As has been reported, the U.S. Census Bureau estimated that the District (D.C.), after years of population increases, lost nearly 20,000 people between April 1, 2020, and July 1, 2021. This is not surprising since moves out of the city accelerated after April 2020, when COVID-19 arrived in D.C. Now, nearly two years into the pandemic, are we seeing people return to the city? The data suggests yes. Apartment vacancy data and change of address data from the U.S. Postal Service (USPS) point to the city’s population reaching a low in the winter of 2020/2021, and then recovering to a large extent by the late summer/fall of 2021.
Apartment vacancy data from Delta Associates, which excludes vacancies in newly built units, shows vacancies in the District peaking the last quarter (Oct-Dec) of 2020 and first quarter (Jan-Mar) of 2021, with the increase in vacancy rate most pronounced for newer apartments with amenities (Class A). The vacancy rate then began to drop in 2021 and was below the pre-pandemic rate in the third quarter (Jul-Sep) of 2021. This is a strong sign of population recovery. Vacancy rates in the suburbs remained relatively flat through the pandemic, which perhaps isn’t surprising since our own research and others’ show the pandemic benefited the suburbs in this way.
Data from the USPS, which provides the number of net moves out of the city, show a similar pattern of population loss then rebound. Net moves out peaked above their pre-pandemic levels in September 2020. They then declined but continued above pre-pandemic levels through early spring 2021. In July 2021, net moves out dropped below pre-pandemic levels for the first time, and remained below pre-pandemic levels through October 2021, signaling a gain in population. This likely gain was due to both more people entering the city and fewer people leaving than pre-pandemic. Unfortunately, it is difficult to get an exact count of population recovered using the USPS data because it almost certainly undercounts people returning.(1)
It is possible not all of this population recovery was captured in the latest Census estimate, since the estimate reflects the population on July 1, 2021, and our data indicates recovery was still happening then.
According to USPS data by zip code, the population rebound appears mostly to have been driven by the neighborhoods that lost the most people earlier in the pandemic: those in and near downtown. However, Navy Yard and south Capitol Hill (zip code 20003) had the strongest in-migration numbers in last summer and fall’s USPS data, despite showing middling population loss earlier in the pandemic. This could be due to the large number of new housing units there. Upper northwest (zip codes 20015 and 20012) also appears to have contributed to the rebound, which is somewhat surprising since it had nearly no pandemic population loss in 2020, according to our prior research.
On the flip side, USPS data shows neighborhoods on the eastern edge of the city still had higher out-migration than pre-pandemic when other neighborhoods were rebounding. East-of-the-river zip codes 20019 and 20020 had roughly the same elevated level of outmigration last summer and fall as earlier in the pandemic, though earlier in the pandemic they didn’t have the stark outpouring of people that we saw downtown.
(Note that zip codes 20007 (Georgetown) and 20017 (Brookland) could be showing higher out-migration due to the presence of universities, though we are not seeing similar patterns in other zip codes with universities.)
While the USPS data shows a somewhat uneven recovery among neighborhoods, apartment vacancy data shows vacancy rates very close to or below pre-pandemic levels among all neighborhoods for which we have data.(2) However, it would be possible for us to see pre-pandemic vacancy rates yet a lower population in some neighborhoods, or even city-wide, if household sizes shrunk due to some members of a household leaving and others staying (as may be the case with roommates or inter-generational households.)
There is also the question of whether those who left the city are similar to those who returned. A decline in school enrollment and evidence of moves to the exurbs could mean families were more represented among leavers than they were among returnees. This migration pattern could also lead to low apartment vacancy rates but a lower population than-pre-pandemic due to smaller household sizes.
There are signs the population recovery is slowing or ending, and even if D.C. has made it back to its pre-pandemic population, there is still the risk of low growth or population stagnation moving forward. In November 2021, USPS data showed outmigration returning to 2019 levels, when D.C. had close to zero total migration and negative domestic migration, according to Census, meaning significant population gains could have ended. Other cities faced the same plight pre-pandemic: Boston, New York City, and San Francisco all had negative domestic migration in 2019, but places like Austin, Raleigh, Denver, and the D.C. exurbs were attracting people and continue to do so. In fact, there were severalreportsnoting that the pandemic increased moves to places already attracting people. Telework of course remains the big wild card; if it continues at substantial levels post-pandemic it could change not only the number of people who live in D.C. but the types of people who choose to live here.
