Over the past year, education and food services have become the largest sources of new jobs in DC

Those sectors’ contributions to wage growth in the District’s economy are, however, quite modest

According to US Bureau of Labor Statistics data, over the course of the last year job growth in DC came increasingly to be dominated by two sectors: education and food services. These two sectors were responsible for 11,800— about 90%—of the 13,133 net increase in all wage and salary employment in DC from September 2016 to September 2017. This high proportion of the job increase over the last year is remarkable because in September 2016 education and food services accounted for just 14% of all jobs in the city.

Graph 1

Rapid job growth in education and food services did not, however, translate into comparable income gains. The latest income data from the US Bureau of Economic Analysis (combined with seasonally adjusted employment data) indicates that the two sectors accounted for just 5.9% of the net gain in wage and salary income in DC from 2016.2 to 2017.2—despite job growth that exceeded the net increase for the economy as a whole.

table 1

Employment in the education and food services sectors. The increase in employment over the past year in both education and food services has been striking. From September 2015 to September 2017 education employment in DC went from 58,433 to 67,333. Of the 8,900 increase over these two years, 76% (6,800) occurred in the past year. Similarly, the last year accounted for 86% (5,000) of the 5,800 increase in food services employment over the past two years. As a share of all employment gains in DC, the combination of education and food services rose from about one-third or less in the year from September 2015 to September 2016 to over 100% during the spring of 2017.

Table 2

Graph 2

Graph 3.PNG

Jobs and wages. The rapid increase in employment in education and food services over the past year appears to have made little impact on amounts earned by people working in DC. Whereas the two sectors accounted for 107.1% of the net increase in employment in DC from 2016.2 to 2017.2, their share of wage and salary growth was just 5.9%. This share of the increase over the year was less than the average share of wages attributable to those two sectors in 2016.2, which was 6.4%.

Part of the reason for the relatively small impact on wages that education and food services have had is that they both are relatively low wage sectors compared to DC average earnings. In 2017.2, the average annual earnings in education, $41,771, was just 45.7% of DC’s $91,405 average. The average wage in food services, $30,187, was one third of the DC average. Over the year 2016.2 to 2017.2, the average annual earnings fell in both education and food services.

Table 3

Table 4

What accounts for the rapid growth in employment accompanied by modest impact on wages in the education and food services sectors? The data on wages and employment do not provide a full answer to this question. However, two reasons probably help to explain why such a development might occur.

The first reason might be part time employment. BLS employer surveys count people who are working on the survey date, whether such people are full time or part time. Even if hourly wages are increasing, average annual earnings could result if the proportion of people working part time was increasing. Both education and food services can be a source of part time work. It is worth noting in this regard that the only other sector of DC’s economy in which average annual earnings fell from 2016.2 to 2017.2 was transportation and warehousing, another sector in which there could be an increasing amount of part time employment.

The second reason could be changes in the composition of employment within the sectors. This appears to be a factor in education where the largest gains in employment have been in the “all other” part of the sector. “All other” could be such things as charter schools or preschools or after school programs where average annual earnings are likely to be less than in colleges, universities, or professional schools.

Table 5

Graph 5Graph 4

It should be noted, however, that the data presented here is preliminary and the picture could change if it is revised in the future. The next significant revision to the employment data will occur in March 2018, and BEA’s income data is revised as more information becomes available.

About the data. The employment data in this analysis is from the US Bureau of Labor Statistics which conducts monthly surveys of employers in DC and throughout the US. The calculations are based on 3-month moving averages of seasonally unadjusted data. The September 2017 data reflects revisions which were included in the October 2017 employment report. Wage and salary data are from the US Bureau of Economic Analysis Personal Income series for DC, accessed through Moody’s Analytics. The most recent BEA data is for the 2017.2 quarter. For consistency in comparing employment and wage data, seasonally adjusted employment data was accessed from Moody’s Analytics. As noted, all BLS and BEA data is subject to revision.

An earlier version of this blog was included in the November District of Columbia Economic and Revenue Trends issued by the District of Columbia Office of the Chief Financial Officer.



The number of occupied apartment units in DC increased sharply last year

The increase tops the peaks of previous years

According to CoStar, a leading commercial real estate information firm, there were an estimated 174,917 occupied units in all classes of market rate apartment building in September 2017, an increase of 5,433 over the prior year. This was the largest one year gain since 2001, the period covered by the data base. The 5,433 one-year gain exceeded the annual gains that occurred before the Great Recession and following the relaxation of sequestration restraints.

graph 1

In 2005 DC’s population started to increase, with the city adding 125,000 residents by 2017, a 22% increase. Not surprisingly, this growth had an impact on the market for apartments. From 2005 to 2017.3 the net inventory of units increased by 35,615. This increase in supply was almost matched by growing demand as occupancy grew by 33,443. The overall vacancy rate rose only slightly—from 4.6% in 2005 to 5.1% in 2017.3.

