The slope of the yield curve is likely to become an increasingly bearish indicator as this year progresses, and recent changes in the calculation of the index of leading economic indicators should not be interpreted as in any way denying that fact.
One of the best-established predictive relations in macroeconomics is the observation that when
the difference in yields between long- and short-term interest rates is low or negative, future GDP growth
tends to be slow or negative. Pretty much any measure of the size of the gap seems to support that
conclusion. Hibiki Ichiue, a student of mine who successfully defended his Ph.D. dissertation
yesterday morning (congratulations!) found that even the 3-year minus 2-year rate does quite well. Most
academic studies have used the 10-year minus 3-month Treasury rates plotted in the top graph on the
right. This spread has averaged 140 basis points over the last 50 years, but it narrowed to a
value much less than that prior to each of the last eight recessions (shown as shaded regions on
the graph), and actually became negative in six episodes, five of which led to recessions.
The spread has dropped dramatically over the last year and is now less than 90 basis points.
The natural question to ask in this situation is, how low does the spread have to go before we
start to get worried about where the economy is headed?
One way you might try to answer that question is with statistical regression analysis, in which
you try to forecast economic growth rates using lagged growth rates and lagged values for the
spread, and see if there’s some cutoff point for the spread above which movements in the spread
don’t seem to make much difference for your forecast. Time-series specifications that allow for
such effects are called threshold autoregressions or smooth transition autoregressions. Empirical
analyses using these methods have found that there does indeed seem to be such a cutoff, but it is
pretty high. John Galbraith and Greg Tkacz, in a study of GDP growth rates published in the Journal of International Money and Finance in
2000, estimated the threshold to be a little under 200 basis points. Ivan Paya, Ioannis
Venetis, and David Peel, in a study of industrial production growth published in the Journal of Forecasting in
2004, found a threshold of around 140 basis points. In other words, according to these
studies, whenever the 10-year minus 3-month spread is below 140 basis points, the lower it gets, the more pessimistic your forecast should be.
The usefulness of the yield spread for economic forecasting led the Conference Board to include it in their popular index
of leading economic indicators in 1996– they use the gap between the 10-year rate and the
overnight fed funds rate. However, beginning with this July’s release, the Conference Board
revised the method by which the index of leading economic indicators is calculated. Previously, whenever the yield spread became smaller, it contributed negatively to the index of leading indicators. Beginning with the July release, the spread will contribute negatively to the index only if the spread itself is actually negative.
The reasons for this change can be found here and here.
The key issue has to do with the way in which the Conference Board combines the information from a
variety of different indicators to form a single index that is growing over time. Some of the individual indicators that make up the index are themselves usually growing (such as stock prices) and others (like the spread) have no clear trend. The predictive relation referred to in the academic studies above is one between the level of the spread and the growth of output. Because the Conference Board relies on changes in the index as a predictor of future output growth, its previous method amounted to imposing a relation between the change in the spread and the growth of output. The Conference Board decided to fix that problem by in effect using the level of the spread itself as something that goes into determining the change in the index. Thus, under the new method, when the spread is positive, that tends to make the index go up, and when the spread is negative, that tends to make the index go down. Because the Conference Board interprets an increase in the index as a favorable signal, this has the practical effect that low but positive values of the spread no longer register as a bearish indicator.
Ritholtz was unreserved in his criticism of the Conference Board for this adjustment, I can see what the Board was trying to do, given the mechanics of how they are trying to combine the information from variables that are trending and nontrending. But none
of that should obscure the reality that, as a statistical proposition, the lower the spread becomes at this
point, the more pessimistic your forecast of output growth should be. This pessimism will not be reflected in the index of leading economic indicators until the spread turns negative, but that does not in any way imply that we should be unconcerned about the implications of low but positive values for the spread.
So where does that leave us right now? The Fed’s expected rate hike next week will narrow the
spread further. My nervousness about that is allayed somewhat by last week’s favorable
economic data. I’ll be more nervous if the Fed raises rates again at the following FOMC
meeting, and even more nervous if they raise rates again in November.
Looking at the
Macroblog projections of future fed funds target changes, I pretty much expect to be nervous
for the rest of the year.