Australasian Science: Australia's authority on science since 1938

Belief Beyond Evidence, Evidence Beyond Belief

By Tim Olds

Will the childhood obesity epidemic condemn young people to a shorter lifespan than their parents?

When I was born in 1955, life expectancy for Australian males was 67. Now it’s almost 80. Around the world, with very few exceptions, life expectancy is increasing at the rate of about 2.5 years each decade, and has been doing so for at least 150 years. So I was indeed surprised when I first read this in 2005:

“If we don’t get this epidemic of childhood obesity in check, for the first time in a century children will be looking forward to a shorter life expectancy than their parents.”

Just where did this idea come from? The first record of such a claim come from a US academic and paediatrician, William Klish, in the Houston Chronicle in 2002. In March 2004 the message was taken up by the US Surgeon-General, Richard Cremona. The head of the Robert Wood Johnson Foundation, the largest US health not-for-profit, reiterated the message in Time magazine in July 2004. President Bill Clinton and Republican luminary Mike Huckabee repeated the claim in the same year.

When asked what evidence he had for his initial claim, Klish replied: “It was based on intuition”. The Surgeon-General’s spokesperson responded that he had not based his statement on empirical evidence, but rather on “some literature that he had read”. The Robert Wood Johnson Foundation clarified that “a lot of policy organisations use soundbites that do not rely on scientific literature because they draw attention”.

The first hard evidence came in 2005, when S. Jay Olshansky published a paper in the New England Journal of Medicine. The media latched onto this immediately: it was covered by the New York Times and virtually every health-related website. It was endorsed by Australian academics in the Sydney Morning Herald. It was cited by the National Institutes of Health and the American College of Sports Medicine.

How much confidence can we have in these claims? Following a study by Katherine Flegal from the Centers for Disease Control which suggested that overweight people live longer than leaner individuals, Olshansky conceded that “my life expectancy forecasts might be inaccurate”. His co-author, David Allison, explained that “these were just back-of-the envelope, plausible scenarios”.

Others were less nuanced. James Vaupel, Director of the Max Planck Institute for Demographics, stated that “the Olshansky piece is seriously flawed. His perspective is that of an advocate making a case rather than a scientist.” An expert from the CDC said he would be shocked if life expectancy fell: “We’ve never seen anything like that”.

In spite of the criticisms, the recantations and the back-pedalling, the claim is still being widely circulated. Ian Olver from the Cancer Council endorsed it in 2011, and the Wikipedia entry on obesity still refers to it. The UK’s Telegraph recycled it in 2011, as well as the Independent and the Daily Mail in 2013.

How can this idea – based initially on pure intuition, cynically or at best uncritically propagated by advocacy groups and supported only by a single discredited study recanted even by its authors – have gained such a foothold even a decade on?

It’s partly due to the way the media, and citations in academic journals, act as a kind of echo chamber where improbable ideas are amplified and made to seem more plausible. Both academic journals and the media have a predilection for attention-grabbing, quotable (and hence more improbable) ideas. Because of the way search engine algorithms work – the websites most commonly referred to by other websites appear at the top of the rankings – the same sites get referred to over and over again. In terms of citations, the rich get richer and the poor get nowhere. Among the first 30 Google listings for the life expectancy claim, only one casts any doubt on the veracity of the claims.

Improbable ideas amplified by the media have something else working for them: what Nobel Prize-winning psychologist Daniel Kahneman calls “heuristics” – rules of thumb for rapidly processing complex data. The “availability heuristic”, for example, occurs when people make judgments about the probability of an idea based on the ease with which examples come to mind. If the media constantly put the life expectancy claim before us we are more likely to believe it. The “affect heuristic” occurs when we take to be true something that arouses emotion in us. For example, an advocacy group may be deeply concerned about childhood obesity.

What’s missing from our considerations is a sense of prior probability – the likelihood that an idea is true before the new evidence is produced. Prior probability is the bedrock of Bayesian inference but is almost entirely ignored, even among academics.

Bayes’ theorem tells us that the posterior probability of an idea – the revised probability in the light of new evidence – depends in part on the prior probability. If the prior probability of the life expectancy claim were 10%, say, then even if the Olshansky study were methodologically flawless (which it most certainly wasn’t) the posterior probability would still only be 35%.

So long as we uncritically fall victim to the media echo chamber, fail to deconstruct heuristics and ignore prior probability, we will have beliefs that go beyond the evidence, and accept as evidence ideas that are fundamentally beyond belief.

Professor Tim Olds leads the Health and Use of Time Group at the Sansom Institute for Health Research, University of South Australia.