Ah, but Ezra, where can we find statistical modeling in the Bible? After all, if knocking on people's doors was good enough for King Herod, it is good enough for us.

we have a vicious fight over the legitimacy of statistical modeling

You've obviously never been in the room when a Bayesian and a Frequentist start going hammer and tongs at each other.

I'm not familiar with the details, but I think it's commonly understood that the statistical models employed by the Census failed in 1990 and may or may not have failed in 2000. That said, I'm in favor for the obvious reasons.

One thing you don't touch on is the dishonest argument against employing statistics in the Census. No one is seriously proposing sampling. There's an argument that once we start using statistical modeling at all, it's a slippery slop to aiming for less than a complete count, but that's not what's on the table. What folks on the left are arguing for is statistical adjustment - determining the inevitable undercount and adding it to the as-complete-as-possible headcount.

I'm 99% certain that this distinction will be lost on pundits 99% of the time as this is debated over the next few years. Note that modeling will be done -- the debate will actually be after data collection concludes and it's time for Commerce to certify the unadjusted or adjusted headcount.

The absurd thing is that the arguments against statistical adjustment have nothing to do with Rahm Emanuel rigging the results in a smoky room. It's not like gerrymandering where folks run all of the models that they think will pass a legal test and pick the best one; there are actual career civil servants involved in every aspect of this. The only politically sensitive question is whether or not to correct for the undercount -- the employment of such a model would be entirely in the hands of professionals and would, I think, be very transparent.

Must. Resist. Urge. To. Reference. West. Wing.

*cough*Mr.WillisofOhio*cough*

The Supremes invalidated a form of statistical sampling in "Department of Commerce v. House of Representatives", and sent a pretty clear signal that they will be strict on enforcing the Constitutional requirement of an actual enumeration.

So are statistical adjustments even legal? Or is the role of the Secretary of Commerce in this census to do what he can to find every single last person living in this country?

If the latter, I can see why Rahm wouldn't want a Republican setting policy on just how hard to look in traditionally undercounted areas.

If the latter, I can see why Rahm wouldn't want a Republican setting policy on just how hard to look in traditionally undercounted areas.

Yeah. Remembering all those Bush departments of regulation nonenforcement and the Federal Emergency Unpreparedness Agency, I don't want a Republican overseeing any government agency where failure to do the agency's job could in any way advance a Republican agenda. The statistical arguments only start after the raw data is in, and the person who decides how hard to look for the raw data, and how much money to spend doing so, had better be a person committed to finding all the data it is humanly possible to find. Given the modern reality of party politics, in which Republican-leaning voters are easy to find, Democratic-leaning ones not so much, that means a Democrat.

@Francis

Current law is that it's illegal to use adjusted numbers as the basis for redistricting; still can be used in all other matters. So yeah, another reason that the anti-adjustment argument is bogus. It's important going forward to avoid calling it "sampling," though, since as I noted above that implies not trying as hard as feasible for a complete count. Here's the relevant part of the decision you cited:
The Census Act prohibits the proposed uses of statistical sampling to determine the population for congressional apportionment purposes. In 1976, the provisions here at issue took their present form. Congress revised 13 U.S.C. § 141(a), which authorizes the Secretary to “take a decennial census … in such form and content as he may determine, including the use of sampling procedures.” This broad grant of authority is informed, however, by the narrower and more specific §195. See Green v. Bock Laundry Machine Co., 490 U.S. 504, 524. As amended in 1976, §195 provides: “Except for the determination of population for purposes of [congressional] apportionment … , the Secretary shall, if he considers it feasible, authorize the use of … statistical … ‘sampling’ in carrying out the provisions of this title.” Section 195 requires the Secretary to use sampling in assembling the myriad demographic data that are collected in connection with the decennial census, but it maintains the longstanding prohibition on the use of such sampling in calculating the population for congressional apportionment. Absent any historical context, the “except/shall” sentence structure in the amended §195 might reasonably be read as either permissive or prohibitive. However, the section’s interpretation depends primarily on the broader context in which that structure appears. Here, that context is provided by over 200 years during which federal census statutes have uniformly prohibited using statistical sampling for congressional apportionment. The Executive Branch accepted, and even advocated, this interpretation of the Act until 1994. Pp. 16—25.

Francis: The Supremes invalidated a form of statistical sampling in "Department of Commerce v. House of Representatives" ... I can see why Rahm wouldn't want a Republican setting policy on just how hard to look in traditionally undercounted areas.

Great description of the real arguments, which will undoubtedly be lost on pundits wailing about lies, damn lies and statistics.

Given that the GOP has shown just how committed it is to ensuring all eligible voters have the right to vote, you can extrapolate just how much time they'd devote to ensuring the accuracy of the census.

The reason that the apportionment of House seats is an issue, though? The idiotic continued capping of its total size. Something that can be changed by law. Add another couple of hundred seats, and you'll frankly not hear the kind of complaints that come now when seats get taken from one state and given to another.

I believe Gregg's issue was that he wanted certain people counted as 3/5 of a person.

It all went downhill from there.

Just put Nate Silver in charge: Problems. Solved.

So let's summarize the true story of the Gregg nomination: Obama foolishly put him up as a hosanna to bipartisanship and possibly to get him as right as possible on entitlements (which is not far; the current Congress will block). Gregg accepted so he could control the census. Various advocacy groups complained about Gregg's control of the census, so Obama took it away from him, no big deal, so far as Obama knew this was not about the census. But it was, so Gregg withdrew suddenly in a way timed to maximally undercut Obama.

There are a lot of people. And some of them -- mainly immigrants and the urban poor -- don't come to the door when an unknown government bureaucrat knocks for an impromptu question-and-answer session (this instinctive mistrust of unfamiliar federal employees is the sort of thing you'd think conservatives would be sympathetic to).

Yes, you would think that. Right you are Ezra Klein.

In fact you're so right that I wonder where exactly you got your evidence that the people who don't answer a census are "mainly immigrants and the urban poor"? How do you know that exactly, since they aren't answering census queries, and since (as you point out) there are other groups who might be just as motivated not to answer government queries?

I guess it's something you "just know", huh? And that's why the numbers for the people you "just know" are being undercounted need to be adjusted upward - to a level that you, and people like you, "just know" to be more accurate. Right?

Yeah, that's kinda the objection.

I wonder where exactly you got your evidence that the people who don't answer a census are "mainly immigrants and the urban poor"?

From the Census Bureau. Since 1940, in fact, when more young black men showed up for the draft.

Sorry to make that extended snarkfest of yours look pretty pathetic.