The balancing by market forces of inventory and demand evident over the past decade is consistent with the modest decrease in the construction of new units that has occurred recently. In September there were 13,022 new units under construction, a decrease of 630 from a year earlier and of 1,687 from the peak pace of 14,709 in March 2017. While occupied units rose sharply over  the past year, inventory grew even more: a 6,727 net increase in inventory versus the 5,443 increase in occupied units. The vacancy rate in 2017.3 for all units also rose somewhat—to 5.1% from 4.7% a year earlier.

As noted in the accompanying chart, new construction began to accelerate in 2010. With some ups and downs, this was soon followed by annual increases in net inventory and in occupied units that have carried to the present time. Most of this activity involved Class A apartments. In the 7 years from 2010.3 to 2017.3, 88% of the net increase in apartment units and 84% of the increase in occupied units were accounted for by152 new Class A buildings. (Class A buildings are new or newly renovated, well located, generally larger buildings with higher rents.)   An increase of 69 Class B buildings accounted for about 15 % of the gains in inventory and occupancy. The number Class C buildings, representing about 38% of the District’s inventory of market rate units, declined over this time. Vacancy rates rose for Class A buildings, which require long lease-up periods, and fell for both Class B and Class C units.


graph 2

table 1

From 2009 to 2012, the three years in which DC experienced the largest annual increases in population over the past decade, the total increase in occupied apartment units for the three years was 4,987 (see the shaded area in the table below). This was less than the increased occupancy that occurred in the last 12 months. This suggests there have been some changes in the connections between a growing population and the city’s housing stock over the past 10 years or so. In the middle part of the decade of the 2000’s, more of the increase in population may have been accommodated by group homes or taking in roommates, by changes to single family or other smaller structures, or by owner-occupied units. Census Bureau estimates of DC’s population in 2017 will not be available until December, so comparison of population and housing unit changes over the past year is not yet possible. The ways in which the District and the owners of its housing stock adapt to changing demographics and housing patterns will no doubt continue to be an area of great interest.

table 2

table 3

About the data. The information is from CoStar’s historical data bases of market rate apartments which goes from the first quarter of 2000 to the third quarter of 2017.  The data includes the total for all apartment buildings as well as for buildings classified as Classes A, B, or C. CoStar Group, Inc. is an American commercial real estate information and marketing provider with headquarters in Washington, DC. Information in the CoStar data base is undated on a continuous basis.

The information here was included in the District of Columbia Economic and Revenue Trends report for October 2017, which was issued by the Office of Revenue Analysis of the Office of the Chief Financial Officer of the Government of the District of Columbia.


The Effect of Self-Employment on Personal Income Growth: Washington DC 2006-2014


The United States’ labor force has experienced great stresses and seen continuous change over the past decade. The recovery from the Great Recession and progressive technological change have combined to reshape how many workers participate in the labor market, the skills and attributes they possess, and their market compensation. As debates about the future of work in the United States swirl, a study by the Office of Revenue Analysis (ORA) looks at the characteristics of the self-employed and implications on their incomes after entering into self-employment.

The Self-Employed Population in DC

This study, which will be presented at the American Economic Association Annual Meeting in January 2018, looks at the self-employed population of DC, defined as people who actively work for money outside of the traditional employer-employee relationship. In order to identify the people who are actively self-employed on at least a part time basis, we use both Federal and DC individual income tax data to identify DC tax filers who meet at least one of these conditions below:

Fig 1_Corey

This identification process finds that the self-employed are about a fifth of all tax filers (between 45,000 and 56,000 tax filers in each of the years studied) in the District. They are spread relatively evenly across the eight wards of Washington, DC, are a bit younger than the overall tax-paying population, and are more likely to be married than the average DC tax filer.

The proportion of the population that is self-employed is also remarkably stable across time. The vast majority of the population is identified by one of the first three indicators shown in Figure 1, and fewer than a thousand in each year are identified solely because they take one of the smaller deductions. Roughly 75% are identified by more than one of the listed indicators.

Fig 2_Corey

A Look at Incomes

Compared to the overall distribution of DC tax filers, the self-employed subset is bi-modal. The figure below divides up the self-employed population into where they fall in the income percentiles for the entire District. There are 50 bars in this distribution, each representing two percentiles of income. All bars sum to 100 percent.
If the self-employed were evenly spread across the population, every bar would be at two percent, but we can see that the self-employed converge around the 11th and 99th percentiles in District income.

