Marking to Markowitz | FT Alphaville

Marking to Markowitz

Harry Markowitz – nobel laureate and father of Modern Portfolio Theory – has outlined a proposal for remedying the current financial crisis – a chief cause of which, he says, is a lack of transparency (H/T to Rolfe Winkler).

Markowitz’s short paper, published in the CFA’s journal,  is available here.

His point is that as long as illiquid, opaque and hard-to-value assets remain on banks’ balance sheets, the crisis cannot possibly be resolved. Japan, is a good case point. Years of unresolved bad-loan problems there led to atrophication and stagnation.

Here are Markowitz’s four points, then, for resolving the transparency crisis, and by extension, the bank crisis itself:

1. Take a census of the institutions that own various instruments—CMOs, CDOs, CDS, and the like—noting the assets, liabilities, and “rules of the game” of each instrument. Although this proposed survey is large, it is no larger than such government efforts as the U.S. Census Bureau’s Annual Survey of Manufactures.

2. Calculate the direct and indirect exposures of each instrument. For example, CDO A contains Tranche B of CMO C, etc., and thus is exposed to these amounts of those underlying mortgages. The mathematics of tracing out these direct and indirect relationships is similar to the techniques by which a search engine (e.g., Google) finds thousands of web pages that match a given phrase.

3. Aggregate the direct and indirect exposures of a given instrument (and the instruments of a given institution) into meaningful categories. Simply knowing that a given instrument is directly or indirectly exposed to a long list of mortgages is insufficient. These mortgages should be aggregated in various ways (e.g., by zip code and late-payment history). The leverage of the instrument and of the institution that holds it should be analyzed, both directly and indirectly (e.g., has the institution borrowed to buy a tranche in an instrument that is itself leveraged?).

4. This information should be disseminated on a need-to-know basis to various parties, including stockholders, counterparties, regulators, and academicians. As with Census Bureau data, the more public disclosures may be aggregated more than those disclosures that are less public.

The problem is that, in spite of what outsiders might say, transparency – as Markowitz identifies it – is not necessarily the problem. A clearinghouse of deep, researched, cross-referenced data into the collateral backing various structured instruments would not do much to elucidate.

Firstly, the issue is not a dearth of information, but precisely the kind of glut that Markowitz proposes to create. Buysiders are, for the most part, already able to access detailed spreadsheets and information on the collateral backing CDO notes – all the way back through the RMBS down to individual mortgages.  They can get that data on Bloomberg; they can get that data from prospectus supplements; and they can get that data from dealers. The problem is that there is mountains of it: too much data, not too little.

Imagine you were buying $5m of senior paper in a subprime ABS CDO. First, you’ve got the data on the underlying securities in the CDO – hundreds of RMBS tranches. Each one of those tranches, in turn, comes with information on the securitisation it was issued from. A securitisation which will contain tens, if not hundreds of thousands of raw mortgages. For your little $5m ABS CDO slice the amount of due diligence is mind-boggling. It’s little wonder many less sophisticated buyers simply relied on ratings.

Secondly, even supposing all this data were needed, and collated, in aspiration to a Markowitzian portfolio-efficient type understanding of it, it would still do no good. Banks deployed (what seemed, at least) some of the most sophisticated modelling available to them to price CDO tranches. Still they failed.

The huge problem for CDOs was – and is – not their relative architectural and structural simplicity, but correlation.

It’s correlation that makes CDOs a mathematical nightmare. And it’s poor understanding of correlation that has dashed attempts to successfully price CDOs. To put it simply: CDOs are not simply the sum of their parts. It’s not good enough just to understand each and every granular little node of collateral – each mortgage – ultimately behind the deals as Markowitz is proposing. You have also to try and grasp at the relationship of each and every granule to each and every other granule.  You need to try and weave those little pieces of collateral into a dynamic, interrelated sheet of information. Tricky.

Banks, though,  thought they had a model for doing this. And in true hubristic style, it was overreliance on this particularly elegant and ingenious model that allowed the system to swell so much – to the point of extreme danger and ultimately, implosion.

Of course, Markowitz should be familiar with this. As a progenitor of modern portfolio theory he is also one of the great intellectual forebears of that model- the gaussian copula framework for default probabilities.