(1) The USPS data counts international moves out, but not in, which are typically a substantial portion of migration to D.C. The absence of international moves in will almost certainly lead to an undercount of people returning, even after normalizing the pandemic data by comparing it to pre-pandemic data.
(2) Delta Associates shows the following stabilized vacancy rates (rates that exclude new units) for D.C. neighborhoods:
*Note that while Cap Hill/Riverfront/SW shows slightly higher than pre-pandemic vacancies, this area had the highest apartment absorption city-wide in 2021, which is not reflected in the stabilized vacancy rate in the above table.
The Class A and B neighborhood data shown above from Delta Associates does not include data for neighborhoods east of the Anacostia River. However, CoStar data shows vacancy rates in “Anacostia Southeast” have dropped to pre-pandemic levels (from 7% in 2019 Q3 to 6% in 2021 Q3).
CoStar data also shows that Class C apartment vacancies city-wide dropped to pre-pandemic levels in 2021 Q2.
When the pandemic restrictions were put in place last Spring, there was plenty of anecdotal evidence that people were leaving D.C. as white-collar workers became remote, others were laid off, and schools shut down. Data from the United States Postal Service (USPS), which records changes of address, provides a clue as to how many people left D.C. and where they went. This data gives us an insight into peoples’ housing choices when work and school are decoupled from where you live, something important to understand if an increase in remote work extends beyond the pandemic.
The USPS data clearly shows more people moved out than in during 2020 than 2019, with moves accelerating after COVID-19 restrictions began in March 2020. In 2019, USPS recorded 11,480 net moves out of the city, while in 2020 that number increased to 29,362, an increase of 17,882 net moves out (or 2.6 times more).(1) Net moves out is the number of moves out of the city that exceed the number of moves in and is a proxy for population loss. The number of moves reflects change of address forms filed by both individuals and families.(2) While there was some decrease in the number of moves into DC in 2020, about 90% of the increase in net moves out of the city results from additional people moving out of the city.(3)
Not all of the moves in 2020 were permanent. When someone submits a change of address form they mark the move as permanent or temporary(4) and the data clearly shows a higher portion of moves than usual were temporary in 2020. Out of the increase of 17,882 net moves out between 2019 and 2020, we estimate 9,335 were permanent and 8,547 were temporary.(5)
The data shows more people moving out of the city than into it in both 2019 and 2020, despite the decennial Census count showing strong population growth between 2010 and 2020. However, D.C.’s population increase from moves within the United States has steadily decreased the last several years, and 2018 and 2019 IRS data and 2019 and 2020 Census population estimates show D.C. lost more people to other parts of the country than it gained. The last couple years, any population increase in D.C. has been entirely driven by births and international migration, according to these estimates.
The USPS data includes at least some, perhaps even most, moves out of D.C to international locations,(6) but does not include moves in from international locations. While USPS tells us that the number of international moves in the dataset for the nation as a whole, from which we extracted data for D.C., is “insignificant”, international migrants make up a sizeable portion of people moving to and from the District. The absence of incoming international moves in the USPS data likely leads to an overcount of net moves out of D.C. each year, but the increasein net moves out between 2019 and 2020 (roughly 17,000, or 9,000 permanent) will not be overcounted unless international migration into D.C. increased between 2019 and 2020, which seems unlikely due to the pandemic. In our end notes we adjust the USPS data with estimates of the missing international moves and take into account families being counted as only one mover to illustrate how the USPS data might translate into net numbers of people moving out of D.C. in 2019 and 2020.(7)
What parts of the city did people leave?
The zip codes close to the downtown core, especially those close to the west end of downtown, appear to have had the biggest population loss during the COVID-19 period, even when accounting for population differences between zip codes. The map below shows this pattern. Specifically, it shows the increase in net outmigration from March-December 2019 to March-December 2020, per every 1,000 residents in each zip code.
