Group items tagged

when a scientific study ends by stating that there’s uncertainty about whether a correlation proves a cause, it doesn’t mean that correlations are meaningless in every circumstance. Everyone knows that creating false correlations between two unrelated elements is easy. But it can be that a correlation is so powerful and reliable that it may actually point to that rare thing in the social sciences, a demonstrable causal relation. As a wise man once said, “Correlation is not causation, but it sure is a hint.” When you can separate out a truly robust correlation between two elements in our social life, it’s a big deal. What makes a correlation causal? Well, it should be robust, showing up all over the place, across many states and nations; it should exclude some other correlation that might be causing the same thing; and, ideally, there ought to be some kind of proposed mechanism that would explain why one element affects the other. There’s a strong correlation between vaccines and less childhood disease, for instance, and a simple biological mechanism of induced immunity to explain it. The correlation between gun possession and gun violence—or, alternately, between gun control and stopping gun violence—is one of the most robust that you can find.

I think most philosophers of science would say that all scientific generalizations are probabilistic rather than logically certain, more so for the social sciences because the systems you are studying are more complex than, say, molecules, and because there are fewer opportunities to intervene experimentally and to control every variable. But the exis­tence of the social sciences, including psychology, to the extent that they have discovered anything, shows that, despite the uncontrollability of human behavior, you can make some progress: you can do your best to control the nuisance variables that are not literally in your control; you can have analogues in a laboratory that simulate what you’re interested in and impose an experimental manipulation. You can be clever about squeezing the last drop of causal information out of a correlational data set, and you can use converging evi­dence, the qualitative narratives of traditional history in combination with quantitative data sets and regression analyses that try to find patterns in them. But I also go to traditional historical narratives, partly as a sanity check. If you’re just manipulating numbers, you never know whether you’ve wan­dered into some preposterous conclusion by taking numbers too seriously that couldn’t possibly reflect reality. Also, it’s the narrative history that provides hypotheses that can then be tested. Very often a historian comes up with some plausible causal story, and that gives the social scientists something to do in squeezing a story out of the numbers.

Despite the astonishing breadth of his interests, Frank’s core obsession in teaching turned on the relation between science and the humanities. More particularly, his concern was with the relation between the causal explanations offered by science and the kinds of humanistic description we find, say, in the novels of Dickens or Dostoevsky, or in the sociological writings of Erving Goffman and David Riesman. His quest was to try and clarify the occasions when a scientific explanation was appropriate and when it was not, and we need instead a humanistic remark. His conviction was that our confusions about science and the humanities had wide-ranging and malign societal consequences.

It would seem that in situations like the current Greek crisis, we should be able to rely on economics to tell us which policies are most likely to work. But does the discipline have sufficient predictive power to play an important role in our debates about public policy?

The problems that we want economists to help us solve are more like predicting how leaves will fall on a windy day than predicting how objects will fall in a vacuum. Economic phenomena are affected by a very large number of causal factors of many different kinds.

A team of researchers walked every street in 228 census tracts around Los Angeles and New Orleans and recorded every outdoor ad they saw. Another group surveyed 2,881 residents of the same census tracts by telephone, paying them to report their height, weight and other information. After analyzing this hard-won data, the authors conclude: “For every 10 percent increase in food advertisements, the odds of being obese increased by 5 percent.” That is, areas with more outdoor food ads have a higher proportion of obese people than ones with fewer ads.

The problem is that their policy recommendations rest on a crucial but unjustified assumption: that any link between obesity and advertising occurs because more advertising causes higher rates of obesity. But the study at hand showed only an association: people living in areas with more food ads were more likely to be obese than people living in areas with fewer food ads. To be fair, the researchers correctly note that additional steps would be needed to prove that food ads cause obesity. But until those steps are taken, talk of restricting ads is premature. In fact, it is easy to imagine how the causation could run the opposite way (something the article did not mention): If food vendors believe obese people are more likely than non-obese people to buy their products, they will place more ads in areas where obese people already live. Suppose we counted ads for fitness-oriented products like bicycles and bottled water, and found more of those ads in places with less obesity. Would it then be wise anti-obesity policy to subsidize such ads? Or would the smarter conclusion be that the fitness companies suspect that the obese are less likely than the fit to buy their products?

When we seek to base policy on evidence, we must remember that not all “evidence” is created equal. Taken at face value, the study on ads and obesity provides some indication that the two are linked, but no evidence that food ads cause obesity. The fact that the causal conclusion may coincide with a moral belief — that it is wrong to tempt people who overeat by showing them ads for food — does not make it valid.

‘“What if?” is a waste of time’ went the headline to the Cambridge historian Richard Evans’ piece in The Guardian last year. Surveying the many instances of public counterfactual discourse in the anniversary commemorations of the First World War, Evans wrote: ‘This kind of fantasising is now all the rage, and threatens to overwhelm our perceptions of what really happened in the past, pushing aside our attempts to explain it in favour of a futile and misguided attempt to decide whether the decisions taken in August 1914 were right or wrong.’

But hold on a minute.

If well-done counterfactuals can help us think them through, shouldn’t we allow what-ifs some space at the history table?

Does bad behavior in movies or other media lead people to behave badly? There’s plenty of research on the link between onscreen media and risky behaviors like unprotected sex, binge drinking, fast driving and even violence. One large meta-analysis of such studies concluded that exposure to risk-glorifying media is associated with risky behaviors by people who consume that media. But causality issues plague most studies in this area: People who engage in risky behaviors may prefer to consume risk-glorifying media. These studies also tend to measure attitudes in controlled lab settings rather than in real life.