Social Interactions and Social Capital
Proceedings
Federal Reserve Bank of Cleveland Hosts Seminar on Social Interactions and Social Capital
While social interactions has been a main focus of sociological research, as recently as the late 1960s, conventional economics followed a more individualistic approach to analyzing how people make decisions. Homo economicus operated within his own bubble, maximizing his individual utility within given constraints. Whether choosing how hard to work or what to consume, decisions were always based on what delivered the largest personal return for given market prices and budgets.
Today, advanced modeling and estimation methods allow economics to join other social sciences in a more rigorous account for nonmarket interactions. Now analysts can recognize the fairly intuitive notion that people’s behaviors are also strongly influenced by their environments.
Are public housing projects rife with unemployment because that’s the sort of people such neighborhoods attract? Or is it because social groups themselves foster certain types of behavior? Models of social interaction and social capital are beginning to address these questions, though many theoretical and empirical challenges endure. All the same, with models of social interaction modern-day researchers can better explain phenomena that traditional theories have failed to capture.
On Friday, November 21, the Community Affairs Department of the Federal Reserve Bank of Cleveland convened a seminar on Social Interactions and Social Capital. Research in social interactions is shedding light on a growing range of development issues – from the value of mixed-income housing developments to the way individuals decide whether to declare bankruptcy. The social scientists who gathered at the Cleveland Fed represent the cutting edge of social interactions research.
In “Social Capital: The Challenge for Research,” University of Wisconsin at Madison economist Steven Durlauf explained the basics of social interactions theory with an eye toward the implications for public policy. “We can’t conceptualize aggregate behaviors in a community as simply the adding up of individually determined outcomes, but as a concatenation of people’s behaviors,” Durlauf said. That is to say, the characteristics and behaviors exhibited in certain groups affect each individual. This in turn leads to social multipliers – where the impact of some characteristic as seen in aggregate level behavior can be much larger than the impact of the same characteristics at the individual level. (If a policy improves one person’s life – such as, in encouraging them to eat healthier – the effect of social interactions suggests that others in the now “healthy-eating” person’s community will also benefit by being influenced to eat healthier themselves.)
In addition, Durlauf said, it is important to think about economic outcomes in terms of “multiple equilibria” – when we observe different outcomes from identical sets of fundamentals. For example, the observation that African-American women – who otherwise share many socioeconomic characteristics of smokers -- tend to smoke at lower rates than other groups suggests that behaviors become reinforcing within groups.
The presence of powerful social influences combined with incentives can stratify entire groups along different characteristics. The result is economic segregation and phenomena like “poverty traps,” in which individuals find it difficult to rise in economic status because of group influences and loyalties. These are behaviors that are not directly market-adjudicated, and as such they represent a domain of activity that falls outside the traditional economic framework, Durlauf said.
Policy prescriptions that fail to account for social interactions may thus end up off the mark. If promoting college attendance is the goal, then it would be important to not only know the prospective students’ socioeconomic background, but also the college attendance rates of his peers – because those peers may be as important an influence as a scholarship. Measurement issues then arise – how do we properly measure social networks?
“The perspective here is not to diminish the importance of thinking about individual incentives or individual decision-making, but in thinking of incentives moving at different levels of aggregation, and melding those together in a common framework,” Durlauf said. “This should make better sociology and better economics.”
Durlauf categorized the notion of “social capital” as a special case of the social interactions model. Most often, social capital is defined as “trust” or as equated with the presence of social networks. But in either case, econometric analysis is tricky with social capital – it is difficult to argue that a group-level variable is linked with a social capital effect as opposed to some other group-level effect. “Social capital is devilishly elusive to define in a way that’s useful in both theoretical and empirical analysis,” he said.
The role that one’s peers have in individual decision-making also appears important in personal finance. For years, analysts have noted that much fewer people declare bankruptcy than would be economically optimal. How to explain this discrepancy? Perhaps the stigma associated with bankruptcy has persuaded individuals who would otherwise benefit from it to do without protection from creditors.
