Reactions to Tett/MacKenzie: Anush Kapadia and the “liquidity crisis hypothesis”

June 16, 2009

Anush Kapadia just posted a fascinating reaction to Donald MacKenzie’s review of Gillian Tett’s book on the credit crisis. I thought that the elaborate comment, buried under eleven previous comments, deserved a lot more prominence. Here it is:

Tett’s is the best account I have seen of the crisis, but she leaves us asking the question: why did super-senior tranches of CDOs and synthetics prove to be the achilles heel of the system? She gives us a few clues, and MacKenzie picks up on one of them, correlation, in his review. While adequately measuring correlation was of course critical, hiking the level of correlation would still fail to account for the magnitude of the disaster. Leaving dollars on the sidewalk, both MacKenzie and Tett provide the materials for a more satisfying answer without putting these ingredients together. Perry Mehrling, ironically cited as an historian of economics by MacKenzie rather than the historically-informed monetary economist he is, has.

The key ingredient for a fuller story is an account of liquidity; it is here that MacKenzie always seems to fall short, (Mehrling has a liquidity-driven account of LTCM as well). The term is of course notoriously difficult to define uniquely, but its frequent pairing with the word “deep” ought to tell us something. Liquidity suggests robustness: a market is liquid if one can buy and sell in significant amounts without affecting the price. But who does one deal with? Typically, a market maker offering a spread. It is thus the ecology of market makers that ultimately determines the depth and resilience of a given market, offering to deal in any amount at their stated prices. The robustness of a market thereby reduces to the robustness of the balance sheets of the key market makers therein. The facticity of the “public fact” of market price thus rests on the ability of market markers to fund themselves. Once these balance sheets start to look anaemic, the prices built on them start to look less reliable.

So it was with the key “public fact” that both Tett and MacKenzie point to: the CDS indices. Oddly, MacKenzie does not think it fit to bring up the key role of the indices that he himself eloquently pointed to in his “End-of-the-World Trade” in his review of Tett. This is unfortunate as Tett is quite categorical about their centrality, and indeed her interpretation differs significantly from MacKenzie’s. For Tett, as for Mehrling, the important thing about the indices was that, being more standardised, their market was more liquid and therefore their prices more reliable. Robust pricing in the liquid derivative ABX index would then enable traders to price the illiquid underlying CDO tranches. Tett:

“Trading in mortgage bonds, let alone mortgage derivatives, was sparse. The only obvious guide was the ABX index, which had been launched in 2006. It provided a gauge of the value of the range of bonds in the CDOs—from BBB to AAA. So what many funds—including Bear Stearns—did was to look at the prices as given by ABS and then use that to deduce the prices of the bonds in their own CDOs,” (pg. 171).

MacKenzie notes this relationship in EOTWT, but for him it is not a matter of trading providing liquidity to a price point but trading providing solidity to the facticity of correlation:

“…trading of index tranches made correlation into something apparently observable and even tradeable. The Gaussian copula or similar model can be applied ‘backwards’ to work out the level of correlation implied by the cost of protection on a tranche, which again is publicly known.”

Correlation is critical to MacKenzie’s story because it goes into the manufacturing of that archetypical public fact, ratings. This is true in both EOTWT and the Tett review. In the latter, he notes that “Essential to the [CDO] assembly line was that the higher tranches of its final products…be able to gain Aaa ratings. A critical issue was the likely correlation of mortgage-backed securities.” This focus on correlation over and above the trading architecture leads MacKenzie to interpret Tett too narrowly, in my view, and thereby to miss a critical functional feature of a public fact that he has himself sniffed out as vital. Of course, Tett leaves herself open to this interpretation because she does not fully spell out the criticality of the CDS indices and their consequent impact on the super-senior tranche prices.

MacKenzie suggests that Tett is praising the Morgan credit derivative inventors for noting an empirical fact, that mortgage default correlations were simply unobservable and therefore the risk in betting on them could not be prudently measured. In the absence of further explication, Tett does indeed give this impression, and MacKenzie’s own commitment to correlation pushes him further in this direction, (although it is indeed strange that, in light of his own observation that CDS indices gave these correlations public facitcity, he does not think the indices warrant a mention in his Tett review). But Tett’s observation that is it trading and therefore liquidity in the CDS indices that enabled the pricing of the underlying CDOs leads us in another direction: not to facticity from modelling and ratings but facticity from trading.