Fig 3_Corey

The self-employed population of DC has a higher mean income than DC as a whole. This helps it account for a larger percentage of the District’s total income than its percentage of the population as shown in Figure 2. At the lower income percentiles (i.e. bottom 25% and lower), the self-employed are generally less well off than the same percentiles for the rest of the population. The self-employed individuals at higher income percentiles (i.e. top 25% and higher) are much wealthier than the rest of the population at those same higher percentiles. Median incomes, however, between self-employed and non-self-employed residents are much more similar.  The impact of the recession is also more evident among the self-employed than among the rest of the population. Mean self-employed incomes fell each year from 2007 to 2010 and median income fell from 2008-2011, while the recession is not evident at all in the trend lines for the rest of the District’s taxpayers’ earnings. This is shown in Figure 4 below.

Fig 4_Corey

Some Determinants of Self-Employment

Using linear and logistic regression models (modeling approach to explain the relationship between a dependent variable and one or more explanatory variables), we examined four different time periods: 2006-2008; 2008-2010; 2010-2012; and 2012-2014 to identify possible determinants of self-employment and how they may have changed through the Great Recession and recovery. During the first year of each time period (i.e. 2006), the tax filer was not self-employed. In the second year of the time period (i.e. 2007) the tax filer met the criteria shown in Table 1 and was considered self-employed. The results compare the tax filer in the final year of the time period (i.e. 2008) to the first year in the period (i.e. 2006), prior to when the filer became self-employed.

We find that on average a wage drop in the first year of each time period is correlated with a roughly 5% increase in likelihood of becoming self-employed in the following year. This effect is greatest in the most financially troubled years from 2008-2012. During that time, the odds of becoming self-employed were twice as great among those who experienced a large wage and salary drop compared to those who did not. Interestingly, taking the mortgage interest deduction is as strongly correlated with becoming self-employed as experiencing a wage and salary drop.

Younger age is also correlated with becoming self-employed. The regression model implies that for every additional ten years of age, an individual is about 1% less likely to become self-employed in a given year.

Surprisingly, a filer’s level of total income had very little effect on their likelihood of becoming self-employed during our study periods. This indicates that self-employment is a phenomena that both high and low income individuals self-select into. This again speaks to the bi-modal distribution of the self-employed population as shown in Figure 3.

What Happens to Income After Becoming Self Employed?

Income gains do occur for those who become self-employed, but it seems that the gains are concentrated amongst those tax filers who were already in the top end of the income distribution prior to becoming self-employed. The rest of the self-employed population experienced an average decline in income after becoming self-employed in all study periods.

Those who became self-employed before the financial crisis and in the 2012-2014 period afterward experienced an increase in income. However, these income gains are confined to the top 10 percent of income earners in all periods. The gains that upper decile earners can expect from becoming self-employed are quite large, ranging from $14,000 to $50,000 depending on the time period. Conversely, becoming self-employed if you are in the lower 75% of the income distribution, led to short term losses ranging from $1,700 to $3,500 in the third year of each time period studied.

Some Takeaways on Self-Employment

Self-employment is a common phenomenon in the District, as over 20% of all tax filers in a given year can be classified as self-employed (on at least a part-time basis). We see that from 2006-2014, the number of self-employed residents has grown in line with the overall population growth in the city. We find that the self-employed subset of the population has a bi-modal income distribution and that the average newly self-employed individual is likely to experience net income decreases unless they are in the higher end of the income distribution prior to taking on self-employment.

One possible factor that explains this phenomenon of the income losses of those who switch to self-employment is that income gains may not be their ultimate goal. Being self-employed for many people may be a lifestyle decision that they make for greater flexibility, the pursuit of entrepreneurial endeavors reasons, or family obligations, among others, and not solely to earn more income. Indeed, a growing number of DC residents appear to be substituting slightly higher incomes to experience more work and lifestyle freedoms.

What exactly is this data?

I used administrative individual income tax records from the IRS and the DC Office of Revenue Analysis to identify self-employed DC residents in the years 2006-2014. Due to data limitations, however, our identification methodology cannot distinguish between an independent contractor, a small business owner or an occasional gig worker. The people we identify as self-employed are likely getting into vastly different lines of work and likely for vastly different reasons.