In “Social Networks and Personal Bankruptcy,” Boston Fed economist Ethan Cohen-Cole examined the mechanisms in which social stigma might influence the bankruptcy choice. Access to information about the ease of filing for bankruptcy might be important, for example. Cohen-Cole aimed to understand whether social stigma or information sharing (learning from neighbors about the ease of filing) were more likely drivers of defaults.
Using credit bureau data, Cohen-Cole confirmed that whether others nearby have declared bankruptcy is a likely determinant of another household’s bankruptcy decision. He then found that while bankruptcies increased in the run-up to 2005’s reform (which imposed stricter terms), the stigma attached to filing actually dropped only among groups that don’t file many bankruptcies in the first place. This points to information costs as having greater effect on the bankruptcy decision than social stigma. “When you suffer a financial crisis, you’re more likely to talk to others,” Cohen-Cole said. “We can’t observe that here, but it’s clearly important in the simple cut we have.”
All the same, social interactions matter. Moreover, it’s clear that U.S. bankruptcy code in some ways relies on social interactions, because as written many more people would seem eligible for protection but still don’t choose it.
Money lending might seem like the most formalized of transactions. But social interactions research helps explain why banks are hardly the exclusive source of loans across the globe. In “Informal Lending and Social Networks,” Iowa State University economist Tanya Rosenblat tried to measure the extent to which social networks generate trust in microfinance lending by providing both information about a borrower’s credit worthiness and enforcement of their efforts to repay.
Rosenblat and her co-authors conducted an experiment in informal lending in a small Peruvian village. In 128 loans, 53 percent went to direct friends, while 26 percent were between friends of friends. The repayment rate was 88 percent. That’s a good outcome overall. The most telling split came between those when the sponsor was either on the hook for 100 percent of the loan, should it default, or just 50 percent. The economists found that, perhaps unsurprisingly, sponsors accountable for the full loan reported higher repayment rates than the lenders accountable for 50 percent. What does this tell us? “It’s that sponsors are doing both screening and enforcement,” Rosenblat said. “The sponsors lending to indirect friends do a lot of information collection, trying to find out if it’s a good risk.” Rosenblat suggested that this model of microfinance could be useful in the real world, as banks can delegate their customary role of information collection and enforcement to the sponsor. In this way, social networks can help lower the cost of capital.
In a discussion of the morning session, Boston Fed economist Mary Burke noted some of the counterintuitive results that social interactions research has produced. Though people tend to define social capital as welfare improving, Durlauf’s research points out that it is not always that way – much depends on the relevant level of economic development. Burke suggested that the role of stigma may still help explain the rise in bankruptcy if the underlying cause is a decline in face-to-face lending, given today’s impersonal and institutionalized lending market. On the flip side, Cohen-Cole cautioned that his studies do find evidence of stigma in “anonymous” lending. “I don’t think we can take lessons learned from one forum and immediately apply them to another,” Cohen-Cole said.
One questioner asked whether social interactions models are suspect because of tweaks that would allow a researcher to generate similar results. In other words, there may be equally useful alternative hypotheses capable of explaining the effects that social interaction models have claimed credit for.
The longstanding difference in educational achievement between races is a major proving ground for social interactions research. In “Desegregation and the Achievement Gap: Do Diverse Peers Help,” University of Wisconsin at Madison economist Jane Cooley studied student accountability policies in North Carolina, where students must perform at a certain level to be promoted to the next grade, to tackle this puzzle. “A considerable amount of attention has been paid to how we might go about marrying the racial achievement gap,” Cooley said. “I’m going to ask the question – by planning racially diverse classrooms, can we help marry the racial achievement gap?”
In short, Cooley’s research found little evidence that desegregating peer groups narrows the achievement gap. The reason appears to be in part that students segregate themselves further within even desegregated classrooms, with whites associating with whites and blacks with blacks. In addition, the effects of desegregation may offset each other, with white and nonwhite students in lower-achieving, predominately nonwhite schools benefiting while students in higher-achieving, predominately white schools experiencing losses. However, desegregation can help raise achievement in low-achieving minorities who otherwise would be surrounded by “low-equilibrium” peers.
Of all the applications for social interactions and social capital theory, one of the most studied has been the potential payoff from mixing people of different socioeconomic backgrounds into single neighborhoods. The premise of mixed-income developments is based in part on the idea that it could help ease the problem of concentrated urban poverty. In “The Possibilities and Challenges of Mixed-Income Development: Emerging Findings from Chicago,” Mark Joseph, a public policy researcher with Case Western University, surveyed residents of several such developments to learn about their experiences. “It’s clear that a cause of persistent poverty is the enduring levels of segregation in our society,” Joseph said. “So if we don’t stop socially isolating the poor from the rest of society, we’ll continue to have this problem.”
The Chicago study was qualitative in nature. One-third of the development’s residents came from public housing, one-third pay at a subsidized rate, and the remainder pay market rate. For the lower-income residents in particular, quality of life was much improved. Increased quality of life, decreased stress, increased sense of possibility and motivation, according to the research, make mixed-income living an overall beneficial experience for former public housing residents who have chosen to live there.
But what stood out in the survey was how little interaction apparently occurs between members of different socioeconomic backgrounds; instead, each third tends to socialize only with members of their same group, if at all. Meanwhile, the highest-income residents exert the most control over how the development operates. “It’s not all kumbayah,” Joesph said. “Sustainability of these types of efforts, will likely require intentional action around governance and community building.”
A quantitative look at the impact of socioeconomic segregation came with University of Wisconsin at Madison political scientist Deven Carlson. In “Long-term Effects of Public Low-Income Housing Vouchers on Work, Earnings, and Neighborhood Quality,” Carlson focused on participants in the national Section 8 program, which provides housing vouchers to 2 million U.S. families. Among the notable findings, 60 percent of voucher recipients ended up moving after a year, compared with 45 percent of match-comparison cases. This suggests that vouchers stimulate mobility, and that the disruption of moving may be offset by the ability to move to areas where employment prospects are stronger.
Carlson used a technique known as “propensity scoring” that aims to reduce selection bias in experiments. Four years out, his study showed that voucher cases live in areas with higher median gross rents, with more children in school and lower poverty and unemployment rates. “Statistically, we do see that over four years there is an improvement,” Carlson said.
Ohio State University economist Bruce Weinberg led a discussion of the afternoon session. Weinberg wondered whether the “winners offset the experience of the losers” in Chicago’s mixed-income developments – is it a zero sum game? “We’re helping some guys clearly,” Weinberg said. “To what extent is this positive, something where we’ve helped lots of people and not hurt others?”
In introducing his own work, “Estimating Social Interactions: Selection Within and Across Groups,” Weinberg described the landmark social interactions study by sociologist William Julius Wilson, in which Wilson identified areas of high concentration of unemployment in Chicago. His work forced the question upon social scientists of all stripes: To what extent are different people selecting into different neighborhoods, or are neighborhoods themselves affecting how people behave? How do we control for self-selection?
Weinberg looked at reallocation of students across schools and found that white students are 25 percent more likely to have white friends than other students with the same characteristics, save for race. For black students, the rate is 57 percent. “That’s a lot of sorting happening with these individuals,” Weinberg said. People tend to associate with people who are like themselves.” If policymakers in fact want to encourage more mixing, they might do so in the context of smaller class or school sizes, since it is more difficult to segregate oneself in a school of 300 students compared with 3,000 students.
During the afternoon discussion period, Steven Durlauf emphasized the potential effectiveness of harnessing market forces in promoting social interaction. Mark Joseph noted the usual problems with markets – assumptions about perfect information and assumptions about transaction costs. In a market-based scenario. You could allow people to self-select into the living situation that works best for them – but do they fully understand what they are buying into, based on the available information? “Transactions-cost-wise, the hurdles that people have to go through can be extreme, in the screening and monitoring,” Joseph said. “It is far from a world where you can say, ‘Let’s hand it over to the market.’”