Commuter Characteristics in US Cities

How do people who commute via public transit differ from those who drive? I’ve been thinking about this question for a while. Based on what I’ve read, the prevailing wisdom today seems to point to widespread demographic differences between public transit commuters and car commuters. There is strong evidence suggesting economic and racial disparities between the two groups. I’ll talk about why one might expect each of those disparities, and then dig into the data to see how widespread these differences might be.

There are clear cost advantages to mass transit that produce an economic disparity between transit and car commuters. According to AAA, owning a car in 2017 cost an average of $8,849 per year (nationally, and this doesn’t take into account parking costs in large cities), whereas an MBTA pass in Boston was $84.50, or $1,014 annually. Given this reality, it makes sense that transit riders would be more economically disadvantaged than car commuters, and that this difference would manifest in any number of other markers: lower homeownership rates, more debt, higher poverty rates, less household wealth.

In addition to an economic disparity between different kinds of commuters, one might also expect a racial disparity. The data seems to confirm this. A report from the American Public Transportation Association, Who Rides Public Transportation, found that 24% of transit riders are African American; in comparison, African Americans make up 12% of the general population.

Why might transit use be tied to race? The story is not as simple as race being associated with income, and income being associated with mode of transit. Rather, generations of government policies throughout the 20th century reinforced racial segregation in urban areas, which in turn contributed to de facto segregation of mass transit. Douglas Rae’s book, Urbanism and its End, describes how New Deal policies singled out minority neighborhoods. In the 1930s, the Federal Housing Administration and Home Owner’s Loan Corporation began publishing “Residential Security Maps” that were designed to downgrade the ratings of nonwhite neighborhoods. This reduced home values in those neighborhoods and made it difficult for nonwhite residents to secure home loans (Rae, Chapter 8, “Race, Place, and the Emergence of Spatial Hierarchy”). Post WWII, federally-subsidized suburban development enabled white city residents to move to newly constructed suburban neighborhoods financed through the GI Bill. Private sector discrimination denied nonwhite residents that same opportunity, deepening the racial and economic divide between cities and suburbs. As cities became increasingly nonwhite in the latter half of the 20th century, there was opposition at all levels of government to investing in urban infrastructure and funding public transit. To learn more about this history, I recommend an article from StreetsBlogUSA, “How Racial Discrimination Shaped Atlanta’s Transportation Mess.”

Clearly, there is evidence to suggest deep economic and racial disparities between different kinds of commuters. I’m interested in understanding how those disparities look across cities. To do that, I compared income, poverty rates, homeownership rates, race, and citizenship status.

To limit the scope of this analysis to large cities, I chose the 100 US cities that had the largest total number of car and public transit commuters, as measured by the 2017 American Community Survey. The following five graphs show different demographic variables, for each city, for car commuters and public transit commuters. Each set of boxplots allows the reader to filter the values by city or state.

For each of the five variables I looked at, the demographic difference between car and public transit commuters was statistically significant and conformed to the expectation that public transit users are relatively economically disadvantaged and more likely to be African American or a citizen of another country.

However, the above box plots and maps also demonstrate the substantial variability and diversity among large US cities. There are many cities on the list where the disparity between car commuters and transit commuters is quite large. In other cities, particularly those in the Northeast and West Coast, the size of the disparity between car commuters and public transit commuters is very small. There are also a handful of cities that do not conform to expectations. In these cities, car commuters actually had lower earnings, higher poverty rates, etc. than public transit commuters.

There are several possible explanations for why this is, and the one I want to emphasize is that the quality of public transit varies substantially across the country. Historical development patterns as well as contemporary practices have how accessible and attractive public transit is to commuters in different cities. In much of this country, public transit is a mess. Many cities on my top 100 list still don’t have light rail in 2019. (If you want to see what decades of disinvestment and political resistance do to mass transit in a city, read about it in the Tampa Bay Times.)

In contrast to the vast majority of American cities, a few have adequate or even preferable public transit. For example, Boston became the first city in the US to have an operational streetcar, as reported by the Boston Globe, in 1897. And today, it joins San Francisco, New York, and Washington, as home to one of the best public transit systems in the country. In those cities and others, taking public transportation every day sometimes means being able to afford living in more expensive downtown neighborhoods that are transit-accessible. In a housing boom that prices residents out of cities and into their surrounding metro areas, daily public transit use is sometimes a privilege. However, looking at large cities across the country, those select few cities are the exception, not the rule.


Data sources: United States Census Bureau

Programs used: MS Excel, Tableau

The Racial Income Gap in American Metro Areas

Just how bad are the current racial poverty and income disparities in cities? The map and scatterplot below show the gap between African American and white poverty rates in 104 metro areas in the US.

If the poverty rates among African American and white residents were equal, that trendline would be at 45 degrees. These metro areas tend to have significant racial disparities among their lowest-income residents. Most of the metro areas with the largest disparities in this map are in the north. The most extreme case of this is Boise, ID Metro Area, where 12% of white residents live in poverty, compared to 61.3% of African American residents.

In addition to poverty rates, income can also help illustrate how black and white residents fare in and around large cities. In this country’s largest metro areas, the disparities are significant and ubiquitous. The following chart shows the average distribution of earnings among black and white residents in the largest US metropolitan areas. The disparity in earnings is consistent and statistically significant; however, the extent of the it varies significantly by metropolitan area. The menu and sliding scale above the chart shows the distribution of White and African American earnings in each of the country’s 104 largest metro areas:

These 104 metropolitan areas are ordered by population size as of 2017. Most of these metro areas follow a similar pattern. In general, a higher percentage of African American residents occupy the lowest two income brackets. For virtually all metro areas in this dataset, the largest racial gap exists among top earners, with white residents typically twice as likely to earn $100,000+.

