There’s several books that have made their way through the blogging world with surprising alacrity. The Tipping Point, Blink, Collapse, and Everything Bad is Good for You have all had their 15 minutes of memedom as they wend from blog to blog. Freakonomics, by Steven Levitt and Stephen Dubner, has endured longer than the others. I resisted reading Freakonomics just as I resisted reading the other books because the myriad reviews disclosed most of their contents through patchwork. Just as I could tell you that a liger is bred for its skills in magic even though I’ve never seen Napoleon Dynamite, I could give a rough accounting of Freakonomics et al. even though I’d never read them. Reading Freakonomics was like déjà vu in that sense. It also confirmed my suspicions that all of these books are probably not worth reading given the effusive praise and uncritical plaudits heaped on them by the average blogger.
It was an extremely easy read. I covered 190 of the pages in about four hours this Friday night. I don’t know if it was the lack of footnotes (I usually read each one as I come across them), the newspaper style, or the simple subject matter but I hardly ever can get through a book that quickly. The remaining 17 pages took another two hours since my wife saw some of its provocative conclusions and kept asking about them. The only other times I’ve zoomed through books as quickly have been children’s books—the first (and only) Harry Potter book took me a day or two—and popcorn fiction. In the end, I think its breezy style was mostly to account for the speed—this would be the perfect book to listen to.
If you haven’t picked up on it yet, I was not impressed by this book. I will admit that I was predisposed to not like this book. First, the hype surrounding it is a big turn-off for me. Hype is one of those things that typically ends in a let down. Most of the population does not share my set of values and my philosophic perspective. When something gets universally lauded, that usually means that it appeals to Christians, conservatives, altruists, and laymen—everything I’m not. Further, I didn’t particularly care for the content that I had already encountered by reading blogs. The analysis seemed glib and facile. Now that doesn’t necessarily mean anything since the average blogger is very glib and facile, so I tried to discard that particular prejudice before I started reading. It turns out that the analysis is glib and facile, but I wanted to read and be surprised.
The book starts out with its strongest cases. The first chapter about teachers cheating and sumo wrestlers throwing matches is an excellent use of data to tease out conclusions. The patterns they found are indicative of cheating since their statistical probability is considerably less than reasonable. They take the reader step by step through the thinking behind the analysis and it is a lesson in how to look for anomalies in data. If the book had continued along these lines, it would have been an amazing resource and would have heralded the beneficial power of statistics. I was really inspired and wished that I had a data set to so analyze; I wouldn’t know what to do but it just made me want to dive into SPSS.
After this great start, the book starts to unravel quickly. I blame the economist’s reliance on incentive as the cause of this slide. Incentives are powerful, to be sure, but they’re not sufficient to explain such complex phenomena as the real estate market and drug dealing. The authors found that real estate agents keep their homes on the market about 10 days longer than their clients. They suggest that this is because the agents want a quick sale to get their commissions but they want the best deal when they’re selling their own homes. While some agents may have exactly that motivation, I find it hard to believe that all do. While focusing on the agent is easy, I think that the real driver of the real estate transaction is the client. Most clients, I suspect, aren’t patient when it comes to selling their houses. They’ve got another house to close on and they can’t afford to carry two mortgages while they eke out another couple thousand dollars on the sales price. They may reject the first offer to cross their bow, but the next one that comes closest to their desired sales price is going to mighty tempting. If there’s any sort of lag between offers, the temptation grows greater with each bid. Agents, on the other hand, are much more experienced with waiting and bidding trends. When selling their own homes, it makes sense that they’d take advantage of their experience and it makes sense that that patience would pay off. I think that’s a more plausible explanation of the pattern than the sinister motivations that the authors posit.
They then survey the terms used in the listings in order to show which words are correlated with the final sales price. They found that certain words were correlated positively and others negatively. Leaving aside their methods (and I have to leave them aside since there’s nothing describing them in the text or the endnotes), they suggest that those words negatively correlated represent words that the agent employs to “subtly encourag[e] a buyer to bid low.” What? Was that in the data? I can just picture the reader of this book firing his real estate agent because she used the term “great neighborhood” in his house’s listing. Levitt and Dubner argue that specificity in description is what positively correlates but how do you specifically describe a great neighborhood, a fact which might be a very real selling point? The authors forget (or don’t consider) that listings are necessarily terse and usually incorporate fuzzier selling points in addition to the specifics.
Further, no one buys a house straight from its listing. If the terms in the listing don’t jibe with what the potential buyer is seeing, then the buyer will pass (and thus never be counted in their data set). But that’s not all: the listing may be perfect and the buyer may still pass (and thus be omitted from the data). In fact, the fuzzy listing may be in an older neighborhood and therefore command a less elastic sales price. Notice that all the specific words they mention as positive correlations are terms that you’d associate with a newer house (or one recently remodeled)—you don’t find “Corian” in a home from the 1950s unless it’s been added in the last ten years. Newer homes don’t exist in older neighborhoods. I don’t if any of the preceding possibilities are at work here. But all of these things represent potential explanations for the data that do not require one to think of real estate agents as bloodsuckers.
Where the pair really go astray is in their chapters on parenting. Their main conclusion is that parents have a negligible effect on children. They admit that bad parenting has a huge effect on children, practically guaranteeing them a ride on the prison express (if they don’t beat the odds). Their point is that, beyond a certain point, all the things parents do to enrich their children’s lives don’t matter. Visiting a museum? No effect. Reading to them each night? Zilch. Stay-at-home mother until of school age? Why bother. These are contrarian, to say the least. But they wouldn’t be making such statements unless they were solidly backed up by data, right? I mean they teased all sorts of interesting conclus ions out earlier by looking at the data in unique ways, so they would do that here as well, right?