To be fair, MacKenzie does note the importance of trading, but it appears to him merely as a “fact-generating mechanism” by way of marking-to-market. Thus he notes towards the end of EOTWT that:

“It has become common to use a set of credit indices, the ABX-HE (Asset Backed, Home Equity), as a proxy for the underlying mortgage market, which is now too illiquid for prices in it to be credible. However, the ABX-HE is itself affected by the processes that have undermined the robustness of the apparent facts produced by other sectors of the index market; in particular, the large demand for protection and reduced supply of it may mean the indices have often painted too uniformly dire a picture of the prospects for mortgage-backed As Carruthers and Stinchcombe note, market liquidity depends on facts. However, today’s financial facts depend on liquidity. The credit markets remain stuck in a vicious circle.”

But if liquidity is so critical why, despite Tett’s corroborating suggestion, does MacKenzie provide no account of it as generative of facticity in addition to the other way round? Tett herself elides the matter.

Liquidity is the missing piece of the puzzle that enables us to understand Tett’s main point regarding the impairment of the super-senior tranches of CDOs. Simply put, these “safer than safe” tranches were so badly hit not merely because all of Wall St. neglected the extent of the correlation of the underlying mortgages but because the markets that priced these tranches had no market maker of last resort. No emergency market-maker, no liquidity, no rational pricing.

When the banks need liquidity, they go to the interbank market and borrow/lend at LIBOR. When they all run out, they go to the central bank’s discount window. As Tett points out, shadow banks had only one liquidity backstop: the absolutely vital “liquidity puts” with the banks themselves, (MacKenzie makes no mention of them at all. See Tett pg. 205-6). Insurance sellers on the ABX were also providing a kind of backstop, and those backing up AAA risks were in effect backing up systemic risk, really the only kind of risk that is expressed in that coveted rating. By making AAA insurance contracts liquid, insurance market makers were implicitly acting as systemic risk providers. Cheap liquidity led them to underprice systemic risk and help create an unsustainable credit boom. When this became clear and everyone ran for the doors, there was no market maker of last resort who the system as a whole could turn to. The system itself melted because the systemic watchdogs were private, profit-driven entities (AIG and the monolines) who, when it comes to systemic risk, are by definition under-capitalized. With the backstops blown out, even the safer-than-safe risks looked unsafe.

This answers the question MacKenzie poses at the end of EOTWT: “Why…have people not been selling end-of-the world insurance when the returns from doing so have jumped ten-fold while the risk of having to pay out remains small?” As noted above, he cites mark-to-market as the paradoxical answer: that fact-generator now blocks the reestablishment of the pubic fact because of…lack of liquidity! This is not paradoxical but circular: what is the difference between mark-to-market as fact generator and fact inhibitor? More generally, when do positive feedback loops turn into negative ones, and why?

A more coherent answer, unavailable in MacKenzie’s vocabulary, is that, in the absence of a liquidity backstop to the insurance market, traders did (do?) not have a sense that the outcomes are bounded in any way. When we are talking about system risk insurance, the only entity capable of providing this backstop is the state (given its super-sized balance sheet) as it does through its central bank in the interbank market, and even it might prove insufficient. We were missing such an entity in the key CDS index markets, markets that form a structural analogy in the erstwhile shadow banking system to the boring credit markets of its “regulated” parent. Given that such devices are only ever the creation of crises, we ought not to be surprised at their absence. But if we want to get these markets starting again, we ought to be agitating for their construction. This is precisely what Mehrling has been doing.

That MacKenzie came so close to this answer but failed to connect the dots might indicate that SSF has a serious epistemological blind spot. So I would like to end this over-long post by briefly reflecting on what this absence of attention to market structure and credit means for the sociology of finance. MacKenzie has made his name by seeking to break open the black boxes of finance and eschew the Parsonian division of labour between sociology and economics. While he has gone further than anyone else in doing this, he has not gone far enough. This is perhaps understandable given his role as a pioneer, but those who have followed in his wake tend to replicate the error, academic markets being acutely prone to herding. While consistently being drawn to the most interesting and critical aspects of modern finance, MacKenzie’s tight focus on particular models and markets has left us without a more general theory of market activity and the pivotal role of the credit markets generally, even when discussing the crisis.

