Hostility in the workplace, including discrimination and sexual harassment, is not only a violation of a workers’ rights but also a threat to their economic security. Research finds that sexual harassment disrupts career advancement and causes financial stress for those who experience it, hampering worker productivity and economic mobility.
Sexual harassment can cost organizations as well. One study finds that sexual harassment in the U.S. military increases the likelihood of staff turnover, even when controlling for factors such as job satisfaction otherwise and organizational commitment. Another study finds that elevated staff turnover costs businesses not only via lost talent and productivity, but also through increased expenditures on recruitment and training.
Sexual harassment, of course, is now prohibited in the United States. Workplace sexual harassment falls under the anti-discrimination protections enshrined in the Civil Rights Act of 1964 and the Equal Employment Opportunity Commission’s 1980 regulation clarifying harassment on the basis of sex as protected under the Civil Rights Act. Additionally, in recognition of the economic costs to individuals who experience sexual harassment, the Civil Rights Act of 1991 established the right to sue for damages related to sexual harassment in the workplace. Yet despite these protections, workplace sexual harassment is pervasive across the U.S. labor market, hindering workers’ overall economic opportunities and serving as an obstruction to broadly shared economic growth.
One of the challenges to effective enforcement of these laws—and thus effective protection against sexual harassment in the workplace—is the prevalence of underreporting by workers. Research shows that a primary reason for not reporting sexual harassment is fear of retaliation, such as being fired, having reduced work hours, or being assigned undesirable tasks. These fears can be exacerbated when workers’ outside employment options are worse—for instance, when there are few available alternative employers or when the unemployment rate is high.
A new working paper, “Why is Workplace Sexual Harassment Underreported? The Value of Outside Options amid the Threat of Retaliation,” investigates the link between external economic conditions and the risk of retaliation for reporting sexual harassment. The researchers—Gordon Dahl of the University of California, San Diego and Matthew Knepper of the University of Georgia—hypothesize that claims are more likely to be found to have merit when underreporting is more likely. In other words, when outside options are worse—for example, when unemployment is high—only the most egregious instances of sexual harassment, and thus those most likely to be found to have merit, will be reported. Other workers who may be experiencing hostile workplaces will “tough it out,” given the risks associated with retaliation for having made a claim—leading to cyclical underreporting.
This initial proposition is evident when comparing changes in the unemployment rate and in the Google search intensity of the phrase “sexual harassment in the workplace.” From 2004 through 2016, Dahl and Knepper find these two track each other, while sexual harassment claims filed with the Equal Employment Opportunity Commission stay relatively level through this period, declining near the end of the time window as state-level unemployment rates also decline. The authors find that a 1 percent increase in the unemployment rate is associated with a 0.5 percent to 0.7 percent increase in the likelihood that the EEOC will find a sexual harassment claim has merit. These findings imply that underreporting of sexual harassment increases as labor market conditions worsen.
To further dig into the circumstances that lead to underreporting, the researchers examine the gender composition of workplaces and of management positions, finding that a greater share of men in a workplace and a higher proportion of men as managers in an establishment leads to more selectivity in the reporting of sexual harassment. Previous research has likewise demonstrated that sexual harassment is more common in male-dominated occupations and industries. As the authors note, “in male-dominated environments, female employees become increasingly reluctant to report despite an increase in [case] volume”—a trend that is further exacerbated by higher unemployment rates.
Dahl and Knepper also look at how policies that influence outside options impact underreporting of sexual harassment. Following the Great Recession of 2007–2009, for example, North Carolina reduced the level and duration of its Unemployment Insurance program and restricted its eligibility requirements, essentially worsening outside options for workers who lost their jobs. After this change, sexual harassment reporting selectivity in the state increased by 7 percentage points, or a 30 percent increase over nearby southern states that did not make changes to their Unemployment Insurance benefits. Projecting from these findings, the authors estimate that a $1,000 reduction in the maximum benefits available reduces sexual harassment reporting by 4 percent.
This research demonstrates how income support programs interact with U.S. labor market opportunities, giving workers more economic security to leave hostile work environments and find jobs that are a better match. It is consistent with research that has found that longer benefits duration improves job matches, with an even greater impact on job-match quality for women workers, non-White workers, and workers with lower levels of education. This further adds to a body of research that shows how income supports, including anti-poverty programs, provide a more stable foundation for workers to move into better-paying jobs, rather than serving as a work disincentive.
Furthermore, in a labor market defined by pervasive monopsony—when workers are constrained in how they can search for and move into new jobs—workplace hostility and limited outside options also may exacerbate an employer’s ability to undercut wages. In a competitive labor market, workers will seek out the jobs that are the best fit for their talents and where they can earn a good living. But if workers make employment decisions to avoid hostility, or if their bargaining power within workplaces is limited for fear of retaliation, employers have more power to set pay below competitive levels at a loss to both workers and the entire economy.
The new working paper by Dahl and Knepper reinforces the existing literature that connects underreporting of sexual harassment and broader U.S. economic and labor market circumstances. The breadth of evidence makes clear that the cost of sexual harassment goes far beyond individual instances of illegal conduct and the impact they have on the affected workers. Hostility in the workplace also leads to a less dynamic U.S. labor market, less stability and security for workers, and lower wages, with lost opportunities for broadly shared growth.
Cash transfer programs are a widespread economic policy tool often used to stimulate spending or encourage savings among consumers. The United States most notably uses direct cash payments as part of its fiscal stimulus programs during economic downturns, such as during the Great Recession of 2007–2009. More recently, the federal government paid more than $850 billion directly to U.S. households during the coronavirus pandemic through three rounds of economic impact payments between March 2020 and March 2021.
Yet these programs that send money directly to households are relevant in other economic contexts aside from acute economic crisis. More than 130 countries around the world use direct cash payments as part of their social infrastructure and anti-poverty programs. Additionally, many developed economies are considering so-called universal basic income policies, which would send cash payments to households on a regular basis.
Whether the goal is to alleviate poverty or encourage consumer spending, our new working paper shows that the success of these programs depends on when households actually receive the funds. Using real-world data from two widely different cash transfer programs, we examine whether the timing of one-time cash transfer payments relative to when they are publicly announced has large impacts on household consumption behavior.
Our study demonstrates that household spending does not merely depend on the size of the transfer but also on the amount of time households anticipate receiving their payment. We find that in both developing and developed country settings, households scheduled to receive cash transfers earlier are more likely to spend and less likely to save, compared to those who receive cash transfers later.
Below, we explain the methodology behind our working paper and its findings, before detailing its implications for policy design.
The timing of cash transfers impacts household consumption and savings behavior
Standard economic theory says that households may respond to “unanticipated” payments by increasing their spending patterns, but should respond little, if at all, to the arrival of “anticipated” payments. This suggests that households spend more when they receive an unexpected cash transfer but do not necessarily increase their spending if the payment was previously publicly announced.
The effectiveness of widely publicized, previously announced fiscal stimulus payments, however, suggests that this story is incomplete. Studies show that households still increase their spending upon receiving a payment even if they know about it in advance. What effect, then, does the amount of time that households spend waiting for the cash to arrive have on their spending decisions?
Our working paper seeks to answer this question by comparing two cash transfer scenarios that played out in different economic contexts. We look at data from both the 2008 U.S. economic stimulus payments that were disbursed amid the financial crisis of 2007 and the ensuing Great Recession, as well as data on the distribution of one-time cash transfers in randomized controlled trials in Kenya and Malawi.
In the first scenario, the timing of federal government tax rebates sent to low- and middle-income households in the United States during the Great Recession was based on the last two digits of the recipient’s Social Security Number. Over a 3-month period starting in May 2008, checks were sent out to households, which also received written notice from the IRS several days prior to their payment dates. We utilize this to compare weekly spending habits across similarly situated households receiving transfers in different weeks. Specifically, we look at three groups who received checks in the first, second, and third weeks of May. (See Figure 1.)
Figure 1
As Figure 1 shows, the spending response for households in each of the three groups receiving transfers differs across different weeks. Households randomly assigned—Social Security Numbers are effectively assigned at random—to receive payments at the earliest-possible date had double the spending increase over the 4 weeks after receiving their rebate, compared to the average household. In fact, we find that an additional 1 week of waiting for a transfer decreases the propensity to spend that transfer as much as would increasing the size of the transfer by $340. This suggests that the timing of payments plays an important role in their effectiveness at boosting spending.
We then examine the relevance of liquidity, or whether a household has at least 2 months of income available either in cash, a bank account, or otherwise accessible in case of an unexpected decline in income or increase in expenses. As seen in Figure 1, we find that these divergent spending habits by payment timing remain across households regardless of their liquidity status. While households that do have some available savings tend to spend less out of their stimulus payments and save more than those that do not have savings, the households with savings that received their payment at the earliest possible date spent about as much as households without savings that had to wait an extra 2 weeks to receive their payments.
