Sunday, October 11, 2009

We Need Open Derivative Models

Before I knew anything about the finance industry, if someone had thrown out the name “Blackrock” I would have conjured up a scene from the Wild West, a cattle rancher hiring a gunslinger to roust out the sheep farmers and take control of the town. But now, of course, I know that BlackRock is a financial behemoth with $1.5 trillion in assets under management -- soon to be over $2 trillion due to its purchase of Barclay’s asset management business. But of more note than its asset footprint is the mantle BlackRock is gradually assuming as the arbiter of value.

BlackRock won a set of contracts to provide analytics for the New York Fed’s trillion dollar mortgage-backed purchase program. Now, BlackRock may end up with an NAIC contract to analyze the mortgage-backed securities in insurers’ portfolios.

I do not mean to diminish BlackRock’s laudable role in assisting in many ways with the financial crisis – coming forward when a number of other large firms demurred. But as one contract is piled on another, their models will become the standard for pricing mortgages, complex derivatives and structured products. Unlike the money management business, which is competitive and relatively transparent, this is a monopoly ready for the making. A monopoly because the more institutions, industry associations and regulatory bodies that employ their services, the more they become the de facto standard. Over time, auditors, clients and equity holders – perhaps even regulators – will start saying, “Well, it is nice to see what your internal models have to say about your portfolio value, but we want your portfolio benchmarked using the BlackRock model.” A BlackRock seal of approval; BlackRock, the JD Powers of portfolio quality.

Here are the problems with this:

First, of course, is the well-known issue of allowing a private enterprise to have monopoly control of a utility – in this case a de facto replacement of the rating agencies (not a bad thing in itself) by putting one firm in the position of providing the benchmark pricing of financial products. Second, there are natural conflicts of interest given that BlackRock is also the asset manager for the New York Fed’s Maiden Lane portfolios and has raised over half a billion in private capital to purchase legacy securities as part of PPIP. (Though I should add that BlackRock is aware of this issue and has stated the firm has strict internal controls preventing any valuation services from being gamed by its investment arm. Which should make us all feel a lot better).

But the most critical problem is that its approach is at variance with the broadly held view that we need to have transparency in the derivatives markets because, unlike, say, RiskMetrics, BlackRock does not share the specifications of the models it employs. We don’t really know what these models are doing. Valuations based on a black-box BlackRock model, or, for that matter, anyone else’s black box model, do not get us the transparency we need. I don’t care what a trading desk uses for its decision making, but when it comes to valuations that carry beyond the firm, we need to be able to see and critique the models that are being used. If a model is to become a standard, if it is going to be used for regulatory or other benchmarking purposes, it should be transparent and subject to peer review.

Which gets to a simple point: If we want to go down the path of standardized valuation and comparability in these complex portfolios, we need open derivatives models. One thing we should have learned from the rating agency debacle is that even if we put aside the issues of monopoly power and conflict of interest, we cannot stop with having the proprietor of such models say, “Trust me, I know what I’m doing.”


  1. On the other hand, if the model isn't open then it's harder to game the system.

    (Right up until the point of your first ex-BR employee hire)

  2. The problem with open derivative models, on the other hand, is their vulnerability to being gamed unless they are extremely restrictive. In particular, they should only allow standardised derivatives traded on exchanges, whose behaviours are designed in and known beforehand. Otherwise, it would not take long before customised derivatives that look good on the model but are actually extremely toxic start getting issued. Just look at what happened with VaR, a somewhat open model, as an example of what might go wrong.

  3. Hmmm. Perhaps the current slump in the finance sector and the trend toward grad school for today's 20-somethings without jobs will help here. It would be interesting to investigate the cycle of academia-vs-industry as Wall Street becomes more and less lucrative. The current crop of grad students may very well decide to stay in academia, providing the thought leadership we need in this area.

  4. standardization of complex derivatives only gives people a false sense of security. we can't price reliably most of those complex products no matter what model we use. doing the same as the industry standard is a great ass-covering exercise for managers involved, but that's about all.
    I run a book that is marked at "consensus" however you can't find a bid at all, even miles below that consensus level and in small quantities. but that doesn't bother anybody, we are "marked to market", nothing can be blamed on the managers. standardizing what we cannot price is dangerous stuff.

  5. Open source valuation models should not only improve transparency, but also improve the quality of the modeling techniques, allowing anyone with abit of quantitative skills to challenge the underlying assumptions.
    It should also improve the consistency and comparability of valuations across firms.