Friday, March 16, 2012

On the Book, "The Spirit Level"

First off, I haven't read it. And, so, yes, I AM going against what I preached in a previous post. But I have in fact examined both the assertions made in the book (that more equal societies have better social outcomes, essentially) and the methodology utilized by the authors and, because of that, I feel that I can make at least some commentary. So, where to start?............................................................................................a) The book deals exclusively with correlations, and, as we all should remember from elementary statistics class, correlations do NOT infer causality. These authors posit that egalitarianism in a society causes lesser in terms of social ills. But the opposite in fact could be true. The social problems could be causing the inequality OR there could even be a third, fourth, and/or fifth variable that's causing them both. Even an elemental understanding of statistics would have clearly helped these individuals.............b) The authors also seem to be lacking in basic understanding of the scientific method. For instance, they seem to think that researchers go about trying to "prove" theories true - this, as opposed to what they actually do; i.e., try and DISPROVE them (they call it rejecting the null hypothesis). The fact that these people seemed so hell-bent on trying to buttress their preexisting beliefs is actually quite ANTI-science.............c) The authors totally cherry-picked their subjects (in this instance, countries). They supposedly set out to examine the 50 richest countries in the world and they ended up only examining 23. They tried to justify this by saying that they didn't want to use countries with an under three million population because of their tax-haven status. But even this excuse didn't make any sense in that most of the countries in question WEREN'T tax-havens. The fact that they were so selective in the countries that they examined is excessively troubling, in my opinion.............d) These same authors also cherry-picked the indicators (AKA, the dependent variables). They analyzed drug use but not alcohol consumption. They analyzed imprisonment but not crime rates. They analyzed homicides but not suicides. They analyzed teenage births but not divorce rates. They analyzed foreign aid but not charitable contributions. And they analyzed neighborhood trust but not volunteerism. They did these things, folks, because they were trying to validate what they already thought and were willing to jig the data in order to accomplish this. Peter Saunders, in his absolutely blistering refutation of the book,showed that, if in fact the authors had analyzed racial bigotry, suicide rates, fertility rates, alcoholism, and HIV, the results would have been the absolute opposite (i.e., the more equal countries showing up poorly). It kind of tells you something, doesn't it?............e) The authors show massive lack of understanding regarding even elementary statistics. Everybody knows that when you do a statistical comparison (whether it be a t-test, an analysis of variance, a regression analysis, whatever), you don't just do a between group analysis. You also do a within group analysis. And one of the ways that you factor that is to disregard the extreme scores. For instance, when the authors were analyzing gun violence, they didn't remove the major outlier; the United States. They didn't, folks, because when you in fact do do that, the correlation absolutely disappears. The same is true with life expectancy. Remove Japan from the equation on this variable and THAT correlation disappears as well.............f) The authors made virtually no attempt to seek out alternative explanations. They never, for example, considered the fact that a lack of effective gun regulation in America might be more of a factor in that country's high level of violence. OR they never once considered the fact that the diet in Japan might be more of a factor for that country's longer life-expectancy....They never considered anything, essentially.......Final score - 2 stars out of 5.

9 comments:

Jerry Critter said...

I have not read the book either, but for the most part your criticisms seem reasonable. I would just like to make one comment about correlation and causation. The fact that correlation does not mean causation is always the first criticism brought up when someone does not like the correlation. But it is never mentioned that it does not mean that the opposite is true. There can be no causation without correlation.

The Prophet Dervish Z Sanders said...

Peter Saunders "debunks" the work of Wilkinson and Pickett via slight of hand... in what amounts to a hatchet-job.

Wilkinson and Pickett denounce Saunders for his reprehensible and prejudiced analysis that "[blames]African Americans for poverty".

Hm, blaming the victim... who'd have guessed a Conservative would do such a thing?

BB-Idaho said...

Not familiar with the book. But having grown up in the 40s-50s,
my gut feeling is that they are correct. A cursory regression analysis would seem to confirm that; as would the Great Depression...

Will "take no prisoners" Hart said...

I'm not a statistician, gentlemen. But I did take Introductory Statistics, Psychological Statistics, 2 courses in Experimental Psychology, and a course in Educational Research. And, I'm telling you, if I had handed in an analysis along the lines of what these particular "researchers" did here, I would have failed each and every time. This is absolutely abysmal.......And Peter Saunders knows more about statistics in his cuticle than these authors ever will, period. They're the ones who did the frigging "slight of hand"; a) setting out to prove a theory, b) cherry-picking the subjects, c) cherry-picking the indicators, d) failing to examine within group variability, and e) failing to consider competing explanations.

Will "take no prisoners" Hart said...

And of course they're going to call Mr. Saunders a racist! That's what liberals "do" (conservatives more so the patriotism card). And he didn't say that race CAUSED the social problems. He just said that if you controlled for race, the correlation disappears. There could in fact be intervening variables (such as the war on drugs, shitty inner-city education, etc.) at play in this scenario AS WELL.

Will "take no prisoners" Hart said...

Jerry, your points are well taken. I would just counsel, though, that if these researchers had instead chose to analyze racial bigotry, suicide rates, divorce rates, fertility rates, alcoholism, and HIV as indicators, the correlation would NOT have supported the authors' assertions....We can all pretty much prove whatever we want to, I guess is what I'm saying.

Jerry Critter said...

You are right about proving a theory. Scientists do not "prove" a theory. They postulate a theory and then try to disprove it. The reason Einstein's theory of relativity has stood the test of time is because scientists have not been able to find an instance where it is not true...yet.

The Prophet Dervish Z Sanders said...

What Conservatives "do" is come up with theories to justify their greed. Here Wilkinson and Pickett have proven (via 30 years of research) that greed is detrimental to society. Of course Conservatives have to "disprove" it.

Will "take no prisoners" Hart said...

The authors have "proven" nothing, wd. They've cherry picked both subjects and indicators and instituted an extremely faulty statistical analysis. Statistics 101 - when you do a regression analysis, you always remove the outlier score. And I don't even know what Mr. Saunders's politics are. You apparently (but, keep in mind, a British conservative is tantamount to gay liberal in Texas).