B

As shown, the bottom 60 percent of those with a college degree still have lower wages than they did in 2000. The 50th-percentile wage among those with bachelor’s degrees was 2.4 percent lower in 2018 than it was in 2000, while the 90th-percentile wage of those with bachelor’s degrees was 9.8 percent higher. (The 95th wage percentile for college graduates is fraught with “top-coding” issues making it difficult to obtain reliable measures of high-end wages and wage growth, as discussed in The State of Working America Wages). What’s clear from the data is that less than 40 percent of college-graduate wages have experienced any growth since 2000.

Between 2000 and 2018, the median high school wage also fell (-1.5 percent), while the 95th percentile high school wage grew 7.1 percent. If we compare (raw) high school and college wages at the middle of each distribution, we see that median college wages are no higher than high school wages in 2018 than they were in 2000. If anything, the gap actually narrowed ever so slightly between 2000 and 2018.

Increases in inequality over the last 18 years clearly cannot be explained away by claims that employers face a growing shortage of college graduates and that, correspondingly, wage inequality is some unfortunate side effect of the positive gains from automation that we neither can nor would want to alter. There are plenty of good reasons to provide widespread access to college educations and skill development, but expanding college enrollment and graduation is not an answer to escalating wage inequality.

That trend was evident in EPI’s new State of Working America (SWA) Wages report, which highlights trends in wages across the wage distribution, by education, as well as by gender, race, and ethnicity.

Overall, the findings indicate wages are slowly improving with the growing economy, but wage inequality has grown and wage gaps have persisted, and in some cases, worsened. In this post, I will highlight one particular worsening wage gap and look at it from multiple dimensions. Since 2000, by any way it’s measured, the wage gap between black and white workers has grown significantly.

The findings here support the important research by Valerie Wilson and William M. Rodgers III, which shows that black–white wage gaps expanded with rising wage inequality from 1979 to 2015. Where their report is incredibly comprehensive, the trends outlined here are rudimentary, but reinforce the same basic truths.

In the figure below, I’ve collected some of the main findings on the black–white wage gap found both in the latest SWA report as well as the SWA data library. Using various measures, I compare wages for black and white workers over the last 18 years, highlighting the gaps in wages in 2000, the last time the economy was closest to full employment, 2007, the last business cycle peak before the Great Recession, and 2018, the latest data available.

Against these benchmarks, I illustrated the growth in the average gap; the gap for low-, middle-, and high-wage workers; the gap for workers with a high school diploma, a college degree, and an advanced degree; and a regression-adjusted wage gap (controlling for age, gender, education, and region).

As always, it’s important to remember the historical and social contexts for differences in black and white labor market experiences and labor market outcomes. We know from a host of economic research that a person’s wages are not a simple function of individual ability. Instead, workers’ ability to claim higher wages rests on a host of social, political, and institutional factors outside their control.

Against that backdrop, and given a long history of excluding black Americans from social and political institutions that boost wage growth, the stubbornness of racial wage gaps is less surprising. Connie Razza provides a powerful framework for examining the systematic social deprivation and economic disadvantage maintained and reinforced by those with economic and political power.

As shown in the figure above, the black–white wage gap is smallest at the bottom of the wage distribution, where a wage floor, otherwise known as the minimum wage, keeps the lowest-wage black workers from even lower wages. Raising the federal minimum wage would also disproportionately benefit black workers because they are overrepresented among low-wage workers and are less likely to live in states or localities that have passed a minimum wage that is higher than the current federal minimum.

Figure A

The largest black–white wage gaps are found at the top of the wage distribution. The 33.4 percent wage gap for the highest-wage workers is explained in part by occupational segregation, the disproportionate likelihood for white workers to occupy positions in the highest-wage professions, and the pulling away of the top more generally.

Occupational segregation is particularly devastating for black women, who face a history of deep-seated racial and gender discrimination. Any number of fields would provide examples, but in my own, Rhonda Sharpe has documented the lack of representation in economics, and the American Economic Association’s Summer Program along with the recently formed Sadie Collective are working to diversify the field.