The zip codes with the largest population loss due to COVID-19 were 20036 (south side of Dupont/Golden Triangle), 20009 (Adams Morgan, Columbia Heights, 14th/U St NW, Dupont), 20024 (Southwest Waterfront) and 20005 (south side of Logan Circle/Franklin Square). All of these zip codes had at least 60 more net moves out per 1,000 people in Mar-Dec 2020 than Mar-Dec 2019.(8) To put those numbers in perspective, 20009, the largest zip code of the bunch with about 52,000 people, lost about 1,700 people from Mar-Dec 2019, even after accounting for people moving in. That number increased to 5,500 people from Mar-Dec 2020. These zipcodes have relatively high numbers of multifamily buildings, indicating that perhaps apartment dwellers were more likely to relocate during the pandemic than those in single family homes.
Which zip code fared the best during the pandemic? Zip codes 20012 (Shepherd Park, Takoma) and 20015 (Chevy Chase, Friendship Heights, Barnaby Woods) had nearly no change in migration between Mar-Dec 2019 and Mar-Dec 2020. In fact, zip code 20015 was the only one where USPS data showed more residential moves in than out during both time periods.
Where did people go?
The data we received from USPS on the destinations of people moving from D.C. was highly redacted for privacy reasons, limiting the conclusions we can make. (This data was received via Freedom Of Information Act (FOIA) request, whereas the data used for the graphs above is publicly available.) Still, with the data we have, we can see that a large portion of movers–at least 31%, but likely many more–moved to another location within the D.C. metro region between March and December 2020. This aligns with other research(9) showing a large majority of moves during COVID-19 were not to far-flung places but instead to locations within the same metro region.
Moves to the D.C. suburbs
For moves within the D.C. metro region, we had a complete dataset for 10 zip codes located in the close-in suburbs of Bethesda, Chevy Chase, Oxon Hill, Silver Spring, Alexandria, and Arlington. This data shows that in a typical year (Mar-Dec 2019), D.C. lost more people to these zip codes than it gained from them. In 2020 this trend accelerated. From Mar-Dec 2020 these suburban zip codes gained, on net, 3 times as many people from D.C. than during the same period in the prior year. (Net moves is the difference between moves in and out).
We can expand the suburban zip codes for which we have complete data if we look only at moves out of D.C. to the suburbs without accounting for people moving from the suburbs to D.C. Looking at only moves out from D.C., we see that suburbs to the north and west of the city seemed to have had the largest increases in moves from D.C. during the pandemic (Mar-Dec 2020) compared to a year prior (Mar-Dec 2019). The Bethesda zip code of 20817 recorded the highest percent increase in moves from D.C. (63%). That zip code received 246 movers from D.C. between March and December 2019 and 400 between March and December of 2020.
The other zip codes to have an increase in movers from D.C. of 50% or higher during the pandemic were 22203 and 22206, both in Arlington.
Moves out of the D.C. region
The redacted USPS data only shows moves from D.C. to places outside the D.C. region for destination zip codes that received more than 10 people from D.C. in a single month. Because this data only shows a relatively large number of people moving from D.C. to a particular place at a particular time, we refer to these moves as “high volume moves.” The USPS data shows 1,608 high volume moves from D.C. to places outside the region between March and December 2020 and 673 of these moves during the same period in 2019.
One distinctive feature of high volume moves in the data we reviewed is that there were simply more high volume move destinations in 2020 than 2019 for locations outside the D.C. region. The map below shows the high volume move destinations outside the region in Mar-Dec 2020 that were not high volume destinations in Mar-Dec 2019. The big surprise from the data is the Delaware beaches were in the top destinations in 2020.
Columbia, MD, and Charlottesville, VA, were the next most common destinations outside the region after Rehoboth Beach and Lewes, DE, that had high volume moves in Mar-Dec 2020 but not Mar-Dec 2019. Notably, these are both small to mid-sized cities not too far outside the boundaries of the D.C. metro region. Annapolis and Baltimore were the top high volume move destinations in both 2020 and 2019.
Moves within the District
In addition to capturing moves in and out of the city, the USPS data shows moves within D.C. The number of moves within the city is close to the number of moves out of the city. From Mar-Dec 2019 USPS data shows 43,074 intracity moves. During the same period in 2020 there were 49,632 intracity moves.
As the map below shows, the general pattern of intracity moves is for people to move from neighborhoods close to downtown, especially those to the north and west of downtown, to outer neighborhoods. Areas east of the Anacostia River also lose residents to other parts of the city. The pattern of people moving from downtown to outer neighborhoods strengthened during the pandemic.
