in pictures

1936 risk assessment map of Philadelphia, used by the FHA in lending decisions


A significant factor in historic housing segregation, the practice of redlining dominated lending practices from the 1930s until the passage of the 1974 Equal Credit Opportunity Act (Massey & Denton, 1998)⁠. The New Deal-era Federal Housing Authority (FHA) was required by law to classify neighborhoods according to financial risk, and to use these metrics as criteria in the processing of mortgage applications, resulting in a racially stratified redistribution of wealth. With the arrival of Levittowns and white flight, a pro-segregation culture began to manifest within FHA publications, which endorsed racial homogeneity at the neighborhood level (Loewen, 2006)⁠. The federal government’s efforts to increase home ownership rates has been described as “one of the greatest mass-based opportunities for wealth accumulation in American history,” but, to the extent that the federal government purposefully tried to maintain the property values of white neighborhoods at the expense of people of color, this opportunity was denied to people of color, and is thought to be a major component in the historic racial wealth gap (Oliver & Shapiro, 2006, p. 18)⁠.


Symbol to declare compliance with fairness in lending laws

While the racial wealth gap and housing segregation would continue to persist, this lending paradigm was shattered following the Civil Rights Era. As political momentum increased for the eradication of both de jure discrimination and the remaining New Deal and Great Society scale social spending programs, the federal government implemented the Equal Credit Opportunity Act of 1974, as well as the Community Reinvestment Act of 1977, which explicitly prohibited redlining and similarly overt racial discrimination in lending practices. While allegations of redlining practices would continue, a distinction between process-based discrimination (in which lending decisions are made according to overtly discriminatory criteria) and outcome-based discrimination (which focuses on the disparate impact caused by lending decisions) ultimately characterized the academic conversation (Ross & Yinger, 2003)⁠. While Ross & Yinger (2003) used these terms to describe the mortgage market, the same concepts could easily be applied to the explosion of consumer credit which would come in the following decade.

First National of Nebraska building; defendant in National Bank of Minneapolis v. First of Omaha Service Corp. (1978). Photo by JonClee86, CC BY-SA 3.0

The following year after the passage of the Community Reinvestment Act, Marquette National Bank of Minneapolis v. First of Omaha Service Corp. (1978) reached the Supreme Court. The plaintiff, Marquette National Bank, alleged that First of Omaha Service was violating state usury laws by offering consumer credit to customers across the country, regulated only by Nebraska’s liberal usury laws. In a unanimous decision, the Supreme Court upheld this practice as legal, effectively legalizing usury nationwide, and “allowing lenders to price in the cost of loans to riskier borrowers” (Trumbull, 2012)⁠. This decision paved the way to a massive increase in consumer debt, as access to credit was, nominally, democratized – however, as the Federal Reserve Board was soliciting testimony on the implementation of the Equal Credit Opportunity Act, the infantile consumer reporting industry made it clear that credit access would be anything but democratic. In their pursuit of quantifying risk, bankers demanded the right to consider marital status, usage of birth control, or whether or not a newlywed took her husband’s surname (Williams, 2004)⁠. While these criteria didn’t all make it through the federal rulemaking process, shockingly, scoring by zip code did, allowing “banks to surreptitiously continue to score race, class, and national origin, using zip codes as their proxy” (Williams, 2004, p. 16)⁠.

Data source: Costly Credit: African Americans and Latinos in Debt, (Silvia & Epstein, 2005)

Data source: Costly Credit: African Americans and Latinos in Debt, (Silvia & Epstein, 2005)











As access to consumer credit demonstrably rose over the coming decades, the scope of the consumer reporting industry’s scoring project increased dramatically, which is now estimated to score consumers, on average, once every calendar week (Thomas, 2000)⁠. These scores are thought to be “far better predictors of outcomes than broad measures of educational attainment or racial classification,” and, yet, despite their individualized sense of objectivity, assertions of outcome-based discrimination persist (Fourcade & Healy, 2013, p. 570)⁠. Instead of overt classist or racist discrimination, it is argued, these credit scoring metrics serve as methods of classification, used by gatekeepers to guard access to material goods – a Foucaultian “dividing practice in which the ‘bad’ are separated from the ‘good’, the criminal from the law-abiding citizen, the mentally ill from the normal” (Burton, 2012, p. 114; Fourcade & Healy, 2013)⁠.


Check cashing stores, like this one in Atlantic City, serve underbanked customers. Photo by Chris Goldberg, CC BY-NC 2.0


Graffiti present on a Bank of America window. Photo by kozemchuk, CC BY 2.0

Yet, despite this rhetoric, academics still argue that the financial system has resulted in a discriminatory disparate impact. Failure to qualify for credit, it is asserted, pushes consumers into a relatively expensive credit market of last resort, where marginalized and underbanked customers are required to pay predatory fees for access to financial services (Carruthers & Kim, 2011; Fourcade & Healy, 2013)⁠. Similarly, access to mortgages, which is now largely governed by credit ratings, is described as a “dual mortgage market,” where individuals without access to credit are required to take on so-called subprime mortgages in order to purchase a home (Immergluck & Wiles, 1999)⁠. Not surprisingly, the dual mortgage market’s impact has been to exacerbate wealth loss in communities of color, via a mechanism not all that different from historic redlining (Rugh & Massey, 2010)⁠.

