Real Interest Rates are predicting an upcoming recession

Best to read this first. Then read this.

Also remember that while recessions can occur without negative real interest rates, whenever negative real interest rates do occur, they are always followed by an eventual recession.

I played around with my spreadsheet even further and averaged out the real interest rate (10 year bond rate minus inflation) over a three month period to see what that would achieve. The results seem far more promising than merely measuring monthly results:

July 1954 - May 1961

Fairly clear here in the fifties. Our first result is a negative reading measured in March 1957 and lasts for three months. A recession follows in October of the same year. A further deterioration in the real interest rate occurs during the recession and it is possible that this may have prolonged the recession already in swing.

May 1961 - May 1971

Nothing here in the sixties to help us. The Beatles were probably responsible.

May 1971 - May 1981

Two huge results for the disco decade. Our second negative result in October 1973 leads to a recession in January 1974. There is a long period of recession and negative interest rates that lasts until October 1975. Rates remain positive for a while but return to negative again in November 1978, which is our third result. A long period of negative interest rates follows before a recession hits in April 1980. As with the previous recession, as soon as the rates return to positive the recession is over.

May 1981 - May 1991

Nothing here in the eighties to help us.

May 1991 - May 2001

The nineties doesn't help us either.

May 2001 - May 2011

The 2000s provides us with the fourth and last result. Rates turned negative in January 2008 and a recession follows in April.

2005 does provide us with a near negative result: 0.06% in October that is most likely associated with Hurricane Katrina (the monthly inflation increase in September 2005 was the 5th highest on record). In fact it is this 2005 result which forced me to reassess my study two weeks ago on real interest rates based upon the 10 year bond rate (which is not averaged out over three months) and caused me to predict a recession in 2012 Q4.

From the four results, I have deduced the following information:
  • Once the results turn negative, a recession occurs, on average, 8½ months later.
  • The median is 6 months.
  • Results vary between 4 months and 18 months.
  • The highest unemployment rate during the recession is, on average, 1.8 times the unemployment rate of the month when real interest rates turn negative.
  • The lowest increase is 1.32 times; the highest increase is 2.03 times

As you can see from the final graph, above, real interest rates have plummeted in recent times. Here is a graph encompassing the past three years:

The final result - May 2011 - has a reading of 0.26%. This is very close to a negative reading and, therefore, another recession.

In order for a recession to be averted, ten year bond rates must increase or annual inflation must drop. Playing around on my spreadsheet shows me that if the Ten Year Bond rate is 3 basis points lower than the annual inflation rate (eg Bond rate of 3.17%; Inflation rate of 3.2%), then US real interest rates (10 year bond rate minus annual inflation, averaged over three months) would sit at 0.01%, still a positive result. Any result of 4 basis points lower or more would end up giving an negative result.

What are the chances of a negative result for June? Fairly high. 10 Year Bond Rates are, according to my stock ticker, at 3.04%. If bond rates stay at May's result (3.17%), then the inflation index would have to read 223.74 for a positive June result. This would imply a 0.5% drop in prices from the previous month.

Realistically, therefore, we are looking at a negative June result. Which would mean the following:

  • A recession starting between 2011 Q4 and 2012 Q4, with 2012 Q1 the most likely.
  • If unemployment remains at 9.1% in June, then the new unemployment peak during the upcoming recession will be between 12.0% and 18.5%, with 16.7% being the most likely result.

Of course I need to point out that all the information I have extrapolated is based upon four results of negative real interest rates occurring between 1957 and 2008. It may be that this time around things might be different: the recession may take a much longer time to occur; the rise in unemployment may be lower than anything beforehand; the length of the recession may only be short. Nevertheless I do believe that the data is on my side and that, barring any miraculous June result that would turn conditions around, another recession is highly likely to occur, with an unemployment peak higher than any other postwar period.

Sources and methodology

For negative real interest rates:

St Louis Fed: 10 Year Bond Rate GS10
St Louis Fed: Inflation Index CPIAUCSL

For Recession:

St Louis Fed: Real GDP GDPC1
St Louis Fed: Population POP

I define a recession as an annual decline in real GDP per capita.

Here's a screenshot of part of my spreadsheet to help make sense of it all (my methodology)

1 comment:

New Deal democrat said...

Great post, but this metric doesn't work for pre-1954 interest rates.

You are using the 10 year treasury as a proxy for long term rates, but of course that only goes back to the early 1950s.

To go back further, you need to use series LTGOVTBD .

I tested LTGOVTBD over the post 1954 era and it almost exactly tracks the 10 year bond series, and makes the same predictions, so we know it is a good proxy.

Problem: from the early 1920s until 1954, you get almost the exact OPPOSITE result as in the more modern era. Negative real long term rates occur during most of the New Deal, WW2, and immediate post-war era. In fact, they are at their trough at about mid-expansion. And these were the biggest expansions of the entire 20th century.

I think you are on to something, though. You might want to play with the direction of inflation, or the direction of yields, or else measuring current rates against the immediately previous 10 year average.

Still, a very nice job.