From Mar-Dec 2019 the neighborhoods that lost the most residents to other parts of the city were 20005 (south side of Logan Circle/Franklin Square), 20036 (south side of Dupont/Golden Triangle), 20037 (West End, Foggy Bottom), and 20001 (LeDroit Park, Shaw, Mt Vernon Triangle), all of which had a net population loss of more than 10 movers per 1,000 residents to other parts of the city. Zip codes 20005 and 20036 had the highest loss at 33 and 32 movers per 1,000, respectively. The zipcodes that gained the most population from intracity moves during this time were 20018 (Brentwood, Langdon, Woodridge, Fort Lincoln), 20017 (Brookland), 20015 (Chevy Chase, Friendship Heights, Barnaby Woods), and 20003 (Navy Yard, south side of Capitol Hill). These zip codes had net gains ranging from 11 to 17 movers per 1,000 residents, with 20018 gaining the most people.
Migration patterns within the city looked much the same during the height of the pandemic (Mar-Dec 2020). The differences that stand out are that during the pandemic several of the zip codes that had gained people in 2019 gained even more people, especially 20003 (Navy Yard, south side of Capitol Hill) and 20015 (Chevy Chase, Friendship Heights, Barnaby Woods), which saw the largest increases. Likewise, many of the zipcodes that lost people in 2019 lost even more during the pandemic. Zip codes 20036 (south side of Dupont/Golden Triangle) and 20009 (Adams Morgan, Columbia Heights, 14th/U St NW, Dupont) had the biggest jumps in movers lost per 1,000 residents. Notably, zip code 20008 (Connecticut Ave. corridor) went from losing people pre-pandemic to gaining people during the pandemic.
It may come as a surprise that zip code 20003, which includes Navy Yard, a neighborhood with many multifamily buildings, had the largest influx of people from other parts of the city during the pandemic, since other areas with a lot of multifamily housing had some of the largest population losses. New apartments coming online and offering incentives for moving in may be why Navy Yard and adjacent areas were able to attract so many residents from other parts of the city.
Will people return?
The question policymakers across the country are asking is will people return to cities they left during the pandemic. Presumably those who filed a temporary change of address request will return, as well as some who left temporarily but filed a permanent move, such as apartment dwellers returning to a different unit. For those who truly moved permanently, it remains to be seen if the city can attract people to replace them, something that could depend on telework policies post pandemic. The good news is that as of May 2021 the USPS data shows net moves out of the city have returned to 2019 levels. But for the city to regain the population it lost, we would need to see an influx of residents into the city at levels we have not seen in several years.
-Ginger Moored, Fiscal Analyst
Thank you to Susan Steward for sharing her knowledge of the FOIA process and to Norton Francis, Lori Metcalf, Daniel Mohammed, and Steve Swaim for their edits and feedback.
(1) USPS redacted data for zip codes with 10 or fewer moves of individuals or families in a given month. While the vast majority of data was left intact, a few smaller zip codes had data redacted for a few months. To test the accuracy of our conclusion that moves in 2020 were 2.6 times those in 2019, we redid the analysis excluding all data for zip codes where some data was missing. That analysis shows 2020 moves were 2.5 times higher than those in 2019.
(2) Families can file a single change of address form for multiple people if all of those people have the same last name and have the same origin and destination. Most change of address forms filed in D.C. are for individuals.
(3) The table below shows the data behind our net move calculations.
(4) A temporary move is valid for up to six months and can be renewed for up to another six months; USPS confirms that renewals are not counted in the data and temporary moves converted to permanent moves are not counted twice.
(5) The USPS dataset gives the percentage of all moves, including business moves, that were temporary and permanent. To estimate the number of permanent residential moves, we applied this percentage to residential moves only. Residential moves greatly outnumber businesses moves in the data. It is possible that some people who filed a permanent move do have plans to return to the city; for instance, a renter who left temporarily might have filed a permanent move since they would be moving back to a different apartment unit upon return.
(6) While domestic movers can file a change of address online, international movers must print out a paper form and bring it to their local post office, which could have an effect on the number of forms filed by international movers. USPS accepts change of address forms for people moving from the U.S. to another country, but not for people moving from another country to the U.S.
