Freakonomics has clearly taken the academic community by storm. Are you aware that there is both a teacher and a student guide to this book? Are you aware that it has become the set text for a number of University programmes, in the USA at least? It's all true. Take a look at the authors' web site and see for yourself.

Does the book deserve such plaudits as additional guides and set text adoption? I can see why; but I wouldn't go so far as to make it a set text. Strongly advise some analysts, practitioners and accountants to read it without a doubt, however.

Freakonomics was published, in the UK anyway, in 2005 and although I bought it a while ago I have only just read it. I should advise anyone who starts in the way that I did that you are misleading yourself if you just use this as a book to dip into for the odd fact or figure.

The book is written by an academic economist and a journalist. Not that they should be scarred for life; but I believe that their backgrounds force them into the analyses they make that others wouldn't necessarily make. This is neither a bad thing nor a criticism!

What is Freakonomics, then? Freakonomics is a book for data analysts in my opinion. I say that because every chapter and section of the book begins with an hypothesis or statement or fact that is then tested with reference to one or more data sets. There are six chapters and an epilogue to contend with and the chapter titles tell you something of what you might expect. For example,

Introduction: The hidden side of everything

1. What do school teachers and Sumo wrestlers have in common?
2. How is the Ku Klux Klan live a group of real estate agents?
3. Why do drug dealers still live with their moms?
4. Where have all the criminals gone?
5. What makes a perfect parent?
6. Perfect parenting, part II, or: would a Roshana by any other name smell as sweet?

Epilogue: two paths to Harvard

Where does my data analysis comment come from? Well, many of the arguments and ideas that Levitt comes up with are supported by data or are derived from a data set. Some of the ideas that he discusses are controversial, too: the idea that the legalisation of abortions was the cause of the reduction in the crime rate in the USA 20 years or so later.

Other controversial topics include a discussion of the Ku Klux Klan, the names of African American people versus the names of Caucasian people and the implications of that. Levitt does say at various stages that some readers may be offended by the ideas and the conclusions that the discusses: so don't read what he writes.

Fair enough: no one makes us read these things and of course we don't have to agree with them. Whether the legalisation of abortion was the cause of the fall in the crime rate is, in my opinion, a matter of opinion although I do think there is some strength in the argument. We should remember that whilst talking about abortions and crime rates, Levitt also talks about regression and correlation analyses and one key aspect of such analyses is that they are suggestive and not conclusive.

What this means is that whilst Levitt has made a good argument for links between abortion and crime I don't think it's the perfect conclusion. One weakness of the book is that he doesn't provide us with his data set and having visited his web site I didn't find any data there either. However, I did look around for the data and came up with the following graph. Now I am not suggesting the Levitt is hiding anything by not providing the data; it's just that I like to play with these things!

This graph is based on data taken from this site; and the American Bureau of Justice.

Maybe you can see that the crime rate really does fall from the mid 1990s following the legalisation of abortion in the USA in the early 1970s. Convinced by that? Possibly; but isn't there something missing from the analysis? What about replicating the experiment? Isn't that one of the keys of such experimental work? In other words, is this result obtained wherever else abortion was legalised? Levitt doesn't mention this.

Levitt also pretty well dismisses the impact of increased spending on the police and on improvements in policing methodologies and so on. They must have a role to play, surely? Maybe not that dramatic but some role I think.

Please note, since I wrote this review I have come across a paper written by Levitt that does contain some of the data and so on that I was looking for. Click here to get that paper (from the Jurnal of Economic Perspectives of Winter 2004, Vol 18 No 1pp163 - 190).

If you become Levitt obsessive, you can download many of his papers from his page at the University of Chicago.

I was much more convinced by the arguments surrounding the analysis of why drug dealers still live with their mothers. That's also a perfect lesson in free market capitalism, short termism and executive pay and incentive schemes for anyone who is thinking about using this book. The student's guide provides us with a table of data emanating from this chapter and it includes data relating to total costs, fixed and variable costs, average total costs and marginal costs. Draw more graphs here, to your heart's delight! I did!

I thought that the chapter on names of different races was gratuitous and not in the least bit convincing. I think that if Levitt had asked me to review his book before publication I would have advised him to get rid of this chapter. His conclusions are interesting but just so difficult to support and so easy to ignore. At the end of it all, don't we all know that names go in and out of fashion? Aren't some people prime movers in choosing names for babies and so on? Look at our own chase to name our babies after the latest addition to the Royal Family. Sad isn't it?

Read about cheating teachers, Sumo wrestlers (I didn't know any of that and am surprised by part of it but not all) and estate agents. Some interesting data and analysis there I must say.

I thought his chapter on perfect parenting was especially weak: chapter 5. Levitt discusses the ideas surrounding what the perfect parent needs to do or to avoid in order to provide little Johnny or Jenny with the perfect start. How does he do this? He does it at least in part by discussing the impact of 16 variables on the ECLS (the Early Childhood Longitudinal Study from the US Department of Education). What this boils down to is that successful parenting is measured by the mark that a child earns at school: I'm sorry but I don't know the ECLS testing system; but it strikes me as being equivalent to saying that if your child had passed the 11 plus exam, then you can be congratulated as being a perfect parent.

Having said that, considering the 16 variables Levitt does discuss is a useful thing to do. Basing his analysis and conclusions on a data set in this is, in my opinion, extremely limiting.

There is no doubt that this is an interesting book and of course don't we all wish we'd written it or something like it? It is possible just to dip into and out of this book as I initially did; but having done that I then read it properly and realised that it is a very serious book both for economists and data analysts alike. Moreover, there is an extensive notes section at the end that is well worth reading if you are interested in using this book as more than just a good read: it's around 20 pages long.

Bear in mind that you will need to concentrate when reading this book and then bear in mind that it's entertaining and a serious data analysis book and you can enjoy it. Some of the topics are controversial however and that might cause you some concern. Finally, if you agree with everything Levitt says, then I suggest you read it again.

 

© Duncan Williamson
12th June 2006

© Webmaster Duncan Williamson 2006