This is ironic, for market making, liquidity, and ultimately credit-money itself are perhaps the most performative aspects of our modern economy. Yet because their performativity has macro-structural predicates—ultimately undergirded by a market theory of money—these objects fall outside the purview of SSF. Yet this is precisely where economic sociology might really take on a faltering mainstream economic paradigm. It is not simply that economics is performative. The critical question is, if the economy is a social entity that does not submit to the scientism of modelling, how is macroeconomic control achieved at all, and how does it break down?

In the conclusion to EOTWT, MacKenzie points out that the power of central banking comes ultimately from the state’s power to tax. True, but this power remains platonic as a control device unless there is a social mechanism for its transmission. Since the inception of central banking, this mechanism has been the credit markets. What does it say about SSF that it was silent on these “boring” markets till after this crisis?

(For Perry Mehrling’s account of the crisis, from which this post is drawn, refer to his interventions here:

13 Responses to “Reactions to Tett/MacKenzie: Anush Kapadia and the “liquidity crisis hypothesis””

  1. danielbeunza Says:

    Anush writes: “Tett’s observation that is it trading and therefore liquidity in the CDS indices that enabled the pricing of the underlying CDOs leads us in another direction: not to facticity from modelling and ratings but facticity from trading.”

    I agree that “facticity from trading” is a fascinating mechanism. Essentially, it entails translating the prices on a traded security into the beliefs implicit in those prices. It’s like telling what people expect about the weather by looking at whether they carry a folded umbrella with them. Economists refer to it as “backing out” and “risk-neutral pricing.” One ex-chief economist of the SEC referred to it as one of the most important financial innovations of the past two decades, alongside Black-Scholes.

    In my own work with David Stark, I have examined at length the possibilities and pitfalls of this technique. What we found is that, in the context of merger arbitrage, this system allows traders to check their assumptions against those of their rivals. We call it “reflexive modeling.”

    But we also found out that this technique can lead to financial disasters. When most arbitrageurs miss a key problem in a trade, reflexive modeling leads the arb community down the path of overconfidence.

    Might this be what happened in the case of the credit crisis?

  2. marthapoon Says:

    Thanks for highlighting this great post Daniel. I’d like to follow up, on behalf of ‘SSF’…

    In his review McKenzie focuses on the ability of the models to properly capture correlation and he closes by agreeing with Tett that the JPMorgan story proves that a more prudent form of banking is possible. By prudent, he means more attentiveness to modeling the underlying phenomenon correctly. Its a moral call for better banking practices.

    This does not, however, seem to detach the story of ratings or models (correct or not) from playing a role in creating liquidity. As William Cronon so elegantly shows in Nature’s Metropolis, if you can’t rate the grain into types then you can’t make it liquid – you can’t pool and move grain from different farmers into the elevators; you can’t get movement.

    The same is true in the credit markets. Take mortgages for example: Without a common rating of LTV and FICO, pooling loans was an authoritative function of government institutions. What makes liquidity pour through mortgages (in a seems like a sudden macro flow) is not just their reputed safety or any quality of the mortgage market as a credit market, but rather innovations in techniques for rating and bundling loan pools in the late 1990’s. These innovations, like writing on packages in the supermarket, is what allows buyers and sellers to enter into exchanges.

    So social studies of finance is attentive, I think, to macroscopic things like liquidity. It just insists that these movements are grounded in and are built upon the spread of tiny particulars. One of the best references touching on this point might be : O’Connell, J. (1993). Metrology: The Creation of Universality by the Circulation of Particulars. Social Studies of Science 23, 129-173…

  3. siva Says:

    Anush: I am not convinced by your account of MacKenzie as having missed “liquidity” as being crucial to a story about CDOs and the credit crisis.

    First, on liquidity, clearly it’s used in numerous different ways in finance. As you rightly point out, liquidity is “notoriously difficult to define uniquely”. You suggest that it has something to do with the ability to trade in size without moving the price in a market, and then later seemingly as something you can go to the interbank market for (so, overnight money, I guess), and then in the usage “cheap liquidity” I suppose you mean the price of overnight money.

    Two questions suggest themselves: is it crucial that one pick and choose between the different uses? It’s not obvious to me that an SSFer(!) should cast themselves in that role at all, but rather simply examine the different uses and their consequences. But you say that they should pick and choose, that it’s a blind spot having to do with an incomplete disavowal of the Parsonian divide. Okay, fine, but in that case how could you impugn MacKenzie for not making liquidity central to his analytic given that you yourself use it so variously? (Correlation as a concept, on the hand, is implicitly treated as suitably stable by MacKenzie and thus available for theory building.)

    Second, taking liquidity strictly as trade order depth, why does this necessarily have anything to do with market making? When electronic markets were first proposed, the great claim was that they’d be fairer to customers because market-makers would be done away with. That is, the claim was that it was better to trade electronically because then everyone could post whatever prices (or hit/lift whatever prices) they wanted.

    No doubt there still exist some traditional market making going on for big orders in most markets. But just look at the volume of computer-generated trading going on these days in, for example, US equities. Some estimates of algorithmic trading are at 80 percent of total daily volume—by both large customers (pension funds, etc.) executing their trades in small chunks so as to manage market impact and, of course, hedge funds and the banks who, arguably, on occasion arrange their program trades /to/ make for large market impact. Doesn’t all of this suggest to you that a theory of market function predicated on “market-making” is likely in serious trouble?

  4. anush Says:

    Many thanks for the comments and references.

    Daniel, I think I mean something quite simple when I’m talking about liquidity as constitutive of price facticity. A price in a liquid market is more “solid” in the sense that it won’t move around when I try to trade at that price, even at significant volumes, because, in an OTC market, the market maker on the other side of the trade has the balance sheet to absorb the inventory. Of course, with contracts like CDS, it’s not so much a matter of “inventory” as solvency: will the counter-party be there to payout? The price of the contract reflects this counterparty risk and thereby ties liquidity and solvency closer in an OTC market for derivatives. Whence all the chat about exchanges.

    Martha, I naturally agree that ratings in particular and standardisation in general can play a role in generating liquidity. As Tett notes, this is precisely why the CDS indices were more liquid than the underlying CDOs. My point was that this argument is exactly what one would expect from a practitioner of SSF: liquidity from facticity. MacKenzie cites Carruthers and Stinchcombe on precisely this point, before noting that the arrows go the other way as well: “However, today’s financial facts depend on liquidity.” And yet, facts-from-liquidity appear only as an afterthought in EOTWT and not at all in the Tett review. So it is not that SSF cannot account for liquidity but that it has no account of it as arising from the regular activity of market making. This, I respectfully suggest, is because the basic microstructure of OTC markets—and most of the concatenated credit markets are OTC—has been passed over because of SSF’s concern with “tiny particulars.”

    Siva, I fully agree that, whether one is an SSFer or not, one ought to be very catholic about the use of a term like liquidity. Yet I can’t see how my point that MacKenzie virtually ignores it in any sense commits me to using it in just one sense! But to clear, I was really using it basically one sense, which is fairly stable viz. the money markets. Market depth in the overnight money market is, in the final instance, provided by the big balance sheet: the central bank. When that big balance sheet says liquidity is cheap, it is. The point is that the structures of what Perry has called the “legacy banking system” have not evolved to transmit the intentions of that big balance sheet to the extant, capital-markets credit system as credit risk and liquidity risk are priced in different markets, making the overnight rate not ineffective but very blunt as a control mechanism. Another big balance sheet was doing the backstop work in the credit markets (hence my reference to a “structural analogy”), but, so the argument goes, it wasn’t, and couldn’t be, big enough when tide turned.

    Regarding your second point, again, I am not saying that marker making is the only market structure out there, far from it. Electronic exchanges abound, of course, but let us note that a) they are basically centralised counterparties with huge balance sheets that eliminate counterparty risk, b) only work for the most standardised contracts, and c) are conspicuously absent from the credit markets (so far).