(7) Census PUMS data shows an estimated 10,606 people moving from an international location to D.C. in 2019. If we add these people to the USPS data for both 2019 and 2020 and assume all families counted in the USPS data have two family members, the net number of people moving out of D.C., or population loss due to moves, would be 2,881 in 2019 and 23,555 in 2020, meaning D.C. lost an additional 20,674 people in 2020. Adjusting for redacted data in the USPS dataset (see End Note #1) would bring that loss down to approximately 19,443 people (12,898 permanent). The actual loss will be higher if the average USPS “family” has more than 2 people and if international migration to DC in 2020 were less than in 2019.
(8) Downtown zip codes 20004, 20005, and 20036 had insufficient data on moves of families due to the low number of these types of moves. Therefore, data for these zip codes in the map “DC Zip Codes that Lost the Most People” reflects only moves of individuals. Data for all other zip codes reflects moves for both individuals and families. Limiting the data in this way did not change the ranking of the top four zip codes with the largest increase in moves out mentioned in our narrative, nor did it change the conclusion that they all had an increase of at least 60 net moves out between Mar-Dec 2019 and Mar-Dec 2020.
(9) See “More Americans Are Leaving Cities, But Don’t Call It an Urban Exodus”, published by Bloomberg CityLab (Patino, Kessler, Holder, Gu, and Rojanasakul) on April 26, 2021, available here: https://bloom.bg/2TLvvDC
What exactly is this data?
The number of net moves out of D.C., including the number of net moves by D.C. zip code, comes from a publicly available dataset on the USPS FOIA (Freedom of Information Act) website. You can access that data here. This dataset separates moves into three categories: business, family, and individual. This analysis looks only at residential moves, which is the sum of family and individual moves. A move is counted as a family move if one form is submitted for multiple people with the same last name. Individual moves are filed for just one person. The data also separates moves into those that are temporary and permanent. A temporary move is valid for up to six months and can be renewed for up to another six months; USPS confirms that renewals are not counted in the data and temporary moves converted to permanent moves are not counted twice.
D.C movers’ destinations and intracity moves by zip code come from a dataset the Office of Revenue Analysis obtained through a FOIA request for USPS data. You can access that data here. This dataset includes residential moves only but does not separate out moves into individual and family or temporary and permanent.
The “D.C. metro region” refers to the Core-Based Statistical Area (CBSA). We matched zip codes to the D.C. CBSA using a crosswalk available on the website of the U.S. Department of Housing and Urban Development. The crosswalk is available here.
Zip code populations come from the 2019 5-year American Community Survey data published by the Census Bureau. That data can be accessed here.
Around 829,000 people work in D.C. (within the city-proper), and about 26 percent of them are immigrants. Today, the Washington Post reports that some of D.C.’s immigrant workers, particularly those working in restaurants and some daycare centers and schools, are going on strike.
Indeed, the industries with people on strike have some of the highest concentrations of immigrants in D.C., as you can see in the graph below. Seventy-one percent of chefs and head cooks working in D.C. (within the city-proper) are immigrants, as are 61 percent of lower-rank cooks. Fifty-seven percent of childcare workers in D.C. are immigrants. The occupation in D.C. with the largest concentration of immigrants is carpenters, 80 percent of whom are immigrants. (In our analysis we only looked at occupations with more than 3,000 workers in D.C.)
Most of the occupations with the highest concentrations of immigrants in D.C. are those with low or middle wages. However, immigrants comprise almost half of D.C. workers in several high-wage occupations: economists (46 percent), mathematicians and statisticians (43 percent), and physical scientists (42 percent).
We define a job as low-wage if its median wage was in the bottom 25 percent of median wages across all jobs in D.C (or below $44,000). High-wage jobs have median wages in the top 25 percent (or above $86,000) and middle-wage jobs are in between.
While low-wage jobs in D.C. have the highest concentration of immigrants (41 percent of all low-wage workers in D.C. are immigrants, compared to 22% of middle- and high-wage workers), the number of immigrants in low-wage jobs in the city is roughly equal to the number of immigrants in high-wage jobs, since the city has many more high-wage workers. There are about 75,000 immigrants in low-wage jobs in D.C. and about 73,000 immigrants in high-wage jobs.
