Gender pay gap among the District’s workforce

The gender pay gap refers to the systematically lower wages women receive relative to men with similar talents and responsibilities.  A White House study tells us that the median wage for women equaled only 77 percent of the median wage for men across the nation–a pay gap of 23 percent. The pay gap exists between men and women with similar degrees, similar work experiences, and similar occupations.  The study mentions various factors that create this gap: family responsibilities such as child rearing generally fall on the women; or men tend to negotiate and seek promotions more aggressively.

But explanations offered for the gender pay gap, more often than not, do not settle the matter; rather they lead to new debates. Economists from the Federal Reserve Bank of St. Louis, for example, point out that broad comparisons are not appropriate for gender gap analysis and one must take into consideration differences in attainment, experience, and occupational choice as well as the intensity with which someone works. They argue that when one accounts for these differences the pay gap between men and women declines to five percent. There are even disagreements of the effects of childbirth: some argue that high-skilled women experience a smaller wage effect from taking time off for childbirth because of the demand on their skills, whereas others argue that the declining the declining birth rate among high-skilled, highly-paid women is evidence that childbirth hurts long-term wages.

We became curious: How about the District’s workforce? In this city with many workers on government pay schedules, do we have a gender pay gap that is significant? How does the gap change with age, marriage, education level, hours worked, or occupation? We looked at the annual wage and salary earnings of men and women.  (In this study, we use average wages, and not the median. The median salary is useful when data has very extreme values. But we are relying on American Community Survey for data, which already caps incomes and therefore throws out extreme values. Even then, the pay gap to some extent reflects the dominance of men among very highly paid workers. For example, among those who receive wage or salary income of $500,000 or more, men outnumber women by more than 3 to 1.) Here is what we found:

Overall gender pay gap

2014 data from the American Community Survey tells us, in that year, women made up 48 percent of the workforce (the nationwide share is 47 percent) and received only 42 percent of the wages. The average salary for women over the age of 21 in 2014 was $71,000 and for men it was $90,000. That is, women’s annual wage and salary earnings were 78 percent of men’s; putting the District’s gender pay gap at 22 percent compared to the national gap of 23 percent.

Gap by type of employer

The pay gap is greatest among workers of the private entities. Among for-profit firms, women, on average, receive 76 percent of the average wages paid to men; among non-profit firms, the comparable number is 77 percent. It appears that the government sector has a smaller pay gap, with a 16 percent gap among state and local government employees (largely D.C. government but also employees at charter schools, UDC, and other local entities); and the gap is 15 percent among federal government workers.

Has the gap been closing over time? To see this we look at the wage differentials among older and younger cohorts of workers. The gap generally widens by age across all four types of employers. For example, among for profit firms, the pay gap is only 7 percent for workers between the ages of 25 and 34; among the workers of non-profit firms, we don’t see a gap at all. However even when the gap is small at this age group, it does not tell us definitively that the gap will continue to be smaller when this cohort ages. This is because a small gap can widen over time even if the two groups received the same percentage increase in their wages. Viewed from this angle, it appears like the government sector, especially the federal government, has promotion practices that are most gender-blind, as the gap does not widen as quickly across cohorts.

image002.png

How about education? The data suggest that in the private sector, having a college degree (or higher) does not necessarily close the pay gap.  This is significant, but is still not a definitive indicator of gender-dependent pay or discrimination. For example, if more men study STEM and gets jobs in science and technology fields, and more women study social sciences or humanities that feed into lower paying jobs, then a pay gap will exist for given degree in education. Data do tell us that 70 percent of STEM degrees awarded at universities and colleges indeed go to men.  This we can also see at the doctoral level: even though more women receive doctoral degrees than men in a given year, women are extremely underrepresented in STEM fields; they are also less likely to get a degree in physical sciences and business programs, which tend to lead to higher paying jobs.  But, even with professional degrees (doctors and lawyers) there is a significant gap of 25 percent between men and women.

Finally, for DC residents who make up a third of the District’s workforce, the gender gap is smallest at the federal government and private, for-profit firms, and widest among DC residents employed in a state or local government job.

