Fed funds rate and probability of a recession

Question

From an article today on Investopedia: Bonds Signal up to 60% Chance of Recession

Earlier in 2019, the Federal Reserve announced a pause in its program of interest rate hikes, re-energizing the stock market in the process. Now the money market appears to be anticipating a cut in the federal funds rate before 2019 is over, which, in turn, suggests that the Fed is increasingly more worried about preventing a recession than combatting inflation. Deutsche Bank projects that the fed funds rate will end 2019 at 2.15%, implying a 60% chance of recession within the next 12 months, Barron's reports.

INDICATORS OF INCREASED RECESSIONARY RISK

• The Fed now seems more concerned with recession than inflation
• The money market anticipates a federal funds rate cut in 2019
• This implies 60% odds of a recession starting in the next 12 months
• Longer-term yields imply 28% odds of recession with next 12 months

How did they calculate the odds of a recession from the fed fund rates?

Answer 1

Calculating the odds of a recession based on the fed funds rate most likely used a dartboard. Or maybe some colorful dice.

I kid. Most likely it was a computational model, which is totally different from dartboards or dice because people won't pay institutional research firms who claim to use methods that are transparently easy to understand to try to predict macroeconomic market timing.

So I want to be clear - what you are reading is a sales pitch for Deutsche Bank, which is a sales pitch provided by Deutsche Bank Research, wrapped inside a sales pitch for Ned Davis Research, wrapped inside a sales pitch for Barron's, delivered as a sales pitch for Investopedia's blog.

It does not say how they calculate such a probability, because that is what Deutsche Bank would like you to pay them for. They have teams of highly educated, highly skilled, highly experienced analysts (who I'm sure are wonderful people) churning out weighted random numbers like this - they can't just give away the secret sauce!

But at a high level, I can tell you how they calculated it:

First you build a statistical model that takes certain variables and assumes or attempts to estimate certain relationships between the variables, which will likely include the fed rate and dozens, hundreds, or thousands of other variables. Then you take some source of data about market performance and fed rates, pulled from somewhere (probably bought it from a vendor, but maybe provided by an internal team). But you can't just put that in the model as-is, so you make a variety of ultimately arbitrary decisions on how to process the data, so you do that and feed it to the model. The model puts out tons of outputs, but no single number, so you have to either cherry pick out what number is interesting or try to shove all the numbers together somehow. If you do a good job you have a number of competing models and processes, but that only means you have even more numbers coming out the other side, so you have to cherry pick one or shove them together again.

The reporting team, or someone's boss, wants a number, such as the chance of the market going into recession in the next 12 months. They don't want a distribution, they don't want a confidence interval, they don't want to hear about your assumptions or your process or any of your mumbo jumbo about information theoretic predictive accuracy, pareto efficiency of estimators, or your concern about autocorrelation in the Kullback–Leibler divergence of the residuals - they want a damned number to put in the sales report, alright?

Fine, 60%.

It's a nice number, no one is going to say you are insane for making it so high, and they won't say you are a Pollyanna who thinks a recession can't happen because it is so low. It also isn't 50%, because if you said that people would point out that's just saying bear vs bull is the result of a coin-flip! What are you, some kind of amateur?

Yeah, 60%. Now that's a number we can work with. But wait, why did we come up with this number? Don't tell me about all the variables you included in the model, so help me if you try to say something about "polynomial" or "multivariate" I'll slap the...right, bonds, let's go with that. Bonds are signalling, there we go - wrap that up and you've got a nice headline!

Bottom-line: news fit only to line the bottom of bird cages. It has no real informational content. Timing the market has long proven to be a fools errand - badly reported, non-transparent, baloney statistics don't change that.

Answer 2

There is actually a lot more behind this number than another answer suggests.

The phrase,

...the money market appears to be anticipating a cut in the federal funds rate...

can refer to either the futures or OIS swap market. Both markets are something frequently used in the market to infer probailities of interest rate hikes, see for example:

• Bloomberg WIRP tool
• CME FedWatch Tool

For stock futures, the price is simply a function of the spot price, interest rates and dividends and as such contains no information that the spot market does not contain (it's a simple no arbirtage argument).

On the other hand, interest rate futures cannot be calculated from another. There is interesting information inherent in interest rate future prices because the future for March 2023 is completely different from Feb 2024 and price setting relies on market expectations.

Another application of the same logic is to use the shape of the yield curve (market parlance for the gap between yields on longer- and shorter-dated instruments) to imply recession probabilities. When the gap between short and long yields narrows, the jargon is that the yield curve “flattens.” This commands a lot of attention for at least two reasons:

1 ) yield curve inversions tend to be a great warning that a recession is coming in a matter of months

2 ) inverted curve tends to make life very difficult for banks, who traditionally make their money by borrowing short term (through deposits) at low rates, and lending longer term at higher rates, and pocketing the difference (classic maturity transformation).

Point 1 ) is what matters here. Formal approaches exists since at least Estrella and Mishkin 1996 but the information content of the yield curve were discussed long before that seminal paper. Generally, one could look at treasuries, fed funds or swaps. Most research on United States business cycles has relied on interest rates for U.S. Treasury securities because the data for many maturities are available continuously from the 1950s to the present in a consistent format.

Some interesting papers:

• The Yield Curve as a Leading Indicator - Fed of New York
• The Yield Curve as a Predictor of U.S. Recessions
• Yield Curve Prediction For The Strategic Investor - ECB Working Paper Series
• Economic Forecasts with the Yield Curve | San Francisco Fed (frbsf.org)
• Information in the Yield Curve about Future Recessions | San Francisco Fed (frbsf.org)

Obviously, this is not a perfect tool, which Michael D. Bauer and Thomas M. Mertens from the Fed of San Francisco describe like this:

Furthermore, when interpreting the yield curve evidence, one should keep in mind the adage “correlation is not causation.” Specifically, the predictive relationship of the term spread does not tell us much about the fundamental causes of recessions or even the direction of causation. On the one hand, yield curve inversions could cause future recessions because short-term rates are elevated and tight monetary policy is slowing down the economy. On the other hand, investors’ expectations of a future economic downturn could cause strong demand for safe, long-term Treasury bonds, pushing down long-term rates and thus causing an inversion of the yield curve. Historically, the causation may well have gone both ways. Great caution is therefore warranted in interpreting the predictive evidence.”

Now, there is an additional effect to consider, namely the FED fighting inflation. So as long as long term yields rise (generally seen as implying stronger growth ahead), one need not be too worried about a recession, even if the curve is flat or inverted. Applying a model fitted to past data may misinterpret the current yield curve shape.

Nonetheless, it is a lot less like we just need a number and claim it is bonds who are signalling it because it is easy and makes a nice headline with no information content that simply makes a good sales pitch.