What exactly is this data?
Our data on immigrants by occupation comes from the 2015 American Community Survey 1-year PUMS data. “Immigrants” include naturalized U.S. citizens and non-citizens. “D.C. workers” are people who live in D.C., Maryland, and Virginia who report D.C. as their place of work. We only look at occupations with more than 3,000 people working in D.C. in order to reduce sampling errors. Because the ACS is based on a sample, there is a margin of error in all of our calculations. Our calculations should be treated as estimates, not precise counts.
Our wage data comes from the May 2015 Bureau of Labor Statistics Occupational Employment Statistics Survey. We define a job as low-wage if its median wage was in the bottom 25 percent of median wages across all jobs in D.C (or below $44,000). High-wage jobs have median wages in the top 25 percent (or above $86,000) and middle-wage jobs are in between. In cases where the occupation code in the ACS data did not match the occupation code in the BLS data, we calculated a median wage using the ACS data.
Nearly 800,000 people work in the District of Columbia, yet only about 30 percent of the District’s workers live in the city-proper. Workers in low-wage jobs are more likely to live in the city than those in middle- and high-wage jobs. Thirty-nine percent of D.C.’s workers in low-wage jobs lived in the city between 2010 and 2014, compared to 30 percent in middle-wage jobs and 27 percent in high-wage jobs.
We define a job as low-wage if its median wage was in the bottom 25 percent of median wages across all jobs in D.C (or below $44,000). High-wage jobs have median wages in the top 25 percent (or above $86,000) and middle-wage jobs are in between.
You can see how this plays out by occupation in the graph below. Cashiers, janitors, childcare workers and others in low-wage jobs are more likely to live in the city than most other workers, though people in a handful of middle- and high-wage occupations, like managers of social and community services, teachers, and chief executives, have relatively high rates of living in the city too. Registered nurses and police officers (which include transit and federal police) are the least likely to live in the city.
People in low-wage jobs tend to live in the city more than others, but that’s been changing over the past decade, as you can see in the chart below. The city is losing construction workers, cashiers, childcare workers, and janitors, and gaining people in high-wage jobs, like managers of social and community services, operations research and management analysts, and economists.
In less than a decade, the workers most likely to live in the city shifted from cashiers, retail salespersons and clerks (50 percent lived in the city in 2005-2009) to managers of social and community services (47 percent lived in 2010-2014).
Meanwhile, over the same time period, the least likely to live in the city switched from software developers (9 percent in 2005-2009) to police officers (11 percent in 2010-2104).
As the graph below shows, this is part of a larger pattern of D.C. workers in middle- and high-wage jobs starting to show a preference for living in the city, and workers in low-wage jobs increasingly living in the suburbs – a trend that’s unsurprising given the District’s increasing cost of housing. The percent of workers in low-wage jobs living in the city decreased from 43 percent to 39 percent between 2005-2009 and 2010-2014, while the percent of workers in high-wage jobs living in the city increased from 24 percent to 27 percent over the same time period. These changes may seem small, but they are statistically significant at the 99 percent confidence level.
What exactly is this data?
Wage data: Our wage data comes from the Bureau of Labor Statistics Occupational Employment Statistics survey of D.C. workers from May 2015. We define a job as low-wage if its annual median wage was in the bottom 25 percent of annual median wages across all jobs in D.C (or below $44,000). High-wage jobs have annual median wages in the top 25 percent (or above $86,000) and middle-wage jobs are in between.
Percent of Workers Living in the City: Our data comes from the American Community Survey PUMS data for 2005-2009 and 2010-2104. For our universe of D.C. workers we started with everyone living in D.C., Maryland, or Virginia who works in D.C., so we are excluding long-distance commuters who work in D.C. but live in places outside of D.C., Maryland, and Virginia. When we analyzed specific occupations, we looked at all occupations with 8,000 or more workers in D.C., with the exception of miscellaneous managers since the category is vague. We grouped some occupations together so they surpassed our 8,000 person threshold.
Map of Where Workers Live: This data comes from the American Community Survey PUMS data for 2014. We only look at workers who work in D.C. and live in either D.C., Maryland, and Virginia. All of the occupation groups in the map have 8,000 or more people working in the city.