Education is not a panacea for closing these wage gaps. Again, this should not be shocking, as increased equality of educational access—as laudable a goal as it is—has been shown to have only small effects on class-based wage inequality, and racial wealth gaps have been almost entirely unmoved by a narrowing of the black–white college attainment gap, as demonstrated by William Darity Jr. and others here.

The figure below shows that across various levels of education, a significant black–white wage gap remains. Black workers can’t simply educate their way out of the gap. Even black workers with an advanced degree experience a significant wage gap compared with their white counterparts. And after controlling for age, gender, education, and region, black workers are paid 16.2 percent less than white workers.

While the wage gaps differ depending on measure, what is obvious from the trends displayed is that the gaps widened in the full business cycle 2000–2007 and continued to grow in the Great Recession and its aftermath. Even though the black unemployment rate has fallen precipitously over the last several years, wage growth has remained particularly weak for black workers.

In a recent paper, Josh Bivens and Ben Zipperer present evidence in support of hopes that a high-pressure labor market can improve relative labor market outcomes for black workers, particularly when it comes to labor force participation and work hours. The data here suggest that wage growth continues to lag progress in employment.

While black workers continue to increase their educational attainment, occupational segregation persists, and employers continue to wield considerable leverage to dictate wage contracts, including ones that perpetuate discriminatory practices. Unfortunately, in lieu of stronger labor standards and worker bargaining power, it takes tighter and tighter labor markets for all workers to reap the rewards of a strong and growing economy.

I’m optimistic that as the economy continues to move toward genuine full employment, black workers will see their wages rise. But it will take more than a couple of years of a full-employment economy to close racial wage gaps and compensate for years of lower wages, lower incomes, and lower wealth.

This report provides a state-by-state breakdown of unemployment rates by race and ethnicity and racial unemployment rate gaps for the second quarter of 2017. It shows that while there have been state-by-state improvements in prospects for black and Latino workers, their unemployment rates remain high relative to those of white workers. Following are some key highlights of the report:

• While the African American unemployment rate is at or below its prerecession level in 14 states (of the 22 states and the District of Columbia for which these data are available), in 15 states and the District of Columbia, African American unemployment rates exceed white unemployment rates by a ratio of 2-to-1 or higher.
• The District of Columbia has a black-white unemployment rate ratio of 6.7-to-1, while Indiana and Louisiana have the highest ratios (2.6-to-1) among the states.
• The highest African American unemployment rate is in the District of Columbia (12.8 percent) followed by Louisiana (9.7 percent). The highest Latino state unemployment rate is in Connecticut (8.9 percent). In contrast, the highest white state unemployment rate is 5.0 percent, in New Mexico.
• While the Hispanic unemployment rate is at or below its prerecession level in nine states (of the 16 states for which these data are available), Colorado was the only state where the Hispanic unemployment rate was lower than the white unemployment rate.
• In three states, Hispanic unemployment rates exceed white unemployment rates by a ratio of 2-to-1 or higher (Massachusetts, 2.3-to-1; Connecticut, 2.1-to-1, and Pennsylvania, 2.0-to-1).

Background

In June 2017, the national unemployment rate was 4.4 percent, a slight decrease from 4.5 percent at the end of the first quarter of this year.2 State unemployment rates in June ranged from a low of 2.3 percent in Colorado and North Dakota to 6.8 percent in Alaska.3 According to a previous EPI analysis of unemployment by state, from March to June 2017, 33 states saw their unemployment rates decline, 15 states and the District of Columbia saw unemployment rates rise, and two states saw no change.

State unemployment rates, by race and ethnicity

EPI analyzes state unemployment rates by race and ethnicity, and racial unemployment rate gaps, on a quarterly basis to generate a sample size large enough to create reliable estimates of unemployment rates by race at the state level. We only report estimates for states for which the sample size of these subgroups is large enough to create an accurate estimate. For this reason, the number of states included in our map and data tables varies based on the analysis performed (unemployment rate, change in unemployment rate since the fourth quarter of 2007, and ratio of African American or Hispanic unemployment rate to white unemployment rate).

