By Michael McCracken, Assistant Vice President and Economist
As the U.S. economy’s current expansion continues in its 11th year, there remain concerns about how long it will last. Recessions may be inevitable, but predicting when they will happen is notoriously difficult. The fact that recessions are difficult to forecast is precisely why there is significant literature attempting to build better models that can predict them.
Since a recession is binary (i.e., either you are in one or you are not), these models typically generate forecasts that take the form of a probability. For example, the most recent value of the “Smoothed U.S. Recession Probabilities” available in FRED reports a 2.06% chance of a recession, as seen in the figure below.
In contrast, the most recent value of the “Probability of US Recession” reported by the Federal Reserve Bank of New York is much higher at 25.2%.See the New York Fed’s page “The Yield Curve as a Leading Indicator.”
In the absence of further information, it is easy to wonder what to make of these numbers. Is one of them high while the other low? Are they both low?
To answer these questions, we need a better understanding about what, exactly, these recession probabilities are. Both values use data available through January 2020, but they measure different recession probabilities:
This implies that the former is a recession prediction for a singular, current month, while the latter is a prediction over a span of the next 12 months. In short, the two probabilities are different because the models are designed to ask different questions regarding recessions.
Now that we have a clearer understanding of what these models are forecasting, we can return to the question of whether the most recent recession probabilities are high or low.
For the value in FRED, it is clear that 2.06% indicates a very low probability of having been in a recession in December. To see why, note that if we were to pick a month at random over the last 60 years, the probability of selecting a month in recession is roughly 13%, or six times higher than the current value.
For the value currently being reported by the New York Fed, it makes more sense to determine the percentage of historical 12-month spans that contain at least one month in which the U.S. was in a recession. Doing so, we find that if we were to pick a 12-month period at random over the last 60 years, the probability of selecting a span with at least one month in recession is roughly 25%, a value that is perfectly aligned with what is currently being reported by the New York Fed.
In summation, it is very clear that the U.S. economy is currently not in a recession. However, looking forward over the next 12 months, the view is cloudier with recession probabilities close to the historical average.
1 See the New York Fed’s page “The Yield Curve as a Leading Indicator.”
2 Keep in mind that due to data release lags, many macroeconomic series associated with December 2019 are not released to the public until some point in January 2020.