Article Correctness Is Author's Responsibility: How Socially Integrated Is Your City? Ask Twitter.

A central feature of contemporary life is the geographic sorting and segregation of people across class, racial, and other lines. In his book The Big Sort, Bill Bishop elucidated how Americans increasingly sort ourselves into different places according to income, education, political ideology, and cultural beliefs. But at the same time, there are aspects of cities—density and transit, trust and social capital—that help push us together and form connections.

A new study by a team of researchers at Harvard and Northeastern universities, written up in an article in Sociological Methods & Research, uses Twitter to examine social connectedness among the neighborhoods of a city. The researchers, who include the eminent urban sociologists Robert Sampson and Mario Small, define the “structural connectedness” of a city as the extent to which residents in each neighborhood travel to other neighborhoods. It is essentially a measure of how socially integrated a city is—the opposite of segregation.

The extent to which city residents are structurally connected to each other is important to urban life for three key reasons, the authors argue. First, the more that people move between neighborhoods, the more potential there is for them to form social ties and generate meaningful connections. Two, more movement across places means more potential for the diffusion of social and cultural attitudes, or tastes in fashion, music, and culture. And, three, higher intra-urban mobility and deeper connections between a city’s places accelerate the movement of ideas and information.

The study treats the neighborhoods of the city as a social network, and uses data from Twitter to assess people’s day-to-day travel patterns to measure the connectedness across this network. The base data was some 650 million geotagged tweets, sent by more than 1 million Twitter users over 18 months, from October 2013 through March 2015. Grouping tweets by census block group for the nation’s 50 largest cities, the research team ended up with a final dataset of more than 130 million geotagged tweets, sent by more than 375,000 people.

The team developed two distinct measures of social connectedness: One, the primary index, gauges the degree to which people move between neighborhoods in roughly similar proportions (the “equitable mobility index,” or EMI). A secondary index looks at the extent to which visits are concentrated in a handful of places (the “concentrated mobility index,” or CMI). The correlation between the measures is quite low (-0.033), indicating that they capture distinct elements of city life.

The researchers identify four types of connectedness, each associated with a different city: New York, San Francisco, Detroit, and Miami. On the graph below, which compares America’s 50 largest cities on the two social indices, you can see that all four of these cities are outliers.

The relationship between cities’ equitable and concentrated mobility indices. (Nolan E. Phillips, Brian L. Levy, Robert J. Sampson, Mario L. Small, and Ryan Q. Wang/Sociological Methods & Research)

San Francisco, with a high score on both indices, appears in the upper right-hand corner. New York occupies the upper left-hand corner, with a high score on the CMI but a low score on the EMI. Miami sits in the lower right-hand corner—the opposite of New York, with a high score on the EMI and a low CMI score. And Detroit falls in the lower right-hand corner, with low scores on both measures.

In highly connected San Francisco, a small number of neighborhoods draw residents together into an urban commons of sorts, oriented around locations like the Financial District and Telegraph Hill. At the same time, many of the city’s neighborhoods are quite closely linked in everyday mobility patterns. Austin, Seattle, and Washington, D.C., show a similar dual pattern—areas that are major social draws, and evenness in daily mobility—though to a lesser degree.

By contrast, New York is a city where connections are heavily concentrated in a small number of neighborhoods (reflected in its high CMI score) but there is much less equitable exposure to others across the city (reflected in its low score on the EMI). Overall visits in New York are centered in Manhattan and especially around the transportation hub of Penn Station, but many other neighborhoods have limited connectedness. Los Angeles, Chicago, Dallas, and Houston also follow this pattern to some extent.

Miami lacks a dominant hub or hubs for social connections, as shown by its low CMI score. However, it does have robust social connections spread widely across the city (high EMI). Miami is the most extreme outlier on the graph, with really no other cities following a similar pattern.

Detroit is more disconnected. Like Miami, it lacks a dominant social hub or hubs. But it also suffers from low levels of connectivity across the city, with few of its neighborhoods connected to one another in meaningful ways. A number of other cities—Cleveland, Baltimore, Philadelphia, and Milwaukee—also follow this pattern, though to a lesser degree.

The study then turns to the features of cities that contribute to their social connectedness. To get at this, the researchers used statistical models to assess factors like size (based on population), land area (a measure of distance to travel), public transit, what they call “cosmopolitanism” (based on the share of college graduates), and segregation by income and race. The authors make sure to point out that their models and findings are not causal, but simply reflect associations among variables.

It turns out that size is the single strongest variable bearing on the social connectedness of cities, accounting for roughly half the explanatory power of the researchers’ models. But it works in reverse (with coefficients that are negative and significant). In other words, larger cities have lower levels of social connectedness, and smaller cities have higher connectivity. According to the authors, this reflects the basic fact that it is harder to visit more neighborhoods in larger cities.

After controlling for size, a number of other factors also come into play, although they are less salient. Social connectedness is higher in more highly educated or “cosmopolitan” cities. Conversely, it is lower in more segregated cities. Surprisingly, perhaps, public transit contributes little to the evenness of social connections. This could be because in most cities, transit lines connect only certain, limited areas.

America’s cities have wide variations in social connectedness. Finding ways to make them more connected—and less sorted and divided—is the challenge.