DC’s unemployment as been increasing over the past six months, with the rate rising to 6.4% in July

More DC residents are working, but resident employment growth has not kept up with that of the labor force

According to the US Bureau of Labor Statistics (BLS), unemployment in the District of Columbia has been rising over the past six months. Seasonally adjusted unemployment rose from 22,376 in January 2017 to 25,706 in July 2017, an increase of 3,330 (14.9%). The rate of unemployment rose from 5.7% in February to 6.4% in July.

graph 1 sept 2017graph 2 sept 2017

The rise in unemployment does not mean, however, that the number of employed DC residents fell. To the contrary, there were 4,313 (1.2%) more DC residents working in July 2017 than there were in January. Unemployment rose because for the past 6 months the increase in jobs for DC residents did not keep up with the even faster growth in the DC labor force.

table 1 sep 2017

As explained below, unemployment can be viewed as the difference between the labor force and resident employment. Unemployment goes down if resident employment increases more than the labor force. This is what happened from July 2016 to January 2017. At that time unemployment decreased by 1,324, following the trend of the prior two years. Unemployment goes up if resident employment increases less than the labor force. This is what happened from January 2017 to July 2017 when the labor force increased more than twice as much as in the prior 6 months, and unemployment rose by 3,330.

Unemployment is defined by BLS as people without jobs who are looking for work. This is calculated each month based on a monthly survey of a sample of households. The survey also counts people who are working. The labor force is then estimated by adding together the number working and the number who are unemployed. Unemployment can therefore be viewed as the difference between the labor force and resident employment, and the unemployment rate expresses unemployment as a percentage of the labor force.

The following charts and table show that for most of the past 3 years DC’s resident employment has grown faster than the labor force, with the consequence that unemployment and the unemployment rate steadily declined. The data does not explain why unemployment has started to rise in recent months. The reasons the labor force can grow more than resident employment include arrival in the city of more workers looking for jobs and existing residents returning to the labor force because of improving prospects of finding work. Whatever the reasons, DC’s unemployment rate over the past 6 months rose from 5.7% to 6.4% as the US rate was falling from 4.8% to 4.3%.

Table 2 sept 2017.PNG

graph 3 sept 2017graph 4 sept 2017

graph 5 sept 2017

About the data. The data discussed here are labor force statistics prepared each month for the US and the states by the US Bureau of Labor Statistics (BLS). The data are derived from household surveys, and are subject to sampling and reporting errors as well as changes in underlying demographic information that is taken into account by BLS in making the estimates. In practice, labor force is constructed by adding together those who say they are working and those who say they are unemployed (this is, not working but looking for work). All calculations are from seasonally adjusted data. The data reflect revisions to the original July 2017 estimates made by BLS in its August report, but the data are also subject to further revision by BLS. Seasonal adjustment is the method BLS uses for removing seasonal elements (such as school graduates seeking to enter the labor force or holiday period fluctuations) from monthly labor market statistics. This is done to reveal underlying trends and cycles in the data.

A version of this blog appeared in the September 2017 OCFO report District of Columbia Economic and Revenue Trends.

Metrorail vs. Uber: Travel Time and Cost

With Uber and other ridesharing services becoming a common transit option for some D.C. residents, we wanted to get a sense of when someone might substitute an Uber trip for a Metrorail trip. To do this, we plotted data on travel time and cost, creating a visualization that shows whether Uber or Metro is faster, and at what cost, for 114 different trips between Metro stations. By adding in the time it takes to wait for a Metro train or Uber, walk to the Metro, or sit on a delayed train, we can see how a person’s decision might change depending on their circumstances.

The trips we analyzed include trips between the city and the suburbs as well as trips within the city. In our visualization below we define a long trip as one that’s more than five Metro stops.

(Click to interact with the graph and change the assumptions.)main graph

We found that for longer trips between the center of the city and the suburbs, Metro tends to be both more cost-effective and quicker than Uber. But for trips within the city that require a Metro transfer, Uber is often quicker than Metro, especially when Metro wait times are long, like on weekends, or when there are delays. While Uber’s regular service tends to be much more expensive than Metro, Uber Pool makes some Uber trips nearly as affordable as Metro.

The data for our analysis comes from the Metro trip planner, the Uber app, and Uber Movement. All travel times are from the weekday evening rush hour, which likely gives Metro an advantage on trip times. We pulled our data after SafeTrack ended.

Short Metro wait times and no delays make Metro an attractive option

The graph above shows a scenario that favors Metro, since there are no delays and relatively short wait times for trains (3 minutes). Under this scenario, Metro provides a trip as fast or faster than Uber for 67 of the 114 trips we looked at. Metro is especially efficient for longer trips from downtown to the suburbs that do not require transfers. A trip from Metro Center to Bethesda, for instance, takes 33 minutes, and that includes time to wait for the train and walk to and from the station. An Uber takes 47 minutes, and comes at a far higher cost. Meanwhile, Uber tends to be faster, even during rush hour, for trips within the city requiring transfers. Going from Union Station to Minnesota Avenue or from Columbia Heights to Cleveland Park, for instance, is about 20 minutes quicker in an Uber than on Metro, even when Metro is running efficiently, without delays or long headways.

