Friday, December 03, 2004

Lies, damn lies - and statistics.

While googling for a site comparing effectiveness of Canadian Charities (which I didn't find yet) I came across this 'Generosity Index' produced by the Catalogue for Philanthropy to indicate the relative generosity of American tax payers when it comes to charitable donations, broken down state by state. The generosity rank is calculated like this:
Having Rank, ranks each state by their average gross income.
Giving Rank, ranks each state by their average donation for tax returns with a charitable donation itemized.
Generosity Index, the ranking of each state based on its Having Rank - Giving Rank.

Looking through the table of relative generosity there are some surprising results. For example the richest states all seem to be clustered towards the bottom of the generosity rankings, Wisconsin, Rhode Island and New Hampshire - all in the top 5 when ranked by average income - are in the bottom 5 of the generosity ranking. At the other end of the spectrum, Mississippi - dead last for average income - was top of the generosity scale.

Now it could be that these results are just striking evidence that people who have money like to keep hold of it. Or that people surrounded by poverty are more likely to give as generously as they can. I am sure both are true to some extent, but I was still surprised at just how dramatic a picture these numbers paint.

Googling around that I found several blogs that interpreted the data in the index in terms of the red state, blue state split (because we haven't heard enough about that particular split). To summarise, some bloggers are using these numbers to show how generous the red-staters are in comparison to their mean blue-state compatriots. According to Technorati there are over a dozen blogs that have linked to the index in the last week, here are a few examples: one, two, three.

Note: Thanks to Frank Lynch for pointing out that the Generosity Index figures depend on the amounts given by donors who give to charity and itemize their donations on their tax return, many people do not itemize donations for reasons explained here.

Now I'm going to try to avoid getting into the politics of this issue, as I have been paying more attention than is healthy U.S. politics of late, particularly considering I don't get a vote there. My point is one about statistics, I am not a statistician (or even particularly good at mathematics) but the way this index is being presented does seem to be a little skewed to me. The calculations only take into account those taxpayers that give to charity. The Generosity Index (GI) calculations do not take any account of the proportion of taxpayers that give to charity, just the amount given by those who do give. Here is an extreme hypothetical example of the gaps this leaves in the information provided by the index: If Texas had only two taxpayers who donated to charity and itemized those donations on their tax return, but each of these generous cowboys donated $5 billion, then Texas would be top of the Generosity Index (no offence Texas, actually you have 1,835,627 charitable, itemizing taxpayers). The problem here is that the GI only takes into account those taxpayers who give and itemize, it ignores those who do not. So a state that has a relatively large proportion of the population who give to charity is actually likely to go down in the GI ratings - not up as you might expect. This is because smaller itemized donations (even though they may be all that lower income people can afford) drive a state's average donation amount down and the GI goes down with it, while the fact that more people are giving does not have any positive effect on the GI.

Back to my original point, this Generosity Index is being used as 'proof' on several blogs that people from the red states give more to charity that those from the red states do. My point is that, as the title of this post suggests, statistics aren't always what they seem. Below is a table that I put together from the very same data supplied by the Catalogue for Philanthropy, but this time sorted by the proportion of charitable, tax return itemizing, donors from each state. If I was playing the political game I could use this to say that it is the red state people who are generous and the blue staters who are mean, but that isn't my point.




Table of Generosity Index by State


Thanks to Mark Twain, Benjamin Disraeli or persons unknown for the quote 'Lies, damn lies - and statistics.'

posted by Patrick ASD @ 8:50 PM

1 Comments:

At 2:12 PM, Anonymous Anonymous said...

I *love* your hypothetical extreme of the two Texans. That was a
wonderfully simple demonstration of one of the major problems.

One important distinction: there is a difference between charitable givers
and itemizers. Your language suggests that only those who itemize have
given to charity. The way US tax laws work, if your itemized deductions
don't pass a certain threshold, it doesn't make sense to go through the
effort of itemization, but to just take the "standard" deduction.
Achieving the threshold may not be a function fo your giving so much as it
is other factors, such as owning your own home and having mortgage
interest (which goes towards the threshold). So non-itemizers may really
be giving something, we just don't know how much.

If we take the GI people at their word (that their source says itemized
deductions account for 60% of all charitable deductions), and figure
perhaps 20% more comes from individuals (with the last 20% coming from
corporations -- not unreasonable, from something else I saw), we see the
error of generalizing to 100% from 30%. We could make assumptions about
how to distribute the remaining 20%, but the whole thing is already
fraught with so many assumptions, it's all very tenuous.

(Another issue I only recently realized is that the GI is calculated not
on taxpayers but on tax *returns* : two taxpayers filing jointly count as
one return, not two people. If a state suddenly achieved a higher marriage
rate and more people filed jointly, it's adjusted gross income per return
would automatically increase without any change in the economic wealth of
the state beyond those associated with household economies of scale. Thus,
states with higher marriage rates should have an "advantage" in the having
ranks. I have no idea whether this would inflate red or blue states
differentially; it might actually work to the benefit of red states, but I
bring it up as another complexity not considered in the GI itself. The
red/blue splitters can do with it what they want, it's all tenuous.)

Frank

 

Post a Comment

<< Home


Technorati Profile