Is there a marriage toll for women?

Among men and women who have never been married, the gender pay gap is only 6 percent. Among men and women who are married, the gender gap is 23 percent. A single woman (whether married or divorced) between the ages of 45 and 54 makes more than a single man in the same age group. A married woman of the same age group makes 23 percent less than married man in the same age group. Why could this be? Perhaps married women in this group work fewer hours, or they took time off to raise children, or took more flexible but lower paying jobs. In fact, the gap is much smaller among married women under the age of 34 (and presumably without children).  We can think of many other possible contributors: married men are happier, and perhaps for this reason more productive. (They make more than unmarried men as well: see here and here). Because their wives take over responsibilities at home, perhaps married men can dedicate more time to their work. Perhaps there are evolutionary reasons: men who have not been picked for a marriage by the age of 45 are just low in productivity. Perhaps it is selection bias in the other direction: women who are aggressive about their careers do not get (or stay) married. Perhaps employers low-ball the salaries of married women expecting that they would take more time off at some point in their career and perhaps just leave. Again, we do not have definitive answers.

image004.png

Work hours

Gender pay gap is a feature of full time work.

When we look at men and women who work fewer than 30 hours, we see more women than men, and their wage and salary earnings are generally greater than men’s. But the size of District’s part time workforce is small—it stands at less than 10 percent of the total workforce in the District. Among those who report working 30 to 40 hours a week, the gender pay gap is 11 percent (464,000 workers report working 30 to 40 hours and half of them are women.)  Among those who report working over 40 hours (approximately 250,000 workers, 41 percent of whom are women) the gender pay gap is 21 percent.

This is a perfect example of how the same information can support exactly opposing policy solutions. One can look at the gap and see the need for interventions; others can look at it and conclude that the labor market can create many types of jobs that require different combinations of effort and wage that fit the needs, abilities or desires of all kinds of different workers. If that is the case, interventionist policies would actually reduce employment and hurt women more.

image008.png

Industry and occupations

Dominance of women in a given industry does not guarantee a better pay.  For example, 70 percent of workers in the health care industry are women.  This includes not just people with medical occupations such as doctors and nurses, but also all other types of workers from managers to computer programmers who work for a healthcare company. In this industry, the gender pay gap is 36 percent. Women in construction—a male dominated industry—make more than men but this is because they are more likely to hold managerial or creative jobs, rather than working on the construction site.

image014.png

So perhaps we should compare men and women doing the same job. Cutting and slicing the data gets precarious when we go down to individual occupation level, but we looked at occupations with more than 2500 men and 2500 women, and found only a handful of occupations where the pay gap is small or favor women. Physicians and surgeons in the District face a gender pay gap that is as wide as the healthcare industry: 37 percent! Looking at the lawyers—the gap is 73 percent. We can string together more explanations: more women go in-house (lower pay) or most women lawyers are younger.  But some other outcomes are puzzling: 63 percent of education sector employees are women; the industry gender pay gap is 23 percent. Among elementary and middle school teachers only (we are excluding administrators, and we don’t know if they are public charter school,  DCPS or private school teachers), the gap is 20 percent. Why?  What explains the gap for cashiers among whom women receive only 63 percent of what men earn? How about waiters and waitresses where the gap is 11 percent? Do they get fewer hours of work or does gender, ethnicity, or other characteristics–of the wait staff or the customers–play a role in how we tip waiters and waitresses?

image015.png

Survey data could be idiosyncratic, especially given the small size of the District’s sample (but in this case we are using data from three states). Still the pay gap is there, even when one controls for age, marital status, or even down to occupation. Is this a sign of flexibility in the labor market? If all jobs required 40 hours a week (no more, no less), had the same degree of flexibility (or inflexibility) would the labor market outcomes be better for women? Or do firms knowingly take advantage of lower reservation wages for women? If so, why do women accept lower salaries when everything tells them to negotiate? There has been research that says women can be penalized for negotiating if they negotiate like men, suggesting that gender roles continue to play a role in the workplace.