Man sleeping on public bench, Atlanta, GA. Photo by Adam Hermann.

Despite these developments, the individualized nature of credit scoring complicates collective action. Increasingly, financial responsibility is equated with personal morality, and financial irresponsibility is stigmatized (Graeber, 2011; Pathak, 2014)⁠. Furthermore, the consumer reporting industry has expanded to offer “thousands of ‘consumer scores,’” expanding both the scope of their surveillance, and the responsibilities of the typical consumer significantly, through the application of these credit scoring strategies in non-financialized domains (Pasquale, 2015, p. 33)⁠. While some political organization exists, such as Occupy Wall Street’s Strike Debt movement, scoring metrics and questionable lending practices persist, and increasingly capture scores of young people via the growing student debt bubble (Draut, 2006; Strike Debt, 2014; Williams, 2004)⁠.

Occupy Wall Street (2011). Uncredited.


Burton, D. (2012). Credit scoring, risk, and consumer lendingscapes in emerging markets. Environment and Planning A, 44, 111–124.

Carruthers, B. G., & Kim, J.-C. (2011). The Sociology of Finance. Annual Review of Sociology, 37(1), 239–259.

Draut, T. (2006). Strapped: Why America’s 20- and 30-Somethings Can’t Get Ahead. New York: Anchor Books.

Fourcade, M., & Healy, K. (2013). Classification situations: Life-chances in the neoliberal era. Accounting, Organizations and Society, 38(8), 559–572.

Graeber, D. (2011). Debt: The First 5,000 Years. Brooklyn: Melville House Publishing.

Immergluck, D., & Wiles, M. (1999). Two Steps Back: The Dual Mortgage Market, Predatory Lending and the Undoing of Community Development. Chicago: The Woodstock Institute.

Loewen, J. W. (2006). Sundown Towns: A Hidden Dimension of American Racism. New York: Simon & Schuster.

Massey, D. S., & Denton, N. A. (1998). American Apartheid: Segregation and the Making of the Underclass. Cambridge: Harvard University Press.

Oliver, M. L., & Shapiro, T. M. (2006). Black Wealth / White Wealth: A New Perspective on Racial Inequality (2nd ed.). New York: Routledge.

Pasquale, F. (2015). The Black Box Society. Cambridge: Harvard University Press.

Pathak, P. (2014). Ethopolitics and the financial citizen. The Sociological Review, 62(1), 90–116.

Ross, S. L., & Yinger, J. (2003). The Color of Credit: Mortgage Discrimination, Research Methodology, and Fair-Lending Enforcement. Cambridge: The MIT Press.

Rugh, J., & Massey, D. (2010). Racial Segregation and the American Foreclosure Crisis. American Sociological Review, 75(5), 629–651.

Silvia, J., & Epstein, R. (2005). Costly Credit: African Americans and Latinos in Debt (Borrowing to Make Ends Meet Briefing Papers No. 5). Baltimore: The Annie E. Casey Foundation. Retrieved from

Strike Debt. (2014). The Debt Resisters’ Operations Manual. Oakland: PM Press.

Thomas, L. C. (2000). A survey of credit and behavioural scoring: forecasting financial risk of lending to consumers. International Journal of Forecasting, 16(2), 149–172.

Trumbull, G. (2012). Credit Access and Social Welfare: The Rise of Consumer Lending in the United States and France. Politics & Society, 40(1), 9–34.

Williams, B. (2004). Debt for Sale. Philadelphia: University of Pennsylvania Press.



I think, for my small scale social network analysis paper, I will be using my dataset from the StudentLoans subreddit (link this). In their content analysis of a peer support forum for consumer debtors, Stanley, Deville & Montgomerie (2016)⁠ argue that sharing information about their debts, within the context of a culture that increasingly associates financial wellbeing with morality, constitutes a source of resistance to creditors. Within their content analysis, Stanley, Deville & Montgomerie (2016)⁠ identified two primary types of posts in these peer support forums: troubleshooting posts, where individuals were looking for technical assistance with their debts, and journeying posts, where debtors seek emotional support. I’ve selected the StudentLoans subreddit because, based off my observations of the various subreddits related to student debt, I anticipate that the StudentLoans subreddit will be more inclined to troubleshooting-type posts, which will help my project stay in conversation with Stanley, Deville & Montgomerie (2016)⁠.

