Looking at the Fed's data release through Needle

Written by Chuck McCallum Friday, 10 December 2010 14:44

On December 1, the Federal Reserve posted details of transactions it undertook to stabilize markets during the financial crisis. The data is available as a series of spreadsheets, and, for the most part, spreadsheets are a good way to organize this data. There is one exception.

With the financial crisis in full swing, the volume of withdrawals was forcing mutual funds to sell large volumes of commercial paper to obtain cash: this in turned forced down prices on commercial paper, putting mutual funds in an even worse spot, and making it likely that withdrawals would only accelerate. Seeing the problem, the Federal Reserve instituted AMLF (Asset-Backed Commercial Paper Money Market Mutual Fund Liquidity Facility). The Fed would lend banks money, with the requirement that the collateral for the loan be commercial paper purchased from the mutual funds. This supported the market for commercial paper, which, in turn, supported the mutual funds.

The first few columns of the Fed's spreadsheet are reasonable: date of the loan, borrower, amount, etc.; Then it gets complicated: Each loan might be supported by multiple pieces of collateral, so the spreadsheet repeats the same set of four columns 29 times, because in one instance there were 29 distinct pieces of collateral behind a loan. The data in this form really isn't usable.

Enter Needle. I first defined a domain with a one-to-many relationship between loans and collateral, and then uploaded and tagged the spreadsheet with the elements of that domain. We can look at the data to confirm that the one-to-many relationship has been captured, and then build on that to see, for example, the volume of lending over time.