Saturday, March 29, 2008

Will, Brooks, and Compassionate Conservatives, er, Religious People

Statistics do not lie. However, unscrupulous people often present statistics in ways that make them appear to support positions that a more thorough, and dare I say honest analysis, would reject.

George Will's article summarizing the findings of Arthur Brooks that conservatives give more to charity than liberals is a classic example. He makes no effort to account for major influences on the data that explain it much better than political affiliation does, even when he is aware of them. Further, he and Will conflate percentages that are not comparable in a way that invites the reader to compare them and draw erroneous conclusions.

-- Although liberal families' incomes average 6 percent higher than those of conservative families, conservative-headed households give, on average, 30 percent more to charity than the average liberal-headed household ($1,600 per year vs. $1,227).

The first rule of making statistical comparisons is to make your groups as similar as possible except for the variable you are studying. Thus, if we are to compare liberals and conservatives (self-defined per Brooks), we would want to account for geographic disparities that would effect income. After all, the fact that we have Red and Blue states proves that conservatives and liberals do not live proportionally in the same places, and consequently, do not have the same cost of living to wade through before they can decide whether or not to give to charity. Per capita income by statevaries tremendously.

When deciding how much is given to charity, disposable income, that which remains after necessities, is the relevant measure, not net income. Who among us is surprised that a conservative living in Iowa gives a greater share of his income to charity than a liberal living in New York making the same income? That's because the New Yorker has more of his income swallowed up by taxes and other costs of living. So Brooks' analysis should have been liberal New Yorker vs conservative New Yorker and conservative Iowan vs liberal Iowan just to get the basics covered. Otherwise it is blatant apples and oranges.

The use of percentages above is also incorrect. Percentages should only be used when comparing items of similar size. Otherwise, the larger percentage cannot be safely assumed to be a higher absolute figure than the smaller percentage. A cursory reading of Will's paragraph above might lead one to erroneously conclude that the conservative gap in giving is 5 times as large as the liberal gap in income. The proper percentage to compare would be the relative giving as a percentage of the annual income, not of the lower charitable giving. Consider that hypothetically, his example could be groups whose average incomes are $40k and $42.4k (6% different), meaning that the relative percentages of giving listed represent 4.0% and 2.9% respectively. That 1.1% gap isn't nearly as impressive as the 30% figure, which is no doubt why it wasn't used.

-- Conservatives also donate more time and give more blood.

This is the sort of statistic whose meaning depends very much on when it was taken, in this case, whether it was during a war or not. Conservatives are overrepresented in the military, which would bias the blood stat if it were taken during an active war (or prolonged police action).

It also falls victim to the single biggest flaw in these statistics, the role of religion. Brooks inexplicably counted all giving to churches as charity. That's ALL giving, whether it was tithing, an extra donation, or volunteering at the 7:00 bingo game. This is flawed in several ways. Church income is obviously used for much more than charitable activities. All that property and gold costs money. Second, tithing in many churches ranges from an act of piety to a requirement. There's no such thing as forced charity. Third, churchgoers are in a real sense paying for a service, whether we buy the god thing or not. I'll gladly give them credit for the charitable work their churches do, but to count 100% of church giving as charity adds a huge mitigating factor in the data. It will clearly bias the data in favor of conservatives.

-- Residents of the states that voted for John Kerry in 2004 gave smaller percentages of their incomes to charity than did residents of states that voted for George Bush.

-- Bush carried 24 of the 25 states where charitable giving was above average.

-- In the 10 reddest states, in which Bush got more than 60 percent majorities, the average percentage of personal income donated to charity was 3.5. Residents of the bluest states, which gave Bush less than 40 percent, donated just 1.9 percent.

Cross state comparisons are going to be flawed at their core if they are not adjusted for standard of living. The average income of Kerry Blue states $34,673 vs $28,559 for the states Bush carried, which at first glance might seem to support Will's point even further, actually can work in the other direction, since higher average income also indicates higher cost of living. If this 21% difference is representative of the cost of living, it would easily overwhelm the 6% liberal-over-conservative income edge that Brooks cites. And once one considers state income taxes, which are much higher in the blue states (6 of 7 states with no income tax are red, including whoppers Texas and Florida), it becomes not only plausible, but likely, that as a percentage of disposable income, liberals give more than conservatives. Brooks and Will simply haven't done their homework.

And yet, apparently Brooks was aware of the flaw in his studies regarding religion, and yet did nothing about it:

The single biggest predictor of someone's altruism, Willett says, is religion. It increasingly correlates with conservative political affiliations because, as Brooks' book says, "the percentage of self-described Democrats who say they have 'no religion' has more than quadrupled since the early 1970s." America is largely divided between religious givers and secular nongivers, and the former are disproportionately conservative. One demonstration that religion is a strong determinant of charitable behavior is that the least charitable cohort is a relatively small one -- secular conservatives.

But this destroys Brooks entire thesis! If secular conservatives are the least charitable of all, and religious liberals are more charitable, then it is religiousity and non-religiousity that are the major factors in determining charitability, not conservatism and liberalism.

This is political hackery at its worst. The data indicate one conclusion (the pious are more charitable than the non-pious), and yet the authors ignore that and instead pursue their preordained agenda of showing conservatives to be more charitable than liberals. Subjective definitions, flawed data analysis, and refusing to follow the evidence where it leads mar this study. If this is indicative of the level of science conservatives practice (and my bet is that Will is above right-wing average), it is no wonder conservatives seem incapable of discerning the difference between science and pseudoscience in the arenas of evolution and global warming. They simply do not understand the process of gathering evidence and objectively evaluating it at all.


Jeff said...

Do you have any statistical evidence to prove otherwise? You claim Brooks didn't do the research thoroughly enough, and yet you fail to do any research yourself to disprove his claim. Not to sound harsh or anything, but you would do yourself credit to provide your own evidence...

ScienceAvenger said...

I spelled out in considerable detail the evidence that proves otherwise in my article. Note again that Brooks concedes my point (that giving is tied to religiousity, not political affiliation) in his own article. In any case, your point is off base, for one does not have to gather any evidence to note statistical flaws in a study.