Trends among whites

In the second quarter of 2017, the white unemployment rate was lowest in South Dakota (1.5 percent) and highest in New Mexico (5.0 percent), as shown in the interactive map and underlying data table, which present state unemployment rates by race and ethnicity. South Dakota has had the lowest white unemployment rate for eight consecutive quarters, and New Mexico has had the highest white unemployment rate for two consecutive quarters.

Interactive Map

Table 1 displays changes in state unemployment rates by race and ethnicity from the fourth quarter of 2007 to the second quarter of 2017. The white unemployment rate remained most elevated above its prerecession level in New Mexico: 1.7 percentage points higher than it was in the fourth quarter of 2007. The white unemployment rate is at or below its prerecession level in 36 states, up from 30 states in the first quarter of 2017. The largest declines in white unemployment since the end of 2007 have occurred in Michigan (down 2.6 percentage points), Oregon (down 2.4 percentage points), Maine (down 1.9 percentage points), and Wisconsin (down 1.9 percentage points). The white unemployment rate is above but within 0.5 percentage points of its prerecession level in 11 states and the District of Columbia.

Table 1

Trends among African Americans

African American unemployment rate estimates are available for 22 states and the District of Columbia. During the second quarter of 2017, among states, the African American unemployment rate was lowest in Arkansas (5.6 percent) and highest in Louisiana (9.7 percent); in the District of Columbia, it was 12.8 percent. The District of Columbia also had the highest black unemployment rate during the previous two quarters, though among states Tennessee had the highest black unemployment rate in the first quarter of 2017.

In the second quarter of 2017, of the 22 states with African American estimates every state had African American unemployment rates below 10 percent; in 12 of these states, the rate was lower than the second quarter national average for African Americans (7.4 percent).

As shown in Table 2, which displays the black-white and Hispanic-white unemployment rate ratios in the second quarter of 2017, California, Missouri, and North Carolina had the smallest black-white rate gaps of the 22 states and the District of Columbia. In those states, black unemployment was 1.6 times the white rate. Delaware had the smallest black-white unemployment rate ratio at the beginning of the year (1.4-to-1). In the second quarter of 2017, as in the previous three quarters, the largest gap was in the District of Columbia, where the black unemployment rate was 6.7 times the white rate. The next highest unemployment ratios were in Indiana (2.6-to-1) and Louisiana (2.6-to-1).

Table 2

The black unemployment rate in the second quarter of 2017 was at or below its prerecession level in 14 states: Arkansas, California, Illinois, Indiana, Michigan, Mississippi, Missouri, New Jersey, New York, North Carolina, Ohio, South Carolina, Tennessee, and Texas. This is an increase from the 11 states that had reached this benchmark by the first quarter of 2017. However, with the exception of Texas, all of these states had black unemployment rates of at least 8.0 percent before the recession. Of the states in which black unemployment rates have recovered, four have black unemployment rates higher than the second quarter national average for African Americans (7.4 percent): Illinois, Ohio, Tennessee, and Texas. The black unemployment rate remains most elevated above its prerecession level in Delaware (3.6 percentage points higher).

Trends among Hispanics

Hispanic unemployment rate estimates are available for 23 states and the District of Columbia, and data on the change in Hispanic unemployment rates since the fourth quarter of 2007 are available for 16 states. In the second quarter of 2017, the Hispanic unemployment rate was highest in Connecticut (8.9 percent) and lowest in Colorado (1.9 percent). Connecticut replaced Pennsylvania as the state with the highest Hispanic unemployment rate, and Connecticut and Pennsylvania were the only states with Hispanic unemployment rates above 8.0 percent in the second quarter. Among states, Colorado also had the lowest Hispanic unemployment rate in the previous quarter.

