10.29.2010
CR Report
The Role of Social Effects in Spreading Subprime Lending
New research from the Federal Reserve Bank of Cleveland investigates the extent to which social connections and context in poor neighborhoods determined the likelihood of high rates of subprime loans.1
This report is based on a working paper by Ben Craig, Senior Economic Advisor in the Research Department and Francisca Richter, Research Economist in the Community Development Department
For much of the 20th century, the practice of redlining choked off credit to people living in poor neighborhoods. By the beginning of the 21st century, that problem had been turned on its head: Where before there was virtually no lending in low-income communities, suddenly there was virtually nothing but subprime lending. Concerns about access to credit were replaced with concerns about the cost and quality of credit.
Federal Reserve Bank of Cleveland economist Francisca Richter has been studying regional housing finance trends for several years. Her focus of late has been on trying to explain the extraordinarily high foreclosure rates seen in Cleveland. While it’s no surprise that high rates of subprime lending have led to high rates of foreclosure, it seems surprising that the problem has been worse in certain Cleveland neighborhoods than demographically similar neighborhoods elsewhere.
Most likely several factors contributed to Cleveland’s elevated levels of subprime lending. Some may have had to do with mortgage regulations and their enforcement. In Ohio, regulations that provided for the supervision of mortgage brokers—who tend to be the largest purveyors of subprime credit—were possibly less strict.
Richter was curious about another possibility: that people’s social context and networks were influencing their mortgage decisions. Perhaps signs posted throughout neighborhoods advertising easy mortgage credit made neighbors less leery of subprime loans; maybe they heard about one family up the street that used a home equity loan as a windfall to pay off other debts. Over time, people might spread subprime lending like a cold.
What if there was something about neighborhoods of concentrated poverty that made social effects even stronger? Added to the “who you are” demographic explanation of why some people were more likely to fall prey to subprime loans—income, race, and education level are among the characteristics associated with higher levels of subprime lending—this would be the “who you know” story.
The anecdotes to support this version of events are plentiful. In a recent paper, Federal Reserve Bank of San Francisco researcher Carolina Reid used interviews with dozens of low-income borrowers to understand how social networks shape loan outcomes. “Over and over again, borrowers indicated they felt pressure from their friends, family, brokers, and the media to purchase a house ‘now,’” Reid wrote.
Richter wanted to take a more analytical approach. “If these anecdotes were just being repeated over and over but the actual facts were not significant, then probably the data would not be able to capture this effect,” she said. But if the facts were significant, then the data would validate Richter’s hypothesis.
Why so much subprime?
The pervasiveness of subprime lending in poor neighborhoods is well-established. Study after study shows that borrowers in low-income neighborhoods are more likely to take out high-cost mortgages to finance their homes. Often lacking traditional banking services, these borrowers turn to independent mortgage brokers, who more often than not are aligned with non-bank lenders that specialize in subprime credit.2
Richter’s earlier research found that the poorest quartile of borrowers in Cuyahoga County (home to Cleveland) were about 36 percent more likely to take out subprime loans than their counterparts in Allegheny County (home to Pittsburgh). Even accounting for similar incomes, credit scores, and education levels, the divergent foreclosure rates in Cleveland and Pittsburgh defied easy explanation.
Richter also noticed that the foreclosure crisis was heavily concentrated in certain Cleveland neighborhoods. In 2006, it wasn’t uncommon for more than half of all mortgages in Cleveland’s poorest neighborhoods to be of the subprime variety.3 A map showing the prevalence of subprime lending across Cleveland neighborhoods tells a story in itself. Most of the activity was near the city’s impoverished core. Extreme rates of subprime lending show up in a semicircle that loops around downtown, enveloping some of the city’s lowest-income neighborhoods.
Additional analysis by Federal Reserve colleague Lisa Nelson observed that mortgages tended to be issued by only a handful of lenders, contrasting with the much more diversified and competitive environment in Pittsburgh.4
Could some of this elevated subprime activity be attributable to social effects? In the realm of subprime lending, social effects might manifest themselves in two ways. First, simply by living among poor neighbors, a borrower is more likely to be exposed to the same marketing efforts as the neighbors—they see the same signs on lawns, get the same flyers stuffed in their mailboxes, walk by the same billboard each day, and so forth. This is what we term the social context. Second, by interacting with their neighbors, nearby friends, and family (their networks), prospective borrowers learn more about product availability and recommendations.
The second social effect is of crucial interest to policymakers because it also has a multiplier effect—friends and neighbors and family pass on information, either directly or indirectly reducing risk aversion to subprime products. This reduced risk aversion is reinforced as information flows back and forth between prospective borrowers, eventually leading to higher demand. An analogy is in the realm of education—students who get tutoring will not only improve their own grades, but may motivate their peers to do the same, which in turn further motivates the tutored students.5 So the impact of tutoring just one student may in fact extend to many more through the multiplied motivation it generates. Likewise, the impact of a neighbor’s recommendation to a mortgage broker might reach others and increase this type of lending.
Quantifying the impact of social effects
Richter’s hypothesis was that social effects leading to subprime lending are stronger in neighborhoods where poverty is acute compared with less-poor neighborhoods. But how to isolate the social effect? Geography alone was not enough.
Even if one were to hypothetically do away with any likely social effects, similar patterns of subprime lending in close-by neighborhoods would still be possible due to similar people living in geographic proximity of each other (similar in education, income, credit scores, etc.).