Turning to the second scenario and context of our study, we analyze data from existing randomized controlled trials distributing one-time cash payments to households in Kenya and Malawi. These programs were not instituted as a response to economic crisis, as seen in the U.S. scenario, but rather as part of randomized controlled trials by researchers and nongovernmental organizations to understand the impact of cash transfers on households in poverty.
We find that the differential spending behavior by the timing of the transfer delivery seen in the U.S. context holds in these two other settings as well. In other words, we find that shorter waiting times between payment announcement and receipt led to increases in household spending rates, while those households that experienced delays in payments increased their total household savings rates.
Implications for designing cash transfer program policies
The results of our working paper have implications for designing cash transfer programs intended to boost consumer spending, as well as those that aim to alleviate poverty. Taken together, the data and our model suggest that a payment delay of even just 1 week can have a large impact on how much of that payment a household spends versus saves—an important distinction for policymakers, depending on their broader economic priorities.
Our study suggests, for instance, that households treat large, one-time payments, such as stimulus checks, differently from a smaller series of regular transfers. Researchers have argued that large payments tend to feel more like wealth and less like disposable income, so households may save more of a large lump-sum payment and spend more of a series of smaller payments. Yet in 2009 and 2010, when the United States implemented a series of small transfers to households with the Making Work Pay Tax Credit, these payments were only half as effective in stimulating U.S. household spending, compared to the one-time stimulus payments in 2008.
Our model addresses this tension by highlighting the crucial role of anticipation and timing. The unexpectedly diminished effectiveness for consumption of the smaller series of payments in 2009 and 2010 can be explained by the fact that households had more time to anticipate receiving those transfers, thus shifting their focus to saving rather than spending.
This implies that different household responses to both the size of a payment and how and when it is distributed must be taken into account when determining a policy’s effectiveness. Household spending behavior following one-time transfers, for instance, might not be an accurate indication of the efficacy of universal basic income policies—which tend to be smaller and delivered on a monthly basis rather than in one larger lump-sum—for reducing poverty.
Our findings also indicate that policymakers must consider their ultimate goals—boosting spending or building wealth—when designing the disbursement of cash transfers. If increasing consumption is the top priority, then cash payments should be rapidly sent out to households with little notice of their arrival. If longer-term investments and savings are the goal, however, then policymakers should consider providing more advance warning of the impending payments to households.
—Neil Thakral is an assistant professor of economics and international and public affairs at Brown University and the Watson Institute. Linh T. Tô is an assistant professor of economics at Boston University.
The U.S. economy added 467,000 jobs between mid-December and mid-January, according to the most recent Employment Situation Summary by the Bureau of Labor Statistics. Even with COVID-19 cases reaching an all-time high in the first weeks of January, employment growth surpassed projections. Upward revisions to previously published employment growth estimates for November and December now put the 3-month average at 541,000 jobs gained.
Almost 1.4 million workers joined the U.S. labor force last month, yet the overall unemployment rate rose slightly from 3.9 percent in December to 4 percent in January. The share of 25- to 54-year-olds currently working—an indicator also known as the prime-age employment-to-population ratio—rose slightly from 79 percent to 79.1 percent. Additionally, because of the omicron-fueled surge in COVID-19 cases, 3.6 million workers were absent from work due to illness in January.
In addition, jobs growth was varied for different demographic groups of workers. Employment gains for most groups was robust, but White workers saw a slight increase in unemployment and the employment rate for men was unchanged while continuing to improve for women. Yet Black women, who have seen the weakest jobs recovery since February 2020, saw a significant increase in their employment rate, alongside a decline in unemployment. (See Figure 1.)
Figure 1
Across sectors, jobs added were greatest in the leisure and hospitality sector and the professional and business services sector, which added 151,000 jobs and 86,000 jobs, respectively. The manufacturing sector also registered a net increase in employment in January, with 13,000 jobs added, but this key sector also suffered substantial disruptions due to the latest wave in the rolling coronavirus pandemic.
As the spread of the omicron variant picked up in December, the country reached a record number of daily cases, and millions of workers stayed home either because they were sick with the coronavirus themselves or because they were caring for someone who was. Understaffed factories and operations troubles—along with the pandemic-driven boost in demand for goods such as furniture and car parts—increased pressure on already-burdened workers and manufacturers.
In the manufacturing sector, job openings and quits are near record highs
Indeed, there are signs that the manufacturing sector is going through an important adjustment. Across industries, the number of workers quitting their jobs and the number of job openings have hit record-breaking highs within the past year, but no other industry has seen a larger increase in either its number of quits or job-openings relative to its pre-pandemic levels than manufacturing. (See Figure 2.)
Figure 2
There are a number of explanations for these record-high trends. One is that the pandemic made manufacturing an even more dangerous industry. Many manufacturing plants were designated critical infrastructure and allowed to continue operating during the peak of the early 2020 pandemic, when many other sectors were in lockdown, yet employers often did not provide personal protective equipment, follow social distancing protocols, or provide paid sick days to workers.
A Centers for Disease Control and Prevention report in 2021 examined high-density workplaces, specifically food processing and manufacturing plants, finding they were at high-risk for coronavirus transmission. Because only about a third of manufacturing workers can telework, the coronavirus crisis made an industry that already had high rates of injuries and illnesses even more dangerous.
Pay and job quality in the manufacturing sector has been in decline for decades
Other reasons why manufacturing workers are leaving their jobs at record rates are difficult work conditions, understaffing, care responsibilities, insufficient pay, and labor market conditions that are providing opportunities for workers to move on to other jobs. Many of these pay and job-quality problems, however, have plagued the manufacturing sector for decades.
A pre-pandemic study by the Economic Policy Institute found that, on average, manufacturing workers continued to have an advantage in terms of pay over comparable workers in other industries, but this premium declined between the 1980s and the 2010s. One of the key reasons behind the erosion of both of these job advantages, the study proposes, is an increased reliance on outsourcing, since workers employed through staffing agencies are paid significantly less than workers employed directly by manufacturing firms.
This decline in the manufacturing earnings premium, and in job quality more generally, set the stage for important disruptions during the coronavirus recession and the continuing pandemic today. In addition, the erosion in pay has been particularly stark for workers without a college degree and coincided with a generalized decline in labor standards, lack of government investment in infrastructure, and falling union membership rates both in the manufacturing industry and across the U.S. economy.
For instance, a recent analysis finds that lack of investment in infrastructure and trade imbalances have been greatly harmful for manufacturing-sector workers in general and for Black workers in particular, whose share of the manufacturing workforce saw a large decline in the 1990s and early 2000s. There is also evidence that the decline in manufacturing has widened Black-White income and employment divides. Research additionally suggests that in recent decades, employers have used automation technologies in manufacturing and other industries not to increase productivity but rather to de-skill jobs and lower wages.
While employment in the manufacturing sector is now near its pre-pandemic level, this erosion of the manufacturing earnings premium and bad-quality working conditions could lead to slow job growth in the sector in 2022 and beyond. While goods-producing sectors such as manufacturing did not see the massive employment losses that service-providing industries experienced in the first months of the pandemic, employment in the sector never fully recovered from the previous two recessions. Indeed, manufacturing-sector employment reached its peak in the late 1970s and has been declining somewhat consistently ever since. Currently, employment in the sector is 35 percent below its 1979 level and 1.8 percent below its February 2020 level. (See Figure 3.)
Figure 3
Boosting job quality and inclusion in the U.S. manufacturing sector is essential for workers and the economy writ-large
As the U.S. labor market continues to recover, it will be essential to boost efforts to protect workers and their communities through the effective enforcement of labor standards, both during the pandemic and beyond. Indeed, manufacturing sites have been a major driver of the pandemic’s spread, especially in the early days of the pandemic, and poor workplace protections continue to place the country’s supply chains at risk. Meatpacking plants in particular appear to have been a significant source of COVID-19 cases and deaths, especially in many rural areas.
More broadly, manufacturing companies need to come to terms with their history of exclusion and create pathways to correctly allocate talent and boost innovation to attract workers and foster economic dynamism. Policies that support organized labor and boost union membership will also create pathways to increased job quality and safety for workers in the sector. In addition, the federal government can support the creation of good jobs by, for example, building out much-needed infrastructure and buying manufactured goods from firms that pay high wages, provide good benefits, and provide quality workforce training. These measures will help create more good jobs for workers in the manufacturing sector, boost productivity, and drive broad-based and resilient U.S. economic growth.