With Metro’s spoke-and-hub configuration, it’s not surprising that trips requiring a transfer that have origins and destinations relatively close to each other are quicker in an Uber than on Metro. Metro accounts for this, and has bus routes connecting the “spokes” of its rail system. You can get between the Union Station and Minnesota Ave Metro stations using the X2 bus. To get from Columbia Heights to Cleveland Park you can take the H4 bus. Some folks have their own work-arounds, and might bike between these locations. All of this is to say that even if Uber is faster than Metro for trips with a transfer, there are other modes of transit a person can use to make this sort of trip.

Long Metro wait times give Uber a large advantage

The scenario under which the case is stronger for Uber is when Metro has a delay or a long wait for the train, like what you’d encounter on a weekend or at night. If we increase the wait time for each train from 3 minutes to 10 minutes, as we do in the graph below, Uber becomes quicker for the vast majority of trips (99 out of 114). For example, the trip from Columbia Heights to Eastern Market is 9 minutes faster in an Uber when there is a 10-minute wait for a train. With a wait of only 3 minutes, Metro would be faster for this trip.

long waits

Factoring in cost

But what about cost? You can save 9 minutes by taking an Uber from Columbia Heights to Eastern Market when there are 10 minute headways, but the Uber ride will cost you about $10 more than Metro. For some people the time savings will not be worth the cost. But with Uber Pool, which allows you to share a ride with a stranger whose route roughly coincides with yours, you can get rides at a reduced cost.

If you take an Uber Pool for the trip between Columbia Heights and Eastern Market, for instance, the ride will be about $1 more than on Metro. The graph below shows that with Uber Pool, 74 of the 114 trips we analyzed have Uber fares that are no more than $5 above what you’d pay on Metro for the same trip.

With the reduced price there is a trade-off: we assume a ride in an Uber Pool will take 5 minutes longer than in a regular Uber.

uber pool.png

It is unclear how long Uber prices will remain this low. Several news outlets have reported that Uber subsidizes its rides with money from investors, meaning current fares might not reflect the full cost of a ride.


We want to caution that we are not sure if our findings are representative of the entire Metrorail system. We looked at 114 trips, but there are over 4,000 possible trips you could take on Metro. The trips we included in our analysis are listed in the graph below. You can use the graph to search for a particular station and see whether Uber or Metro is faster for a trip involving that station, given assumptions you enter about wait times. The trips we included fall into two categories: 1) trips from either Metro Center or Chinatown to all other stations in the system, with the origin being the station that makes the trip one without a transfer (though we did exclude trips of only one stop in length); or 2) trips within DC city limits requiring a transfer, with one end of the trip including a jobs center (Foggy Bottom, Navy Yard, etc), nightlife center (Shaw), or other heavily-used station (Columbia Heights). In the second case we purposefully included some trips for which we thought Uber would be faster than Metro in order to broaden the range of differences in trip times.

(Click to interact. You can use this graph to search for particular stations.)bar graph

What exactly is this data?

You can download the data we used for this post here. Below we describe how we gathered the data.

Metro: Metro trip times come from the Metro trip planner. We did not include in the trip times the time it takes to wait for a train or to transfer trains, since we made our own assumptions about wait times. We also used the trip planner for the cost of Metro trips. The trip times and costs are for weekdays at 5:30pm. We pulled data from future dates, meaning the times do not capture unexpected delays, and the trip times did not appear to be affected by planned track work. The dates we entered to pull the data were after SafeTrack ended.

Uber: Uber trip times come from Uber Movement. The times are the three-month average for April, May, and June, 2016, during the weekday evening commute. It is unclear whether the times from Uber Movement include Uber Pool trips, which we believe could take longer than Uber X trips. (We asked Uber whether Pool times were included, but did not receive a response.) When we accessed Uber Movement, the tool was in Beta form. It is now no longer in Beta and is available to the public.  We gathered Uber cost data on the Uber app between 5pm and 6pm on the following dates: July 5, July 6, and July 19, 2017. To the best of our knowledge, prices were not surging on these dates nor was there rain during the times we collected the cost data.