Modeling subprime lending rates in, say, Neighborhood A, Richter and co-author Ben Craig controlled for neighborhood variables such as race, education level, credit score, and income not only in A, but in its adjacent neighborhoods as well. Then they used geographic proximity as a proxy for social networks. The idea is that after eliminating the effects of these characteristics in Neighborhood A and close-by neighborhoods, what is left over in the differences in subprime lending between nearby communities is likely attached to social effects.
The authors looked for differences in subprime lending by non-depository institutions6 between poor and less-poor neighborhoods (census tracts) in Cleveland between 2004 and 2006, when subprime lending was peaking. They found a clear difference.
“The rate of non-depository subprime lending taking place in the poorer tracts (those with more than a 20 percent poverty rate) is significantly higher than that in less poor tracts,” the authors reported in a May 2010 working paper, “Lending Patterns in Poor Neighborhoods.” The increase attributable to social interactions is small but statistically significant.
What’s not clear is how much of that effect can be traced to the demand side of social networks—that is, from people altering other people’s behavior through word of mouth or other interactions. Borrowers may have grown less concerned about the risk of taking out a subprime loan upon seeing their neighbors enjoy the initial fruits of refinancing, for example. But some of the effect is most likely associated with the supply side of social networks in neighborhoods with concentrated poverty—the amount of advertising, for example, that people see in common by virtue of living in the same community.
“Both of these effects speak to the impact of living in areas with concentrated poverty,” Richter says. “Your networks are restricted in a way, but in the same way your exposure to financial products is also restricted.”
To better capture the spatial dimensions of subprime lending, the researchers also looked at refinancing and home improvement loans, since purchases of new homes would likely encompass borrowers moving from outside the neighborhoods under study. The effect was still evident but not as strong as with the full model.
Street-level view
To be sure, the factors that drive people to take out subprime loans vary and certainly go beyond social networking effects. In some cases, borrowers may have been the victim of fraud or misinformation; other times, it was simple unawareness of the terms and conditions they were agreeing to.
Yet Richter’s and Craig’s results generally square with what community development workers on the ground have learned about the decision-making behind subprime borrowing. Social ties do have an impact. Frances Cintron, a counselor with Neighborhood Housing Services of Greater Cleveland, described it this way: “A friend would say, ‘Guess what, I bought a house, and I didn’t have to put a penny in it! If I could do it, with my credit score, you could do it too.’ It’s mostly [word of] mouth—your family, your cousin, your uncle...go buy a house, I did it, why not you?”
Or as Anita Brindza, executive director with the nonprofit Cudell Improvement Inc., puts it: “When someone tells you they got a loan and their payment is just ‘x,’ that has a very powerful pull to it.” Brindza noted a one-block stretch of West 89th Street in Cleveland as a case in point. Eight homes on the same block were recently in the foreclosure process. Is all of that explainable by demographics?
The ease with which borrowers could take out subprime loans may also have been a factor, community development officials say. Brokers papered neighborhoods with signs and flyers, and once connected to a borrower, they could provide all the next steps—a deal with a lender, an appraiser, and so on. This might be seen as a part of the social networking effect, in so much as the geographic concentration of marketing by subprime brokers was such that it made subprime borrowing a major context of living in poor neighborhoods. (However, this kind of social network effect lacks the multiplier effect that happens when friends talk to friends. From a policy impact perspective, this kind of an effect may call for different responses compared to the one involving the multiplier.)
The flip side is the opportunity for policymakers and community groups to influence social networks. “Homebuyer awareness training is the other side of this,” says Matthew Lasko, director of housing with the Detroit Shoreway Community Development Organization in Cleveland. “If people hear us, they learn about good loans versus bad loans.” Indeed, one of the main policy implications of this research is the potential importance of effective financial education efforts in mitigating the socially driven component of subprime lending. There may also be an expanded need for consumer protection policies that apply specifically to lending by mortgage brokers. Finally, the low reliance on traditional banking institutions in poor neighborhoods needs to be addressed—if low-income borrowers perceive they have nothing but payday lenders and pawn shops in their communities by way of financial services, where else are they to turn?
Next steps
The data that Richter and Craig analyzed provides a bird’s-eye perspective. It does not connect borrowers with loans and as such reveals only differences in subprime lending patterns between census tracts, not within them. Next up is an analysis of this loan-level data, which should provide a more precise estimate of the social network effect.
Poverty’s double burden is an economic fact. Not only are the poor limited in their ability to escape poverty because of their financial constraints, but also because of the places they live. In a time when most people liken “social networking” to Facebook, it’s worth remembering that a lot of social interactions still take place the old-fashioned way—face-to-face, neighbor-to-neighbor. Richter’s and Craig’s research provides a useful reminder about the myriad sources of the housing crisis and why one-size-fits all policy responses are likely to fall short.
Notes
- F. G.-C. Richter and B. Craig. Lending Patterns in Poor Neighborhoods. Federal Reserve Bank of Cleveland Working Paper 10-06 (2010).[Return]
- P. Calem, J. Hersha, and S. Wachter. Neighborhood patterns of subprime lending: Evidence from disparate cities. Housing Policy Debate,15(3):603{622, 2004.[Return]
- Richter and Craig, p.15, figure 3.[Return]
- D. Campbell, L. Nelson and F. G.-C. Richter. “Foreclosure Differences Across State Lines,” Federal Reserve Bank of Cleveland CR Report, April 2009.[Return]
- J. Cooley, “Can achievement peer effect estimates inform policy? a view from inside the black box.” Technical report, Department of Economics, University of Wisconsin–Madison Working Paper, 2009.[Return]
- These are loans issued by an independent mortgage company or a subsidiary of a bank, and likely facilitated by a mortgage broker.[Return]