Am I color-blind? How can this tax map be correct?

This mind-boggling Tax Foundation map came through my RSS feed this morning.

Although the American public might well be inclined to accept it at face value, I don't see how this map can possibly be right, unless there is something extremely weird going on in residential segregation patterns when broken down by zipcodes.

Perhaps I am colorblind, but it appears to me there are many zipcodes showing up in maximally bright blue on the map.  I'd really like to see the data behind this map, because it is very hard for me to discern the variations in subtle colorations.

According to the map legend,  taxpayers in each of those bright blue zipcodes paid an AVERAGE of over 25% of their adjusted gross income (AGI) in federal income tax in 2008.

But--according to the IRS, in 2008, every single income bracket paid an average of LESS than 25% of its AGI in federal income tax.  The bottom 50% paid 2.6% of their AGI in federal income tax.  The top 10% paid 18.7% of their AGI in federal income tax.  The top 1% paid 23.3% of their AGI in tax.  The top 400 taxpayers in the country paid 18.11% of their AGI in tax.

Now, it is true that there are wide variations in effective tax rates WITHIN each income category, for a variety of reasons.  For example, at any given income level, there are some taxpayers with more tax preferences than others (some have more children than others; some have a large proportion of their income in lightly taxed capital gains and dividends rather than in more heavily taxed labor income, some have large amounts of itemized deductions such as mortgage interest, state & local taxes, charitable donations, etc.)

So, for sure, there must be a few individuals who paid more than 25% of their AGI in federal income tax.  For example, President Obama and his wife paid 29.6% of their AGI in tax in 2008.

But it is completely inconceivable to me that there could be a large number of zip codes where all taxpayers in the zipcode collectively paid total taxes averaging over 25% of their AGI.  This is especially true for the apparently bright blue zipcodes in New York and New Jersey, because itemized deductions tend to be large there (high state taxes and big mortgages lead to big Schedule A deductions in those places.)

A maximally bright blue zipcode could only happen if all taxpayers in the zip were clustered as follows:   (1) extremely segregated by income (no low-income working families living in the zip to pull the average down) AND also (2) extremely segregated by unwillingness (or inability) to take advantage of tax preferences taken by most folks in their respective income brackets.

This seems unlikely to me--as I said above, I'd really like to see the source data behind the map, because it is very hard to discern shades of bright blue.

(It's also important to note that a graph of total federal taxes paid divided by AGI would look VERY different from the above!  And a graph of total taxes divided by AGI by zip code would look even more different!  And, if we used a theoretical ideal denominator, one based on Haig-Simons income rather than AGI, it would look still more different.)

UPDATE:  Originally, all that came through my RSS feed was the unadorned graph.  I have now found a blog post giving more details.  Using their cited datasource, I tried to come up with some numbers for zipcodes near me (where I have some idea of the income distributions), but for some reason the NY file does not list any zipcodes higher than 10502.  (Albany-Schenectady-Troy zips are up in the range starting with 12xxx, so I'll have to figure out why they are missing.)

UPDATE 2:  I received the following response from the IRS statistics division replying to my query about why there were no zipcodes above 10502 reported for New York.

In response to your request, there was a change in the number of returns needed to be included in the Tax Year 2008 ZIP code data.  If a particular ZIP code has fewer than 250 individual income tax returns, data from that particular ZIP code is not included.  

This response is completely mystifying, since there are plenty of zipcodes above 10502 in New York with more than 250 returns.  For example, my own zipcode (12309) had over 14,000 returns in 2008.

UPDATE 3:  Still no further word from the IRS statistics division, but thanks very much to Nick Kasprak at the Tax Foundation, who helpfully provided a complete datafile of the entire database, including upstate New York.  I'll see if I can make sense of the patterns that emerge in interpreting the data in nearby communities I know well.

Update 4:  A short amount of investigation quickly turned up a very bizarre anomaly in the IRS data for a Manhattan zipcode.  For more details, see my next post, here.