Police Officers: Police officers in this case includes more than just people employed by the Metropolitan Police Department; it also includes transit police, federal police, and police who said they work for private organizations. In 2014, the Metropolitan Police Department released data showing 17 percent of its officers live in the District.
Errors: The data in this post have various margins of error since the data comes from surveys. In most cases we used a five-year data set to reduce the errors, and only looked at occupations with 8,000 people or more. The errors are highest for the map of where people live because for that we had to use a one-year dataset (geographic boundaries changed within the five-year dataset, making a map more difficult to produce). The map is intended to give readers a general sense of where people live; we discourage people from using it for direct area-to-area comparisons. Our findings on the loss and gain of workers of different occupations and wage levels are in many cases statistically significant and we have noted this in the post.
Recently we reported that the total number of children in D.C. has been increasing since 2010 after about fifteen years of decline. Since 2008 the increase has been driven by young children under 6, and in 2012 the number of school-aged children (aged 6 to 17 years) began to increase as well.
Now we look at how the number of children in different income groups has been changing. We define lower-income as households making less than $50,000; middle-income as households with incomes between $50,000 – $150,000; and higher-income as households making over $150,000
Both higher-income households and lower-income households are driving the boom in young children under 6. The average number of children in both income groups increased between the two five-year periods for which we have reliable data: 2006 to 2010 and 2010 to 2014. The higher-income group, though, grew more.
A different story emerges for school-aged children. For children between the ages of 6 and 11, there was a net increase only among higher-income kids. The number of lower-income and middle-income children in this age group declined slightly, but because the declines are small and the data has a sampling error, we are less confident in these trends. (We discuss the reliability of the data at the end of the post.)
For older children between the ages of 12 and 17, the average number of lower-income kids declined between the two five-year periods that ended in 2010 and 2014. The data shows a slight increase in the average number of higher-income children and a slight decrease in the average-number of middle-income children, but we have less confidence in these trends since the changes are small.
Lower-income children still outnumber middle- and higher-income children of all ages. But the higher-income group appears to be catching up with the middle-income group, especially among younger children.
Is this proof of gentrification? The general trends suggest yes. When we look at children of all ages, only the higher-income group has, on net, added children. The total number of lower-income and middle-income children has remained about the same between the two five-year periods that ended in 2010 and 2014. This means that higher-income children now make up a bigger portion of all children in the city. But it could be that because we’re looking at five-year groupings of data, we’re missing year-to-year trends that could show more nuanced trajectories for children in different income groups. It’s quite possible, for instance, that the number of school-aged lower-income children has been increasing the past few years, but losses in the late 2000’s and early 2010’s have outweighed recent gains. Or it could be that the number of school-aged lower-income children has not begun to increase yet, but will soon, following a trajectory similar to that of higher-income children.
What exactly is this data?
Our data comes from the American Community Survey (ACS) 5-year data sets for 2006-2010 and 2010-2014 for the District of Columbia. We could not use ACS 1-year data sets because the small sample size makes them unreliable for this analysis.
Errors for each of the data points in our graph range from +/- 6% to +/- 11% for 90% confidence intervals. For the bolded lines in our graph, trends (population loss or decrease) still hold even if we assume the largest errors (at 90% CI) in the least favorable direction. For the non-bolded lines, trends reverse when we factor in the largest errors in the least favorable direction.
The median income for a household in D.C. is about $73,000, but household incomes vary widely depending on the type of household you live in. The households with the highest incomes are those headed by married couples. Married people with children at home have a median household income of $161,000, and married people not living with children have a median household income of $136,000. The lowest-income households tend to be those headed by an unmarried person with children. Among these households, the median income is $36,000.
These extremes in income don’t appear to be influenced much by household size once there’s more than one person in the household; 2-, 3-, and 4-person households in the city all have median incomes between roughly $90,000 and $100,000.
The large differences in income are likely due to a number factors, like number of earners in a household; higher incomes for people farther along in their careers; and higher marriage rates for people with college degrees, who are likely to earn more money than those without a college education.
What exactly is this data?
We analyzed the 1-year American Community Survey Public Use Microdata Sample (PUMS) for 2014. We converted all incomes into inflation-adjusted 2014 dollars.
Here are more details on the household categories we formed for our analysis:
Married with kids: Married couple with children present in the household.
