Don’t Drive On Tax Day! You Might Die

Tax day crashesThis is it. The day to traipse to the post office and mail in your income tax papers. In farm lingo, this is the time swill the pigs. To feed the beasts. To oil the machine. To encourage the greedy sons-of-you-know-whats. To unwilling fork over what the politicians think belongs to them.

Unless you’re one of the lucky few—and by “few” I mean half of all Americans—who don’t have to pay federal income taxes, today is the day you’ll have to donate over a large chunk of your income to insatiable Washington bureaucrats. This can lead to a certain amount of “stress.” And that modern malady can cause you to drive distractedly. Which in turn can cause you to die in a bloody, fiery crash.

Or so says Donald Redelmeier and Christopher Yarnell who published a peer-reviewed research letter in JAMA which claims to have statistical proof that traffic accidents increase on Tax Day.

Taxes, our pair says, “influence the long-term health of the economy,” but how do taxes affect the health of people?

We investigated the number of individuals involved in a fatal road crash on tax day under the prespecified hypothesis that stressful deadlines might increase the risk of road trauma by impairing drivers or by compromising surrounding individuals from making compensatory adjustments.

They looked at each Tax Day from 1980 to 2009 (30 years; though the data extend to 1975; why leave it out?), and the two days a week before and after Tax Day, and counted the number of fatal crashes, defined as at least one person crapping out because of and within 30 days of the crash. There were 6,738 fatal crashes on tax day and 12,758 fatal crashes on the other two days. Conclusion?

Comparisons of tax days with control days yielded an odds ratio of 1.06 (95% CI, 1.03-1.10; P < .001), equivalent to an absolute increase of 404 individuals in fatal road crashes on tax days over the study interval or about 13 individuals during the average tax day.

So if Redelmeier is right, driving to the post office in a state of high dudgeon on Tax Day kills about 13 Americans each year. As the good doctor said in a news report on this study, “Who looks to tax day in a joyful way? Even those who are getting refunds are stressed. Refunds are never as big as they thought and there is always the chance they could be reviewed at a later date.”

That 1.06 appears to arise from ( 6738/30 ) / ( 12758/60 ), incidentally. Ignore the p-value, because with sample sizes this large publishable p-values are free for the asking. Focus instead on the reasons our researchers gave:

One explanation is that stressful deadlines distract drivers and contribute to human error (a national poll suggested that tax day was the second most stressful day in 2011)…Driving patterns may be altered on tax day. Although
electronic submissions might be expected to lessen driving on tax day, we observed an increase rather than a decrease in fatal crashes in recent decades.

Now isn’t that odd. E-filing—that is, the non-necessity of driving to the post office—accentuates the effect. Somehow. Who knows how. But, as the saying goes, the numbers don’t prevaricate.

Redelmeier says,

We always knew road crashes destroyed the lives of thousands of people each year in the U.S., and driver error contributes to 93% of the crashes. Stress is often speculated to contribute to driver error, but stress is usually impossible to study, [but] tax day is synchronized, repeated, and tremendously onerous…

Almost every one of these fatal crashes could have been avoided by a small change in driver behavior.

Tax Day isn’t the only danger, no sir. The good doctor, who is a kind of crash aficionado, has in other work also identified Super Bowl Sunday and Election Day1 as killers. Yes, I know what you’re thinking, but you’re wrong: the risk of crashing on Election day “was similar, regardless of whether a Democrat or Republican was elected.” His research also “does not indicate which candidate could best help save the 100 lives lost each day in fatal crashes in the United States.”

Redelmeier has also, in the same statistical manner, “discovered” that emergency department visits decreased during an Olympic gold medal television broadcast. This reminds me of the apocryphal story I used to hear as a kid: that water use surged during commercials in the I Love Lucy program.

The way to criticize these studies is so obvious that I leave it as a homework exercise in the comments. (Plus, I may shortly have a way to investigate the numbers.)


Thanks to DAV for suggesting this topic.

1This paper was published with Rob Tibshirani, a good statistician and aggressive if not unskilled basketball player. But nobody bats 1.000.


  1. Bruce Foutch

    Regarding the photo:

    As Dave Barry might say; ‘the driver was released after producing a valid Florida driver’s license’…

    Happy tax day to all! 🙁

  2. DAV

    I drove to the post office to send out my returns on Friday the 13th and I didn’t get killed even once. Good thing I didn’t wait until the 17th, eh?