Sorry to burst your bubble, Chester. They rely on test scores. No let’s-look-at-what-real-estate-agents-do-when-selling-their-own-homes or how-do-sumo-wrestlers-compete-when-they’re-at-the-crucial-7-wins-and-7-losses-stage on this one. They used test scores and academic progress combined with an invasive questionnaire to draw their conclusions. Not very rogue, if you ask me. It’s also completely irresponsible. How you can make sweeping generalizations about the power of parenting from academic performance of any sort is beyond me. If you’ll bear with me, I’m going to really delve into this issue because this is where the book totally fails.
I say that his statements are irresponsible because far too many parents don’t put enough thought or effort into parenting as it is. Telling them that parenting doesn’t matter is like giving them a carte blanche to slack even further. Even if so-called “obsessive parenting” didn’t result in higher test scores, that doesn’t mean that it’s not worthwhile to take your children to cultural events, read to them, or otherwise participate in their rearing. Sure, there’s no general agreement as to the effects of various parenting styles and you could find hundreds of books claiming to be the One True Way to effective parenting but that doesn’t mean that the pursuit of perfect parenting isn’t worth it. Some might have bristled at my use of “perfect parenting” but the reality is that you should also strive for perfection in parenting—as in the rest of your life. You’ll make mistakes and missteps, but that doesn’t mean you should adopt the posture of “well, perfection is impossible” while you beat your kids. I am not a perfect parent, but I believe that I should try to be one. That means that I should take mistakes seriously and work to stop them from recurring.
The actions that he lists as not having an impact on test scores reads like a laundry list of the characteristics of a working parent (or single parent). It strikes me as an apology for not staying home with the kids. While it may be possible that the effects are negligible—I sincerely doubt it—it seems like those parenting magazine articles of the same form that make working parents feel better about their decision to not commit fully to raising a child. I don’t know enough about the lifestyles of the authors so I can’t really say that this was some sort of rationalization on their parts, but it wouldn’t surprise me a bit if it were.
The fact of the matter is that performance on a standardized test may not be capable of measuring or indicating the benefits of staying home with your children. Off the top of my head, staying at home could produce better adjusted children that feel more love and trust for their parents. That could make a criminal future more unlikely. Staying at home could also shelter children from the social shock attendant with day care facilities until they are better prepared emotionally to deal with the cliques, taunts, and ostracism that often comes in social situations. That could avert many of the aberrant behaviors that accompany them. Or it may not. What is certainly true is that the authors haven’t proven anything one way or the other. I’m not sure if we even have means to measure and analyze these circumstances adequately.
Reliance on declarations by study participants is another fatal flaw. People lie on questionnaires, even more so when they’re delivered by an actual person from the government. They say what they think they ought to say, they misremember things, and I would suspect that some people actually lie outright. Sure, people who admit to spanking their children probably aren’t lying but how does that establish the study’s accuracy? It doesn’t, but it provides a nice distraction from the issue as the authors move on to the next shocking revelation.
In the end, the authors want to have their cake and eat it too. They repeatedly assert that people who value education have children with higher test scores. (168, 172, 174, 175) However, they also assert that “by the time most people pick up a parenting book, it is far too late.” (175) As mentioned earlier, they sincerely believe that, beyond a certain point, parenting doesn’t matter. But if parents value education, how does that carry over to the child? Is there an education gene? They argue no such thing. The actual causal chain that they never explicitly reveal is that parents who care about their child’s education will impart that sentiment by example. The academic performance of Asian children in America is well-documented and the reason typically given is that their parents want their children to do well in school. That desire is translated into a myriad of behaviors: rules about studying before watching television, allocation of resources to educational support, saving scrupulously for college, and vigilant monitoring of performance—to name a few. But if parents don’t matter, then how can their attitudes affect test scores? And if parents do matter, perhaps they matter all the way down? The authors, unfortunately, don’t bother to delve into these other possibilities.
There’s a number of other bones I could pick with the book, especially about the inane chapter on baby naming that reads like an afterthought. I’ll merely state that the chain they craft to link educated, upper-class naming to uneducated, lower-class naming is completely bogus. People may not get names from celebrities, but they do get them from the people they encounter in real life, the people they read about or see in the media, and the children their children encounter—all of whom have names themselves, fodder from which to draw. The educated upper class don’t associate with the uneducated lower class by and large. They travel in different circles and usually encounter each other indirectly (e.g., the dock worker might know the company president by name but he certainly doesn’t know the president’s kids’ names). How the name choices are transmitted from one to the other doesn’t stand up to scrutiny.
The book has some interesting examples, but it’s wrongheaded by and large. Worst of all, it’s insanely popular right now and the ease of reading it would suggest that it won’t just sit, binding uncracked, on the coffee tables of America. Its insights into the “hidden side of everything” will fit the conspiracy theorist’s mentality very well and may turn otherwise optimistic Americans instead skeptics and cynics. Who knows, the 20th anniversary edition of the book may have a chapter on the book’s unintended consequences. That may be ascribing it too much cultural power, but small things can have big effects.
[UPDATE (8/4/2005): Here is a similar, though much condensed, opinion.]