These extreme gaps in earnings among large-metro-area residents are symptomatic of even more entrenched racial inequalities. In 2015, an analysis by the Federal Reserve Bank of Boston revealed that, while the average net worth of white households in the Boston metropolitan statistical area was $247,500, the median net worth of black households of Greater Boston was only $8. African-American residents were significantly less likely to own homes, cars, and other assets than white residents, and much more likely to be in debt. Last year, the National Bureau of Economic Research released a report suggesting that even after controlling for education, neighborhood effects, and family characteristics, African American residents continued to trail white residents on income mobility. The statistics I’ve presented on the poverty and income gaps are troubling, but they are also only the tip of the iceberg when looking at racial economic inequality in metropolitan areas.


Data sources: United States Census Bureau

Programs used: MS Excel, Google Fusion Tables, Google Sheets, Tableau

Population Size, Urban Density, and Party ID: How Liberal is Massachusetts?

According to a Pew study, in 2014 Massachusetts was ranked the second most liberal US state, with 56% of adults identifying with or leaning towards the Democratic Party. Even though the majority (54%) of the Commonwealth’s voters in are unaffiliated with any party, registered Democrats still outnumber Republicans 3:1, and Hillary Clinton received twice as many votes as Donald Trump did in 2016.

What I’m trying to understand is whether and how this widely-recognized liberal identity varies across different cities and towns. There is good reason to think that geography and political affiliation are not independent from one another; in fact, there’s an extensive literature on the relationship between the two. Democrats cluster in cities, and they do it so much so that, as The New York Times put it, “Republicans don’t even try to win cities any more.” Books like The Politics of Resentment by Katherine Cramer document a deeply ingrained political, cultural, and economic divide between rural and urban areas.

It’s pretty clear that Massachusetts has some degree of an urban-rural political divide. Somerville — my hometown, the densest city in Massachusetts and the 16th densest city in the country — is a widely hailed as a liberal haven. Somerville is part of MA’s 7th congressional district, which recently held an election that WBUR identified as a “battle between progressives and Democrats.” MA-7 is so thoroughly blue that whoever won the primary was virtually guaranteed to also win the House seat. However, west of Somerville, Massachusetts does have a handful of towns that vote Republican, as the Boston Globe observed in late 2016. And perhaps the clearest signal of a geographic political divide is how the 2016 Democratic primary turned out. Bernie Sanders outvoted Hillary Clinton in rural counties across the country, and MA was no exception. As the Boston Business Journal reported, “the results [in Massachusetts] generally fell along urban versus rural lines.”

In order to illustrate the urban-rural divide in Massachusetts, I made an interactive map of statewide voter enrollment data, shown below:

Those who are familiar with maps of Massachusetts will notice that this one looks pretty similar to the state’s population density map.

Using the following two scatterplots, I show this association between the party ID of voters and the population size of those voters’ municipalities. This dataset contains all cities and towns in Massachusetts, omitting statistical outliers:

A few things jump out about these graphs. First, yes, the cities and towns tend to have a much higher percentage of Democrats than they do Republicans. Second, the data points in the Democrat plot are upward sloping. A higher percentage of Democrats is associated with a larger population size. To be clear, causation can’t be inferred from these data. Others have tried to link party ID to population size using much larger datasets, and it’s hard to determine which way the causation runs.

Although population size can be a useful way to categorize rural areas, exurbs, suburbs, and urban areas, it is not the only relevant metric. Population density, too, is an important tool used to distinguish urban areas from rural ones. As The Atlantic’s City Lab summarized in an article in 2012, “politics are inseparable from density.” Let’s take a look:

In the first graph, the percentage of Democrats in MA cities and towns is positively correlated with population density. In the second graph, the data are essentially flat and the correlation coefficient is near zero, indicating that there is almost no association between the percentage of Republicans and the population density.

The first scatterplot is also hereroskedastic, meaning that, when a trend line is fitted, the size of the error is correlated with population density, the x-variable. This simply means that there’s more variation at the left end of the distribution than there is on the right. Areas of medium-to-high population density follow the pattern better than areas of low density. As with the Democrat scatterplot, the Republican scatterplot suggests that low-density cities and towns have more variation in party ID than high-density ones.

In summary, the differences between these two graphs indicate that Massachusetts municipalities’ party enrollment data is highly related to how urban or rural each municipality is. And, in general, cities and towns in Massachusetts conform to expectations that denser areas are more liberal and less dense ones are more conservative.

Of course, I wouldn’t be me if I didn’t qualify these statements with their potential limitations.

City or town-level population density assumes uniformity of composition within each municipality, which isn’t always the case. My measure of density is the number of people per square mile in each municipality. Although widely-used, this is an imperfect measure of how people actually live, since the composition of cities and towns varies. For example, a municipality with a dense town center surrounded by low-density agricultural land would have a much lower overall density rating than an otherwise similar town without the surrounding open space. As there is no clear way to avoid this measurement issue, assessments of any individual city or town’s population density should be viewed as an approximation rather than an exact measure of how urban or rural the municipality is.

So, how liberal is Massachusetts? I think the simple answer is, pretty liberal, but it depends on where you live. My hope is that, when trying to understand the geography of political attitudes, we can more away from the “red state, blue state” narrative. State-level statistics matter enormously in studies of political representation in the Senate or Electoral College. However, when trying to understand the geography of political attitudes, maybe the more important question isn’t which state you live in, but rather, what does the building with your closest polling place look like?


Data sources: Mass GIS, Secretary of the Commonwealth of Massachusetts Voter Registration Statistics, United States Census Bureau

Programs used: MS Excel, Google Fusion Tables, Tableau