When Does a Recession Become a Depression?
November 24, 2008
Recessions and depressions are each parts of the business cycle when demand, production, and employment are contracting. Depressions are considerably more severe. Economists do not agree on objective quantifiable criteria for deciding when a recession is bad enough to constitute a depression. The United States has experienced 32 business cycle downturns since 1857. Six of them have gone down in history with the depression label, and two others were called “panics,” implying an event that’s worse than a recession but not quite bad enough to be a depression. It’s been 70 years since the last U.S. depression, a credit to greater understanding by economists, central bankers, and politicians about what transforms a recession into a depression and how to minimize the risk of that happening.
I submit two criteria to be used for deciding if a downturn warrants being called a depression. The first characteristic involves the peak to trough magnitude of the contraction, and the other factor is the length of the event. If output declines by 10% or more, that ought to qualify. The deepest post-WW2 recessions only reached about 3%, while the Great Depression saw output plunge about 30%. If a downturn lasts at least two years, a drop in production of somewhat less than 10%, say 6%, probably ought to be considered a depression or panic. The six depressions and two panics had an average length of 30.5 months, but the result was skewed upward by the depressions of October 1873 – March 1879 and August 1929 – March 1933. The average downturn for all the recessions since 1857 was 12.5 months.
If the present downturn began in April of this year, activity would have to be still dropping in 2Q10 to be considered a depression under the above criteria on length alone. Reasons why the possibility of either such a long uninterrupted downturn or a slump of at least 10% include 1) the failure of all efforts so far to quell the problem, 2) the potential for a considerably deeper decline in housing prices and financial asset prices, 3) feedback into consumption and investment from a massive loss of wealth and the resultant effort of households to rebuild savings, 4) lower interest rates and direct cash injections stimulating money growth but not credit growth (Keynes’ infamous liquidity trap), and 5) the temptation for governments to increase trade barriers as they did in the 1930’s.
A different way to measure the severity of economic down-cycles is by their deviation below the trend level of economic activity. Similar to the duration and peak-to-trough magnitude, no economic downturn since the 1930’s has come close to equaling the shortfall from trend that occurred in the Great Depression or in other depressions. It’s been 15 months since the first tremors of the credit crisis. It’s still too early to project the totality of the present cycle.
ShareThis
nowhere in your submission is there any mention of employment metrics. i wondered whether that might be implied in the other metrics you have discussed…but in each case there was significant elasticity in the relationship between employment and the metric. but “peak to trough of contraction” could be a couple of different things. do you mean GDP? the DJIA or the S&P? i assume since you mention “output” that you are talking about GDP, and not the equity markets. or is it some combination of a number of factors you find significant, and not just GDP? i think one could understate a recession if only looked at in terms of one or two outputs. likewise, one could overstate a recession. presently, if i were to look at the current punditry and the macroeconomic data, i would almost want to assume unemployment would be MUCH higher than was recently reported. to my amateur eye, unemployment is conspicuously low — or perhaps we have just seen the beginning of a long round corporate retrenchments, with Citi being among the first to bolt from the herd. or perhaps we will return to the “underemployment” of the 80s, with its own damaging effects on our economic structures — some of which have lasted to today. in any case, given the data, my feeling is unemployment is under-reported…not the 15% or 20% or 40% I associate with a depression…but higher than 7%. when i was taking macroecon courses back in 1991/2, i vaguely remember that full employment was considered 7%! but what proof do i have that it is under-reported? should i even trust an expert econometrician at this point? i am not looking forward to seeing our place on the UNDP’s human development index in the next ten to fifteen years. but i’m strangely confident that i’ll be okay. is that weird?
Did you mean that the the results of the average length of a depression was skewed upward by the depression of October 1873 – March 1979 or March 1879?
Stephanie, Thanks for catching this typo, which I just fixed. It was a longer depression than the one in the 1930’s but not more than a 100 years long.
My remark about peak to trough was referring to real GDP. You make good points here. The determination of start and end dates to the business cycle is based upon several indicators, not one or two. GDP is a good proxy, however. It’s the rare recession that doesn’t encompass at least two consecutive quarters of falling GDP, and there had never been a period with so many consecutive months of net falling employment that was not associated with a recession. I too am suspicious that the unemployment is not higher. It peaked at 9.0% in the mid-1970’s recession and at 10.8% in the early 1980’s. Inventory control was not as good back then, accounting for why unemployment might peak at a lower level now for any downturn that is equal in severity to one a generation ago, but I would not expect the disparity to be as wide as 2 to 4 percentage points due to that factor.