The Test That Predicted Who Would Die Young, on the Road Too, Better Than Medical History

A study of Australian ex-servicemen pitted dozens of variables against each other to predict who would die young. Medical history, behavioral record and education level all came in behind a score on a mental aptitude test, taken years before any accident.

The result stops looking mysterious once you see what those men were dying of: mostly, on the road.

A study on an unusual scale

In The g Factor (1998), psychologist Arthur Jensen summarizes the work of Australian researcher Brian I. O’Toole, published in 1990 and 1992 (the latter with Lazar Stankov).

• The starting cohort was Australia’s former national service conscripts from the Vietnam War era.
• The studies focused on those who had died from causes unrelated to combat before the age of forty.
• Every one of them had a score on the Army General Classification test (AGC), a test heavily loaded on the g factor, along with other tests of speed, accuracy and mechanical comprehension.

Alongside those scores, the researchers gathered dozens of other variables, grouped into broad blocks: education level, personal conduct (alcohol use, arrests, unauthorized absences from military service) and health history (number of days in hospital, among other medical data). According to the review by Linda Gottfredson and Ian Deary (2004), the 1992 analysis compared the men who had died with a sample of survivors, about 2,300 men in total, and used 56 variables in addition to IQ.

Dozens of variables in the same race

The statistical analysis, a regression with all the variables entered at once, was built to answer one very specific question: which of them independently predict who would still be alive years later?

The result, as Jensen describes it, was clear-cut: the AGC, the most g-loaded of all the tests administered, made the clearest independent contribution to predicting mortality. One standard deviation higher on the score was linked to a 16% reduction in the mortality rate, which Gottfredson and Deary sum up as roughly 1% lower risk per IQ point.

Health history, with all its real medical information, did not turn out to be as good a predictor as that mental test taken years earlier. It’s worth reading this carefully: it doesn’t mean health doesn’t matter, but that among young and almost entirely healthy men, the medical information barely told one man apart from another.

What those men were dying of

The reason for this seemingly strange result becomes clear once you look at the causes of death. In that age group, most deaths were due to external causes, above all traffic accidents, not illness. And IQ was also the best predictor of that specific cause.

The dedicated analysis of motor vehicle accident mortality (O’Toole, 1990) produced this table, organized by IQ level:

IQ level	Death rate per 10,000
Above 115	51.3
Between 100 and 115	51.5
Between 85 and 100	92.2
Between 80 and 85	146.7

The traffic-accident death rate nearly triples from the highest-IQ group to the lowest. The table has a detail that often goes unnoticed: above 100 the curve is almost flat, and the difference is concentrated in the lower part of the distribution.

It isn’t an illness that explains the gap, but something to do with behavior in fast, shifting risk situations, like driving.

What came next: IQ and life expectancy

The Australian study was one of the first, but not the only one. Over the following two decades, the field known as cognitive epidemiology, with Ian Deary and David Batty among its main drivers, followed large cohorts from childhood or military recruitment into old age.

• In Scotland, a follow-up of the Aberdeen children tested in the Scottish Mental Survey of 1932 found that those who had scored higher at age 11 were more likely to still be alive at 76 (Whalley and Deary, 2001).
• In 2011, the meta-analysis by Catherine Calvin and colleagues pooled 16 cohort studies with follow-ups of 17 to 69 years: one standard deviation higher intelligence in youth was linked to a 24% lower risk of death from any cause.

So the effect isn’t limited to the road or to young men. What changes with age is which cause of death weighs most at each stage of life.

Why a mental test predicts better than a medical report

There’s no single explanation, and the researchers themselves treat them as hypotheses that add up:

• Education and socioeconomic status. A high score opens the door to more schooling, less dangerous jobs and better living conditions. How much of the effect runs through this path is one of the open questions in the field.
• Health behaviors. Smoking, physical activity, diet and sticking to a treatment are all related to IQ. Gottfredson describes looking after one’s own health as one more cognitive task, with new information to absorb, risks to weigh and decisions to keep up over time.
• Handling immediate risk. Reading a traffic situation in fractions of a second, anticipating what another vehicle will do or reacting to the unexpected depend in part on processes that the g factor captures well and that a medical report, focused on illnesses and hospital stays, doesn’t capture at all.
• System integrity. Deary has suggested that the same overall good functioning of the body could favor both cognitive performance and health, so the test would work partly as a marker of that state.

Jensen extends the same logic to other social outcomes (dropping out of school, poverty, crime), which in the data he reviews show an inverse relationship with IQ. That part of his book is far more contested, because it’s hard to separate IQ from each person’s social background and opportunities.

That same way of reading a fast, changing risk situation is exactly what the risk tolerance quiz explores, in a much lighter way.

The limits of such a striking finding

This study supports some conclusions and not others:

• It’s a difference in rates between groups of thousands of people, not an individual prediction.
• The vast majority of people, at any IQ level, never have a serious accident.
• The sample was young Australian men who went through the military: it doesn’t represent women or other ages, and traffic at the time was very different from today’s.
• Any one person’s real risk depends on a great many factors (driving experience, mood, road conditions, luck) that a population study can’t capture.

What the finding does leave is an idea that’s rarely framed this way: part of what keeps us safe has less to do with the body we bring to the doctor and more to do with how quickly and accurately we process what’s happening around us.

If you’re curious how that ability to process patterns quickly in complex situations gets built, chess grandmasters don’t have a better memory than you explains what really sets an expert apart from a beginner.

Sources

• Calvin, C. M., Deary, I. J., Fenton, C., Roberts, B. A., Der, G., Leckenby, N. and Batty, G. D. (2011). Intelligence in youth and all-cause-mortality: systematic review with meta-analysis. International Journal of Epidemiology, 40(3), 626-644.
• Gottfredson, L. S. and Deary, I. J. (2004). Intelligence predicts health and longevity, but why? Current Directions in Psychological Science, 13(1), 1-4.
• Jensen, A. R. (1998). The g Factor: The Science of Mental Ability. Praeger.
• O’Toole, B. I. (1990). Intelligence and behaviour and motor vehicle accident mortality. Accident Analysis and Prevention, 22(3), 211-221.
• O’Toole, B. I. and Stankov, L. (1992). Ultimate validity of psychological tests. Personality and Individual Differences, 13(6), 699-716.
• Whalley, L. J. and Deary, I. J. (2001). Longitudinal cohort study of childhood IQ and survival up to age 76. BMJ, 322(7290), 819-822.