The Frontal Cortex

Economics, Statistics and Autism

Last year, a Cornell University economist named Michael Waldman noticed a strange correlation: the more precipitation a region received, the more likely children were to be diagnosed with autism.

[This] soon led Prof. Waldman to conclude that something children do more during rain or snow — perhaps watching television — must influence autism. Last October, Cornell announced the resulting paper in a news release headlined, “Early childhood TV viewing may trigger autism, data analysis suggests.

The resulting paper was a nifty trove of complicated statistics and unexpected correlations. But it was the rumor of causation – the possibility that television might trigger autism – that made the paper so instantly notorious. An interesting article in the WSJ explores whether or not economists should even be asking such questions:

Prof. Waldman’s willingness to hazard an opinion on a delicate matter of science reflects the growing ambition of economists — and also their growing hubris, in the view of critics. Academic economists are increasingly venturing beyond their traditional stomping ground, a wanderlust that has produced some powerful results but also has raised concerns about whether they’re sometimes going too far. …

Such debates are likely to grow as economists delve into issues in education, politics, history and even epidemiology. Prof. Waldman’s use of precipitation illustrates one of the tools that has emboldened them: the instrumental variable, a statistical method that, by introducing some random or natural influence, helps economists sort out questions of cause and effect. Using the technique, they can create “natural experiments” that seek to approximate the rigor of randomized trials — the traditional gold standard of … research. …

But as enthusiasm for the approach has grown, so too have questions. One concern: When economists use one variable as a proxy for another — rainfall patterns instead of TV viewing, for example — it’s not always clear what the results actually measure. Also, the experiments on their own offer little insight into why one thing affects another.

My own opinion is that, as long as we recognize the limitations of these economic approaches, other fields (like neuroscience) should welcome them. Detecting statistical correlations among vast data sets is a valuable tool, especially when it comes to generating surprising hypotheses. (Of course, we need to remember that correlation is never causation.) Whether or not these hypotheses turn out to be true is a completely separate matter.* But the worst thing that can happen is the falsification of an intriguing idea.
What do you think? Should scientists welcome the statistical speculations of economists?

*Perhaps I’m being too sanguine about the empirical potential of economics. After all, economists still think humans are rational agents, a psychological premise that was debunked decades ago.

Comments

  1. #1 Agnostic
    February 27, 2007

    Assuming the studies are conducted properly, what’s to lose? We should be as imaginative as possible (well, within the bounds of reason*) when coming up with hypotheses. Finding confirmation, replication, fitting it in with existing theory (if possible), etc. — that’s the hard part.

    In this case, the precipitation data, if true, are pretty interesting: to me it’s one more datum suggesting that autism is infectious (even if there’s genetic variation in susceptibility).

    *I guess you could allow any old hypothesis, and the bad ones will be shot down quickly.

  2. #2 bigTom
    February 27, 2007

    I don’t see a problem, as long as the results are treated with the appropriate degree of uncertainty. In the cited case, I’d say nothing is proven, but some interesting conjectures can be proposed, and this might be useful in selecting other avenues of search. Think of this as data mining used to propose possibly interesting conjectures.

  3. #3 mcewen
    February 27, 2007

    I am so glad that hundreds of years ago when I was a young person I read “How to Lie with Statistics’ by Darrell Huff. It has given me cause to pause and think many a time, which is no bad thing if you’re not as technically minded as the experts.
    Best wishes

  4. #4 Jason Anastas
    February 27, 2007

    “This] soon led Prof. Waldman to conclude that something children do more during rain or snow — perhaps watching television — must influence autism. Last October, Cornell announced the resulting paper in a news release headlined, “Early childhood TV viewing may trigger autism, data analysis suggests.”

    As a Cornell Alum and someone with a Stats and Biology graduate degree, I hope that Waldman took into account factors like in rainy places (Boston, New York, San Franscisco, Seattle etc) there also happens to be more hospitals and more doctors that specialize in autism. Autism also has a genetic and environmental component depending on the particular disease you’re looking at. Problems of cell division that lead to trisomy (having an extra chromosome) that causes Down Syndrome for example, has been shown to be stimulated by various environmental toxins in addition to other factors such as age of the mother and father etc. In general, you tend to find higher levels of children with autism in places that have higher levels of toxic waste. To assume that it’s “early television watching” sounds pretty retarded to me. Usually I trust economists because they have better training that psychologists or political scientists in data analysis and mathematics, but this conclusion that Waldman has jumped to sounds sounds pretty off the wall. After living in Ithaca for too much of his life, I think the rain perhaps might be affecting his cognitive functions, as it seems to do to many that inhabit that tiny rainy dystopia.

  5. #5 Steve Silberman
    February 27, 2007

    In general, you tend to find higher levels of children with autism in places that have higher levels of toxic waste.

    I’d love a citation for this, Jason.

  6. #6 Jason Anastas
    February 27, 2007

    Hi Steve, here you go:
    Chromosomal congenital anomalies and residence near hazardous waste landfill sites.
    http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=pubmed&cmd=Retrieve&dopt=AbstractPlus&list_uids=11830202&query_hl=6&itool=pubmed_docsum

  7. #7 CA
    February 27, 2007

    Check out the october 2006 blog on this article at MixingMemory
    http://scienceblogs.com/mixingmemory/2006/10/autism_and_television_1.php

  8. #8 Terry
    February 28, 2007

    “Should scientists welcome the statistical
    speculations of economists?”

    Hey knowledge is valuable where ever it comes
    from. On the other hand we now have two strong
    examples of television related studies by
    economists that are pure B.S.

