Thursday , November 14 2019
Home / S. Sumner: Money Illusion / Sahm’s Rule and mini-recessions

Sahm’s Rule and mini-recessions

Summary:
A business cycle forecasting tool developed by Claudia Sahm has recently attracted some media attention.  This is from a WSJ article: Sahm suggested that she got the idea from Jason Furman, who heard about the idea from Doug Elmendorf.  Back in 2011, I had a similar insight in a blog post on mini-recessions (or more specifically the lack of mini-recessions): I searched the postwar data, which starts at 1948 and covers 11 recessions.  During expansions I found only 12 occasions where the unemployment rate rose by more than 0.6%.  In 11 cases the terminal date was during a recession.  In other words, if you see the unemployment rate rise by more than 0.6%, you can be pretty sure we are entering an recession.  The exception was during 1959, when unemployment rose by 0.8%

Topics:
Scott Sumner considers the following as important:

This could be interesting, too:

Menzie Chinn writes Neel Kashkari at Madison: “Kashkari says he takes comfort from yield curve shape”

Tyler Cowen writes Tuesday assorted links

Tyler Cowen writes Television facts of the day

Tyler Cowen writes Monday assorted links

A business cycle forecasting tool developed by Claudia Sahm has recently attracted some media attention.  This is from a WSJ article:

Sahm’s Rule and mini-recessionsSahm suggested that she got the idea from Jason Furman, who heard about the idea from Doug Elmendorf

Back in 2011, I had a similar insight in a blog post on mini-recessions (or more specifically the lack of mini-recessions):

I searched the postwar data, which starts at 1948 and covers 11 recessions.  During expansions I found only 12 occasions where the unemployment rate rose by more than 0.6%.  In 11 cases the terminal date was during a recession.  In other words, if you see the unemployment rate rise by more than 0.6%, you can be pretty sure we are entering an recession.  The exception was during 1959, when unemployment rose by 0.8% during the nationwide steel strike, and then fell right back down a few months later.  That’s not called a recession (and shouldn’t be in my view.)  Oddly, unemployment had risen by exactly 0.6% above the Bush expansion low point by December 2007 (when the current recession began) and by 0.7% by March 2008, and yet many economists didn’t predict a recession until mid-2008, or even later.

What’s my point?  That fluctuations in U of up to 0.6% are generally noise, and don’t necessarily indicate any significant movement in the business cycle.  But anything more almost certainly represents a recession.

Now here’s one of the most striking facts about US business cycles.  When the unemployment rate does rise by more than 0.6%, it keeps going up and up and up.  With the exception of the 1959 steel strike, there are no mini-recessions in the US.  The smallest recession occurred in 1980, when the unemployment rate rose 2.2% above the Carter expansion lows.  That’s a huge gap, almost nothing between 0.6% and 2.2%.

Sahm’s formulation is superior to mine because the month-to-month unemployment rate is noisy. By taking a three-month average in measured unemployment she gets a better view of the underlying trend in actual unemployment. In my 2011 post, I hypothesized that any rise in the actual unemployment rate, no matter how small, was an indicator of recession. Because the data is noisy, I suggested that you’d need more than a 0.6% rise in the measured unemployment rate to be certain that the actual unemployment rate had risen at all:

To understand mini-recessions we first need to understand the monthly unemployment data collected by the Bureau of Labor Statistics.  This data is based on large surveys of households.  It seems relatively “smooth,” rising and falling with the business cycle.  Month to month changes, however, often show movements that seem “too large” by 0.1% to 0.3%, relative to the other underlying macro data available (including the more accurate payroll survey.)  So let’s assume that once and a while the reported unemployment rate is about 0.3% below the actual rate.  And once in a great while this is followed soon after by an unemployment rate that is about 0.3% above the actual rate.  Then if the actual rate didn’t change during that period, the reported rate would rise by about 0.6%.

Sahm says that you need more than a 0.4% rise in the 3-month moving average of measured unemployment.

The FT argued that the Sahm indicator doesn’t really “predict” recessions, as it’s a coincident indicator. But even that is highly useful, as we often don’t know that we are in a recession until 6 to 9 months after it began. The most recent recession began in December 2007, but as late as August 2008 the Fed didn’t even know we were in a recession. By that time, the Sahm rule had been indicating a recession for months.

The Sahm Rule has a deep connection to the mini-recession mystery. The rule works in the US precisely because we never have any mini-recessions (for some unknown and deeply mysterious reason.) It doesn’t always work in foreign countries because they do have mini-recessions.


Tags:

 
 
 
Scott Sumner
Scott B. Sumner is Research Fellow at the Independent Institute, the Director of the Program on Monetary Policy at the Mercatus Center at George Mason University and an economist who teaches at Bentley University in Waltham, Massachusetts. His economics blog, The Money Illusion, popularized the idea of nominal GDP targeting, which says that the Fed should target nominal GDP—i.e., real GDP growth plus the rate of inflation—to better "induce the correct level of business investment".

Leave a Reply

Your email address will not be published. Required fields are marked *