Researchers are increasingly interested in better understanding the determinants and effects of social connectedness (Jackson 2014). However, systematic analyses of social connectedness and social networks have traditionally been complicated by the absence of high-quality and large-scale data on which individuals are connected to each other. More recently, the emergence of data from online social networking services such as Facebook, LinkedIn, and Twitter has helped researchers overcome this measurement challenge, and has expanded our understanding of the determinants and effects of social networks across a large number of settings.
A new ‘Social Connectedness Index’
In recent work along these lines, we introduce a new measure of social connectedness between US county-pairs, as well as between US counties and foreign countries (Bailey et al. 2017a). This measure, which can be shared with other academic researchers, is called the Social Connectedness Index (SCI), and is based on anonymised data on the number of friendship links on Facebook, the world’s largest online social networking service. Given Facebook’s scale (over 2 billion active users globally and 236 million active users in the US and Canada) and the relative representativeness of Facebook’s user body, these data provide the first comprehensive measure of friendship networks at a national level.
We first use these data to better understand the geographic structure of social networks. The probability of friendship links is strongly declining in geographic distance, with the elasticity of the number of friendship links to geographic distance ranging from about −2.0 over distances less than 200 miles, to about −1.2 for distances larger than 200 miles. For the population-weighted average county, 55.4% of all friendship links are to individuals living within 50 miles, and 70.3% of friendship links are to individuals living within 200 miles. Conditional on distance, we find social networks to be substantially stronger within US states than they are across state lines.
In addition to distance and state lines, social networks within the US are substantially shaped by historical events. Consider Figure 1, which shows heat maps of the geographic distributions of the social networks of Kern County, CA (Panel A), and Cook County, IL (Panel B). Kern County has strong social connectedness to Oklahoma and Arkansas (in addition to strong links throughout the state of California). These connections are likely related to past migration patterns: Kern County was a major destination for migrants fleeing the Dust Bowl in the 1930s, and half of the residents of the San Joaquin Valley in Kern County have ancestors who migrated from the affected regions. Similarly, Cook County, IL, the home of the city of Chicago, has strong connections to the South (in addition to connections throughout the Midwest). This pattern is likely explained by the ‘Great Migration’ of southern African Americans to Northern and Midwestern cities throughout the earlier and middle parts of the 20th century.
Figure 1 Heat map of social connectedness
Panel A: Kern County, CA (Bakersfield)
Panel B: Cook County, IL (Chicago)
While, on average, US friendship networks are quite local, there is substantial heterogeneity across counties in the number of friends that live close by. The 5th-95th percentile range of the share of friends living with 50 miles is 38.1% to 70.3%. Some of these differences are driven by differences in population density across the US. To construct measures of the geographic concentration of social networks that are less affected by population density, we calculate, for each county, the share of friends that live within the nearest 50 million people. Figure 2 shows the distribution of this measure across the US. Social networks are most concentrated in the South, the Midwest, and Appalachia.
Figure 2 Share of friends among nearest 50 million people
Social networks and socioeconomic outcomes
The geographic dispersion of social networks is correlated with a large number of socioeconomic outcome variables at the county level. The populations of counties with more geographically dispersed social networks are generally richer, better educated, and have higher life expectancy. Interestingly, they also have higher social mobility as measured by Chetty and Hendren (2015), suggesting that social networks might play an important role in facilitating social mobility. We hope that the availability of the SCI data will encourage more research into the causal determinants of these observed empirical relationships.
We also document that counties that are more socially connected see more bilateral economic and social activity. For example, we show that region-pairs that are more socially connected have larger trade flows, even after controlling flexibly for the geographic distance between counties. This suggests that social networks might help overcome some of the informational and cultural frictions that can inhibit trade. We also find that when county-pairs have higher social connectedness, they are likely to have more cross-county patent citations. These results point to an important role of social interactions in the process of innovation, providing empirical evidence for a class of theories of economic growth that have focused on knowledge spillovers. Finally, we find that more connected region-pairs see more bilateral migration and job flows, highlighting the potential of social networks to overcome frictions to moving across the US.
International social networks
We also explore the international dimension of social networks. We first document that past international migration patterns are important determinants of present-day social connectedness, but with elasticities that are declining in the time since the peak of the respective primary migration wave. For example, Figure 3 shows the number of friendship links between each US county and Norway. The friendship links to Norway are strongest in the Upper Midwest (particularly in North Dakota and Minnesota), a region that saw significant immigration from Norway in the second half of the 19th century and the early 20th century. We also show that, just as we observed in the within-US data, the degree of social connectedness to foreign countries is correlated with higher trade with these counties.
Figure 3 Social connectedness abroad: Norway
Earlier research using Facebook data
This research, which aggregates Facebook friendship links to geographic a level that enables access to SCI data by the broader research community, complements earlier research that highlighted how data from online social networking services such as Facebook can help researchers better understand the causal effects of social interactions on economic decision making (Bailey et al. 2016).
In that paper, we first documented significant systematic heterogeneity in the structures of social networks at the individual level. We then analysed the role of social interactions in the housing market. We showed that the house price experiences within people’s social networks have large effects on their housing market investments, including their decision whether to buy a house and how much to pay for a given house. These effects are due to the role of social interactions in influencing people’s perceptions about the attractiveness of buying local real estate. In related work, Bailey et al. (2017b) used a similar setting to explore the role of house price beliefs in driving individuals’ mortgage leverage choice.
Bailey, M, R Cao, T Kuchler, and J Stroebel (2016). “The Economic Effects of Social Networks: Evidence from the Housing Market”, Journal of Political Economy, forthcoming.
Bailey, M, R Cao, T Kuchler, J Stroebel, and A Wong (2017a). “Measuring Social Connectedness”.
Bailey, M, E Davila, T Kuchler, and J Stroebel (2017b). “House Price Beliefs and Mortgage Leverage Choice”.
Chetty, R, and N Hendren (2015). “The impacts of neighborhoods on intergenerational mobility: Childhood exposure effects and county-level estimates”.
Jackson, M O (2014), “Networks in the understanding of economic behaviors”, The Journal of Economic Perspectives 28(4): 3–22.