Friday, May 27, 2005

The Last Chapter

As I mentioned, I just managed to read the last chapter of Freakonomics. There's some good stuff in that book. Using a combination of thought experiment, theory, and snippets of evidence, they establish believable explanations for many interesting things - most of which apparently turn out to be true, given their further actions on a number of them.

The book is written for the armchair intellectual. It's pretty good!

EXCEPT THE LAST CHAPTER.

I don't know exactly WHAT those two were thinking, but they decided to exclude thought experiments and snippets of evidence from the last chapter. They talk about how their data "shows this" or "shows that", but they don't give us any evidence of that data - no hard numbers. Unlike the earlier chapters. They also don't have the clean, crisp thought experiments I had come to expect.

So, yeah, their results are pretty interesting. In actuality, I am inclined to believe them, since they coincide closely with my own thoughts on the matter.

But without data it is SENSATIONALISM, pure and simple. Why did they suddenly starting writing DOWN to their audience? The rest of the book is so GOOD!

They use the phrase 'regression analysis' as a holy word to explain the fact that they aren't giving us any data. I've used regression analysis. Give us the cropped data set. I would have been happy with that.

Without the data, I can only assume you're TALKING OUT YOUR BUTTOCKS. Because 90% of scientists are liars, and 10% of scientists are wrong. So give me the data, so I can see for myself.

Here's an example of the further sloppiness of the last chapter.

One of the things they study is the effect of names on people's success. There is a definite correlation between the two but, as I always shout whenever 'correlation' is spoken, "correlation does not prove causation!" In this case, according to their invisible data, there is no causation at ALL. They determined that ghetto names (such as "Shanique" or "Shithead") are, in fact, usually given IN THE GHETTO. Hence most people with such a name are from lower socioeconomic brackets and, therefore, more likely to live less-educated, less successful lives.

HOWEVER, their regression analysis determined (at least, so we are told) that people who grow up in the ghetto but have LESS ghetto names, such as "Sara" or "Dick", do just as poorly as people from the ghetto with ghetto names. Obviously, this means the correlation is entirely an artefact, and the cause of both is, in fact, the low socio-economic bracket.

Okay, fine. Cool. Wish I had some DATA.

EARLIER in the chapter, they talked about how they ran a test using two identical resumes and applying for identical jobs... but giving one of them a ghetto name and the other a more upper-crust name. The resume with a 'rich' name got a LOT more callbacks and interview opportunities.

I don't know the exact parameters or extents here, but it is unlikely that this is merely correlation. Given all other factors are equal, it's clear that they get fewer callbacks BECAUSE THEY HAVE A GHETTO NAME. That's CAUSATION. In fact, this experiment is a good example of HOW TO PROVE CAUSATION.

Logically, less callbacks means less job offers, less job offers means less chances to do well socioeconomically, meaning that, LOGICALLY, a ghetto name DOES cause a negative socioeconomic stigma.

Now, of course, there's a major factor here missing, and the difference can be explained. I will, in fact, go on to give one probable explanation. But the sloppiness is that the authors DON'T cover this. They let the dichotomy STAND. No attempt to explain.

They do, in fact, give the explanation in a different study - a study of schools. It turns out that (again, NO FREAKING DATA) going to a better school doesn't actually make you a better student (although, like a freaking KLUTZ, he contradicts THIS, too, in this chapter, no explanation given). It's just correlation, again.

When students are allowed to TRY to enter a better school (chance to be chosen is by lottery), those who opt to try but don't get in have the same performance as those who opt to try and DO get in. This implies that it's the TRYING that matters, not the SCHOOL.

This is, in my mind, the obvious solution to the naming conflict. A ghetto name will reduce the DENSITY of opportunities. If two indentical people, Shithead and Molly, both try to get a job, Molly will probably get a job first. Shithead will get fewer callbacks, even though it's pronounce "sha-THEED". But that doesn't mean Shithead WON'T get a job. It'll just take him/her longer.

The fact that they are SEEKING opportunities is, in the end, dramatically more important than the actual number of opportunities per unit of time they get. This definitely gives creedence to "you make your own luck".

Or, it WOULD, if I could rely on the studies. Grr.

No comments: