Trump’s Executive Order on the Census

Hi. Please don’t consider this a preview for the coming week: most of the posts will be a lot shorter. I promise.

Until yesterday, the official government reason for getting citizenship information in connection with the decennial census was the Voting Rights Act. The Supreme Court affirmed that that was a lie pretext. Yesterday’s Executive Order laid out a few more ostensible reasons for getting citizenship information in connection with the census. A few people have asked or wondered aloud why Commerce didn’t just offer the reasons now laid out in the EO, when they were hunting for a pretext before the March 2018 decision.

If the Administration’s real goal was to put a citizenship question on the decennial enumeration (I offer some thoughts here on why that might have been so), and only as a secondary matter to come up with a reason to justify that goal, there’s a answer to the question that doesn’t depend on incompetence. The EO rationales are only surface-level plausible.  Either they don’t actually explain the need for putting the question on the enumeration, or they don’t fit the pre-existing OMB preference for uses reflecting strong federal enforcement needs. Or both.

I have no insight into Commerce’s actual decision-making process, beyond what’s been revealed in litigation.  And so what follows represents the attempt to reverse-engineer a rational search for a purpose to put forward.  YMMV for whether you think this resembles any plausible consideration in the decision making process that actually went on, in whole or in part. 

  1. Despite the assertions in yesterday’s press conference, the Secretary’s March 2018 decision was not a decision for the Census Bureau to start collecting information on citizenship rates from the public.  The census’s American Community Survey already collects citizenship information from the public on a rolling (and hence regularly updated) and fairly detailed basis (including data like country of birth and year of entry).

    Instead, the March 2018 decision was a decision to collect information on citizenship (alone) from the public on the decennial household-to-household 100% enumeration form.

    That sort of collection risks fomenting inaccuracy (inaccuracy quantified by Census Bureau experts in what they explicitly designated as an exceptionally conservative estimate, and undersold by SCOTUS as a predictive figure, without that same caveat).  The size of the inaccuracy was disputed, but not the fact of the inaccuracy.  So the decision to forge ahead despite that inaccuracy required at least some articulable benefit from asking for the information on the enumeration, over and above either administrative records or surveys.  The ostensible benefit Commerce settled on was the illusion of block-by-block precision purportedly needed for Section 2. 

    Beyond redistricting, none of the other uses in the Executive Order (overall national immigration policy, overall national benefits expenditures, various broad legislative purposes concerning the undocumented population) seem to offer a plausible meaningful incremental benefit (over surveys or administrative records) for household-by-household collection of the one incremental citizenship question sufficient to justify the lapse in accuracy.  That is, the other uses beyond redistricting in the EO are pitched at state or national estimates (where the ACS margins of error are really quite small) rather than block-level estimates, and often require information about year of entry, country of origin, or immigration status that the single citizenship question wasn’t designed to get.

    Or more simply still: except for redistricting, these other applications might explain why the Administration would want ACS data it’s already got.  But they don’t really purport to explain why they’d need one extra household-by-household question on citizenship.  The VRA pretext offers an excuse for block-level data on citizenship (without the extra questions on year of entry or status or the like).

  2. I don’t know offhand the process for reviewing content on the household-to-household enumeration, but there’s a rather detailed OMB-fostered process for reviewing content on the ACS.  It privileges information to be used by federal agencies, and particularly privileges with the following priority: 1) data explicitly required by statute to be gleaned from the census, 2) data required by statute or case law and naturally derived from the census, and only thereafter 3) data useful for programmatic planning.  Data without mandatory or required uses by federal agencies, and only programmatic uses at geographic levels larger than counties, were lowest priority, and essentially flagged for likely deletion from the ACS.  And would have been tricky to sell as prime candidates for elevation to the household-to-household enumeration.

    This means that in the existing OMB procedure, the gold standard (at least for ACS content) was information needed by a federal agency in small levels of geography in order to perform a function specifically required by statute or caselaw.  (I’m assuming that the process for enumeration content – to the extent there was a process — was at least as restrictive.)  “The statutes/courts make me do it” is the best reason to collect information.

    That explains why Commerce first went shopping to federal agencies – DOJ, DHS, DOJ again – for some requested justification, rather than just saying “hey, we’re adding the question because this stuff might be useful for someone.”  The VRA pretext met that threshold of caselaw requirement for data in small levels of geography needed for federal agency enforcement.

    None of the other uses in the EO do the same thing.  The first few have to do with changing laws and appropriations and broad national policies, but not enforcing existing statutes at the local level.  These rationales in the EO seem pitched primarily at helping Congress pass better laws.  But if Congress needs information to inform its policymaking, Congress can add whatever information it wishes to whichever Census instruments it wishes (which may well be why, in the absence of more specific legislation, the OMB process privileges data to assist executive functions).   And the redistricting rationale in the EO isn’t a federal agency rationale either – it’s entirely dependent on the cited interest of “some State officials.”

I’m not suggesting that there aren’t other hypothetical pretexts that would both 1) fit the preference for data required for federal agencies to do their existing jobs and 2) make the case for a single question on citizenship collected at the block level, yielding information ostensibly more suitable than the stuff already collected on the ACS.  But in that respect, the VRA nonsense did deliver a two-fer.  (To be clear, the VRA pretext doesn’t really hold up on the merits either.  But on these two axes, it might hold up a bit better than the EO alternatives.)


Comments are closed.