We wonder what our readers think.

What exactly is this data? We use the American Community Survey micro data for 2014. We combine District of Columbia, Maryland, and Virginia data for respondents who report working in the District. Wage and salary income is what respondents report earning in a year.

 

Abstract abilities and skills are the best predictors of high wages in the District

District workers are handsomely paid. The median salary in the District was $64,890 in 2014 or 1.8 times the median U.S. salary. One explanation usually offered for this high pay is the presence of the federal government (see here, and here).  Not only does the government pay higher than the private sector in the District, it also supports the kinds of jobs (lobbyists, lawyers, contractors) with compensation packages they would not get anywhere else.

In this post, we show that District workers receive high salaries, not just because the federal government is here, but because District jobs (including administration, lobbying, and professional services) attract people who have skills and abilities highly rewarded anywhere.

We begin with some background: The District’s labor markets survived the great recession rather well. Between 2005 and 2014, the District added 64,880 new workers – a 10 percent increase in total employment. A one percent growth per year may seem meager, but during that period, national employment increased by 4 percent only, so the District did 2.7 times better than the nation.

But equally important is the change in the District’s occupational mix. Managers and professionals (lawyers, doctors, business and finance people, scientists, and engineers) made up 60 percent of the District’s workforce in 2014, up from 53 percent ten years earlier at the expense of sales and office jobs. While a shift towards management and professional jobs is a feature of the national economy, the pace of change is much faster in the District: Management and professional jobs increased by 3 percent in the US compared to 7 percent in the District.

image012image003The District has one of the highest concentrations of management, professional and technical jobs in the country.  Within the Washington metro area, more of these jobs are in District proper. The share of management, professional and technical jobs in the greater metro area is 47 percent–higher than other metro areas with similar workforce composition.  For example, in the Boston-Cambridge-Quincy, MA metropolitan area, the comparable share is 44 percent. In the San Francisco-San Mateo-Redwood City, CA metropolitan area, it is 41 percent.

Here is our question: Can we explain salaries by looking at what people do, without really worrying about where they work or what their job titles might be?  

To answer this question, we used detailed occupational data from the Department of Labor and borrowed from the work of two economists. The U.S. Department of Labor’s O*NET program, the nation’s primary source of occupational information, collects data on each occupation including information across 35 different skills (including things like programming, active listening, and persuasion), 52 different abilities (for example, arm-hand steadiness, stamina, or originality), 42 tasks (collecting information, staffing, inspecting equipment, structures, or material) and 57 work contexts (for example, contact with others, frequency of decision making, time pressure). O*NET scores each occupation along each of these dimensions on a scale of 1 to 5 based on peer evaluations. This gives us 186 different scores for each occupation. Acemoglu and Autor further group these different skills, abilities, and tasks into three broad areas: abstract, routine, and manual (they have further subgroups, but to keep things simple, we used these three). Their methodology yields scores for abstract, routine, and manual dimensions of each occupation.

Abstract tasks and skills include things like data analysis, creative thinking, interpreting data for others, coaching, guiding, etc. Lawyers, teachers, physicians and managers–occupations that constitute the largest source of jobs in the district–score high on the abstract dimension.  Occupations that score the highest in routine tasks (structured and repetitive work that relies on accuracy) include meter readers, bookkeepers, and cashiers—the office and retail workers, who are losing ground in our city.  Finally, manual skills include operating machinery, or work that involves hands or body, so you will find in this group lots of construction and manufacturing workers—types of occupations that are relatively rare in the District.

To be clear, the scores do not necessarily reflect the skills of persons holding those jobs. Herman Melville worked as a customs inspector, and Einstein logged hours in the patent office. A parking attendant might be very good at math puzzles and could spend his weekend designing dungeons and dragons games. But he would not use these skills in his daily job.

