Dueling "recession probability" indicators

In chronological order...

  •  1 March 2019: James Hamilton, The long expansion continues

The Bureau of Economic Analysis announced yesterday that U.S. real GDP grew at a 2.6% annual rate in the fourth quarter of 2018. That’s below the 3.1% average for the U.S. economy over the last 70 years, but better than the 2.2% average rate since the recovery from the Great Recession began in 2009:Q3...

The solid growth numbers kept the Econbrowser Recession Indicator Index at 1.5%, among the lowest levels we ever see. That means the U.S. economic expansion has now been under way for 9-1/2 years, 2 quarters shy of the longest expansion on record (1991:Q2-2001:Q1).

GDP-based recession indicator index.
The plotted value for each date is based solely on information as it would have been publicly available and reported as of one quarter after the indicated date, with 2018:Q3 the last date shown on the graph. Shaded regions represent the NBER’s dates for recessions, which dates were not used in any way in constructing the index, and which were sometimes not reported until two years after the date.

  •  24 March 2019: Tim Duy's Fed Watch, Fed Needs to Get With The Program

Everyone has their pet recession indicator; many are probability models based on some combination of yield spreads and other leading indicators. Most will be raising red flags like this estimate of the probability of recession in six months based on the 10s2s and 10s3mo spreads and initial unemployment claims:

... The risk of recession has risen to levels that demand attention from the Federal Reserve. In the two cases of similar spikes in the 1990s, a recession was avoided by the rapid response of the Fed in the form of rate cuts. The times that response was lacking, a recession followed.

So now I switch from analyst to commentator: The above leads me to the conclusion that the Fed needs to get with the program and cut rates sooner than later if they want to extend this expansion.

Once again I took Duy's graph, erased the background, and used it as an overlay. This time I didn't erase the recession bars on his graph. I used his recession bars to resize and position his graph over Hamilton's.

This is probably not the must useful graph I ever made, but I wanted to see it:

Graph #3: Hamilton's (black) and Duy's (blue) Recession Probabilities

Duy's three light gray recession bars hide Hamilton's recession indicator during the recessions. But you can see it on the first graph. And anyway, I'm more interested in recession probabilities when we're not already in recession.

The interesting piece of the graph is the right end, where Duy indicates recession and Hamilton does not. Hamilton's data ends before Duy's; as Hamilton notes, "The GDP data were a month late being released due to the government shutdown." Another month or two, or three, we'll see what happens on Hamilton's graph.

My first thought was that when the data's available, Hamilton's graph will show an increasing recession probability because of the interest rate spread, as Duy's does. But Hamilton's index isn't based on the interest rate spread. As described in The Econbrowser Recession Indicator Index, it is based on GDP growth rates and the probability (based on experience through 2005) that a given GDP growth rate occurs during a recession or during an expansion.

It's a different and, as Hamilton says, an objective approach. And it has nothing do do with interest rate spreads. So there is a chance Hamilton's graph will not spike upward as Duy's does. Ooh, this just got interesting.