Recessions are notoriously difficult to predict. Like timing the stock market or predicting next year's weather, the ability to time the next recession is generally held to exceed the limits of human -- or machine -- intelligence.
There are two good reasons economists and portfolio managers rarely try to call the business cycle. First, the cycle is vulnerable to shocks -- such as wars, weather or new technologies in which chance -- pure dumb luck -- would appear to play the dominant role. Second, the cycle is hostage to the actions of people, namely central bankers and politicians who can prevent a recession -- or cause one -- by managing or mismanaging the money supply and government spending.
This does not mean, however, that trying to predict the business cycle is fruitless. Rather, like the weather and stock market it is susceptible to probabilistic analysis.
For example, we know that in the U.S. since 1854, over the course of 34 expansions, the duration of recoveries -- the period from trough to peak -- has averaged 39 months with a median of 30 months. In modern times, recessions have been rarer and recoveries longer probably due to improved information, a better understanding of the economy on the part of policymakers and structural changes such as the growth of services and reduction of manufacturing inventories. Since 1961, recoveries have lasted about eight years on average. The longest recovery on record is the Bush-Clinton era one that ran from 1991 to 2000. It was followed by a seven-year recovery from 2001 to 2007. We are currently six years into a recovery from the Great Recession.
If the new norm is eight years, does this mean that our current recovery has two more years to go? Could it be said of the business cycle as with sports teams or a win or loss that we are due for a recession?
In a world of pure chance, the cards have no memories, nothing is ever due and the past should not prejudice the future.
But in the real world where chance interacts with causality, each passing year does weigh on the next. In the same way that actuaries predicting lifespan must acknowledge that age will influence years left on earth, the age of a recovery also affects its future. A person who has lived 70 years has a longer life expectancy than a newborn, but a far lower expected remaining life as a result of the effects of age. Similarly, we would expect a recovery in its sixth year to have less runway than at its outset.
In the same way that actuaries don't know exactly what may befall a person applying for life insurance but do know that risk rises with age, so any recovery is influenced by its age. Actuaries can precisely estimate average years remaining for a pool of customers at a certain age and improve their predictive accuracy if they have access to variable such as smoking. It would certainly be nice to be able to predict the time remaining for a recovery in its sixth year.
But in the case of the business cycle in the United States, there are simply not enough observations to attempt this analysis. After all, there have only been five recoveries since 1900 that have lasted six years. (For what its worth, the average time remaining was 2.25 years for this tiny group suggesting a recession in 2017.) It turns out there is one variable however that has repeated itself enough times to produce enough observations to yield statistically significant results. That is the presidential election every four years.
In an election year, there is an obvious disincentive to those in office to have a recession. However, the year after the election the President has a political incentive to get the recession over with. And, indeed, this is what the evidence suggests happens -- on average.
Since 1792, a period of 222 years encompassing 54 election cycles, there have been 47 recessions. A suprising 19 of them have begun during the first year following a presidential election far more than chance would predict. In contrast, only 9 have occurred in the election year itself, also a statistically significant variation. This trend grows even stronger if we move forward in time at the expense of observations.
Since 1928, a whopping nine of 14 recessions have begun in the year following a presidential election, far more than then the three that should have occurred randomly. Since 1969, four of seven (or four of six if one excludes the brief 1980 Carter recession) have occurred in the year following a Presidential election more than twice what chance would predict.
There is a debate among economists about the role played by policymakers in recessions. Rudy Dornbush famously remarked that none of the post-war expansions died in bed of natural causes: they were all murdered by the Fed. Others attribute some recessions to shocks. It is also true that in the US, the President cannot tell the Federal Reserve or Congress what to do. Nonetheless, A fair reading of the data suggests that the general pattern has been for politicians to spend money in an election year and for the Fed to tighten the following year, triggering a recession, giving rise to the pattern observable in the data.
So when will the next recession occur? The odds say it will occur in 2017 following the next Presidential election.
This analysis does not include confounding variables of which there are certainly many. But it does suggest that 2017 is a reasonably good guess for the year of the next recession.