On February 4, the U.S. Bureau of Labor Statistics released new data on the U.S. labor market during the month of January. Below are five graphs compiled by Equitable Growth staff highlighting important trends in the data.
The prime age employment rate edged upwards to 79.1 percent as the economy added more jobs than predicted in January.
Unemployment decreased for Black workers in January, converging slightly with other groups of workers. White workers had a slight increase in unemployment and Hispanic workers’ unemployment was unchanged.
Leisure and hospitality continued to lead employment gains, but job growth occurred across industries.
The employment rate for women increased 0.3 percentage points to 54.6 in January, while the employment rate for men was unchanged.
Among the unemployed, there was an increasing proportion who had left their jobs in January, signaling continued confidence in the labor market.
Rising income inequality and more unstable earnings in the United States today results in more precarious lives for low-income U.S. workers and their families. This earnings and income instability leads to swings in spending as well. Weekly spending on food, for example, jumps up when families enrolled in the Supplemental Nutrition Assistance Program receive their benefits electronically and then falls 20 percent the following week.
This type of food insecurity is linked to lowered health outcomes and lowered educational performance. When families face such consumption volatility, it prevents them from investing in their own well-being and future economic mobility, which, in turn, constrains overall U.S. economic growth across the country. This is why policymakers need to understand that spending volatility is central to the economic well-being of U.S. workers and their families.
In a recent working paper, we examine both the level and volatility of consumption across U.S. income and socioeconomic distributions to understand how consumption volatility and income intersect. Using quarterly spending data from the Consumer Expenditure Survey, we look across several categories of consumption central to the daily lives of families.
We use quarterly volatility because annual volatility may miss important changes within a year. While families may spend more on food in one year compared to the next, this may miss times when food insecurity within that year impacts families’ well-being and could have other effects, such as on children’s development amid, say, the school year.
Concerns over the downside consequences of consumption volatility are warranted for several reasons. Low- and moderate-income families—the latter of whom are often living on the edge of low-income earnings—might simultaneously consume lower levels of goods and services on basic needs (think food and apparel), as well as expenditures on entertainment and alcohol. To discern these spending patterns, we assess differences in average expenditures across income deciles for food, clothing, entertainment, and alcohol. We find clear income gradients across all categories. (See Figure 1.)
Figure 1
Figure 1 demonstrates that higher-income households spend more, but that’s not the whole story. We turn next in our working paper to understanding how consumption volatility differs across the income distribution and by family socioeconomic characteristics. We define volatility using a measure of transitory volatility. To calculate transitory volatility, we first find the average spending for each family over the four quarters. Next, we subtract each quarter’s spending from the average for that family. We then square this value and add up the squared differences for each quarter. Lastly, we divide this sum of the squared differences by three. This measure captures how spending in one quarter differs from average spending over the four quarters we observe among U.S. families. Larger values indicate higher spending volatility.
We find that food volatility is higher at the bottom than the top of the income ladder. When compared to base levels for households at the top, food volatility is more than twice as high for those households at the bottom of the income distribution. The bottom 20 percent experience higher consumption volatility than those with higher income, indicating that there is potentially worrisome food volatility, not solely for the poorest of the poor but also among the bottom 20 percent of U.S. households by income. (See Figure 2.)
Figure 2
Differences in spending on different things by category also are telling, as are the reasons for the differences. Apparel represents clothing for all household members, including footwear, watches, and jewelry. Lower-income households may be able to forego apparel spending during lean times, which would lead to higher observed volatility. The apparel category follows the same pattern as the food category, insofar as consumption volatility for apparel is highest among the poorest families. For entertainment, the bottom decile exhibits higher consumption volatility relative to the top income decile, followed by a gradual decline in volatility.
Alcohol represents a departure from all previous results. Alcohol volatility is not highest for households at the bottom of the income ladder. Relative to those households at the top, alcohol volatility is lower at all deciles and lowest at the bottom. That said, alcohol consumption is a tiny fraction of spending for the bottom of the income distribution, as shown in Figure 1.
Our working paper sought to examine consumption volatility across the income distribution, but our findings also have important consequences for a broader understanding of how sociodemographics, such as educational attainment and race, and consumption are related. Broadly, those with less than a high school degree face greater volatility across all spending categories, except alcohol. Those with a college degree or higher face the lowest volatility across all consumption categories.
The link between consumption and educational attainment, the latter of which is highly correlated with income and race, can be viewed as a proxy for exposure to employment risks in an increasingly polarized U.S. labor market, where higher-level credentials lead to greater employmentstability. Lower-wage workers have higher volatility in hours worked, reflecting lowered union bargaining power and worker protections.
Similarly, households headed by Black individuals or by other non-White individuals have higher consumption volatility than households headed by White individuals, except for alcohol. Importantly, this is consistent with findings from other economic research on income volatility, which generally finds a similarsociodemographic pattern. This evidence is a useful complement to work showing that Black families, on average, have loweraccess to credit and lower wealth to buffer against labor market volatility. They are also more likely to be exposed to broader labor market risks to hours stability associated with low-wage work.
Consumption volatility is higher for Black individuals at the middle and top of the income distribution. Consumption volatility is relatively lower for Black individuals in the bottom of the distribution. This finding, alongside other research highlighted above, suggests that broader characterizations of well-being beyond income level are required to more fully capture economic security across race. Importantly, many middle- and high-income Black families lack the cushion from wealth to absorb income fluctuations and smooth consumption.
Overall, our findings provide convincing evidence that some of the most essential categories of consumption exhibit the highest volatility among lower-income households—households that were already consuming at relatively low baseline levels. Consumption volatility among the nation’s lower-income and socioeconomically disadvantaged families pose potentially serious consequences not just for their own overall economic security, such as worse health outcomes and lower educational performance, but also for their future well-being and economic mobility.
This damaging interplay between income inequality, consumption inequality, and economic security harms the broader U.S. economy and its potential for sustained, long-term growth. Policymakers need to take this evidence into consideration when designing income supports and social infrastructure investments to improve the livelihoods and well-being of U.S. workers and their families and the broader U.S. economy.
As more and more job-seekers in the United States went online to look for work in the late 1990s and early 2000s, several scholars wrote about the potential of the internet to change the process by which workers search and find jobs. By allowing candidates and recruiters to access an unprecedented amount of information about each other, technologies such as online job boards could lower barriers that make it difficult for workers and employers to meet and establish valuable employment relationships.
Moreover, if these new job-search methods led to the creation of more and better worker-employer matches, then their widespread use also could have implications for economywide dynamics. The impact of these new technologies could affect, for example, the aggregate level of unemployment, wage inequality, and U.S. economic growth.
Fast-forward to today. Online job searches are the main way in which U.S. workers look for jobs. The effect of this search method on workers’ economic outcomes and on broader labor market dynamics, however, continues to be considered and discussed. On the one hand, online job searches shorten the amount of time it takes job-seekers to find work, and it appears to make job switching easier and more likely. On the other hand, advertising job openings and applying for open vacancies is much faster and cheaper now than it was two decades ago, creating new frictions in the U.S. labor market, including barriers that make it difficult for job-seekers and employers to find each other and create good matches.
Focusing on the perspective of job-seekers, this issue brief examines how online job boards and other information technologies change the job-search process, affecting workers’ labor market outcomes and influencing other dynamics, such as aggregate unemployment and wage inequality. We also examine how online job-search experiences are changing and the obstacles job-seekers face when looking for work, including racial and gender discrimination, information asymmetries, and the lack of transparency in and around the job-search process.
We close the issue brief with an overview of policies that could render online job searches and application processes more effective, more equitable, and more capable of matching the best workers to the best jobs by, for example, more effectively protecting workers against discrimination in the hiring process and ensuring all U.S. households have access to high-quality internet services. These key social infrastructure investments and more efficient enforcement of labor standards in the United States would help boost more broad-based, and thus more sustainable, U.S. economic growth.
Job-search frictions can make finding work costly and time consuming
In a hypothetical perfectly competitive labor market, workers seamlessly match into open vacancies when wage offers are higher than their reservation wage—economic parlance forthe lowest wage at which any given worker will be willing to accept a job. In turn, the theory is that the wages that employers offer are determined by economywide competitive forces of supply and demand for labor. The sum of these dynamics leads to market clearing, where all workers have jobs at wages equivalent to the value they contribute to production.
This way of understanding the labor market relies on the obviously spurious assumption that both workers and employers have perfect information. In this theoretical framework, unemployment ostensibly only exists inasmuch as it is a temporary phenomenon while workers are switching jobs.
