Read an Excerpt
Cultures without Culturalism
The Making of Scientific Knowledge
By Karine Chemla, Evelyn Fox Keller Duke University Press
Copyright © 2017 Duke University Press
All rights reserved.
ISBN: 978-0-8223-7309-4
CHAPTER 1
On Invoking "Culture" in the Analysis of Behavior in Financial Markets
DONALD MACKENZIE
How does one represent other cultures? What is another culture? Is the notion of a distinct culture ... a useful one, or does it always get involved either in self-congratulation (when one discusses one's own) or hostility and aggression (when one discusses the "other")? — EDWARD W. SAID, Orientalism
In recent decades, the concept of "culture" has had a strangely bifurcated history. The critique implicit in Edward Said's questions has been deeply influential, especially in social anthropology, for which culture was the single most central concept. By 1996, an encyclopedia of the discipline suggested the possibility of "abandon[ing] talk of different 'cultures' altogether, because of its taint of essentialism" (Barnard and Spencer 1996, 142), in other words, because of its connection to the simplistic idea that a culture was a kind of "package" that was "coherent inside and different from what is elsewhere" (Mol 2002, 80). Simultaneously, however, invocation of culture has increased sharply in areas in which it had not been hugely salient previously, such as in the social studies of science in the form of, for instance, the "local scientific cultures" of Barry Barnes, David Bloor, and John Henry (1996), the "epistemic cultures" of Karin Knorr Cetina (1999), and the "epistemological cultures" of Evelyn Fox Keller (2002). Indeed, "culture" has escaped the boundaries of academia: it has become common to find the term used in everyday language in a sense roughly similar to its social science usages. When teaching students thirty or so years ago, it was necessary to explain that the term did not refer simply to high culture such as opera; now such a warning is hardly necessary.
Of course, suspicion of the concept of culture arises above all in relation to Euro-American representations of non-Euro-American peoples (the ellipsis in the quotation from Said conceals a reference to race, religion, and civilization), and the cultures invoked in the social studies of science are not, for example, national cultures but far more local. Nevertheless, Said's questions are worth asking: is the notion of culture a useful one in other contexts? With the partial exception of Knorr Cetina (1999), invocations of "culture" in the social studies of science have not tended to devote much discussion to the polysemic, politically treacherous aspects of the term. In this chapter, I ask whether a notion of culture broadly inspired by these usages in the social studies of science can productively be applied to a specific aspect of behavior in financial markets: evaluation, that is, efforts to determine the economic worth of financial instruments. (I include efforts to estimate the intrinsic value of financial instruments, either in absolute terms or relative to that of other instruments, as well as efforts simply to judge whether the price of an instrument is likely to rise or to fall.)
There are two reasons for shifting the focus from science to the financial markets. First, evaluation in financial markets is of enormous importance yet very poorly understood. For example, a crucial role of financial markets is to channel investment capital to some activities and not to others, and the amounts of capital involved are huge. The evaluation of financial instruments — of shares, government or corporate bonds, and so forth — is a crucial aspect of this channeling role. Second, shifting attention to the financial markets highlights an issue that tends not to be prominent when "culture" is applied to science: the theory of action associated with the concept. In their ordinary social-science usage, invocations of culture are often associated informally with an implicit theory of action as based on habit, belief, and routine rather than on rational choice. There is nothing inherently necessary in this association, however: culture is still relevant even when one views actors as reflexive and rational. Since rational, reflexive action is to be found in the financial markets if it is to be found anywhere, those markets provide a useful arena in which to explore the usefulness of the concept of culture in contexts in which such action is prevalent. This chapter proceeds as follows. First, I explain my empirical focus: two similarly structured types of financial instrument important to the 2007–2008 global financial crisis, asset-backed securities (ABSs) and collateralized debt obligations (CDOs). The next two sections then describe how ABSs and CDOs were evaluated. The contrast between how this was done in the two cases shows how clusters of practices can differ consequentially even when similarly structured instruments are being evaluated. I then turn to the question of what would be needed to justify moving beyond the notion of clusters of practices to a stronger notion of evaluation cultures, suggesting four criteria that would need to be met. The conclusion then returns to Said's question of the usefulness of the notion of culture, and I express what are inevitably rather personal views on that.
The empirical research on which this chapter is based is a detailed study of the evaluation and governance of complex financial instruments such as ABSs and CDOs. This research draws upon two main sources. First is a set of 101 interviews conducted with analysts, managers, traders, and such who worked with these instruments. These interviews took a broadly oral-history form, with interviewees being led through their careers in respect to the instruments in question, with a view to examining the main developments in the markets for those instruments and in the ways in which they were evaluated. Second, the trade press and technical literature were searched to assemble a corpus of documents relevant to analysis of those market developments and evaluation practices; such documents include, for example, the technical reports in which the credit rating agencies outlined the models they employed in generating their ratings.