Married no kids: Married couple with no children present in the household.
Two or more unmarried adults living together: Un-married householder living with other adults; no children present. People in household can be related.
Living alone: Adult living alone; no children present.
Unmarried with kids: Un-married householder with children present in the household. Other adults may be present in the household too.
Last year, 48 percent of students attending D.C. Public Schools were at-risk, meaning they were homeless, in foster care, qualified for food stamps or TANF, or were high school students who’d been held back a grade. The share of at-risk students in a DCPS school varies greatly, from 1 percent to 90 percent.
DCPS students are assigned to a neighborhood school but school mobility is high. In 2012, 43 percent of DCPS students attended their assigned neighborhood school. For charters, though, there are no attendance areas. So we wondered: can charter schools, which base admissions entirely on a lottery instead of a student’s address, better mix at-risk students than DCPS and create more egalitarian schools? Our answer is a qualified no.
Though charters last year enrolled about the same percentage of at-risk students as DCPS (44 percent), most charters (47 out of 53 elementary and elementary/middle schools) have lower concentrations of at-risk students than nearby DCPS schools serving similar grades—and by nearby, we mean within the same DCPS elementary zone. The difference in concentrations of at-risk students between charters and DCPS is greatest in areas of the city where high numbers of at-risk students go to school. For DCPS schools where 75 percent or more students are at risk, the at-risk population in nearby charters is lower by at least 10 percentage points. In areas where DCPS schools have smaller at-risk populations, some nearby charters have at-risk populations closer to those of DCPS, while others have some of the lowest concentrations of at-risk students in the city. Charters tend not to locate near DCPS schools with the lowest concentrations of at-risk students.
We found similar results when we plotted the percent of at-risk students in a school versus neighborhood home prices. In neighborhoods with the lowest home prices, charters have considerably smaller portions of at-risk students than DCPS schools. As home prices increase, there are charters that have at-risk populations more similar to those of DCPS schools, though other charters in these areas have very small at-risk populations. There are few charters in the areas with the highest home prices.
What does this look like across the city? The map below shows that the concentration of at-risk students in both DCPS and charters generally increases as you move from northwest to southeast (and here we show all schools, not just elementary and elementary/middle schools). West of Rock Creek Park, where the DCPS schools with the lowest concentration of at-risk students are located, there are no charter schools, and the charters with the lowest concentrations of at-risk students are located in the center of the city. East of the Anacostia River, charter schools tend to have lower concentrations of at-risk students than DCPS schools.
Put another way, both DCPS and charters have some of the least economically diverse schools in the city—they’re just located in different parts of the city.
The map below gives you another way to see what’s going on in different parts of the city by comparing charters and DCPS side-by-side. The map is interactive and it lets you zoom in on different neighborhoods and filter schools by the size of their at-risk populations.
One thing that strikes us in the map is the marked difference in at-risk populations in charters and DCPS schools east of the Anacostia River. Since only two charter schools east of the river had more than 25 percent of their students cross the river to go to school last year, it seems likely that charters here attract non-at-risk students from neighborhood DCPS schools or keep non-at-risk students in the public school system who would have otherwise left for private schools or moved.
The map below also shows that in the north and northeast parts of the city, plenty of charters have at-risk populations similar to those of DCPS schools, while other charters have much smaller at-risk populations.
We’re curious to hear what you conclude from these graphics and what patterns you see that we may have missed.
What exactly is this data?
Data on at-risk populations is from the Office of the State Superintendent of Education and is for the 2014-2015 school year. Readers should note that some types of students that might seem at-risk might not be counted as such, including:
undocumented students who qualify for food stamps and TANF but do not receive benefits because of their immigration status;
adult students; and
students enrolled in alternative education programs.
School locations are from the Friends of Choice in Urban Schools. We define a charter’s “nearby” DCPS school as the DCPS elementary school for which a charter’s location is zoned in the 2015-2016 school year. The boundaries for the DCPS elementary school zones are from the Office of the Chief Technology Officer.
We have excluded from our analysis schools that serve alternative, special education, and adult populations.
Three charter schools in the graphics above closed after the 2014-2015 school year: CAPCS, Tree of Life, and Options.
Thank you to Yesim Taylor for her comments on this post.