Because of the clear limitations to this framework, in the 1960s and 1970s, several economists started to develop a “job search theory” to model a more realistic labor market—one in which workers may search for jobs and not immediately find them. Along with the insight that in virtually all markets, buyers and sellers incur costs due to searching to find one another, one of the key contributions of job-search theory is that it takes time and effort to find an employment opportunity that is a good fit because, among other obstacles, workers have imperfect information about jobs and wages.
By incorporating these barriers into their analyses of the labor market—obstacles that social scientists call “search frictions”—researchers are able to account for the often costly and time-consuming processes in looking for a job, as well as looking for workers. Information asymmetry, lack of information, and belated informationare all examples of search frictions.
Workers, for instance, might not realize a workplace environment is hostile when looking or applying for positions, might not have the know-how on looking for good employment opportunities, or might be unsure if a vacancy is a good fit for their interests, skills, and desired compensation. Some groups of workers also can be especially likely to face obstacles during their job searches due to discrimination, hostility, or social norms.
A recent study using French labor market data finds that even in the absence of outside options—other job opportunities that are available and appropriate for a given worker—women are more likely to quit their jobs due to sexual harassment, resulting in differences in labor market flows between women and men. Similarly, circumstances that impact the supply of labor also may reduce workers’ ability to look for work. Research shows, for instance, that the combination of caregiving responsibilities and transportation costs mean low-income workers tend to face big time constraints that limit their abilities to search for jobs.
Search frictions also have important implications for workers’ wage outcomes. Workers who look for new opportunities while employed, for example, have greater wage bargaining power with their current employers and other employers’ offers. In contrast, significant search frictions lower workers’ so-called labor supply elasticity—the extent to which they will respond to a wage change by leaving jobs with diminishing value of wages or going into jobs with better pay. When workers face a hard time finding new jobs and are less responsive to lower wages or deteriorating working conditions, those dynamics, in turn, give individual employers the power to undercut pay.
These barriers in the U.S. labor market also have implications for economywide trends and dynamics. While standard economic theory proposes that aggregate levels of employment and unemployment are determined exclusively by the supply and demand of labor, incorporating search-and-match frictions into analyses of the labor market helps address questions, such as:
How public policies, such as robust social insurance programs, help workers toward better-paid employment relationships and improve the functioning of the labor market
Search theory also is applied as a tool to model dynamic monopsony in an imperfectly competitive labor market by giving economists the measurement tools for the mechanisms that reduce competition for workers and give space for employers to undercut wages without losing their entire labor supply.
Even with modern internet technologies, all of these search frictions are present in the dominant way in which workers now seek new jobs in the U.S. labor market—through online job searches. For instance, over the past decade or so, the proliferation of online job searches also created new types of informational frictions, including phantom vacancies—job advertisements that remain posted online even after the position is filled. This new development makes it difficult for workers to know which listing is a true vacancy and which is a phantom, turning job searching into a less efficient process.
The promise of online job search
In the early 2000s, many economists wrote about the potential of the internet to smooth search frictions and allow more workers and firms to meet and establish productive employment relationships. Compared to the “help wanted” ads in newspapers, online job posts are easier to search for, may contain more information about a position (especially for those job-seekers who previously did not have access to job information through their networks), and can be found by a far larger pool of candidates.
For employers, too, the benefits of the internet seemed promising. Posting a job opening online offers more tools to screen candidates, is much cheaper than advertising a vacancy in print, and makes it possible to edit and update ads as needed. Compared to referrals, job-matching technologies could be an equalizing force, given that there is evidence that searches based on who employees or employers know can exacerbate racial and income disparities.
The widespread use of the internet, therefore, was expected to smooth search frictions and allow more workers to land positions that were a good fit for their interests and skills. As both the search and recruiting processes became faster, cheaper, and more efficient, the rise of online job search also was expected to have implications for broader labor market dynamics and possibly reduce economic inequality. In the early 2000s, a number of economists argued that online job search had the potential to drive greater productivity, lower frictional unemployment, and aggregate wage gains.
In a 2001 essay, for instance, economist David Autor at the Massachusetts Institute of Technology proposes that online job search should—in theory—make the process by which workers and job openings match more efficient:
Job boards and other Internet labor market connections should increase the efficiency with which workers are matched to jobs … Because workers and firms can consider more potential matches more rapidly, their reservation match quality—the minimum productivity an employer will tolerate, or equivalently, the minimum wage a worker will accept—both rise. Higher match quality raises output, and worker earnings and firm profits rise accordingly. In general, lower search costs will also reduce unemployment. (See page 27.)
Autor and other scholars also point out, however, that the rise of online job searches could create a new set of frictions and inefficiencies in the labor market, including greater wage inequality and an excess of low-quality information. If online job searches led to more valuable employee-employer matches, then the disparities in internet use and access along the lines of race, income, and formal educational attainment could create further disadvantages for already-marginalized U.S. workers and households. In addition, if the cost of applying for jobs and advertising vacancies fell, then more information could come at the expense of its quality.
Indeed, in the early 2000s, some employers were already reporting receiving a large—and often unsustainable—number of job applications. At the same time, job-seekers began to encounter screening and tracking technologies that arbitrarily remove many candidates from the hiring process. Those looking for work also started running into a new set of barriers in their job search. For instance, the aforementioned phantom vacancies make the job-search process more costly, discouraging, and difficult to figure out.
Job-seekers using the internet to look for work skyrocketed, but there exist important disparities in access to online resources
Over the past two decades, there has been a massive increase in the share of workers using the internet for job-seeking purposes. Online search is now the most widely used job-search method in the United States. While only 26 percent of unemployed job-seekers and 11 percent of employed job-seekers reported using the internet to look for work in 2000, by 2011, those numbers had climbed to 76 percent and 38 percent, respectively. More recently, a 2015 survey by the Pew Research Center finds that 54 percent of all U.S. adults had used the internet to look for job information and 45 percent had applied for a vacancy online. When narrowing respondents to those who had looked for work over the previous 2 years, 84 percent reported having applied for a job opening online.
While online job searches are much more common now than in the early 2000s, there are important differences in the likelihood of using the internet to look for work across demographic groups. For instance, the same Pew Research Center survey shows that younger adults, Black adults, adults with higher levels of formal education, and adults with higher incomes are all especially likely to use online resources for job hunting. (See Figure 1.)
Figure 1
Some of these differences are driven by disparities in access to high-speed internet since job-seekers without broadband are much less likely to go online as part of their search. According to a recent study, in 2018, almost 45 percent of U.S. households with children and an annual income of less than $25,000 did not have high-speed internet at home. About a third of households with an income between $25,000 and $50,000 lacked access to high-speed internet. Conversely, only 8 percent of households with an annual income of more than $150,000 lacked this service. Across race and ethnicity, disparities are also stark, with American Indian and Alaska Native households experiencing the lowest rates of high-speed internet access. (See Figure 2.)
Figure 2
But job-seekers do not necessarily need either high-speed internet or even a desktop or laptop computer to go online to look for work. In 2015, more than a quarter of U.S. adults had used their smartphones for job-seeking purposes. Of those smartphone job-seekers, 94 percent said they had used their phone to search for job listings, and at least 74 percent had used it to connect with a potential employer.
In addition, a substantial share of adults also reported using their phone for more complicated tasks. Among job-seekers using their smartphone, 50 percent had used it to complete a job application online, and 23 percent had used it to create a resume or write a cover letter.
Adults with lower levels of formal education were more likely to rely on their smartphone for these more intricate steps in the job-search and application processes—a finding that reflects a disadvantage for job-seekers with lower levels of education since smartphone job-seekers are likely to run into problems, such as trying to access online content that is not smartphone-friendly or having trouble uploading supporting documents for a job application.
Online job search seems to have become a more effective method to look for work
As internet use rose and online job search became an increasingly popular method to look for work across the United States, a quickly rising number of job-seekers believed that online job searching was a good way to find and apply for employment opportunities. An early survey finds that the share of employed workers reporting that the internet was the primary method they used to get their current job skyrocketed from 6 percent in 2000 to 22 percent in 2002, lowering job-seekers’ reliance on more traditional search methods. As such, over those 2 years, the share of workers pointing to personal referrals as their main job-search method dropped from 56 percent to 44 percent, while reliance on newspaper ads fell from 27 percent to 24 percent.
While early studies generally find that online job searches did not improve unemployed workers’ chances of landing a job, there is evidence that it was an effective job-search method for at least some candidates. In a study using data from the early 2000s, economist Betsey Stevenson at the University of Michigan finds that internet use may have reduced the cost of searching for other opportunities for already-employed workers, thus making job switching more likely, reducing the probability of experiencing unemployment in the first place, and increasing their bargaining power.