    Why is it that fixed income markets tend to be OTC and equities on exchanges? It didn’t start like this: the London Stock Exchange, famously, started as a bond market. What is the relationship between instrument and market structure? It is this a sociological question? Yes, if market construction and (mal)function are on the agenda.

  5. marthapoon Says:

    Anush, What is ‘the microstructure of the OTC markets’. (I’m afraid I can’t fully understand your reply with out unpacking the particulars of this phrase!) Martha

  6. […] agreement against your point, you may want to reconsider. I also want to point out this excellent Review of a Review of Tett’s Fool’s Gold by Anthropologist Anush Kapadia: When the banks need liquidity, they go to the interbank market and […]

  7. anush Says:

    Sorry for the jargon! Microstructure simply refers to the trading architecture of a market, the rule-structure governing the execution of trades. It is tremendously important for price formation, liquidity and so on and is an important topic in market design. See:

    OTC is “over-the-counter”, which is a form of trading that is basically bilateral rather than on an exchange. See:

    The OTC vs exchange debate goes back at least to the LTCM debacle. In its current form, it is an argument about how CDS should be traded. There is a political economy to this as banks like the more non-standard and therefore lucrative OTC format whereas regulators like the stability of an exchange…

  8. siva Says:

    Anush, I’m not sure what the reference to OTC/Exchange Traded Markets gets you. There’s plenty of algorithmic trading in both. The reason I raised the latter, especially customer uses of it, is simply to underscore an important fact about market structure—it’s not just about providing liquidity. That is, I don’t see how one could have a nuanced understanding of market structure that just talks about small and big liquidity providers.

    To be sure I think Mehrling’s argument that the government, or some relevant agency like the Feds or the FDIC, ought to intervene in the credit risk market as a matter of policy is fine. What I baulk at is that it’s built out of a picture of money has no role for liquidity taking at all. It’s as per classical neoclassical finance—liquidity taking is some kind of random “noise” function at best.

    To take your example of money markets and the Feds. The Feds take liquidity every time they change rates. (More precisely, the more unexpected a rate change, the greater the liquidity withdrawn from the market.) A model of credit that is incapable of addressing this point seems straightforwardly errant.

    As for MacKenzie’s use of fact-from-liquidity. It seems to me that it is sensitive to the context in a way that your is not. That is, yes, sure, the appropriate thing to do at the height of the crisis was for a big balance sheet to sell insurance. But the /reason/ for that surely can’t be located in a model built only out of liquidity provision regardless of context.

  9. anush Says:

    Of course market structure is not all about liquidity, but it is a critical predicate of market liquidity, more so in OTC markets.

    On Mehrling, I don’t think his view is one-sided in theory, although it might be in practice. Taking liquidity is completely comprehensible in his view: the big balance sheet moves up the outside spread, forcing all the “downstream” dealers to contract their balance sheets.

    If by context you mean where we are in the credit/business cycle, then no, I am most certainly not insensitive to it. We have to had the big balance sheet take liquidity at the top of the cycle, for sure. The question is, when it suspects we are getting to the top of the cycle, how much does the big balance sheet have to take in order to slow things down? My hesitation on this question is what makes me say that there might be a difference between theory and practice. But this is an old question:

  10. siva Says:

    You say in the main post that “Liquidity is the missing piece of the puzzle that enables us to understand Tett’s main point regarding the impairment of the super-senior tranches of CDOs. Simply put, these ‘safer than safe’ tranches were so badly hit not merely because all of Wall St. neglected the extent of the correlation of the underlying mortgages but because the markets that priced these tranches had no market maker of last resort. No emergency market-maker, no liquidity, no rational pricing.”

    You then use this point to assert something about SSF and its alleged blindness. But now you appear to be saying that there’s more to the market than just liquidity, much less just liquidity provisioning. This is no trivial matter because opening up the story about how markets work to bring in other “predicates” also opens up your critique of MacKenzie. Why is he wrong to not emphasize liquidity provisioning when there are other factors at play?

    Did MacKenzie mistakenly think only of a mispricing of credit risk rather than thinking of a lack of liquidity, as you argue, or is it that you, despite the disavowal you make in the comment above, implicitly think any mispricing in any market must stem from a lack of liquidity in that market?

    More to the point, why is this question relevant to the other claim you make about SSF and MacKenzie not having rejected the Parsonian divide? It seems to me that the particulars of money markets matter and that the theorizing about economics ought to be slowed down just a little.

  11. anush Says:

    Clearly I think that, in this particular case of the CDS index markets, what is at stake is liquidity, how it is predicated on market structure, and how these two interacted to misprice systemic risk. It is only against this assertion that critiquing MacKenzie for overlooking these aspects makes sense. This implies nothing about a general theory of mispricing, market structure, and so on.

    My limited point was that, despite pointing out some key features about the CDS index markets, MacKenzie didn’t make the jump to market structure, how it was connected with liquidity, and how that might have led to the crisis.

    Neither has mainstream economics. This is a problem for theory, given the economist’s monopoly on these kinds of topics. And this is a problem for policy, given the access economists have. Opening black boxes means stepping on their turf. Or it ought to.

  12. Alex Says:

    I don’t think that liquidity (or the lack of) is “generative” of facticity, although given how one loops into another that might seem to be the case. Instead, it seems more likely to me that a breakdown in facticity gave rise to the drying up of liquidity. I see this as a matter of risk and uncertainty. When discrepancies occur in “normal” markets, small spreads are usually arbitraged away and performativity more or less ensures that the markets follow theory. In such situations, risk assessment is possible, which in turn means that spreads are usually kept in line, and the virtuous cycle continues. However, crises or black swan events screw things up. After a certain point in the current crisis, there was a rupture in performativity. The apparatus of finance – models, indices, credit ratings, etc., were no longer aligned with the social reality of the market. The resulting uncertainty led to a drying up of liquidity. In fact, I think the ongoing market rally is a similar phenomenon. There is still considerable uncertainty and distrust in the facticity of indices, valuations, etc. It’s just that this time around, there is excess liquidity flowing into the equity markets.

    Huault and Montagner have an interesting paper in Organization Studies (2009) where they assess that market makers for credit derivatives struggled and ultimately failed to “performatise” theory with practice. There was considerable scepticism over pricing of credit derivatives, even after indices like ABX were in place. This explains why the majority of trades in credit derivatives were concentrated among a small number of banks, despite the growth in the overall market.
    (Their paper is here:

    Actually, I think credit ratings agencies played a large role in the drying up of liquidity. They provided the classificatory system for debt tranches and turned credit derivatives into the “boundary objects” mentioned by Millo and MacKenzie – i.e. facilitating communication across institutions in the financial field. This enabled liquidity. Ratings provided a convenient proxy for valuing credit derivatives, which have been acknowledged as notoriously complex to price. As mentioned, the ABX index used ratings on debt tranches and that information, in turn, helped others price their own CDOs. When the subprime crisis started brewing, some banks started shorting ABX as a hedging measure. Credit spreads in ABX had been increasing since Jan 2007 and surged up around July 2007. There comes a point when performativity breaks down. I think this came when the ratings agencies proceeded downgrading their debt ratings from 3Q07 onwards. This reassessment was no doubt prompted by what they could clearly see happening in the subprime market and reflected in ABX. But their downgrades, instead of having the effect of more accurately reflecting “reality”and enabling risk management, had the opposite effect. Instead of communicating risk, it transmitted uncertainty. The facticity of the ratings as a measure of the quality of debt was no longer taken for granted. Liquidity dried up. This is evidenced by the huge spread in A-AAA corporate bonds in 2009 compared to 2007.

    I think the breakdown in facticity is a useful area of research. It exposes the “taken-for-granted” world we live in and analysing these transitional, often violent, moments can shed some light on how social actors deal with risk and uncertainty.

  13. Your site is pretty interesting to me and your subject matter is very relevant. I was browsing around and came across something you might find interesting. I was guilty of 3 of them with my sites. “99% of blog owners are guilty of these 5 errors”. You will be suprised how easy they are to fix.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: