Paul Krugman. November 5, 1996. Extract from his debate with Galbraith on Slate on "Who is the Real Economist?"

      "Economics is at least partly about quantities and their relationships; so you can't make sense of it unless you are willing to do some arithmetic and even some algebra to make sure that the stories you tell hang together--and that they are consistent with the evidence. This doesn't sound like much, but experience shows that there are many influential intellectuals who are prepared to make sweeping pronouncements on economics without doing the arithmetic. And by the way, throwing around lots of statistics is not the point: It's a question of thinking hard about how the statistics fit together.
       Maybe the only way to explain what I am talking about is to give an example. It involves Michael Lind of The New Yorker, against whom I have no particular grudge--in fact, part of the point is that he is, undoubtedly, a very smart guy. That is what makes this particular example so revealing: It shows what happens when someone who is very bright but does not understand the role of arithmetic in economic analysis tries to make pronouncements on the subject.
       Two years ago, Lind wrote the following in Harper's:

Many advocates of free trade claim that higher productivity growth in the United States will offset any downward pressure on wages caused by the global sweatshop economy, but the appealing theory falls victim to an unpleasant fact. Productivity has been going up, without resulting wage gains for American workers. Between 1977 and 1992, the average productivity of American workers increased by more than 30 percent, while the average real wage fell by 13 percent. The logic is inescapable. No matter how much productivity increases, wages will fall if there is an abundance of workers competing for a scarcity of jobs--an abundance of the sort created by the globalization of the labor pool for U.S.-based corporations.

       Now what should Lind have done before publishing this passage? He should have had an internal monologue--something like this: "Hmm, do these numbers make sense? Well, historically, compensation of workers has been around 70 percent of national income. So let's say that initially, output per worker is 100, and the wage is 70. Now if productivity is up 30 percent, that means that output is 130, while if wages are down 13 percent, that brings the wage down to around 61, which is less than half of 130--wow, that means that the share of labor in national income must have fallen more than 20 percentage points. Let me check that out in the Statistical Abstract. ..." Of course, if he had, he would have found out that the share of compensation in national income, far from declining 20 percentage points, was about the same (73 percent) in 1992 as it was in 1977, offering a clear warning bell that something was wrong not only with his numbers--for example, he turns out to have confused productivity in manufacturing with productivity in the economy as a whole--but with his story. (This was not a throwaway passage marginal to his main argument; the claim that globalization has shifted the distribution of income drastically in favor of capital was central to his article.)
       How could Lind have failed to go through this little monologue? Well, I have had several conversations with impressive, highly articulate men, who believe themselves sophisticated about economic matters, but who simply do not understand that if productivity is up and wages are down, this must mean that labor's share in income has fallen. These conversations are not pleasant: They want to discuss deep global issues, and end up being given a lesson in elementary arithmetic. But that is precisely the point: All too many people think that they can do economics by learning some impressive phrases and reciting some gee-whiz statistics, and do not realize that you need to think algebraically about how the story fits together."