Learning Curves: Identifying a Recession

Includes: DIA, QQQ, SPY
by: Ira Artman

This post will do three things:

  1. Review the National Activity Index’s (“NAI”) ability to identify the starting date of recent recessions;
  2. Discuss how the yield curve can predict recessions; and
  3. Show that if a recession did begin in the fall of 2007 (as suggested by the NAI) , then it was predicted by the shape of the yield curve in  the fall of 2006.

National Activity Index ("NAI")

On October 25th, The New York Times’ Floyd Norris examined recent performance of the NAI in his “Off The Charts” column. The news is not pretty. Maintained by the Federal Reserve Bank of Chicago, the grim index data through September 2008 suggests that we are in a recession and that this recession began sometime between October 2007 and February 2008.

Below is a copy of the chart that accompanied the Times’ article (click to enlarge images):

Source: F. Norris, New York Times - Off the Charts: More Evidence That the Recession Has Already Arrived, 25 Oct 2008

As Norris notes, the NAI reached recession levels in Dec 2007, and virtually

every time it has gone below the recession level a recession has occurred, … starting … almost always … within two months of the indicator reaching that level.

The precise date of the recession will be determined by the NBER, but their determination typically lags the actual timing of the recession by many months.

Yield Curve as a Predictor of Recessions

If, as Norris suggests, the recession began in the fall or winter of 2007/2008, then this supports the view that the slope of the US Treasury yield curve can predict recessions.

More specifically, the difference between the UST 10 Year Yield and the UST 3 Month Yield can be used to predict recessions one year in advance.

Yield curves are normally “upward sloping”, i.e. long term rates are higher than short term rates. When this “normal” relationship is disturbed, i.e., long term rates are the same or lower than short term rates, this usually means that the Fed is trying to slow down the economy with restrictive monetary policy, and the real economy usually “obeys” (for want of a better term), but it takes about a year for this to occur.

The Federal Reserve Bank of New York researched this topic in depth, and presented major findings in the form of a FAQ - Frequently Asked Questions.

The FAQ Page includes pdf copies of two research papers (see links below) that are relatively easy to read. The Fed also provides data and charts.

Since we are probably in a recession, it would be nice if there was a similar tool or convention that could be used to predict the end of a recession.

Unfortunately, there is not, other than the observation that recessions “typically” last a year or so. But just because the NBER determines, after the fact, that a recession has ended, does not mean that the hard times don’t continue well after the recession end-date.

Using data from the Federal Reserve’s own Fed H15 “mini site”, you can easily download the necessary 10 Year and 3 Month yields, and with a little manipulation you can build own “recession predictor”, based on the relationships described in the Fed papers - or you can use the Fed’s tables and charts on the FAQ site, described above (If anyone has any questions about this, contact me a note, and I’ll walk you through it) .

If you use US Treasury yield data, through year-end 2006, you can then (based on relationships in the Fed papers, see below) predict the probability of recession through year-end 2007, i.e. one year later.   The chart looks like this, (I’ve also overlayed “recession bars” for the start and end of recent recessions):

Source: Author’s calculations, based on NBER & Federal Reserve data, and Federal Reserve publications.

Current Cycle

To my eye, it seems pretty clear that the year-end 2006 data suggested that a recession would begin by year end 2007, and that this is completely consistent with the “NAI” timing reviewed by Norris. That’s it for today.


Disclosure: none