An Apple a Day—No, Wait 
by
Gustavo Bamberger
Posted to the eraider.com commentary page, 1 August, 2001.

Frank Patrick Herbert (1920-1986): newspaperman, cameraman, radio commentator, oyster-diver, judo instructor, jungle survival instructor, lay analyst, Navy photographer, serious amateur scientist, ecologist, Tacoma native, Seattle Mariners fan and second-tier science fiction writer.
I don’t read it often, but I find that there’s often an odd, surreal quality to what gets reported in USA Today. Many years back, I remember a boxed story in the sports section with a picture of a forgotten pitcher (a Seattle Mariner, I think). The heading was something like “Joe Blow’s Five Favorite Frank Herbert Novels.” For the non-science fiction fans out there, Frank Herbert is best known for his novel Dune (but see link for many other books), which was followed by a series of sequels (Dune Messiah, Children of Dune, etc.). Turns out the pitcher’s five favorite Frank Herbert novels were, in order: 
  • Dune
  • Dune Messiah
  • Children of Dune
  • another Dune sequel
  • another Dune sequel
  • With the possible exception of Arrakis, the desert world on which much of Dune takes place, on what planet does that qualify as news, let alone sports news?
    Seattle Mariners argue over the best Frank Herbert novel
    I was reminded of the top five Dune novels last week when I unexpectedly found myself in Columbia, South Carolina for three days. The hotel where I was staying delivered USA Today to my room every morning, and I awoke one morning to find that the lead headline in the paper was “Food Pyramid Challenged.” Under the headline, it was reported that “Harvard nutrition expert says the government’s model is outdated and perhaps ill-conceived.” You can see why that would be the lead story for the entire country—there’s nothing more electrifying than a claim that a government policy is “perhaps ill-conceived.”

    Take two bananas and call me in the morning
    Then again, I don’t pay much attention to the latest “what food’s good for you” news stories anyway since we seem to get a fair amount of conflicting advice: should I drink more red wine? put olive oil on everything? To be fair, there seems to be general agreement that the basic western diet—lots of fats and sweets—is bad for you, and that news seems to have had limited impact on eating habits. (And not just bad for us: studies of baboons in eastern Africa found that troops that discovered garbage dumps near tourist lodges consumed more calories and matured faster, but also had substantially higher levels of cholesterol and insulin than their savanna-foraging cousins.)
    Why can’t medical researchers get their stories straight about what we should be eating? I don’t actually know, but I’m sympathetic. Economists get criticized for a similar inability to nail down the answers to some basic questions: what is the effect of a minimum wage on unemployment, for example. Economists have a good answer: the world is complicated, and it’s often hard to isolate individual effects in a complicated world (I suspect medical researchers have a similar answer).
    How to identify with economists
    One of the most difficult issues that economists who analyze data have to contend with is known as the “identification problem.” Here’s a standard version of the problem. Suppose that you want to know what will happen to sales of a product if its price were to increase by five percent. You can go out and collect information on, say, monthly sales and prices of the product over the last five years, and use that to estimate the relationship between price and units sold. You’d probably expect to find a negative relationship—if prices are high, sales will be low. Unfortunately, the world tends to be more complicated than that. Think of the standard-issue supply and demand chart: price on the vertical axis, sales on the horizontal axis, demand slopes down, supply slopes up; market price and quantity is where the two curves intersect. (You may want to find a piece of scratch paper and pencil at this point.) Suppose that during the period you’re studying, costs are shifting. Cost shifts mean that the supply curve is shifting left or right over time. If the supply curve shifts left, the intersection points occurs at a higher price and lower sales—what you were probably expecting.
    But suppose instead that costs are unchanged, but demand for the product increases.  Then the demand curve shifts right, so the intersection of supply and demand takes place at a higher price and higher sales. So now higher prices are associated with higher, not lower, sales. The problem is that if the supply curve is shifting, points on the demand curve are traced out, but if the demand curve is shifting, points on the supply curve are traced out. In the first case, the demand curve is “identified;” in the second, you can “identify” the supply curve. In general, though, both supply and demand curves can be shifting at the same time, and it can be difficult to disentangle the two effects. That’s the “identification” problem.
    Identification problems show up throughout empirical economics. Another classic example arises in the “human capital” literature. Suppose you want to know the effect of an extra year of school on wage rates. You could collect information on individuals’ earnings and years of schooling and estimate the relationship between the two. But there’s a problem—the people who got a lot of schooling are not a random sample of all workers—in general, people with more schooling stay in school for a reason. And one reason is that school may be a better investment for “high ability” workers than for “low ability” workers. If that’s the case, then the premium that workers with high levels of schooling receive over others will reflect both the effect of additional schooling as well as higher average ability levels—in other words, those workers may have earned more even if they had stayed in school no longer than the low-schooling group. 
    So how do you disentangle the effects of ability on wages from the effect of schooling? How do you solve the identification problem? The statistical issues that have to be dealt with can get extremely technical and complicated, but economists also have been ingenious in finding “natural experiments” that allow them to deal with the identification problem. The problem for school-earnings analyses is that for any individual, years of schooling likely is related to ability, which likely will influence wages. But in certain circumstances, years of schooling probably is not related to ability. One study, for example, looks at US students who received only the minimum amount of schooling. Suppose that schooling is mandatory until the age of 16, and that a child must begin school if he or she turns six before September 1. Now consider the group of kids who left school at 16. Those kids born before September 1 will have received 10 years of schooling, but those born after September 1 will have stayed in school only nine years. To the extent there’s no relationship between ability levels and date of birth, the kids born after September 1 will have the same average ability level as those born before, but one less year of schooling. This difference can be exploited to estimate the effect of an additional year of schooling on wages. Because years of schooling is unlikely to be related to ability in that group of workers, the effects of schooling and ability can be disentangled—the schooling effect is identified. (Another approach is to use identical twins with different levels of education to estimate the effect of schooling.)
    Fortunately, new statistical techniques continue to be developed, and new and better data sources continue to become available, so we’re likely to witness continued progress in the economic analyses of empirical questions in the years ahead. Not to mention continued progress on the food pyramid front. In the meantime, eat sensibly and exercise regularly.
    References
    The effect of diet on wild baboons is discussed in Robert M. Sapolsky, “Junk Food Monkeys,” in The Trouble with Testosterone and Other Essays on the Biology of the Human Predicament, 1997. The identification problem in human capital and other studies is discussed in Mark R. Rosenzweig and Kenneth I. Wolpin, “Natural ‘Natural Experiments’ in Economics,” Journal of Economic Literature, December 2000.