The IRS and a statistics problem

by Andrew Gelman on July 22, 2013 · 5 comments

in Bureaucracy,Data

John Mashey writes:

Have you or do you know if anyone has looked at this hypothesis:
“The IRS unfairly targeted conservative organizations for scrutiny in getting 501©(4) status.”

I’d think one would want to have:
N years of rates of application (because I conjecture there was a big burst after Citizens United)
Some grouping of those into “conservative,” “progressive” or other.
Rates of investigation, time taken.
Rates of rejection.

I haven’t followed this carefully enough, but it actually looks like an reasonable social science question, if only to identify the
data actually required to reach any real conclusions.

I have no idea.


Scott Monje July 22, 2013 at 5:24 pm

About a dozen people in Cincinnati processed some 70,000 applications a year, and apparently a lot of them (the applications, that is) were pretty vague. Given the dimensions of the problem, it’s hard to believe we will ever have detailed knowledge of the entire set. In addition to Citizens United, I recall something about a change in the rules prompting a lot of existing 501(c)(4)s to reapply for tax-exempt status. But doesn’t it seem likely that there would have been a surplus of conservative applications? This was the breakout period for the Tea Party, so there would have been many of them forming, and I would imagine that they would have been more likely than progressives to take cues from Karl Rove’s Crossroads GPS, the model for using 501(c)(4) organizations for political purposes.

jonathan July 22, 2013 at 6:46 pm

The IG was asked to do this – and not to investigate whether other groups (like progressive ones) were similarly investigated and then that part was left out of the report, apparently by the initiative of GOP Congressman Issa. But beyond that, none were denied, some were delayed, so what exactly would a study show?

denguyfl July 23, 2013 at 5:59 am

What percentage of all applications are conservative and what percentage are conservative would be useful as well. saying that X number of conservative groups where reviewed while only Y number of liberal groups is useless if there is way to judge if that is unreasonable. If X and Y are in proportion to their total numbers in all filings there ( i.e if 3 out of 4 being reviewed are conservative, that would be completely within the realm of reasonableness if 3 out of 4 applications were conservative). Anecdotal evidence of under 300 groups is useless when the pool of total applications is as large as it is. This is simple statistics and doesn’t require Nate Silver.

Andrew Gelman July 23, 2013 at 9:17 am


Don’t sell Nate short. Simple statistics are good too. What’s important is not so much how fancy the analysis is, but rather what data are included.

JG July 23, 2013 at 8:57 am

Isn’t this a non-issue now? Moreover, is a hypothesis test even necessary? We know that 0 conservative groups were denied 501c4 status and 0 conservative groups were stripped of their 501c4 status while at least 2 progressive groups lost theirs.

Small numbers, sure, but as with the Congressional investigation “there is no there, there.”

Comments on this entry are closed.

Previous post:

Next post: