Why couldn’t experts see the coming financial crisis in their data?

In the years leading up to the financial blowout in 2008, the two most important sets of missing data that could have helped experts spot looming trouble were measurements of (1) mortgage quality and (2) repo volumes and collateral quality, according to an April 15, 2010, report from Federal Reserve Board economists.

The report is “Financial Statistics for the United States and the Crisis: What Did They Get Right, What Did They Miss, and How Should They Change?” by Matthew J. Eichner, Donald L. Kohn, and Michael G. Palumbo.

Still today, no one is collecting comprehensive information about repo volumes and the quality of repo collateral, even though top U.S. regulators have said the high leverage and interconnectedness of the repo market created the systemic risk most responsible for forcing the taxpayer bailout of the mega-banks in 2008.

The Federal Reserve economists note that in the lead-up to the crisis, analysts needed to be able to see (1) that mortgage quality was declining – in spite of what credit rating agencies were saying – and they needed to be able to see (2) that a funding mechanism was developing that could carry and amplify the mortgage shock throughout the financial system.

The funding mechanisms that have gotten the most attention by economists and the press since the crisis, and thus have become the best understood example of how the mortgage bubble was fed by short-term money that could quickly withdraw, are structured investment vehicles (SIVs) and other asset-backed commercial paper (ABCPaper) programs.

But SIVs and ABCPaper are not the chief mechanisms that carried and amplified the mortgage shock, according to the writers.

Repo was.

From the paper:

Despite the considerable attention paid to SIVs and other ABCP programs in many narratives of the crisis, SIVs actually had quite a small footprint relative to the size of the money markets as a whole. Rather, a much more important source of short-term funding for mortgage-backed securities  (as well as many other classes of structured financial products) in the years preceding the crisis was the market for repurchase agreements (repos). Repos serve as a collateralized source of shortterm lending and borrowing that began as a way for institutional investors and broker-dealers to earn small returns by lending Treasury and agency securities on a short-term basis. But, as we describe in more detail below, in the years preceding the financial crisis, the repo market evolved slowly, but substantially, to become a major funding source for structured financial products, including some of the more complex and systemically risky instruments (such as senior tranches of ABS CDOs).

As emphasized by Brunnermeier (2009) and Gorton and Metrick (2009), repo funding was a major factor behind the expansion of the shadow banking system prior to the financial crisis and the sudden withdrawal of repo funding—and a pullback in repo transactions with certain  counterparties (including Bear Stearns in March 2008), against specific collateral classes (ABS CDOs), and a rise in repo haircuts played a major role in propagating the subprime/alt-A mortgage shock throughout the financial system.

The report discusses the inadequacies of data collection. For example, broker-dealers and bank holding companies report repurchase transactions to the Securities and Exchange Commisssion, and commercial banks report them to bank regulators, but these volumes are netted, meaning that a financial firm tallies its repo lending and its repo borrowing and reports the difference, when the deals are done with the same counterparty or involve the same type of security as collateral.

In times of normal market functioning, it makes sense for brokers and dealers to net such positions in their financial statements and regulatory reports. However, in times of stress, information about gross positions, and thus about the overall volume of outstanding short-term funding trades, are more relevant for sizing the potential market frictions that could become destabilizing, amplifying and propagating even small disturbances.

As repo volume grew, the quality of the collateral deteriorated, and there was no data collection to expose that decline, the authors note.

Over time, however, the repo collateral pool evolved to include a much wider range of instruments, many of which were not likely to benefit during a flight to quality. According to some estimates, by 2007, the fraction of the more pristine collateral types had fallen to about one-third of the overall funding market. Particularly insidious, of course, were structured products tied to mortgages. Not only were these instruments, it is now clear, less likely than debt issued by the Treasury or agencies to remain liquid during a stress scenario, but, because of the assumptions about default correlations that were fundamental to their design, they were particularly exposed precisely to the sort of systemic risk event that would also negatively affect financial institutions that were funding these instruments on a secured basis.

Though the authors want to see better data collection, they caution that just collecting data based on the last crisis will probably not capture what will be needed for the next crisis. Analysts will need to dig deeper.

For example, the writers note, one reason broker-dealers began using the repurchase market to finance so much of their business in the past two decades was because the lesson they learned from the failure of Drexel Burnham Lambert in 1990 was to avoid unsecured debt.