Further, more recent research generally finds that those who use the internet for job-seeking purposes have higher job-finding rates than those who do not. Peter Kuhn at the University of California, Santa Barbara and Hani Mansour at the University of Colorado Denver find that the effectiveness of online job search increased between 1998–2000 and 2008–2009. While internet job searches appeared to be counterproductive in the former period, the authors show that by 2008–2009, young jobless workers using the internet to look for opportunities experienced unemployment spells 25 percent shorter than otherwise-similar workers using more traditional search methods.
Similarly, research using 2011 data by Jason Faberman and Mariana Kudlyak at the Federal Reserve Bank of Chicago finds that internet job search increased unemployed workers’ chances of finding a job within a year by 25 percent, compared to jobless workers who did not use the internet in their job searches.
The effectiveness of the internet as a job-search method that seemingly increased throughout the 2000s and early 2010s could be associated with improvements in the design of job boards, as well as with what economists and other social scientists call “network externalities” or “network effects.” In other words, a greater number and a wider variety of job ads got posted online and the user-experience of job-matching sites improved internet job searches such that they became a more valuable and efficient method to look for work.
The effect of online job search on broader labor market trends and dynamics
As online job searches became an increasingly prevalent way to look for work, scholars also wrote about the potential of the internet to have an effect on overall labor market trends and dynamics, such as aggregate unemployment and inequality in earnings and employment. Let’s look at each of these in turn.
U.S. unemployment rates
By reducing search frictions and making it less costly for workers and firms to find each other, online job searches also were expected to influence economywide dynamics, such as the aggregate unemployment rate. For example, in 2000, UC Santa Barbara economist Kuhn writes:
Unless the increased efficiency of an Internet search draws large numbers of new workers into unemployment to look for new jobs, and unless both firms’ hiring standards and workers’ reservation wages rise by so much as to completely eliminate this first-order effect, the equilibrium unemployment rate should fall. (See page 43.)
In other words, by reducing search frictions and making it less costly for workers and firms to find each other, online job searches also were expected to influence economywide dynamics, such as the aggregate unemployment rate. Surprisingly, however, over the past 20 years or so, researchers have not found evidence that job boards or other improvements in information technologies have had an effect on overall U.S. joblessness.
At least not yet. Research examining the internet’s effect on the Norwegian labor market, for example, finds that the roll-out of broadband infrastructure lowered the average duration of vacancies and lowered the percent of establishments with unfilled positions. It also led to higher job-finding rates and starting wages, and to a decline in the country’s unemployment rate.
Still, evidence points to U.S. unemployment rates not being particularly affected by online job searches. In one of the first studies analyzing the effect of online job search on local labor markets, Kory Kroft at the University of Toronto and Devin Pope at the University of Chicago examine whether the expansion of Craigslist—a website in which users advertise jobs, apartment rentals, and items for sale—made the matching process in the job and apartment-rental markets more efficient. The authors find that as Craigslist was introduced into a city, there was a substantial decline in the number of vacant rental units.
This decline, Kroft and Pope propose, reflects that the introduction of the website led to an improvement in the matching process in the apartment rental market. In contrast, the expansion of Craigslist did not affect the labor market. While the website caused a shift away from print and toward online job ads, it did not have a meaningful effect on local unemployment rates.
There are a few potential reasons why online job search may lead to fewer unemployment episodes and shorter unemployment spells but not to a decline in aggregate joblessness. A team of researchers finds, for example, that use of broadband at home or at public locations, such as libraries, reduces the likelihood that unemployed workers become discouraged and drop out of the labor force altogether. As such, information obtained online could keep unemployed job-seekers from believing there are no employment opportunities available.
As a potential reason why Craigslist led to greater matching efficiency in the apartment and housing rental market but not in the labor market, Kroft and Pope propose that while online job posts are less expensive than posting a vacancy in print, online job ads might not substantially improve the information available for job-seekers. The reason, the authors suggest, is that the most valuable information about a vacancy—specific details about the position, such as salary level, benefits, company culture, and opportunities for growth—tends to be communicated during interviews or in other stages of the hiring process. Alternatively, any reduction in the unemployment rate could be offset if Craigslist led to greater competition for any given job opening by reducing the time and effort needed to search and apply for open positions.
Another potential explanation is that if there is not much frictional unemployment in the economy to begin with, then the effects of the internet on the jobless rate will be limited. Then, there are the findings by Christine Fountain at Fordham University, who proposes that while the proliferation of online job boards makes it easier for employers to reach more candidates and for workers to find more vacancies, the proliferation of information may come at the expense of its quality.
Contrary to UC Santa Barbara’s Kuhn and U-Colorado’s Mansour, Fountain argues that as more U.S. workers went online to look for work, online search became less effective as a job-search method. She argues that online search lost its usefulness among job candidates to signal their technological literacy to potential employers. Further, she posits that as the use of the internet for job-search purposes jumped, it created a potential problem for firms hoping to hire—too many applicants.
Earnings and employment inequality
There are a few ways in which researchers had expected online job searches to either ameliorate or exacerbate economic inequality. First, if using the internet to look for work resulted in higher rates of workers finding new jobs and in more valuable employee-employer matches, then disparities in access to the internet would create further disadvantages for job-seekers who did not have access to high-quality online resources.
In addition, if more and better information made it cheaper for workers to switch jobs, then some researchers expected firms might have to increase pay for “star” workers in order to keep them from moving on to other jobs. Employers would therefore need to tie pay more closely to each worker’s productivity—a dynamic that would drive greater wage inequality. Or, if online job ads led to either an under- or over-qualified pool of candidates, then employers might have to rely more on personal referrals to hire, exacerbating the disparities associated with using personal networks to look for and find work opportunities.
Conversely, other researchers proposed that internet search could ameliorate disparities in economic outcomes if online job platforms and other websites encouraged more employers to post wage and salary information for their open positions. Through greater pay transparency, the internet could lead to greater wage compression since there is evidence that policies that encourage or mandate salary disclosure can lead to narrower gender pay divides, and can pressure employers into increasing pay for workers who are unfairly undercompensated.
Currently, there is little empirical evidence on the relationship between inequality and online job searches specifically. But over the past few years, some studies have examined whether the expansion of internet services has an effect on labor market disparities. In a 2021 paper, for example, economist George Zuo at the RAND Corporation examines how access to broadband affects low-income families. Zuo studies the roll-out of Comcast Corporation’s Internet Essentials—a service that offers subsidized broadband to low-income households—finding that in places where the service is available, eligible individuals saw an increase in earnings and employment rates.
Similarly, a study by Hilal Atasoy at Rutgers Business School finds that during the late 1990s and early 2000s, federal policy programs aimed at expanding broadband internet access led to a substantial increase in the employment rate, especially in rural areas. An important chunk—about 40 percent—of the estimated employment gains were driven by greater labor force participation, suggesting that a decline in job-search costs was one of the drivers of the increase in employment.
New research also suggests, however, that the returns to internet access can be unequal. In the same study, Atasoy finds that, all else being equal, the positive effect of broadband on employment was greater in counties and industries in which a larger share of the workforce had a college degree. Similarly, in a recent paper, Paolo Martellini at the University of Wisconsin-Madison and Guido Menzio at New York University propose that search frictions have declined due to advances in communication and information technologies, leading to higher-quality worker-employer matches and greater labor productivity.
Yet Martellini and Menzio argue that the now-lower barriers to finding an appropriate job have benefited some workers more than others. Specifically, they find that:
For workers who are specialists—in the sense that their productivity varies a great deal across different jobs in their labor market—the decline in search frictions leads to high productivity and wage growth. For workers who are generalists—in the sense that their productivity is similar across different jobs—the decline in search frictions leads to minimal productivity and wage growth. Thus, declining search frictions leads to lower growth for “jacks of all trades” (the generalist workers) and higher growth for “masters of one trade” (the specialized workers). (See page 1.)
Online job search created new avenues for employment discrimination
Prior to the passage of the Civil Rights Act of 1964 and other landmark anti-discrimination laws, such as the Age Discrimination in Employment Act of 1967, employers could—and often did—overtly discriminate against candidates by specifying the age, race, gender, national origin, and other demographic characteristics required of job applicants. While U.S. law now bans employers from posting job ads that either discourage or show preference for candidates due to their race, gender, or another protected characteristic, hiring discrimination remains a pervasive feature of the U.S. labor market. In addition, job-search platforms and other hiring intermediaries create new avenues for employers to discriminate when advertising job openings online.
Social media platforms, for example, have a wealth of information about individuals and their online activities, and can enable hiring discrimination by either highlighting or hiding job ads from different groups of workers. A 2017 investigation by The New York Times and ProPublica finds that big employers, such as Verizon Communications, Inc., The Goldman Sachs Group, Inc., and Target Corporation, used Facebook to promote job advertisings only in the feeds of younger users, excluding potential candidates age 40 or older.
Similarly, a recent analysis by The Brookings Institution finds that while Facebook’s online ad service no longer allows for the targeting of members of specific affinity groups, the platform allows advertisers to target ads according to users’ online behavior and other interests. For instance, the social media platform offers the option to target “people who have expressed interest in or like pages related to African-American culture,” which the analysis found could be even more accurate in targeting African American users than previous demographic-based targeting options.
As the volume of online job postings grows, job-search platforms also play a more active role in the job-search and application processes, shifting from more passive job boards to providing more dynamic matching and recruiting services. Specifically, online job platforms actively mediate the job-search process by recommending postings to candidates based on their resume or other attributes, suggesting recruitment targets to employers, and selecting which applicants to show or recommend to the company that is hiring. At each of these steps, the use of artificial intelligence and other predictive technologies can reproduce, exacerbate, or introduce algorithmically based discrimination and bias by race, ethnicity, gender, age, disability, or other characteristics throughout the search and application process.
Automated recruitment and hiring systems can exclude qualified applicants and inhibit quality job matches
While this issue brief focuses on the use of third-party job-search platforms and is limited to the job-search stage (as opposed to the application, interview, and hiring processes), the issues discussed here are also present in employer-developed recruitment and hiring processes. Indeed, exclusionary practices by employers, whether algorithmically driven or not, are likely reducing the likelihood of quality job matches and exacerbating harmful credentialism.
A recent report by Joseph Fuller and Manjari Raman at Harvard Business School and Eva Sage-Gavin and Kristen Hines at Accenture suggests that automating recruiting and hiring systems that screen out qualified candidates is a widespread issue among middle- and large-sized companies. The authors note that such automated systems were designed to sort through large numbers of potential candidates quickly and efficiently, reducing costs and recruiter time, and often filter out or ignore qualified job-seekers based on overly restrictive, unrelated, or inflexible criteria.
The need to self-market on online job boards also can lead to the exclusion of and discrimination against some groups of workers. A report by Data & Society finds that platforms such as Care.com, Handy, UrbanSitter.com, and Uber—platforms that serve as intermediaries between clients and workers—require that job-seekers have self-branding and social media skills in order to be visible to potential clients, creating challenges and barriers for workers who might not be as digitally fluent. These workers are disproportionately likely to be older workers, lower-income workers, Black or Latino workers, and workers who do not speak English as a first language. As such, the skills, tools, and resources needed to navigate online platforms can exacerbate disparities in the job-search process.
Policies to address the digital divide are necessary but insufficient to address disparities and discrimination in the job-search process
As recently as 2015, the McKinsey Global Institute, the research arm of the business consulting firm McKinsey, released a report claiming that online job-search platforms, such as LinkedIn and Monster.com, had the potential to boost economic growth by allowing for more and better employee-employer matches, faster job creation, and by drawing more people into the labor force. Both because U.S. workers tend to switch jobs more often than workers in other high-income countries and because online “talent” platforms give job-seekers tools to search for opportunities more efficiently, the report argues, the productivity gains of online job search could be especially large for the United States.
So far, however, there is no clear evidence that either online job search or the internet more generally have made the worker-employer matching process more efficient or led to more competitive, better-functioning labor markets. The widespread use of online job search therefore seems to have ameliorated only some of the obstacles that make it difficult, costly, and time-consuming for workers to look for work while exacerbating other barriers.
That being said, most job-seekers in the United States now go online to look for work. Even when jobs do not require computer skills, the job-search and job application processes likely do. Because internet and computer access are now central to finding a job, guaranteeing access to affordable, high-speed internet for all workers and families is essential for ensuring more equitable access to employment opportunities. It is especially important to ensure that low-income households and households of color have access to high-quality internet services. The Infrastructure Investment and Jobs Act, which was signed into law in late 2021, is an important step in the right direction, with billions of dollars in funds appropriated for broadband infrastructure.
In addition, the enforcement of anti-discrimination laws needs to account for recent developments in the job search, application, and recruiting processes. The greater volume of online job ads and applications has led to the increasing use of algorithms and artificial intelligence tools, allowing, in turn, for practices that both replicate existing discriminatory and exclusionary practices and introduce new ones. The federal Equal Employment Opportunity Commission—the country’s most important anti-discrimination enforcement agency—is taking essential first steps by, for example, launching initiatives that look to educate employers, offer technical assistance, and identify promising practices to tackle algorithmic discrimination.
Beyond policies designed to improve technological integration into job-search processes, research also shows that a variety of other factors related to a worker’s bargaining power and economic security can lead to higher-quality matches and more equitable outcomes. In addition to increasing worker power and updating laws and regulations to account for recent technological developments, social infrastructure remains one of the most effective tools for improving job searches in the presence of persistent frictions.
The broader structural frictions in the U.S. labor market that hinder competition also need to be addressed. Collective actions, such as strikes, and labor unions improve efficiency and social welfare in the U.S. labor market by reducing firms’ ability to set wages. But collective actions need the support of legislation and institutions, such as the National Labor Relations Board, in order to be most effective in leveling the playing field between workers and employers. Collective action also needs to be matched with antitrust enforcement and a pro-competition policy agenda to push back against labor market monopsony, alongside policies that enable the rebuilding of an inclusive labor union movement. Only then will advances in job-search technologies deliver gains for workers and the U.S. economy writ-large.
Every month the U.S. Bureau of Labor Statistics releases data on hiring, firing, and other labor market flows from the Job Openings and Labor Turnover Survey, better known as JOLTS. Today, the BLS released the latest data for December 2021. This report doesn’t get as much attention as the monthly Employment Situation Report, but it contains useful information about the state of the U.S. labor market. Below are a few key graphs using data from the report.
The quits rate ticked down slightly to 2.9 percent in December, as the number of quits fell 161,000 to 4.3 million.
With job openings fairly steady at 10.9 million and hires decreasing 333,000 to 6.3 million, the vacancy yield declined to 0.57 in December.
The ratio of unemployed-worker-per-job-opening remains low, falling further from 0.63 unemployed workers per job opening in November to 0.58 in December.
The Beveridge Curve continues to be in an atypical range compared to previous business cycles, as the unemployment rate declined slightly in December and the job openings rate remained elevated.
Job openings slowed in several sectors—such as education and health services, manufacturing, and trade, transportation, and utilities—that have seen elevated job openings relative to pre-pandemic levels.
The U.S. Bureau of Economic Analysis, in late 2021, updated its data series on income inequality in the United States. This update adds data for 2019 and extends the data back to 2000, making this a very useful series for understanding how economic inequality evolved over the past two decades.
Below, we’ve assembled six charts that show how high-, middle-, and low-income households experienced inequality over the first two decades of the 21st century.
Inequality is rising
The new data show that economic inequality continues to rise. Our first chart shows growth in disposable personal income in each year between 2001 and 2019. Each bar is subdivided to show where growth went. The blue sections indicate growth that went to households in the bottom half of the U.S. income distribution. That is, blue bars represent all households in the country that earn less than the median household income. The warmer colors—yellow, red, and orange—represent groups with higher incomes. Notably, significant percentages of growth in every year flow to the top 10 percent of households, which are represented on the graph by the yellow and red sections of each bar. (See Figure 1.)
Figure 1
One easy way to see that inequality is increasing is to compare average personal income growth—which is the number most commonly reported by the media—to growth for particular groups of U.S. households. Our second chart breaks the population into deciles of income, with the lowest-income households on the left side of the graph, showing that for the vast majority of Americans, “headline” growth in disposable personal income overstates income growth for households in their decile. Only the very highest decile beats average income growth, and those in the top 5 percent and top 1 percent of the distribution beat it by large margins. (See Figure 2.)
Figure 2
Inequality rose at a similar rate in the two most recent economic expansions
The next chart shows how growth was subdivided in the 2002–2007 economic recovery from the dot-com bubble of 2001 and the 2009–2019 recovery from the Great Recession of 2007–2009. In both of these expansions, growth patterns were similar.
In both recoveries, the bottom 50 percent of households received around 20 percent of economic growth in the expansion, despite representing 50 percent of the population. The “upper 40” group, which includes households above the 50th percentile of household income and below the 90th percentile, received about 42 percent of total growth in both periods, suggesting that this group is receiving a relatively fair share of growth. (See Figure 3.)
Figure 3
Low-income households had poor wage growth while high-income households registered strong business profits and asset-price rises
Next up are three charts that show how specific components of income contributed to the economic fortunes of the bottom 50 percent of households, the upper 40 percent, and the top 10 percent by income.