Calamari, Cocktails, and Cars: A Glimpse into the District of Columbia’s 10% Sales Tax Collections

In the District of Columbia in 2015, a whopping $361.46 million was generated by the 10% sales tax rate, which is one of five rates used to tax the retail sale of various goods and services in the city. This rate, commonly known as the “Restaurant Tax”, encompasses the sale of not only restaurant meals, but also other prepared foods, alcohol/liquor sales for on-premise consumption, and the renting of vehicles. With so much revenue generated by this tax rate (over 35% of total DC sales tax revenue), we set out to examine the business and economic trends affecting the city’s 10% sales tax collections. To do so, we analyzed the annual sales tax receipts of all businesses in the city for all years from 2005 to 2015. We then focused on the top 100 taxpayers in the 10% tax rate during each year and categorized each taxpayer into one of nine sectors to conduct further analysis: Fine Dining, Casual Dining, Fast-Casual, Fast Food, Food Service/Catering, Supermarket, Hotel, Transportation, and Other, utilizing them to help draw general conclusions about the business outlook of the city as a whole. For example, we found that while there has been significant growth and development of the restaurant industry in the city, when we adjust for inflation between 2005 and 2015 for the top 100 taxpayers, many sectors actually see declining revenues in real terms. The full report can be found here.

I. DC’s 10% Tax Rate: A Sector Analysis

First, we examined each of the top 100 taxpayers within the scope of their sectors in order to track the changes in each aggregate. Overall, we found that all but two of the sectors (excluding “Other”) increased their revenue from 2005 to 2015 (Table 1). Surprisingly, this growth was led by two of the more “unusual” sectors in the city’s “Restaurant Industry”: transportation and supermarkets. While comparatively small in 2005, these two sectors have grown tremendously, so that they outpace even Fast Food in sales tax revenue in 2015. While we aren’t too sure why car rental and car sharing services have done so well, especially since there are plenty of options to travel around DC, including taxis, the Metro, the Circulator bus, and many ridesharing options, this growth may suggest robust demand from 1) the city’s business sector; and 2) the city’s residents, who may rent vehicles for short trips and excursions out of the city given the trend for a growing number of new residents to choose to forego car ownership.


The two categories that experienced negative revenue changes from 2005 to 2015 (again excluding “Other”) were Fine Dining and Fast Food. This finding may be the result of the growing number of city restaurant patrons becoming more health conscious and/or more price sensitive. Therefore, they are largely turning away from fast food, which is considered less healthy, and fine dining, which can cost hundreds of dollars per meal. These customers are likely then turning to restaurants in the middle of the “dining spectrum”, fueling growth in fast-casual and casual restaurants, which offer relatively healthy and lower-priced dining options.


After adjusting all figures for inflation according to the District of Columbia Consumer Price Index, however, the figures tell a much grimmer story. Instead of the top 100 business mostly seeing increased business and consumer activity, most sectors are not able to keep pace with inflation, with only four sectors able to do so (Figure 1). Of these four, only two (transportation and supermarkets) were able to increase revenue at a rate more than 3% faster than inflation, showing the difficulty most companies face while trying to achieve growth.

II. Business Size vs. Tax Impact

As displayed in Figure 2, the top quintile accounted for over 80% of the market share in 2005, while the lowest quintile accounted for a meager 0.2%. Even though in 2015 the top quintile still accounted for the lion’s share of the overall market share (76%), revenue is slowly being redistributed towards the lower quintiles over time, with the share of all four lower quintiles increasing from 2005 to 2015, suggesting an increasingly diverse and competitive marketplace in the city.


More clearly displayed in Figure 3, the vast majority of growth in sales tax receipts is visible in the middle three quintiles, which include taxpayers with liabilities that range from $4,500 to $150,000 in 10% sales tax in 2015. Each of these middle quintiles grew over 30% from their original market share, equating to $80 million in total tax payment growth. Additionally, we see an 18.5% growth in the number of businesses filing a 10% sales tax within the city from 2005 to 2015, signaling a positive environment for small/mid-sized business development and growth.


III. Business vs. Household Tax Impact

Every day, hundreds of thousands of people commute into the nation’s capital, and many more are in the city on business trips, for leisure, and for other reasons. Therefore, unlike most taxes, such as the individual income tax and property tax, the 10% sales tax is not drawn entirely from District of Columbia residents. Rather, it stems from the entire District’s population at any given time, which during working hours is approximately 79% larger than the number of DC residents, according to the US Census Bureau as of 2015. Thus, we wanted to estimate the proportion of the tax falling upon business consumers compared to DC households.

To do so, we grouped four of the aforementioned categorizations as “Business”: Fine Dining, Hotels, Food Service/Catering, and Transportation, while we grouped another four as “Household”: Casual Dining, Fast-Casual, Fast Food, and Supermarkets. We defined the Business industry as sectors where a majority of sales would likely stem from businesses and/or governmental organizations, while we defined the Household industry as sectors where the majority of sales would likely stem from individuals’ retail purchases.