You can click on this link to explore scores for occupations we find in the District.

image006

Here are the same scores, this time grouped by broad occupation groups.  Each dot represents an occupation and the gray lines give the median score for that broad group in each area.  Management and professional occupations—the greatest source of employment in the District, score highest on abstract tasks (the median score is 5.4) and lowest on manual tasks.  Many occupations have a routine element and most occupations demand little in terms of physical labor (the exception is construction and repair jobs). You can click on the image to see the details.

image010

So how do these scores explain the median wages in the city? It turns out that salaries and abstract task skills are strongly correlated. A one point increase in abstract task scores increase median salary by $8,230.  This is 13 percent of median salary paid in the District. The manual task scores, on the other hand, are negatively correlated with pay.  Median salaries across different occupations decline by $3,000 per one unit increase in manual task scores, but notice from the graph that the variation in manual scores is smaller across occupations (and the relation is not as strong, you can see here).  Routine tasks scores cannot explain salary differentials at all.

image019

Is this strong relationship between abstract skills and abilities and wages unique to the District? It turns out, not. We find a similarly strong relationship between abstract tasks and skills and salaries in other jurisdictions. To compare, we provide the same information for Boston, San Francisco and Honolulu.  Boston and San Francisco are relatively similar to the District in their labor composition, with a large concentration in management and professional occupations.  Honolulu is different, with a larger concentration in service occupations.  In all these locations, abstract skills are the most important determinants of pay.

image029When we think of incomes, we generally think of who we work for and what we do, but we can also think in terms of skills and competencies. Within every broad occupation group, incomes generally increase as one takes on more tasks that require abstract skills such as data analysis, problem solving and interpreting information.  This matters, both at a personal level, as we choose to invest in our own or our children’s education, at for the entire city, as we consider workforce development options for many who have a hard time finding a job.

What exactly is this Data?

Occupation and wage data are from the U.S. Department of Labor’s May 2014 estimates. Description of O*NET’s occupation scoring is here. The data for the Acemoglu and Autor paper is available for download here.  Here is where they explain how they construct the task measures.

The data we used for this post can be downloaded here.

District’s labor market and workforce are intertwined with Maryland and Virginia

In 2014, nearly 774,000 workers reported working in the District of Columbia and they collectively earned $63.5 billion in wages and salaries. Of these workers, only 251,000 or 32 percent were District residents. The remainder were commuters from Virginia or Maryland, accounting for 68 percent of people employed in our city. The District’s share in total wages earned was even lower: District residents accounted for $18 billion of salaries and wages earned in the District. This is about 28 percent of all wages and salaries earned in the city.

Wage and Salary.jpg

In addition 89,000 District residents reverse-commuted to Virginia and Maryland, mostly working for private entities (76 percent including non-profits) and the federal government. This group collectively earned $6 billion in wages, compared to the $45 billion Maryland and Virginia commuters earned in the District.

Reverse commuters.jpg

The data reveal other trends. District residents who work in the District hold a disproportionate share of the lower-paying jobs: 44 percent of jobs that pay a wage of $30K or less are held by DC residents, compared to 32 percent of all jobs in the District. Virginia residents, on the other hand, tend to hold a larger proportion of higher paying jobs: 28 percent of jobs in the District and nearly 40 percent of all jobs that pay $100K or more.

image003.png

The data also show that District residents dominate employment in the non-profit sector, one of the lowest paying sectors in the District.  Commuters from Virginia and Maryland, on the other hand, typically come to the District to work in the private sector and the federal government.

image004.png

District’s labor market and workforce are tied deeply with those of Maryland and Virginia. If salaries are any indicators, the most educated and productive residents of our neighboring jurisdictions work in the District. In 2014, District residents who worked in the District reported wage earnings of $63,700 compared to $69,400 for commuters from Maryland, and nearly $95,000 for commuters from Virginia. But even within the same sector, District resident’s wages could be low: In the non-profit sector, District residents earned, on average, $68,500 in wages—13 percent less than Maryland workers and 20 percent less than VA workers.

image005.png

Here are the data, in greater detail, for you to explore:

Resident employment and workforce.jpg

What exactly is this data?

The data is from the single-years PUMS release from American Community Survey for 2014. The analysis was done in SAS, and SAS files are available from the author.