    This autism study and this study:

    http://www.slate.com/id/2136372/

    Notice that neither were not published in a
    peer-reviewed journal. Instead a couple
    of economist did a bit of a thought experiment
    and then found some statistics only marginally
    related. But because the “results” were
    so counter-intuitive and shocking, the
    press ran with it.

    In the study described by Slate, the economists
    assumed that as soon as TV was introduced,
    children instantly started watching as much TV
    as they do today.

    According to this Neilson News Release:

    http://www.nielsenmedia.com/newsreleases/2005/AvgHoursMinutes92905.pdf

    If you then download their excel spreadsheet of
    historical levels, you’ll be able to see the
    huge increase in the amount of time the tv
    is on (actual early viewership not available).

    Also their “results” contradict much more legitimate
    studies such as:

    http://www.otago.ac.nz/news/news/2005/05-07-05_press_release.html

    http://www.usatoday.com/tech/gaming/2006-10-02-gaming-grades_x.htm?csp=34

    http://www.msnbc.msn.com/id/8471072/

    As for the autism study, why correlate with rain?
    Why not instead just buy some Neilson statistics by
    region and see if there is any kind of correlation?
    I have never seen any evidence that there is a
    huge correlation with the amount of rain and the
    amount of tv watching.

    Also, since autism can be diagnosed as early as
    18 months, that means that it would have to be
    infants watching tv that would lead to autism.
    I can’t imagine that infants watch substantially
    more tv because of rain. On the other hand, with
    the popularity of “Baby Einstein” videos and similar
    if we now see a new big increase in autism, that
    would be more suggestive.

  9. #9 Reality Bytes
    February 28, 2007

    It’s certainly not a bad idea for economists to provide suppositions. However, it is bad when it gives rise to misunderstandings in the media (including blogs) “TV causes autism in children”. In this case, it could be any number of environmental factors related to rain. It should be clear that there isn’t necessarily causation here. It is only beneficial to the degree that direction is provided for science to investigate.

  10. #10 Steve Silberman
    February 28, 2007

    Jason, thanks for the link. While it’s certainly a matter of deep concern that kids who are born near toxic landfills seem to show higher levels of chromosomal abnormalities in general, I think it’s a fairly big leap from that study to the statement that “you tend to find higher levels of children with autism in places that have higher levels of toxic waste.”

    It may seem like I’m putting too fine a point on it, but I believe that the assumption that autism is caused by a toxic trigger in the environment — vaccines or pollution — while appealing psychologically, may obscure other more likely mechanisms. So I don’t take the toxic hypothesis on faith, and until a specific environmental trigger for autism is identified, I’m not comfortable with the assumption. The whole field of autism research and treatment has been plagued by misdirection and unfounded assumptions, most notoriously Bruno Bettleheim’s “refrigerator mother” theory. I’m not aware of any controlled studies of populations of autistic kids where a specific substance in the environment has been identified as the cause. I’d love to know about such a study if it’s out there. (Media stories about the possible link to vaccines don’t count, I’m afraid — that link has never been proven, and there are a number of reasons why a genetic disorder that doesn’t “present” until two or three years could masquerade as a disorder triggered by vaccines given at the same age.)

  11. #11 Terry
    March 2, 2007

    Here is an article published in the
    The Cornell Daily Sun:

    http://cornelldailysun.com/node/19539

    called “Science by Press Release” by
    Matthew Belmonte, an assistant professor
    at the Department of Human Development
    at Cornell University

    “The controversy surrounding Waldman
    et al.�s unpublished and incompletely
    substantiated study illustrates the
    danger that arises when the conduct
    of science by peer review is supplanted
    with science by press release.”

    As for the rainfall/autism correlation,
    people living in overcast regions with
    high rainfall also tend to be low
    in Vitamin D, maybe lack of Vit D is
    the culprit. That’s assuming that the
    rainfall correlation is even correct,
    I’m skeptical of even that.

    On the other hand, if this study had
    been peer-reviewed and still concluded
    that the rainfall/autism correlation
    was substantiated, then I would have
    much more confidence in the study.

    I am no fan of TV, and it could very
    well be that it does trigger autism.
    But if in the future some substantial
    study came out showing a link, it will
    tend to get dismissed because of this
    flighty study.

  12. #12 John Carragee
    March 9, 2007

    The study, as reported here, raises an immediate concern that is common to all such statistical “fishing expeditions.” If you test 100 random variables against the incidence of autism, with no reasoned trail of causation, you should expect 5 of those variables to show a “statistically significant” correlation. That’s just a tautology since we usually define “statistically significant” as meaning only a 5% chance the result occurred at random. You can then work backwards and try to conjure a correlation (in this case, “rainfall reflects TV viewing which causes autism”) but you then have run way beyond the realm of science — whether hard science or social science like economics.

    The issue is not that economists are running amok, but rather that statistics are being misused.

    The same technique was used to “prove” years back the electric power lines cause brain cancers. Nothing of the sort is true; but if you test enough variables indiscriminately you will find lots of completely bogus correlations.

    I’m surprised this “study” even gets a post on a science blog.

  13. #13 bail
    November 17, 2011

    Undeniably think that that you stated. Your preferred justification seemed to be on the web the simplest thing to be aware of. I say to you, We definitely obtain annoyed while people consider worries they just donít learn about. You were able to hit the nail upon the very best as well as described out the event without having side-effects , people may take an indication. Will probably be back to get more. Many thanks

The site is currently under maintenance and will be back shortly. New comments have been disabled during this time, please check back soon.