Audio Book Review: Freakonomics /Steven D. Levitt & Stephen J. Dubner
Steven D. Levitt is a University of Chicago economist who is more interested in economic choices and the psychology of decision-making, rather than dusty macroeconomic studies, stock market prediction or studies of inflation. In his own words, his main interests are cheating, corruption and crime. This is good, but academic studies are unlikely to make a good audio book, since technical material, especially mathematics is not well put across in the audio book domain.
Fortunately, his co-author Stephen Dubner (a journalist, who also reads this audio book), is a dab hand at presenting complex statistical economic data in straightforward, plain English. Hence the book is an addictive listen, and very well suited to the format. Its an unabridged, unaltered (as far as I could tell) 7 hour long reading of the original printed book, published by penguin.
He addresses questions as varied as whether Sumo wrestlers cheat (apparently they do), to more controversial questions such as whether legalizing abortion prevented a crime wave in the 1990s. The results are presented with a degree of certainty that probably is not appropriate to an academic text, but which makes the book a lot more interesting!
There are nine chapters, each of which follows a similar pattern. An attention-grabbing title question is posed which embodies or questions an issue of our time, and which likely has already received a fair amount of media discussion and attention. Even more likely, the received wisdom of the answer to the question, as framed by a slanted and self-censored media viewpoint, is completely inaccurate. This summary makes the book sound like a million other collection of journalistic articles, in which a fearless, maverick investigator brings you “the hidden truth behind x”. But hold on, it gets more interesting for two reasons: 1. The questions are sometimes unexpected (what do schoolteachers and sumo wrestlers have in common?) and interesting in themselves. 2. There is some hard scientific analysis backing up the suggested alternative viewpoint. The book seeks to show the actual behavior of human beings rather than the behavior prescribed by the moral framework we supposedly subscribe to. In many respects it succeeds, far more than you might have expected.
Its not easy to pin-point the exact reasons why the book is less satisfactory, but the review will briefly try to do so. Lets look at chapter 1 (on Schoolteachers and Sumo wrestlers). Some techniques for detecting cheating Chicago schoolteachers are discussed, based on various data-driven algorithms being applied to the answer sheets, tracking the same class year by year. (Cheating is here defined as a teacher changing the students answers after tests are handed in). Sure enough, the year a bonus incentive is announced for teachers, there is a spike in cheating. The statistical tests used to detect cheating look for suspicious patterns in student answers (eg getting lots of easy questions wrong, but hard ones right, many students giving the same wrong answer (because the teacher does not know the right answer), or rapidly swinging performance from year to year). Sport cheating is most often about cheating to lose, as this will allow manipulation of the results for purposes of betting. And indeed analysis of sumo wrestlers performance shows that they will often trade losses at times where its crucial to the winner, but less so for the loser (he has already qualified for the next stage of the tournament, for example). In closing the chapter Levitt makes the point, using data supplied by a man who sold bagels on the honor system, that “people are honest 87% of the time”, even with no one watching. So indeed not every one is crooked, but in fact the sumo wrestlers only cheated when it “didn’t matter”, in the sense that the ultimate contest result was unaffected. However, this was not the case for the Chicago schoolteachers, who clearly were undermining the entire system for their own gain. So the parallel is not exact in this case. The deeper point, which is perhaps critical to the books weaknesses, is that the cheaters were only detectable against a background of statistics gathered from people the majority of whom were not cheating. If one cannot make this assumption, then the entire dataset becomes meaningless. So the “most people are honest” assumption is required not just for public morality to make any sense, but also to have any hope of understanding the data!
The statistical weakness of the book is perhaps to not make this latter point clear enough. There are also problems with some of the domains chosen: some are “open”, but others are “closed”. For example the proportion of population of the united states having measles can be clearly and accurately estimated – this is a closed domain where all the facts (medical and demographic) are in, and probabilities can be confidently assessed. Other domains are open, that is incomplete in the sense that not enough data has been collected to confidently estimate probabilities. Probability of a crack dealer living with their mother is hard to measure without good data on total numbers of crack dealers, for example. The latter is a minor carp however. More serious question marks float over chapter 4, which links the drop in crime seen in the 1990s to the legalizing of abortion in the 1960s. Clearly crime is linked to demographics, but then share prices are clearly linked to broader economic performances. However, the link is not guaranteed in the case of an individual share! A drop in street crime might be as much about fashionability, and/or a criminal’s cost/reward analysis of a particular type of crime, as a sudden dearth of muggers or drug dealers. For all this, its a thought-provoking chapter whose thesis is well argued, and supported by considerable evidence. Very definitely worth a listen if you have some dead time on the way to work!
Full of fascinating snippets of information (Listerine mouthwash was originally sold as floor cleaner, and a cure for gonorrhea for example) and unexpected links (eg Estate Agents and the Ku Klux Klan), the book is at all times interesting even if some the conclusions are a little too pat. The author is clear to distinguish “causes” from “is associated with” early in the book, but later on “is strongly correlated with” is clearly rhetorically used to mean “is the cause of”, or the associated chapter would lack any bite. Its not clear that all other possibilities have been eliminated. None the less its a great book, both entertaining and challenging – a rare combination in an economics book!
As later chapters depend on points made in earlier chapters its best to work through from beginning to end, and not hop about.