Popular Economics - Mini Book Reviews Part I: Freakonomics
I've read three economics books for the general reader in the past year and a half or so. While I've enjoyed all three, I thought I'd pass on slightly more specific thoughts and recommendations. The three titles are Freakonomics, The Undercover Economist, and The Origin of Wealth. I'll give my mini-reviews in (at least) three installments, in ascending order of my ranking/relative enjoyment.
Freakonomics, I believe, sold the most copies and received the most buzz by a wide margin. Enjoyable, but still my least favorite of the three. At least a couple of people close to me were decidedly unimpressed with the methodology used to make some of the more controversial claims. In particular, the result linking legalized abortion to the drop in crime seems to set a lot of folks' teeth on edge. I cannot speak with much authority on the methodology, perhaps my anonymous economist buddy will chime in sometime. It does seem like multivariate regression analysis is tricky business. On the one hand, these economists (Levitt, et al.) do seem to do their homework. They go out of their way to acknowledge that correlation does not prove causation, but then attempt to show that other plausible explanations don't show the same correlations. After a while the case seems to build and at least approach convincing, at least to a layman such as myself.
I guess the main caveat I'd throw out is that I am still dubious about much (most) of the methodology of the social and behavioral sciences. Proving things about human behavior still seems like an immensely daunting and complicated task, and I'm not convinced that any manageable set of data points is up to it, at least yet. In the specific example, sure they weave an impressive tale about how all the other variables they can think of (and have data for!) do not account for (all of) the drop in the crime rate. But what about something they haven't thought of or do not have data for? There are so many possible variables--human brains and social constructs are just so vastly complex! I suppose in some scenarios, you can try to account for this sort of thing with an "everything else" bucket, which would include all the things currently unspecified, but that seems like a dubious technique. Reminds me of the unrealistic assumptions about perfect information and perfect "rationality" in classical economic theory (but more on that when we get to The Origin of Wealth) . My two bit, armchair amateur critique.
Freakonomics, I believe, sold the most copies and received the most buzz by a wide margin. Enjoyable, but still my least favorite of the three. At least a couple of people close to me were decidedly unimpressed with the methodology used to make some of the more controversial claims. In particular, the result linking legalized abortion to the drop in crime seems to set a lot of folks' teeth on edge. I cannot speak with much authority on the methodology, perhaps my anonymous economist buddy will chime in sometime. It does seem like multivariate regression analysis is tricky business. On the one hand, these economists (Levitt, et al.) do seem to do their homework. They go out of their way to acknowledge that correlation does not prove causation, but then attempt to show that other plausible explanations don't show the same correlations. After a while the case seems to build and at least approach convincing, at least to a layman such as myself.
I guess the main caveat I'd throw out is that I am still dubious about much (most) of the methodology of the social and behavioral sciences. Proving things about human behavior still seems like an immensely daunting and complicated task, and I'm not convinced that any manageable set of data points is up to it, at least yet. In the specific example, sure they weave an impressive tale about how all the other variables they can think of (and have data for!) do not account for (all of) the drop in the crime rate. But what about something they haven't thought of or do not have data for? There are so many possible variables--human brains and social constructs are just so vastly complex! I suppose in some scenarios, you can try to account for this sort of thing with an "everything else" bucket, which would include all the things currently unspecified, but that seems like a dubious technique. Reminds me of the unrealistic assumptions about perfect information and perfect "rationality" in classical economic theory (but more on that when we get to The Origin of Wealth) . My two bit, armchair amateur critique.
posted by wRONgainey at 7:59 AM
0 Comments:
Post a Comment
<< Home