In the bottom 50 percent of households, wages and government transfer programs—economic parlance for social infrastructure programs, such as Unemployment Insurance, that underpin the economy during downturns—make up the vast majority of all income. Accordingly, these categories matter far more than others for determining income growth in this group.
Wage growth was relatively weak for the bottom 50 percent of households in the first two decades of the 21st century. The vast majority of income growth for this group came thanks to government transfers, such as the Affordable Care Act, more commonly called Obamacare and more formally known as the Patient Protection and Affordable Care Act of 2010. Obamacare boosted this group’s income significantly when it was implemented in 2014 and 2015. Health insurance provided what is known as a social transfer in-kind, meaning it doesn’t give money but gives a valuable service to people that they would otherwise have to pay for. That’s why Obamacare was a boost to incomes overall. (See Figure 4.)
Figure 4
The story for the upper 40 percent of households is similar. Households in this group also get most of their income from wages, but they still receive some government transfers.
Unlike the bottom 50 percent of households, the upper 40 percent holds assets, and interest and dividend income from these assets represents about 8 percent of their total income in 2019. This group experienced robust wage growth in most of the expansionary years, providing the bulk of income growth for this group. (See Figure 5.)
Figure 5
Then, there’s the top 10 percent of households, which boasts diverse sources of income that include wages, interest and dividends earned on assets, and business income. Notably, too, the BEA data series does not include capital gains, which would significantly increase the incomes of the top 10 percent of households by income. So, this should be considered a low estimate of top 10 percent income growth over the period.
This group enjoyed very strong wage growth, which was further supplemented by sources of capital income that are mostly concentrated in this group, even when capital gains are excluded. (See Figure 6.)
Large employers in the United States can have significant impacts on the local labor markets in which they operate, not only affecting their employees directly but also spilling over to other workers living and laboring nearby. Recent research by Ellora Derenoncourt of Princeton University and Clemens Noelke and David Weil of Brandeis University, for example, finds that voluntary minimum wage increases among large employers, such as Amazon.com Inc. and Target Corp., spilled over and led to wage increases for other workers in the same commuting zone.
Yet a new Equitable Growth working paper by grantee Justin Wiltshire of University of California, Davis finds the converse is also true. He finds that the expansion of a particular “low-road” employer—Walmart Inc.—offering low wages leads to a negative spillover in wages and decreased employment in a local labor market. This new evidence features clever empirical analysis of the opening of Walmart Supercenters around the country, compared to locations where a Supercenter was proposed but blocked, adding to a growing body of evidence on the prevalence of monopsony dynamics across the U.S. labor market suppressing wages and limiting employment.
Walmart has long been the largest private-sector employer in the United States. According to the new analysis by Wiltshire, the expansion of Walmart—and, in particular, the opening of Walmart Supercenters—accounted for half of retail employment growth since their beginning in the late 1980s, accounting for 2.5 percent of all employment growth in this time period.
Walmart also is notable for remaining a low-road employer—not following the lead of other retail giants in establishing a higher minimum wage. With low wages and poor job quality, Walmart faces an estimated employee turnover rate of 70 percent per year. Walmart’s demand for labor to match their high turnover rates, paired with its persistent low wages, indicates that Supercenters may be monopsonistic employers that set wages below the rates that would exist in a more competitive labor market.
While it’s intuitive that a large low-road employer such as Walmart may negatively impact local labor markets, previous research found mixed evidence due to methodological limitations in the research and what is known as endogeneity biases. Endogeneity biases occur when examining the impact of, say, a Walmart opening in a particular location makes it difficult to delineate between the impact of that Walmart itself and whether Walmart’s expansion was itself impacted by local conditions ripe for the expansion of a big box store and low-road employer.
Wiltshire seeks to overcome this bias in his new working paper. Using an illuminating empirical design, he is able to better test the direct impact of a Walmart expansion by constructing a so-called stacked-in-event-time synthetic control estimator of locations, economic parlance for identifying the counties where a Walmart Supercenter was proposed and blocked—making up the “donor pool” in the “synthetic control”—compared to those locations where a Supercenter was successfully opened. The group of counties with a proposed, but not realized, Supercenter presumably have similar attributes to counties in which a Supercenter was opened.
To conduct the empirical study, Wiltshire uses data from the Quarterly Census of Employment and Wages from the U.S. Bureau of Labor Statistics for county-level labor market outcomes, Dunbar Market Indicators data for Walmart locations and its employment and sales, and Tax Policy Center data on minimum wages. Using these data sources, Wiltshire finds that the entrance of a Walmart Supercenter increases retail employment by 2.2 percent in a county during the year of entry and by 1.4 percent by the fifth year.
Yet a Supercenter entrance into a county also reduces aggregate employment levels by 2.9 percent and reduces labor force participation by 1.4 percent by the fifth year. Furthermore, the simultaneous retail employment growth with aggregate county employment decline means that the Supercenter increased employment concentration in a county over time. Retail earnings increase by 1.5 percent at the time of Supercenter entry then fall back to pre-entry levels, while county level earnings decline by 5.2 percent by 5 years after entry.
Because earnings decline for both goods-producing and service-providing industries, Wiltshire hypothesizes that perhaps wages in goods-producing industries are impacted by monopsony buyer power vis-à-vis suppliers. In short, labor market conditions got worse, with fewer people working and lower average wages in the years following the opening of a Walmart Supercenter, compared to other, similar counties that were successfully able to block expansion.
One implication of monopsony is that increases to the minimum wage or wage increases driven by collective bargaining among monopsonistic employers can lead to a simultaneous increase in earnings and employment levels at employers operating in monopsonistic labor markets. This can be the case because noncompetitive labor markets are operating under capacity with deadweight loss, defined as the inefficient allocation of resources since the economy is underproducing and undercutting the wages of workers.
Wiltshire continues his analysis by examining the impact of the 1996–1997 federal minimum wage increase on counties that already had a Walmart Supercenter. Five years after the minimum wage increase, Supercenter counties experienced increases in both retail and aggregate employment of roughly 5 percent, alongside an increase of 5 percent to 8 percent in aggregate earnings and 4 percent to 5 percent in retail earnings. These results are consistent with Walmart exercising monopsony power in these counties and the ability of labor market policy to reduce the monopsony power of employers to keep wages lower than market rates.
This study of Walmart clearly quantifies the negative impact of monopsony for workers and the local economies where they work. While previous research on monopsony has found more evidence of monopsonistic wage-setting pressure for relatively high-wage workers who are more likely to be specialized and have fewer outside options, such as registered nurses in the highly consolidated healthcare industry, large dominant low-road employers in local labor markets are a case in which monopsony also impacts low-wage workers.
In these instances, policy tools such as increasing the minimum wage are good for workers and good for the entire economy, with the ability to increase earnings and employment in a win-win. Furthermore, improving income supports for low-wage workers may improve their mobility out of poor-quality jobs. And in a monopsonistic labor market, ensuring that workers have the ability to exercise voice and bargain for higher wages is a critical tool in balancing employer wage-setting power that has plagued the labor market—which, in turn, will lead toward broadly-shared growth.
Economic inequality in the United States continued to rise over the first two decades of the 21st century, according to new data from the U.S. Bureau of Economic Analysis—a trend that builds upon the sharp divergence between the fortunes of the truly rich and the rest of society that began around 1980. As we enter the fourth straight decade of rising income and wealth inequality and their attendant social inequities—such as the widening gulfs in college attendance and life expectancy—U.S. policymakers, now more than ever, need more accurate data to deal with the baleful consequences of inequitable growth.
Most of this rise in inequality occurred before 2000, but the latest data show that U.S. households are still slowly drifting apart from one another. December’s BEA data release expanded the bureau’s distributional data series to cover 2000 through 2019, where it previously only covered 2007 through 2018. These more complete data encompass the entirety of two business cycles: the dot-com bubble-induced recession of 2001 and subsequent expansion, and the Great Recession of 2007–2009 and subsequent expansion, before the arrival of the coronavirus pandemic and ensuing recession in 2020.
Until the Bureau of Economic Analysis extends the series backward further, it’s difficult to say how increases in inequality now compare to those that happened in the 1980s and 1990s—but two things are clear. First, inequality continues to rise. Second, these data are critical to understanding the consequences of economic inequality in the United States.
This issue brief provides analysis of the most recent release. The evident utility of the data series, as shown here, demonstrates why the Bureau of Economic Analysis should request more resources from Congress to expand and improve this dataset, especially by adding more current data that would allow us to analyze trends now, rather than 2 years in the past.