ABSs and CDOs
Let me begin by explaining what ABSs and CDOs are. That is most easily done by explaining how the creator of an ABS or CDO — typically, a major investment bank — goes about setting one up. First, it creates a separate special-purpose legal vehicle, such a trust or special-purpose corporation. The vehicle then buys a set of interest-bearing debt instruments (in the case of ABSs, typically mortgages or other forms of consumer debt; in the original CDOs, loans made to corporations or bonds issued by them). It raises the capital to buy those mortgages, loans, or bonds by selling investors securities that are claims on the cash flow — the interest payments and eventual capital repayments — from the pool of debt instruments.
In all but the simplest ABSs, several different classes (usually called tranches) of securities are sold (see figure 1.1). The highest tranche of securities (called senior or sometimes super-senior) is the safest: the purchasers of these have the first claim on the cash flow from the pool of mortgages and such (at least after management fees have been deducted). Only after these more senior claims are met are the claims of investors in lower tranches of an ABS or CDO met. Tranches at the bottom of the hierarchy are thus riskiest. If the debt instruments in the pool suffer defaults (e.g., if mortgagors stop repaying the loans made to them), then there may be a shortfall in the payments due to investors in the lowest tranche. If there are larger numbers of defaults, investors in that tranche may lose the entirety of their investment, and holders of the next most senior tranche may start to be hit. Because the higher tranches are safest, they can be sold at lower spreads (i.e., with interest rates only slightly above Libor [London interbank offered rate] or other benchmark interest rate), while lower tranches need to offer higher spreads.
The motivations for setting up ABSs or CDOs varied. Originally, the goal was often for a bank or other financial institution to raise money for further lending by selling loans it had already made, or to remove those loans from its balance sheet to reduce the capital that regulators required them to hold in respect to those loans. Increasingly, though, ABSs andCDOs began to be set up simply because it was profitable to do so, for example, because of the management fees that could be earned. Institutional investors bought tranches of ABSs and CDOs because they offered higher rates of return than simpler products such as bonds, with the same credit ratings and thus apparently the same level of risk. We now turn to how these complex products were evaluated.
Evaluating ABSs
Consider the most important class of ABSs, mortgage-backed securities. One particular historical contingency strongly shaped their evaluation: the fact that the securitization of mortgages in the United States (in other words, the packaging of them into pools and the sale of securities that are claims on the income from those pools) began its modern history in 1970 as a government-backed program (for its historical origins, see Quinn 2009), in which investors were made good by government-backed agencies in the event of defaults on the underlying mortgages. Because investors could thus ignore the risk of default, they focused primarily on a different risk: prepayment. Deliberate government intervention in the U.S. mortgage market after the Great Depression (at the peak of which, "nearly 10 percent of homes were in foreclosure" [Green and Wachter 2005, 94–95]) led to the dominance of a specific form of mortgage that Richard Green and Susan Wachter call simply the "American mortgage": a long-term, fixed-interest-rate mortgage with no penalty for prepayment (i.e., for redeeming the mortgage early). The American mortgage thus both protects mortgagors from interest-rate rises and gives them the valuable option of redeeming the mortgage early and refinancing if interest rates fall. The obverse of that benefit to mortgagors, however, is a risk to the investor: that he or she will receive his or her money back early at a point at which (because of low interest rates) it cannot be reinvested as profitably. The evaluation of mortgage-backed securities was thus primarily a matter of determining by how much the borrower's option to prepay reduced the value of those securities.
The focus on prepayment risk in the evaluation of U.S. mortgage-backed securities continued even after private-label (i.e., not government backed) securities with tranched structures such as that shown in figure 1.1 (the early government-backed securities had simpler structures) began to be issued from 1977 onward, and also after mortgage lending moved beyond the prime mortgages that the government-backed agencies would purchase or insure to subprime (e.g., loans to mortgagors with impaired credit histories). Investors in the more senior tranches even of subprime mortgage-backed securities continued largely to ignore the risk of default and still focused primarily on prepayment (see MacKenzie 2011).
