Stock market falls and fatal road accidents

It is pretty well established that distracted driving is dangerous - according to the NHTSA, it claimed nearly 3000 lives in the US in 2018. There are many things that can distract a driver - mobile phones, eating or drinking, or talking to passengers. But how about emotions? The Reduce the Risk website notes that "83% of drivers think about something other than their driving when behind the wheel", and that could be a distraction.

So, I was interested to see this recent paper by Corrado Giulietti (University of Southampton), Mirco Tonin (Free University of Bozen-Bolzano), and Michael Vlassopoulos (University of Southampton), published in the Journal of Health Economics (ungated earlier version here). In the paper, Giulietti et al. look at the relationship between stock market returns and fatal car accidents in the U.S. Specifically, they use data on daily stock returns (S&P500 Index, and some other measures) and daily numbers of fatal vehicle accidents from the Fatality Analysis Reporting System (see here) over the period from 1990 to 2015. They find that:

...a one standard deviation reduction in daily stock market returns increases the number of fatal accidents by 0.6% (that is, by 0.23 accidents over an average of 37.4 daily accidents occurring after the stock market opens).

The relationship is statistically significant, but notice that the size of the effect is pretty small. The standard deviation of the daily stock market returns is 1.04 percent, with an average of 0.04 percent. So, if the stock market loses one percent on a particular day, these results suggest that there would be 0.23 more fatal road accidents that day in the U.S. 

Of course, that attributes a causal interpretation to these results, which are essentially correlations. Giulietti et al. do undertake some interesting robustness checks and falsification tests on their results, as they explain:

In the first set of tests, we exploit the timing of accidents. If the relationship that we find is due to uncontrolled-for events affecting both stock market valuation and driving behavior, we would also expect the relationship to be present for accidents happening before the opening of the stock market. However, we find no relationship in this part of the day, thus providing support for a causal interpretation of the link between stock market returns and accidents. With a similar logic, we show that there is no relationship between car accidents and lead stock market returns...

 In the second set of falsification tests, we pursue multiple approaches to compare the effect of the stock market on groups of drivers with different likelihoods of owning stocks... One approach to isolate drivers who are unlikely to hold stocks is to zoom in on accidents involving only people aged 25 or under. For this group, we do not find a statistically significant relationship between accidents and stock market performance, while we see the effect on accidents involving at least one driver older than 25... In another approach, we exploit differences in the geographical distribution of income, with the idea that people with a higher income are more likely to invest in the stock market. We consider average income in both the county of the accident and the drivers’ zip code. In both cases, we find no relationship between stock market and accidents for the lower tercile of income, while there is a strong significant relationship for the upper tercile.

Those results provide some confidence that the results aren't spurious, but still fall short of demonstrating definitive causality (and falls short of convincing Andrew Gelman as well). Underlying moods affect the stock market ("animal spirits", as John Maynard Keynes termed them), as well as affecting driving. If moods are different in higher income than lower income people, and higher income people's moods affect stock prices more than lower income people's moods, then the falsification tests based on income differences are not valid.

Probably one aspect that should be concerning is that the results appear to hold for stock market falls, but not for rises. Are people likely to be more distracted on bad stock market days than on good stock market days? I think the mechanism needs some further explanation in this respect, and until we have that, the jury is certainly out on the causal relationship between stock market returns and fatal accidents.

[HT: Marginal Revolution, last year]