Five Second ... DiNardo: What is causality?

One of the best things I pulled out of my stack of reading last month was the third part of DiNardo's review of Freakonomics, published in the Journal of Economic Literature, Dec 2007. This was less about what he had to say about the book as what he had to say about causality and randomized controlled trials (RCTs). It's an excellent instructional piece to get students thinking about correlation vs. causality. Some of his better points are in the footnotes [which is also where I finally found a professional picture of him. His website features pictures of DiNardos he is not]. A smattering of the points he makes:

"Rather than hold up the RCT as a paradigm for all research, I review it here because it represents a single case in which we sometimes have some hope of evaluating (limited, context dependent) causal claims, and because what constitutes a severe test is somewhat clearer. ... The RCT often provides a useful template to evaluate whether the causal question is answerable. ...

"For any individual ... we can never be certain that some unobserved determinant of the outcome y is changing at the same time we are assigning the person to treatment or control."

Never. But there are uses.

• A doctor studied the 1590 siege of Paris. "He was led to conclude that one of the 'effects of insufficient food' was that the lethality of diseases such as typhoid was much greater. Nonetheless, 'hunger' or 'lack of food' was rarely cited as a 'cause' of death, although he identified undernutrition as an 'underlying potential cause.'"
• From Heckman (2005): "Two ingredients are central to any definition [of causality]: (a) a set of possible outcomes (counterfactuals) ... and (b) a manipulation where one (or more) of the 'factors' ... is changed." This is how we usually approach both RCTs and economics in general when we talk about regressions. Holding other factors constant, the effect of x is beta.
• Commenting on Moffitt (2005): "[The argument ... that race can not be a cause because it can not be manipulated results from] ... a mistaken application of the experimental analogy.... It does not make conceptual sense to imagine that, at any point in the lifetime of (say) an African-American, having experienced everything up to that time, her skin color were changed to white.... Although it is a well-defined question, it may nevertheless be unanswerable, and it may not even be the main question of interest...." DiNardo writes: "If I were to wake up tomorrow and discover that my skin color had changed dramatically, one possible reaction might be a visit to the Centers for Disease Control to learn if I had acquired an obscure disease! ... If that response were typical of other white folks who woke up one day to find themselves "black," I would nonetheless hesitate to say that the "causal effect of being black" (or white) is an increase in the probability that one makes a visit to the CDC."  Brilliant.

His discussion of obesity is quite illuminating. It seems a shame to paraphrase it but it's a touch long. The key question is how the weight is added or dropped. If weight is lost by starvation, mortality and other disease incidence may increase, suggesting that obesity causes good health. Further, different people have different ideal weights and it may be that lowering one's weight too much is in fact unhealthy. "The presumption that obesity is a cause of ill health made it virtually impossible to debate whether nonobesity was the cause of the increased heart problems." In fact, Gronniger (2003) finds that some income groups are benefited by being heavier. Thus, the salient policy question according to Gronniger "is not what obesity does to people, but what removing obesity would do to people" or, according to DiNardo identifying "the effect of Reddux or exercise."