As the 2007–2008 crisis showed only too clearly, there is a potential agency problem in securitization (i.e., in the packaging of loans into pools and the selling of securities based on those pools): if the risk of default on these loans is thus passed on to external investors, then the originators of the loans that will go into the pool have a much reduced incentive to monitor the capacity of borrowers to repay. Indeed, that agency problem undermined all the pre-1970 waves of mortgage securitization in the United States (see Snowden 1995). For around twenty-five years after the rebirth of private-label mortgage securitization in the United States in 1977, however, the agency problem was held at bay, in part because of the activities of two sets of gatekeepers.
The first was the credit-rating agencies: Moody's, Standard & Poor's, and Fitch. Ratings were essential to the successful sale of mortgage-backed securities. It was very hard to find buyers for securities without an investment-grade rating (i.e., BBB– or above); indeed, the lowest externally sold tranche of a mortgage-backed security was often the "mezzanine" tranche with a BBB– or BBB rating. By far the largest demand was for AAA-rated securities at the very top of the hierarchy.
This made the evaluation of mortgage-backed securities by ratings agencies a crucial matter. The agencies were concerned exclusively with default: they considered prepayment to lie outside their ambit. The evaluation practices they employed gradually evolved from the analysis of the overall characteristics of mortgage pools (e.g., the average loan-to-value ratio of the mortgages in the pool) to logistic regression or hazard-rate models of default on individual mortgages (using a wider range of variables, e.g., creditworthiness scores of the borrowers; see Poon 2007, 2009). Crucially, there was no explicit modeling of the phenomenon that the evaluators of CDOs were to call "correlation" (see below). Interdependence among defaults was handled by other means, such as the use of historically based "stress scenarios," above all the mortgage defaults of the Great Depression; at Standard & Poor's, for example, the criterion for a rating of AAA was that the tranche in question could withstand Great Depression default rates or their equivalents for the pool in question. The use of stress scenarios then made it possible mathematically to treat mortgage defaults as independent events, because it could be argued that correlation was already taken into account (at least implicitly) in the adverse macroeconomic circumstances crystallized in the scenarios. Another source of what CDO specialists were later to call "correlation"— the exposure of a geographically limited pool of mortgages to local economic conditions — was also handled procedurally rather than by explicit mathematical modeling: pools of mortgages considered insufficiently diversified geographically were subject to ratings penalties.
The second set of gatekeepers was the investors in the mezzanine tranches of mortgage-backed securities (as noted, those were typically the lowest of the tranches to be sold to external investors). Their role was pivotal, because the mezzanine tranches were the hardest to sell, and the investors in those tranches were those with their capital most immediately at risk. Investing in mezzanine tranches was typically a specialized activity, conducted by institutional investors with considerable experience in the mortgage market. They would frequently ask for the "loan tapes" (the electronic records of the mortgages in the pool), which investors in more senior tranches almost never did, and inspect the tapes in detail, for example, looking for clusters of particularly risky mortgages. If they found such clusters, they would sometimes demand that the composition of the pool be changed before they would invest. The creators of mortgage-backed securities had to take such demands seriously, because failure to sell the mezzanine tranche would typically mean that a mortgage-backed security could not successfully be created.
Evaluating CDOs
CDOs were a later development than mortgage-backed securities. While, as noted, the first modern private-label U.S. mortgage-backed security was issued in 1977, the first CDO was created only in 1987. The structure of CDOs was typically similar to that of mortgage-backed securities and other ABSs (see figure 1.1), but the composition of the pool of debt instruments differed: instead of mortgages or other consumer debt, the pool of a CDO would typically be corporate debt. Originally, the evaluation of CDOs was broadly similar to the evaluation of mortgage-backed securities, with the exception of the fact that prepayment was a relatively minor issue (with no equivalent of the deliberate government action on behalf of mortgagors, loans to corporations are often floating rate or involve substantial prepayment penalties). For instance, the way in which analysts at rating agencies evaluated CDOs was originally quite similar to how they evaluated mortgage-backed securities. Stress scenarios were prominent, and poor diversification (e.g., too much of aCDO's pool coming from one industry) was again penalized procedurally. When evaluating a CDO, analysts at Standard & Poor's, for example, would "notch" (i.e., reduce by one or more ratings grades) the debt instruments issued by corporations in a given industry if that industry formed more than 8 percent of the CDO's pool (MacKenzie 2011).
(Continues...)
Excerpted from Cultures without Culturalism by Karine Chemla, Evelyn Fox Keller. Copyright © 2017 Duke University Press. Excerpted by permission of Duke University Press.
All rights reserved. No part of this excerpt may be reproduced or reprinted without permission in writing from the publisher.
Excerpts are provided by Dial-A-Book Inc. solely for the personal use of visitors to this web site.