What the latest economic data show about rising economic inequality
The U.S. economy is in its fourth decade of expanding inequality. One way to see this is to compare average growth in incomes to the growth realized within particular income brackets. Average growth in disposable personal income over the period from 2000 to 2019 was about 2.3 percent but ranged as high as 2.7 percent for those in the top 10 percent of household income and as low as 1.6 percent for the lowest-income households. Even within the top 10 percent, there is significant dispersion: The top 1 percent had an average growth rate of about 3.3 percent. (See Figure 1.)
Figure 1
As Figure 1 shows, “headline” personal income growth figures—those that are most commonly reported by analysts, the media, and other sources—overstate the actual growth experienced by virtually all U.S. households. Only the top decile exceeds headline growth. Headline growth is simply an average, and in an age of inequality, the average is largely determined by very rapid income growth for the highest incomes.
These differences add up over a 20-year window: A household experiencing the same income growth rate as top 1 percent of households would end the 20-year period with income 20 percent higher than if they had instead experienced just average growth. The difference between experiencing the income growth of a bottom 10 percent household in this period or experiencing average growth is a 15 percent income premium over the period. And this represents just a small part of the run-up in inequality that started sometime around 1980. The Bureau of Economic Analysis should continue to extend the series back in time, so comparisons to this period can be made.
The federal statistical system needs to be resourced to expand and continue reporting on inequality
Four decades of rising inequality calls for a more robust policy response to ensure broad-based growth in the U.S. economy. An important first step is to develop the data infrastructure to track growth in inequality over time, so that policymakers can monitor and respond to the problem, and voters can hold them accountable to producing strong growth for all U.S. households.
Existing data series are insufficient. One of the few reports by the federal government on economic inequality is the U.S. Census Bureau’s September Income and Poverty Report. The Bureau of Economic Analysis has already significantly improved on this report by using a more comprehensive income concept, releasing more granular income groups (the Census Bureau releases quintiles), and fully accounting for income in one national account (the Bureau of Economic Analysis fully accounts for all Personal Income). But there is much more to do.
Congress must give the Bureau of Economic Analysis the resources to continue developing this work. There are three important avenues for further development of this product. First, the current BEA product underestimates top incomes because it does not account for capital gains. This is not an intentional omission but rather a reflection of the fact that capital gains are not included in any national accounts.
Economists Thomas Piketty at the Paris School of Economics and Emmanuel Saez and Gabriel Zucman at the University of California, Berkeley, in their pathbreaking 2018 article on distributional national accounts, attempted to mitigate this omission by using National Income as their income concept. National Income includes retained corporate earnings, which Piketty, Saez, and Zucman argue can act as a proxy for unrealized capital gains.
Distributing retained corporate earnings, however, requires making several simplifying assumptions that may be prone to error. Statistical agencies should research a separate series of the distribution of capital gains that is compatible with the distributional personal income data. Economist Jacob Robbins at the University of Illinois at Chicago defines a measure of Gross National Capital Gains that could be a model for such a data series.
Second, these data must be released on a shorter delay. Currently, the Bureau of Economic Analysis plans to release these data once a year in December, adding data for the period 2 years prior. This is what happened last month—2019 data were released in 2021. This is a useful tool for understanding the near past, but it will not help policymakers make real-time decisions, and it will not help households understand and respond to the economic conditions they currently face.
There are challenges with publishing more current data, but they are surmountable. With proper resources and a willingness to impute and model current data, BEA staff can release this data series at a higher frequency and with less latency.
Finally, expanding this new data series back in time to include the 1980s and 1990s would enable policymakers to have an even better grasp of rising economic inequality trends over generations. This year’s data release demonstrates how much policymakers can learn about inequality in the U.S. economy—and how much more they could learn if the data were released more frequently and collated over longer periods of time.
Examining U.S. income inequality over the past 20 years
U.S. households at different levels of earnings had significantly different experiences of the past 20 years, with the top 1 percent experiencing more income growth. Those with high incomes saw more immediate and deeper drops in income during recessions, while those with lower incomes were partially supported by government “transfers,” economic parlance for social infrastructure programs such as Unemployment Insurance that underpin the economy during downturns. But the flip side is that low-income households experienced years of stagnation after recessions had subsided, with very little of overall economic growth accruing to those in the bottom half of the income distribution despite moderate headline growth. (See Figure 2.)
Figure 2
Figure 2 shows income growth in each year subdivided into four groups that some researchers and academics commonly use. They are the bottom half of all income earners in the distribution, the next 40 percent of earners (50th to 90th percentile), the next 9 percent at the top (90th to 99th percentile), and the top 1 percent. The blue portion of each bar shows the amount of growth that benefitted earners in the bottom half of the distribution. Warm colors—yellow, red, and orange—show the amount of growth in each year that accrued to the top.
U.S. households move between deciles frequently. The Bureau of Economic Analysis does not follow people over time, so those in the bottom 50 percent of income in one year may not be there the next year. This means these data do not chart economic mobility, only the shape of the overall income distribution. But the overall shape of the distribution is important. When wages for the bottom 50th percentile aren’t increasing, it means that there is little wage growth in the kinds of jobs occupied by those with low educational attainment, young workers, and those in vulnerable populations. The U.S. economy used to deliver real growth to this end of the distribution, but for the past four decades, it has been increasingly stingy.
Households in the bottom 50 percent of the income distribution experienced relatively little growth in incomes over the past 20 years: Over that entire period, the bottom 50 percent of the distribution captured just 20 percent of all growth, even though they represent half the population. Meanwhile, the top 10 percent—a group just one-fifth the size of the bottom 50th percentile—captured 37 percent of overall growth. (See Figure 3.)
Figure 3
This pattern was stable across the two economic expansions in this time period. The expansion after the Great Recession was slightly more equal, with the top 1 percent of income earners benefitting less, compared to other groups, but differences were slight.
The BEA data series also provides detail on how particular components of income, such as wages, transfers, or business income, changed for households in each decile of income. Looking at fluctuations in these components is a useful way to understand how the economy works for U.S. households at different points on the economic ladder.
Households in the bottom half of the income distribution, for example, are largely dependent on wages and government transfers. In 2019, wages represented 49 percent of positive income for this group, and government transfers made up another 40 percent (you can see all the components that make up personal income here), with most other components contributing very little to incomes of the bottom 50 percent of income earners. So, fluctuations in overall income for this group are largely due to wages and transfers. (See Figure 4.)
Figure 4
The top panel of Figure 4 shows annual growth in each component, while the bottom panel shows cumulative growth. The majority of all growth for this group came from growth in government transfers. Wage growth contributed 30 percent of all growth for the group, while the contribution of all other income sources was negligible.
The “upper 40” group of households derive most their income from wages, but government transfers still provided about 13 percent of income for this group in 2019. Unsurprisingly, the relative importance of these two categories for income growth is flipped for this group of households, compared to “bottom 50” households. Over the 2000–2019 period, 64 percent of income growth for this group came from growth in wages, but the Great Recession hit the wages of this group particularly hard. Subsequent years saw a rapid recovery. Transfers accounted for about 26 percent of all growth for this group over this time period. Yet this group did earn some interest and dividend income, accounting for about 8 percent of income. (See Figure 5.)
Figure 5
The top 10 percent of households by income have the most diverse sources of income. Although wages are still important for this group, making up about 51 percent of positive sources of household income, these households also have significant amounts of business income and interest and dividends income. About 27 percent of all growth over the period came from interest and dividends, while about 20 percent of growth came from business income. (See Figure 6.)
Figure 6
Conclusion
The BEA distributing personal income data series has progressed rapidly in just a couple of years. Every year has seen significant improvements in the statistical methodology and the amount and types of data available.
The current product allows for useful analysis of the recent past, with important applications to current policy debates. Knowing how growth in the economy is distributed is just as important as knowing how much the economy is growing. Congress must resource this effort, so the Bureau of Economic Analysis can expand the data series. This should include:
Improve frequency and latency: The bureau should investigate ways to decrease the lag in the release of estimates. Right now, data are released in December for 2 years prior. In December 2022, we will get data for 2020, giving us our first glimpse of the coronavirus pandemic, yet this lag is too long. Ideally, estimates would be released quarterly, putting this product on the same footing as GDP growth.
Extend the time series back: There also is significant value in having a complete history of inequality in the modern U.S. economy. The bureau should make efforts to extend the data series back in time to allow comparisons with earlier eras.
Account for capital gains: Although this has traditionally not been the purview of the bureau, capital gains are increasing and contribute significantly to economic inequality. The Biden administration or Congress should consider tasking the bureau or another agency with tracking this trend.
Together, these three steps would vastly improve our nation’s economic statistics and enable policymakers to act on the harmful consequences of four decades of rising economic inequality.
post
post