The Hispanic unemployment rate is at or below is prerecession level in nine states: California, Colorado, Florida, Georgia, Illinois, Nevada, New York, North Carolina, and Washington. The Hispanic unemployment rate is above but within 0.5 percentage points of its prerecession level in Arizona, New Jersey, and Utah. The Hispanic unemployment rate was most elevated above its prerecession level in New Mexico (1.8 percentage points).

Colorado was the only state where the Hispanic unemployment rate was lower than the white unemployment rate (with a Hispanic-white unemployment rate ratio of 0.8-to-1). This is the second consecutive quarter Colorado had a Hispanic-white unemployment ratio of less than one. The ratio of Hispanic unemployment to white unemployment was highest in Massachusetts (2.3-to-1), followed by Connecticut (2.1-to-1) and Pennsylvania (2.0-to-1).

Trends among Asians

Asian unemployment rate estimates are available for 11 states and data on the change in Asian unemployment rates since the fourth quarter of 2007 are available for  eight states. The Asian unemployment rate was lowest in Massachusetts (2.1 percent) and highest in Washington (4.5 percent). Massachusetts has had the lowest Asian unemployment rate for two consecutive quarters. The Asian unemployment rate was below 5 percent in every state for which estimates are available. The Asian unemployment rate was at or below its prerecession level in California, Hawaii, Nevada, New York, and Texas. In Illinois, it was above but within 0.5 percentage points of its prerecession level. The Asian unemployment rate was most elevated above its prerecession level in New Jersey (1.2 percentage points higher).

Methodology

The unemployment rate estimates in this report are based on the Local Area Unemployment Statistics (LAUS) and the Current Population Survey (CPS) from the Bureau of Labor Statistics. The overall state unemployment rate is taken directly from the LAUS. CPS six-month ratios are applied to LAUS data to calculate the rates by race and ethnicity. For each state subgroup, we calculate the unemployment rate using the past six months of CPS data. We then find the ratio of this subgroup rate to the state unemployment rate using the same period of CPS data. This gives us an estimate of how the subgroup compares with the state overall.

While this methodology allows us to calculate unemployment-rate estimates at the state level by race and ethnicity by quarter, it is less precise at the national level than simply using the CPS. Thus, the national-level estimates may differ from direct CPS estimates.

In many states, the sample sizes of particular subgroups are not large enough to create accurate estimates of their unemployment rates. We report data only for groups that had, on average, a sample size of at least 700 in the labor force for each six-month period.

Endnotes

1. EPI analysis of Bureau of Labor Statistics Local Area Unemployment Statistics (LAUS) data and Current Population Survey (CPS) data.

2. “Labor Force Statistics from the Current Population Survey: Unemployment Rate,” Series Id. LNS14000000 [online data table], U.S. Bureau of Labor Statistics, August 28, 2017.

3. “State Employment and Unemployment—June 2017” [news release], U.S. Bureau of Labor Statistics, July 21, 2017.

The decline of the job seekers to job openings ratio to 2.0 continues the overall downward trend since the high of 6.8 to 1 in July 2009 (see Figure A). The ratio has steadily declined, falling by about 1.0 over the last year.

While this is clearly a move in the right direction, the 9.6 million unemployed workers understate how many job openings will be needed when a robust jobs recovery finally begins, due to the existence of 5.9 million would-be workers  (in August) who are currently not in the labor market, but who would be if job opportunities were strong. Many of these “missing workers” will become job seekers when we enter a robust jobs recovery, so job openings will be needed for them, too.

Figure A

Even further, a job opening when the labor market is weak often does not mean the same thing as a job opening when the labor market is strong. There is a wide range of “recruitment intensity” with which a company can deal with a job opening. For example, if a company is trying hard to fill an opening, it may increase the compensation package and/or scale back the required qualifications. Conversely, if it is not trying very hard, it may hike up the required qualifications and/or offer a meager compensation package. Perhaps unsurprisingly, research shows that recruitment intensity is cyclical; it tends to be stronger when the labor market is strong, and weaker when the labor market is weak. This means that when a job opening goes unfilled when the labor market is weak, as it is today, companies may very well be holding out for an overly qualified candidate at a cheap price.

