Every now and again, something is filled with so much wrong that it cries out for a response. Much of the time, we are obliged to let such errata pass unnoted, lest fisking what is wrong becomes a round-the-clock job. As the old saw goes, choosing your battles wisely is the way to exhibit wisdom and win the war.
Still, now and again, something demands a response — in this case it’s the reappearance of an old meme that refuse to die: “Uncertainty” as an explanation for some pattern of behavior, especially the reluctance of investors and businesses to make plans for the future.
The latest spasm of uncertainty concerns the presidential election, and the idea that an abundance of said uncertainty in years when there is no incumbent running holds back financial markets and the economy.
Let’s examine some of the evidence cited in support of this idea: Since 1928, in election years with no second-term incumbent running, the Standard & Poor’s 500 Index has fallen an average of 2.8%.
We can assume that this fact is true. But so what? (I don’t even understand why 1928 was picked since you have to go to 1952 to find an open election, with Eisenhower versus Stevenson.) Here’s the real problem, though. With just six elections without an incumbent since World War II, the sample size is way too small to draw any meaningful conclusions. Even if we consider the entire 20th century (there were five open elections before 1928) that’s a data set of just 11 — still a rather small sample set to use in any sort of analysis.
And if we look at the last year of an incumbent’s eight-year term — the only year that averaged negative returns — it’s an even smaller group (1904, 1920, 1960, 1988, 2000 and 2008).
These results are indistinguishable from randomness. I asked Salil Mehta, author of “Statistics Topics,” and former director of analytics for the Troubled Asset Relief Program, about this. His said that there can be a number of problems with drawing predictive conclusions from a small data set. There can be spurious anomalies, issues with causation and the presence of confidence intervals. “Seeing a short-term pattern but for the wrong reason,” he said, is often the result of small samples. Perhaps we can revisit this in a few centuries when we have a more meaningful data set.
But the bigger issue is the uncertainty meme itself, which gets trotted out on a regular basis. We have addressed this before, but we are perhaps due for a revisit.
Uncertainty is a condition that exists when the set of possible future outcomes is unknowable. Note that this is very different from an event which is merely unknown.