    Bet there’s a great story behind that photo. Clearly shows the value of mastering parallel parking skills. I wonder what the parking fine was.

  3. Ray

    “Almost every one of these fatal crashes could have been avoided by a small change in driver behavior. ”

    This sounds like something a lawyer would say. Lawyers will always tell you after the fact what you should have done and how everything would have worked out perfectly fine. Lawyers are professsional prigs. Are you sure he isn’t a lawyer?

  4. JH

    Happy Tax Day?

    How did the car manage to crash into one side of the roof and come out the other side? Did the driver forget to descent to the ground level before stepping on the gas? The house doesn’t look like it has a car elevator though.

    A large sample size doesn’t guarantee a publishable probability-value in a nonparametric binomial test. Anyway, is a sample size of 30 large?

  5. JH

    It just occurred to me that the author might soon publish another research letter titled
    “Driving Fatalities on Memorial Days.” *sigh*

  6. SteveN

    Does it matter that Tax Day is always chosen to fall on a weekday, whereas the “control group” could consist of any days in the week? I believe that evening rush hours are the high point for traffic fatalities, so if they did not adjust for that then one would expect the result that they have announced.

  7. Uncle Mike

    I don’t understand the alleged need to tax anybody. The central bank, aka the Federal “Reserve”, electronically creates $billions every day, completely out of thin air. It’s called “quantitative easing” because it so unreservedly easy to gin up quants and quants of free money. Why take mine, or yours, when they can and do invent gadzillions with the push of a button?

    They’re forever blowing bubbles,
    Pretty bubbles in the air.
    They fly so high, nearly reach the sky,
    Then like my dreams they fade and die.

    They’re forever blowing bubbles,
    Pretty bubbles, no more troubles,
    Pretty bubbles in the air.

  8. Matt

    Actually, the 15th wasn’t tax day this year. While tax day in the US is nominally the 15th of April, taxes can not be due on a Saturday, a Sunday or a Holiday for federal employees in Washington DC. Since the 15th was Sunday, and today is Emancipation Day, taxes aren’t due untill midnight tomorrow, Tuesday April. 17th.

  9. Greg Cavanagh

    I successfully predicted that Obama would win the last election, based solely on the premise that he was the least desirable person to be elected.

    I’ll be making the same bet next time for the same reasons. Obama will win, simply because he’s the worst possible choice.

  10. Greg Cavanagh

    Ack sorry; had two windows open, posted to the wrong thread.

  11. Here in Western Australia, the “authorities” decided some years ago that it was more dangerous to drive during a holiday period than any other day, so they doubled the number of “demerit points” for heinous transgressions such as exceeding the speed limit by more than 10 km/h; even on a country road.

    There was of course no statistical justification. The Australian Bureau of Statistics did an analysis and published that there was no significant increase over such periods.

    Country roads in W.A. are often isolated with sparse traffic. Experienced drivers of such roads understand that any serious crash will probably be fatal because help will probably not arrive for several hours. Therefore almost any injury requiring hospitalisation is potentially fatal under those circumstances.

    WA Police get really, really upset when drivers exceed the posted speed limits which have been set demonstrably far too low.

    It’s not new. The obedience training has been going on for decades and results reported by Police in 2000.

    Increased policing resulted in increased compliance with speed limits. But no reduction in KSI crashes. WHich should have come at no surprise because a previous programme (ESEP – Enhanced Speed Enforcement Programme) run in conjunction with the Main Roads Department some 5 years before that, had similar results. No significant reduction in KSI crashes.

  12. David L. Hagen

    Does the Post Super Bowl fatalities study pass your statistical sniff test?

    We observed a 41 percent relative increase in the average number of fatalities after the telecast (24.5 vs. 17.3, P0.20) and a marginal decrease during the telecast (15.7 vs. 17.7, P=0.036). The increase in fatalities after the telecast was evident for 21 of 27 years and amounted to about seven added deaths on the average Super Bowl Sunday as compared with the average control Sunday.

    Driving Fatalities on Super Bowl Sunday
    N Engl J Med 2003; 348:368-369 January 23, 2003
    Donald A. Redelmeier, M.D., Craig L. Stewart, M.Sc.

Leave a Reply

Your email address will not be published. Required fields are marked *