I recently completed Freakonomics by Steven D. Levitt and Stephen J. Dubner. I found it to be a truly provocative read, but it should be taken with a rather large generous portion of salt. This is due to the fact that although Levitt often claims that economics is a science, it most definitely is not. In full disclosure, I am a massive science snob. But I am for good reason. I have looked real science in the face, and have also been punched in the face by real science. And as I lay crying on the floor, real science has also kicked me several times. So I am a science snob. I do not even honestly believe that biology is a “true science”. I am immensely fond of, and often cite, Earnest Rutherford’s famous quotation, “All science is either physics or stamp collecting.” But even if I am not in a particularly snobby mood, economics is maybe 1/5 science 4/5 reading of entrails (well below even biology). So realize that when Levitt talks about the “science of economics”, he is really talking about some “some solid mathematics and a lot of philosophy”.
That is not to say that I do not value economics (or biology for that matter), but I believe accurate categorization is needed. Indeed I am fond of many things that are not, by definition, science: philosophy, writing, music, art, biology, economics, etc. Let me show an example from the book,
On pages 9-12 Levitt addresses the issue of money in politics. He “proves” that the conventional wisdom that the amount of money spent on a campaign directly affects the amount of votes received. His proof comes by citing that from 1972 to 2005 “nearly a thousand” congressional elections were held between the same two candidates (what he calls the ideal comparison). From these 1,000 or so elections he draws his results. He does not show you the mathematics (though in fairness due to the current mathaphobia in our country every equation he printed would probably mean he would loose a lot of readers, and the economic goal of a book is, presumably, to make money so… well do the econ yourself).
My blog, on the other hand, makes me exactly negative $20 a year, so I fear loosing no readership (obviously: I mean, my last entry was about gay penguins). So I will do the math and deal with the reader loss:
There is a 33 year interval that he measured (2005-1972 = 33 years)
Thus, the number of congressional elections during that time period is (ideally) 7,940 ~= 8,000
House of Representatives seat goes up for grabs every 2 years, a total of 465 seats.
Each House seat went up for grabs 16 times (33 years / 2 years per elections ~= 16 elections).
The total number of seats is 465 multiplied by the number of elections (16) gives you a total of 7,440.
Now for the number of Senatorial elections that took place in that time interval (100 total seats * 33/6 number of times the total seats were up for grabs in the given time period) = 500 elections.
Sum the House and Senate numbers to find the total Congressional elections, 7,440 + 500 = 7,940 or about 8,000 seats.
Levitt pulls data from “nearly a thousand” of those. Assuming he defines “nearly a thousand” reasonably, I will say he had 1,000 data points. So he only really looked at 13% of the total elections (1000/8000 * 100% = 12.5% ~= 13% ).
Thirteen percent is not a whole lot. It is a good first order approximation. In most sciences that would not be considered enough empirical data to draw conclusions from (unless it yielded a very accurate model, but in this case it does not). It is a decent amount of data for a first order approximation, and enough to make you want to know more, but certain not enough to prove anything. I doubt there are many chemists or physicists that would publish data with only 13% sampling: particularly when considering the other variables involved (see my procedural notes below).
Procedural notes: Levitt does not specify if he is dealing with only Federal level elections, but I will assume he was. It he was dealing with all levels of congressional elections, than the total number of elections would drastically increase, and, thus, further weaken his argument. It is also unclear how he handled cases where a representative went un-opposed for multiple elections. In essence he would be facing the “same” candidate twice (no one), but would not have to spend any money on his campaign. Not having data on this myself, I will assume he handled this in a way that did not benefit his argument. It is also unclear how much public’s name recognition plays. One candidate may be voted for simply because he is the incumbent and has name recognition, the other may be a known “looser” and not have any votes. Undoubtedly these numbers would be small, but when your data set is only 1,000 entries even a hundred such cases would be 10% of your entire set.
The point of it all:
This calculation was easy. It contained either obvious data (that the average person should just know) or data that is readily available from the internet, library, reference books, etc. Many calculations in Freakonomics were not so easy, and much harder to find. This points to my biggest problem with the book. Levitt goes to great lengths to tell us that experts are not to be trusted simply because they are an “expert”, and that they often use their knowledge advantage to bully you into agreeing with them. This is what Prof Scofield called “proof by intimidation”. It is also true. But Levitt does not really provide his calculations, which would allow you to check his work and see any of his unstated assumptions first hand. Like he says, numbers do not lie. Throughout the book he says that experts hide data from you and asks that you demand that they reveal it, all while he, in essence, is guilty of the same intellectual sin. Though, to his credit, his endnotes are well detailed and offer a wealth of cross references.
At the same time economics does use calculus and high order mathematics. I imagine most people can’t tell the difference between an integer and an integrand. I wish this book was available in a “math literate” version, it would be very revealing.
But the point of the book was really that you shouldn’t trust conventional wisdom and that there is more going on in the world than a surface examination. That is true. It was also about what we think are cause and effect may, in fact, just be coincidental, or it may be difficult to discover which is the cause and which is the effect. Levitt uses the example of snow. If you knew nothing of weather and saw it snowing you would not likely know if it was snowing because it was cold, cold because it was snowing, or just happened to be cold and snowing. Ultimately we need books like Freakonomics to remind us that knowledge is a wonderful thing.
You really should read Freakonomics, I am focusing on the bad points, because I just tend to do that. Don’t be confused, I do not knock economics for its dubious “science” quality, I knock it when its practitioners claim it to be science. The fact that economics is not as rigorous as science is a great of strength. Physics could never begin to measure some of the most basic ideas present in society, simply because it would try to take too much in. Our mathematical models are not good enough yet. By cutting out a lot of the rigorous analysis economics can tease out results that no real science could. Until we have the tools, and (god forbid) the experimental freedom, to really test society, economics is in many ways our best shot at understanding things. It is certainly more precise and accurate than pure philosophy. And let’s face it, it’s more likely to be right with a little bit of mathematics in it than no math at all. Until we have the mathematical tools and experimental freedoms to really do it right, I am more likely to trust a little mathematics and a lot of philosophy and voodoo, over just a lot of philosophy and voodoo. But maybe one day we will have everything we need: then physics can step in and start finding the absolute answers, and economics will go the way of the natural philosopher.