We found that among the top 100 taxpayers, on average, sales tax revenue generated by the business sector represented 62% of total sales, while the household sector only generated 32% of total sales. Since such a large proportion of the tax burden falls upon the business sector, it is clear that rather than individual purchases accounting for the majority of 10% tax revenue, the revenue is largely generated from business purchases of catering, hotel, car, and fine dining services.

IV. Conclusions

By way of our sector analysis of the top 100 taxpayers from 2005 to 2015, we first found that although all but two of the classifications (excluding “Other”) saw nominal revenue increases over the time period studied, this has largely been due to inflation: after an inflation adjustment, only four sectors saw real revenue increases, while only two saw increases significantly above the inflation benchmark.

Next, although we observed the top quintile of businesses in the city accounted for over 75% of the 10% tax revenue during all times between 2005-2015, we have noticed a gradual redistribution of restaurant retail activity towards the lower quintiles, with the vast majority of growth concentrating in the middle three quintiles. This, coupled with an 18.5% increase in businesses paying a 10% tax, suggests that small/mid-sized business growth is prominent in the District.

Finally, even though the city is experiencing steady population growth, gentrification, and other factors that should increase the tax payments of the household sector, the majority of the 10% tax burden still falls on the business sector, with their overall market share actually increasing during the time period studied. This is can actually be viewed as a positive, as many businesses from regions outside the District are contributing hundreds of millions of dollars every year towards D.C. tax revenue, instead of the tax burden entirely falling upon city households and residents.

About this data: Sales tax data are from confidential DC monthly tax filings, compiled by the Office of the Chief Financial Officer, Office of Tax and Revenue. Inflation adjustment data are from the DC Fiscal Policy Institute. Population adjustment data are from the Office of the Chief Financial Officer. Commuter data are from the United States Census Bureau.

The Elephant in the Boom: Global, U.S., and District income inequality in an era of general economic expansion and globalization

The “Elephant Chart”

Across most developed countries, including the United States, income growth has stagnated for low and middle class workers over the last several decades. Real GDP per capita for the U.S. grew merely 36 percent cumulatively during the 20-year period from 1988 to 2008. This period also saw some of the greatest growth stories in world history for China and India, as Chinese and Indian real GDP per capita, as measured in constant international dollars, has increased by 560 percent and 230 percent, respectively, during the same 20-year period, according to data from World Bank.

The stark contrast in global income growth between developed and developing nations is captured succinctly by the graph known as the “Elephant Chart,” as shown in Figure 1. This much-discussed chart was produced by the former World Bank economist Branko Milanovic last summer. We borrowed the chart from Branko Milanovic’s blog.  The chart ranked the world’s households from the poorest 1% to the richest 1%. At each percentile, the chart shows the growth in income between 1988 and 2008, an era of increasing globalization.

The chart, nicknamed the “elephant chart” because of its peculiar shape, shows that in the last few decades the “global middle classes,” mostly from China and India (point A) and the world’s elite (point C) have gained income significantly in the era of globalization, while income for the middle classes in the richest countries (point B) have stagnated. According to Milanovic, from 1988 to 2008, relative income growth was 75 percent for the developing world middle classes and 65 percent for the global top one percent, but only 0-5 percent for the rich countries’ middle classes. Despite acknowledging that correlation is not equal to causation—many economists nevertheless point to the chart as proof that globalization allowed developing countries like China and India to grow at the cost of middle class workers in rich countries.

Figure 1: Change in Real Income Between 1988 and 2008 at Various Percentiles of Global Income Distribution (Calculated in 2005 International Dollars)image001

Source: Branko Milanovic’s blog

How does the U.S. compare?

The “elephant chart” is an interesting and remarkable chart. One may wonder how the contours of inequality for the U.S. and Washington DC look when compared to the “elephant chart.”  Using the Public Use Microdata Sample (PUMS) data from the Census Bureau, we have created our own “elephant charts” for the United States and for the District of Columbia between 2001 and 2015. We will also compare the DC chart based on Census data, with the chart based on personal income tax data collected by the District government. We chose Census data for this specific period mainly for comparison purposes because DC income tax data is available only for this period.

The pattern of inequality in the United States is much different from the world’s. Figure 2 illustrates something that is all too familiar: for the past 15 years, which includes a mild recession and a financial panic, the country’s middle class and the poor have seen their incomes fall. In fact, households in the bottom 75 percent of distribution have experienced income losses of 8 percent on average during the 15 years from 2001 to 2015. Households in the top 25 percent of distribution, on the other hand, have experienced real income gains averaging 5 percent. The upper-class households in the top 5 percent of the distribution have been doing particularly well, with income gains averaging about 14 percent during the same period. The pattern does remind us of the so called “Matthew Effects,” where the rich get richer and the poor get poorer.