Today’s report also marks the first time that the level of economic output has surpassed its pre-recession (fourth quarter of 2007) peak. Since quarterly data began being tracked in 1947, the U.S. economy has never taken this long to regain pre-recession output levels.

Final demand—a measure of output growth that strips out the influence of inventory investment (a particularly volatile component of GDP)—grew at a 3.6 percent rate in the quarter, as the rate of change in private inventories fell. While final demand has outperformed overall GDP growth in many recent quarters, this indicator also does not suggest a fundamentally healthy economy. Growth in final demand over the past year has been just 2.4 percent—still too low to lower the unemployment rate.

Growth in personal consumption expenditures—the single largest component of GDP—essentially matched the overall growth rate and contributed 1.7 percentage points to the quarter’s growth rate.

Business investment in equipment and software rose at a 17.4 percent rate in the quarter, its ninth straight quarterly increase and the largest increase since the second quarter of 2010. Business investment in structures rose for the fifth time in the last six quarters, suggesting that this component of business investment may have finally turned the corner after being depressed by a large inventory of corporate real estate produced by overbuilding before the recession.

Residential investment grew slightly (at a 2.4 percent rate) for the second straight quarter, the first consecutive quarterly increase since the last six months of 2005.  While this is encouraging, because the growth rates have been slow, and because this component of GDP shrank so much in the aftermath of the housing bubble, it is hard to see this sector making a major contribution to GDP growth for some time.

Net exports contributed 0.2 percentage points to the overall growth rate, the second straight positive quarterly contribution and the first two consecutive quarters of positive contribution since the first six months of 2009—during the recession. A key potential danger in the coming year is that very slow growth in the economies of major trading partners in Europe and Japan could be transmitted to the U.S. economy through slowing exports. So far, this does not seem to have provided a drag on overall growth, and this is a positive sign. Still, it seems likely that trade could be a major headwind in the coming year.

While the business investment and trade numbers are slightly more encouraging than in recent reports, the data on personal disposable income are not. Real disposable personal income fell at a 1.7 percent rate in the third quarter. This decline was led by a fall in real compensation and wages—the first such real decline since the last quarter of 2010. It should be noted that this real decline was not driven by a one-time price spike. The deflator for personal consumption expenditures (PCE) rose at an annualized rate of 2.4 percent in the quarter—down substantially from the growth rate of the previous six months.

The decline in disposable personal income meant that the rise in consumer spending in the quarter came largely from a 1 percentage point decline in the personal savings rate, which fell from 5.1 percent to 4.1 percent. If there is any lesson to be taken from the last three years, it is that economic growth resting on a foundation of ever-lower personal savings rates is inherently fragile. This is a prime concern raised by today’s report.

Lastly, the report shows that inflationary pressures remain tame. The one-year rise in the market-based PCE deflator excluding food and energy—a closely watched indicator of potential future inflation—rose only 1.6 percent. This low inflation rate, combined with only a 1.6 percent GDP growth rate over the same period (third quarter 2010 to third quarter 2011), argues that this remains an economy plagued by weak demand. Measures to boost demand are by far the most effective tools to bring the economy back to health.

Over this same period, the ranks of the unemployed swelled by 7 million to 14.5 million jobless workers, meaning that in May there were 12 million more unemployed workers than job openings – or 5.7 job seekers per available job (see Chart). The large increase in unemployed workers in May swamped the small gain in job openings, so the ratio of job seekers to job openings continued to climb sharply.

Although employment and unemployment numbers for June became available on July 2, JOLTS data are released with a one-month lag. Given last Friday’s announcement that the unemployed population increased in June by 218,000, along with the backsliding from May to June in the payroll survey, the ratio of job seekers to job opening most likely rose to more than 6 to 1 in June.

Unemployed workers are currently facing record high rates of long-term unemployment: in June 29% of the unemployed had been unemployed for more than six months. With nearly six times as many job seekers as available jobs, this number comes as no surprise.