A report issued this week by the Federal Reserve Bank of New York debunks the algorithm that has been historically used to track Fed Funds. This is a serious matter, and also throws into question many years of Fed Funds rate statistics published on the Fed’s own website.
According to the report, “What Do We Know about Federal Funds Activity?” by Olivier Armantier and Adam Copeland, “Recent empirical analyses of the federal funds (fed funds) market, historically one of the most important financial markets in the United States, rely on individual transactions inferred indirectly from an algorithm.” A test of the algorithm however was very discouraging. “We estimate the average type I and type II errors from 2007 to 2011 to be 81 percent and 23 percent, respectively. Furthermore, we argue that these errors 1) apply to almost half of the algorithm’s output, 2) introduce systematic biases, and 3) may not subside when the algorithm’s output is aggregated. Our results therefore raise serious concerns about the appropriateness of using the algorithm’s output to study the fed funds market.”
(Editor’s note: we received a notice about this report on Wednesday, October 10. It was online and available here. However, when we went to check something in the report this morning the link had been taken down.)
The algorithm in question uses Fedwire data and infers which pairs of payments are for a Fed Funds transaction. It is the leading algorithm for tracking Fed Funds and results have been published in major academic journals.
The failure of this FRBNY algorithm to accurately track the market gets worse. The authors say, “While our work focuses on the FRBNY algorithm, slightly different versions of this algorithm are used by researchers outside the FRBNY. Economists at the Board of Governors of the Federal Reserve (FRB) use the same proprietary Fedwire data and a similar algorithm to create measures of overnight fed funds activity. We therefore expect the FRB algorithm’s output to suffer from the same large type I and type II errors.”
We interpret these “measures of overnight fed funds activity” to be the data that shows up on the Fed Funds Rate pages. According to a footnote on that page, “The daily effective federal funds rate is a volume-weighted average of rates on trades arranged by major brokers. The effective rate is calculated by the Federal Reserve Bank of New York using data provided by the brokers and is subject to revision.” We may be wrong and invite readers to tell us so, but this is what this looks like to us.
This is serious business; this algorithm provides the underpinning for academic, theoretical, regulatory and practical arguments for why Fed Funds is or isn’t a valid benchmark in financial markets. We ourselves tackled the issue in our August 2012 report, “Repo Indices, Overnight Index Swaps and Other Alternatives to LIBOR.” The Fed report’s authors are saying that the algorithm itself is largely invalid, and there are no accurate data today on the market. This is a dismal state of affairs; we predict that regulators will take action shortly to make banks produce their accurate statistics for market consumption.