Figure 2: Real Income Gains Between 2001 and 2015 for U.S. Households at Various Percentilesimage002

            Source: U.S. Census Bureau

How about the District?

In contrast to the income trend nationally, the majority of households in Washington DC saw their incomes rise, especially among middle-income percentiles. The DC contour, as shown in Figure 3, is “elephant-like:” Middle class households and top earners in the District have experienced the largest income growth.

Figure 3: Real Income Gains for DC Residents at Different Percentiles Between 2001 and 2015image004

            Source: U.S. Census Bureau

Since the Census data for DC is survey based on a sample size of only about 3,600 DC households, the sampling error is relatively large. On the other hand, the number of tax filers in DC averages about 360,000 per year.  The blue line in Figure 4 shows the distribution of before tax income growth for DC based on DC tax filer data. Consistent with the chart based on census data, most DC residents have experienced income gains even after taking inflation into consideration. This is a clear contrast with the national pattern. One notable feature of Figure 4 is that residents at around the 21st to 25th percentiles experienced stagnant income growth. Low income tax filers from the 5th to 20th percentiles fared better. We believe this may be attributable to DC minimum wage policy. Since 2001, the DC minimum wage has grown much faster than inflation. As shown in Figure 5, the DC minimum wage had increased by 71 percent while the cost of living had increased by 41 percent. Thus, residents who have benefitted from the rising minimum wage (those in the 5th to 20th percentiles) have seen their income growing faster than residents working at wage rates just above the minimum wage (those in the 21st to 25th percentiles).

Figure 4: Before and After Tax Real Income Gains for DC Residents at Different Percentiles Between 2001 and 2015 Based on Income Tax Dataimage006

            Source: Office of Tax and Revenue

Figure 5: DC Minimum Wage VS the Consumer Price Index (CPI)image008

            Source: U.S. Department of Labor

The red line in Figure 4 shows an income growth pattern after adjusting for the District government’s tax policies such as taxable deduction, exemption and refundable and nonrefundable tax credits. The area between the blue line and the red line represents increase in income growths that was due to tax policies. The poor apparently benefited significantly from this tax policy. The DC earned income tax credit (EITC) did most of the job in helping the low-income residents (at or below the 30th percentile), as the District’s EITC credit has increased from 25 percent of the federal credit amount in 2001 to 40 percent of the federal amount since 2009.  For reference, Table 1 shows how much income DC tax payers earned in 2015 at certain percentiles both before and after tax.


Source: Office of Tax and Revenue

Table 2 takes another look at how taxpayers at different income percentiles fared from 2015. The share of total income for the bottom 30 percent is much lower despite their relative increase from 2001 to 2015. The bottom 30 percent taxpayers earned only 5 percent of total before-tax income in 2001, and this percentage dropped to 4 percent in 2015. The DC Income tax schedule did help the taxpayers in the bottom 30 percent by maintaining their after-tax income share at 5 percent from 2001 to 2015.

The middle class, typically defined as households with income between the bottom 30 percent and the top 20 percent, fared slightly worse, with the before-tax and after-tax income shares declining by one percent and two percent respectively between 2001 and 2015.

The top 20 percent upper class households fared much better than the middle class by gaining both before-tax and after-tax income shares. The top 1 percent received a lion’s share of 23 percent of total before-tax DC income in 2001, and this percentage increased to 24 percent in 2015. The top 1 percent actually benefited more from the DC tax policy as their share of total after tax income increased from 20 percent in 2001 to 23 percent in 2015.

Table 2. Income as a percentage of total taxable income for taxpayers at different percentiles of income distribution


In conclusion, we have gained substantial knowledge about inequality in DC and in U.S. from reproducing the “elephant chart” for the period between 2001 and 2015. We found that households in Washington DC on average experienced faster income growth than their counterparts in the nation for this period. Although the top one percent earners in DC and in the U.S. have both done well, the middle class in DC actually experienced real income gains, while the middle class in the U.S. experienced real income losses. We also want to emphasize that DC’s minimum wage and tax policies have helped lower income DC residents gain income faster than their national peers.

What exactly is this data?

Our data on personal income tax is from Office of Tax and Revenue personal income tax returns for tax year 2001 to 2015. The minimum wage and inflation data is from the Bureau of Labor Statistics, U.S. Department of Labor. Public Use Microdata Sample data (PUMS) from the Census Bureau was also used to calculate the distribution of income growths for the U.S. and DC.

Fitzroy Lee, Stephen Swaim and Bob Zuraski contributed to this post.