Part of this project is a small-scale content analysis, ideally into a binary category that can be depicted as edge attributes on a social network analysis graph. I’m anticipating that I will probably categorize posts into the categories of “troubleshooting” and “political discussion,” but if I feel that the sampled threads are better represented with different categories, then I do will so. Primarily, I’m interested in the capacity of online peer support as a form of resistance against financial debt, so my research questions will be as follows: 1) what is the structure of the social network formed by interactions in the StudentLoans subreddit?, 2) do prominent members of the community favor discussion in one kind of post over the other?, and, 3) do newcomers to the community favor discussion in one kind of post over the other? Essentially, if these peer support forums are instruments of resistance, I’m interested in testing their resilience – do posters get burnt out telling individuals when they should consolidate their loans and when they should challenge something, or maybe it’s just all held together by a core group of highly motivated individuals?

Example Subreddit SNA

Methodologically, I will be utilizing Reddit data which I gathered via the Reddit API – I’ve written up the process here (link this). I might actually pull this data again now that I’m more comfortable working with json files, because I think there are some data fields available that my script didn’t pull, but I doubt that will change my methodology for this paper significantly. I’m not aware of too many explicitly sociological papers using Reddit data, although I’ve definitely seem some that use it as a component of a broader social media analysis, although one study that does utilize a similar methodology for data collection is Kilgo, Yoo, Sinta, Geise, and Suran’s (2016)⁠ “Led it on Reddit,” which investigates opinion leadership on Reddit. Similarly, they also pull data from the Reddit API, and while they don’t employ social network analysis, I think that they might have really benefited from it as much of their analysis centers around popularity. Regardless, I think that my data collection and analysis will be effective without relying on a model, since I already tested some of it out on a previous practicum for this class – more pressing is if my methodology will be effective for answering my research questions. Similarly, I’m sitting on the fence on how I should define edges in my social network. On the aforementioned practicum, I considered “thread interaction,” or being present within the same thread (which seems reasonable, as this is the level at which information is transferred), which I think I’ll probably do again. I could definitely use “direct replies” as a form of edges too, as Reddit employs nested comments where users can directly reply to each other, and while this design feels like it offers some more precision, I’m not sure how that precision would help with my specific research questions (see how these nested comments look in a typical Reddit thread below).

Kilgo, D. K., Yoo, J., Sinta, V., Geise, S., Suran, M., & Johnson, T. J. (2016). Led it on Reddit: An exploratory study examining opinion leadership on Reddit. First Monday, 21(9).

Stanley, L., Deville, J., & Montgomerie, J. (2016). Digital Debt Management: The Everyday Life of Austerity. New Formations, 87(87), 64–82.


Centrality measures in SNA

The four types of centrality used in social network analysis, degree, betweenness, closeness, and Eigenvector centrality, have a variety of different applications. Briefly, degree centrality is the total number of direct connects one node has with another, closeness considers the average distance from one node to the other nodes in the network, betweenness considers a node’s position within the totality of all shortest paths between any two given nodes within a network, and Eigenvector centrality rates the importance of the other nodes connected to the specific node in question. The application of these modes of centrality depends on the research question. For example, if one is interested in specifying individuals within a network that act as gatekeepers, betweenness might be the most appropriate metric, as it considers a node’s position in important paths within the network. Alternatively, if one is interested in the spread of a pathogen, Eigenvector centrality might help identify node that could compromise the entire network. If one is interested in determining who’s on the “inside” and who is relatively isolated within a network, closeness centrality would be a good choice. Finally, there’s degree centrality, which is somewhat descriptively weak as it doesn’t consider the broader network, but if a node exhibits a high level of degree centrality, it’s probably fairly important within its network.

Measures of centrality don’t feature prominently within every analysis section that utilize social network analysis, but researchers frequently use these metrics to describe a network. For example, in Grandjean’s (2016)⁠ article, “A social network analysis of Twitter: Mapping the digital humanities community,” Grandjean constructs a directional network using Twitter following/er relationships, and ranks them according to degree centrality, betweenness centrality, and Eigenvector centrality – note, Grandjean uses in-degree and out-degree centrality measures, which indicate the direction of the relationships which the degree centrality ranks, because Grandjean’s data is directional in nature. Simiarly, Murthy & Lewis (2014)⁠ also construct social networks based off of social media data. They utilize a survey instrument which also informs much of their analysis, but one of the metrics which they utilize to find “hubs” within their network is degree centrality – they also analyze this population of individuals with high degree centrality and find that they are more likely to be women. Personally, I find this methodology surprising – given the authors’ interest in finding hubs, ranked centrality metrics might bolster the conclusions made from their survey analysis, and I think that it is a very practical form of analysis for social media projects.

Grandjean, M. (2016). A social network analysis of Twitter: Mapping the digital humanities community. Cogent Arts & Humanities, 3(1).

Murthy, D., & Lewis, J. P. (2014). Social Media, Collaboration, and Scientific Organizations. American Behavioral Scientist, 59(1), 149–171.

Privacy Statement