I occasionally re-read my recent posts, just to check that I wasn’t talking U.R. I have just done that with last week’s. Glad to say, I didn’t come to that conclusion, and the talk of the staffroom this week has appeared to bear out what I was saying. But I was drawn to Dylan Wiliam’s quote of Lleras-Muney (2009) who “estimated that one additional year of school adds 1.7 years to one’s life”. At the time, I was more concerned with constructing my own argument, but on reflection, does this not have to be edu-rubbish of the most utter kind?
I think the give-away is in the word ‘estimated’, and I’d be interested to know what methodology said worker used to reach this conclusion, particularly given the precision of the figure.
We in education have been subjected to some drivel over the years, but this takes the biscuit. I concede I know little of this ‘research’ but common sense is surely all that is needed to debunk it. To begin with, it is impossible to know what age any individual ‘would have lived to’ in the alternative scenario with or without education. While mass-statistics may yield some insight, as always the disconnect between macro-data and the specific experience of any one individual is too great to make any meaningful claim of this sort. Secondly, causal density is so high that isolating any one factor with the necessary precision is impossible. I can accept that educated individuals may make certain choices or have certain opportunities that increase their chances of a longer life – but does holding a PhD really reduce the risk of getting cancer or motor neurone disease, or of being run over by a bus?
Within my own circle, my highly-educated mother was dead of cancer at the still-young age of 67, while my in-laws who had limited education and who smoke and drink heavily are still going strong in their eighties. My former academic tutor was killed in a car crash that was not his fault, and Stephen Hawking’s cerebral abilities can only indirectly have impacted on his remarkable survival.
It is risible that such claims are still being made as a serious attempt to justify specific educational practices, and very concerning that people like Dylan Wiliam are prepared to cite them in support of their own work.
If I am missing something essential here, I will happily stand corrected!
teaching personally 
Reblogged this on The Echo Chamber.
this is just an average treatment effect for a population, not a claim that for every single person who receives an extra year of education will live exactly 1.7 years longer. It is no different to saying that on average a drug that counteracts breast cancer will add on average an extra 1.7 years to the lifespan of people who have contracted breast cancer. Clearly some people who get the drug will die long before that point, and some will live for long after, but the average treatment effect is 1.7 years, and we estimate that (in medicine) through the randomized controlled trial.
In education we do not have randomized controlled trials for questions of this sort, but we do have natural experiments that can function as an acceptable replacement (observation studies are not an acceptable replacement, however). In this instance, the natural experiments are driven by 20th-century laws that at various different times in different countries extended the length of compulsory schooling. In some nations these laws were not all implemented simultaneously across all regions, meaning that we can use this exogenous regional variation as a natural experiment.
Having said all that, what is the validity of this estimate? Ironically enough a quick review suggests that it is probably overstated, with the caveat that this research field is still quite young (it seems that not enough people have died yet). http://www.ncbi.nlm.nih.gov/pubmed/26340596 updates Lleras-Muney with more fine-grained data and finds virtually no causal effect. http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3365223/ uses Swedish registry data & a strong natural experiment to find some effects across various subgroups (though I always put huge caveats next to any subgroup analysis for obvious reasons).
In conclusion, therefore, I think there are some indications that Dylan & Lleras-Muney are wrong, but I very strong suggest that such average treatment effects are not drivel, or any other kind of incoherent nonsense. They are important to the policymaker and the scientist, if not to the individual, and can be reliably estimated using well-known analytical methods.
Thanks for your comment; I was waiting for a response like that! I try hard to be considered in my comments, but being human do occasionally fall prey to hyperbole… Since I wrote this I have investigated Lleras-Muney a little further and found her to be a worker at UCLA. One might therefore assume she does not generally indulge in cod science.
In the interests of brevity, I did not explore the validity of her work itself. As a geographer, I am more than happy to accept that education (but ‘achievement’?) does lead to observable macro-scale effects in populations, indeed I teach my students this.
My issue is with the use of such data to justify specific, small-scale interventions, because the precise *reason* for such effects is so obscure. I checked Wiliam’s wording carefully: “…estimated that one additional year of school *adds 1.7 years to one’s life*”.
While the macro-effect is entirely plausible, to couch it with such certainty in specific, individual terms – and then to suggest that classroom practices should be driven by such information – is disingenuous to say the least. It also occurs that ‘school’ in American parlance may refer to graduate school, but such issues are not clear in Wiliam’s context. Time limits one’s ability to follow up all such leads, but I will follow your links when I have a moment.
I agree the estimate seems high, but I can envisage, for example, that the effect might be larger in developing countries where general life expectancy is still somewhat lower. I am curious to know whether Lleras-Muney originates from such, and perhaps grounded her work there. There remains the problem of not being able to know the alternative scenario in terms of individual longevity though!
Big Data can yield no more than generalisations, and my beef is that they are of little use below policy level.
More generally, as my previous post discussed, there are many more flaws in such arguments, such as definitions of ‘achievement’, and its relation to the whole gamut of educational experience, where specific achievement is both hard to quantify and difficult to apply at all, let alone in terms of its effect on longevity. Proxies such as exam results can never capture the true life-changing effect of education on an individual.
I stand by my doubts that it is possible to identify causality from one end of someone’s life to the other in the way that Wiliam implies. And I also find it utterly depressing that the inherent rewards to be gained from intellectual discovery are being ever-further subordinated to such utilitarian thinking. I would argue that the quantity of one’s life is less important – at the margins – than the quality.
It’s probably the relationship between education and health and the impact this has on people in a global setting, i.e. in LICs. Global stats are always saying stuff like the more education a girl receives, the better the chances of her economic wellbeing, her health, that of her children and so on. And the better her chances of living longer. So worth looking up this academic before saying their research is utter rubbish.