In 1991 Halpern and Coren published a famous study in the New England Journal of Medicine which appears to show that left handed people die at much younger ages than right-handed people. Halpern and Coren had obtained records on 987 deaths in Southern California–we can stipulate that this was a random sample of deaths in that time period–and had then asked family members whether the deceased was right or left-handed. What they found was stunning, left handers in their sample had died at an average age of 66 compared to 75 for right handers. If true, left handedness would be on the same order of deadliness as a lifetime of smoking. Halpern and Coren argued that this was due mostly to unnatural deaths such as industrial and driving accidents caused by left-handers living in a right-handed world. The study was widely reported at the time and continues to be regularly cited in popular accounts of left handedness (e.g. Buzzfeed, Cracked).
What is less well known is that the conclusions of the Halpern-Coren study are almost certainly wrong, left-handedness is not a major cause of death. Rather than dramatically lower life expectancy, a more plausible explanation of the HC findings is a subtle and interesting statistical artifact. The problem was pointed out as early as the letters to the editor in the next issue of the NEJM (see Strang letter) and was also recently pointed out in an article by Hannah Barnes in the BBC News (kudos to the BBC!) but is much less well known.
The statistical issue is that at a given moment in time a random sample of deaths is not necessarily a random sample of people. I will explain.
Over the 20th century, left handers have increased as a fraction of the population. Left handedness may be relatively fixed as a genetic matter but in the earlier decades of the 20th century children were strongly discouraged from exhibiting left-handedness. As a result, many “natural” lefties learned right-handed behavior and identified as right-handed adults. Over time, however, the cultural suppression of left-handedness declined and the proportion of adults exhibiting left-handedness increased, as the figure, at left, illustrates (fyi, I believe British data).
Now suppose you take a random sample of people who died in 1990. In this sample, some people will have died old and some young. Among those those who died old, however, fewer people will be identified as left-handed because the old grew up in a time when left-handedness was suppressed. As a result, the old deaths in your sample will tend to be have more right-handed people and the young deaths will tend to have more left-handed people causing you to incorrectly conclude that left-handed people die younger. Studies show that this statistical artifact can easily explain a 9 year difference in apparent mortality rates.
To make this crystal clear consider the following thought experiment (offered by Chris McManus). Imagine you take a sample of people who died recently and asked their surviving family members, Did the deceased ever read the Harry Potter novels? One would clearly find in such a sample that those who died tragically young (at age 12 let’s say) would have been much more likely to have read Harry Potter than those who died in their 90s. Despite what some might argue, however, we should not conclude that Harry Potter kills.
Hat tip: Tim Harford.