This week, the Census Bureau released the results of its formal test of the once-proposed, and now prevented, census citizenship question. The results are a striking rebuke to a recent leftist orthodoxy: Adding a citizenship question to a test-run of the census had no statistically significant effect on overall response rates. Even in the communities most sensitive to the question, such as neighborhoods with many Hispanics, response rates dropped only slightly.
To understand why this report is significant, step back to March 2018, when Secretary of Commerce Wilbur Ross announced that the 2020 census would include a question about citizenship status. Leftists were immediately outraged, seeing the addition as a racist conspiracy.
They made three main arguments: The question addition was motivated by racial animus; adding a citizenship question would dramatically reduce Hispanic response rates to the census, resulting in a large undercount; and anything resulting in a foreknown undercount would be an unconstitutional denial of fair representation to citizens of affected states. So their theory ran that the Trump administration added the question knowing or suspecting it would cause a Hispanic undercount, which would result in blue state populations being underestimated, and thus their congressional apportionment reduced after the 2020 census.
I and many other conservatives were initially skeptical of these arguments. I argued here at The Federalist and at Vox that calling a citizenship question unconstitutional in principle was ludicrous, since similar questions have been asked in almost every census of at least some Americans. Additionally, I argued concerns about Hispanic undercount were wildly overblown.
As for motive, who could say? I did raise a concern that it might technically be too late to add a question to the census, but few commentators focused on that issue.
But today, with the release of the formal test of the citizenship question, we can finally count up the score: Who was right?
First, credit to leftists where credit is due: Trump administration officials, in a veritable festival of incompetence, failed to complete many of the basic administrative tasks required by the Administrative Procedure Act for alterations to the census. Meanwhile, troves of emails were discovered suggesting the administration’s explanations for why they wanted the question were bald-faced lies.
When the Supreme Court ultimately heard the challenge to the decision, it shocked many leftists and conservatives alike. Leftists thought their arguments about racial animus were water-tight, while conservatives felt sure a citizenship question couldn’t be unconstitutional.
In the end, the Supreme Court unanimously struck down the citizenship question because Trump administration officials, through both incompetence and dishonesty, violated numerous provisions of the Administrative Procedure Act. Most strikingly, all nine justices identified Ross’ personal dishonesty about the origins, aims, and intent of the policy as a key reason to strike it down.
In other words, while asking about citizenship is reasonable in its own right, the Trump administration has put such fools in positions of authority that they are bungling even the simplest policy tasks.
The Major Progressive Arguments Were All Wrong
But what of the other key leftist claim? Although it was constitutionally irrelevant, leftists heatedly declaimed that a citizenship question would reduce response rates. The exact forecast varied, but a group of analysts at the Census Bureau eventually produced a series of papers arguing a citizenship question would reduce response rates to the census by 2.2 percentage points.
If true, this would have been remarkable. The citizenship question would have dissuaded about 7 million U.S. residents. Could one question really discourage so many millions from answering the census? When Justice Neil Gorsuch raised a question during oral arguments about the Census Bureau’s research, noting it wasn’t an actual empirical test of the question being considered, leftists mocked him on Twitter and treated him as an innumerate bumpkin.
Well, guess what. Those of us who, like Gorsuch, treated the Census Bureau’s experts with a degree of skepticism were completely correct. Rather than a 2.2 percentage point decline in response rates, the Census Bureau found a less than 0.5 percentage point decline in response rates, and even found that the decline in response rates was statistically insignificant — that is, it could have resulted from random chance rather than a real difference.
In the key, high-risk subgroups where the Census Bureau’s experts projected a 5-8 percentage point decline in response rates, the eventual test revealed a far smaller 1 percentage point decline in response rates. In other words, the actual effect, to the extent it existed at all, was only about one-tenth to one-half as large as experts suggested. Realistically, it was about one-fifth or one-fourth as large. The much-vaunted expert reports were simply wrong.
What makes this particularly galling is that the reports were predictably wrong. I had predicted this error from about day one of the controversy, and as early as last spring had pointed out a list of specific methodological concerns with the papers being cited.
As usual, partisanship is a heckuva drug. Many conservatives were simply unwilling to accept that Ross is woefully unfit for his job and is running a calamitously shoddy operation that would eventually draw the Supreme Court’s ire. And many leftists were deeply wedded to a patently dubious set of unfounded claims about how survey response rates work.
Today, however, we can see how things have shaken out. The left made a decent point: The Trump administration is full of incompetents. But the left also trotted out a panoply of nonsense arguments that have since been proven wrong: Missing a single reporting deadline didn’t make the question unconstitutional, collecting basic demographic information isn’t a privacy violation, the possibility of an undercount can’t nullify the addition of a census question, and adding one fairly mundane question will not cause catastrophic population undercounts.
Conservatives can learn from this experience that if we want to win, we’re going to need people who can tie their administrative shoelaces without falling face-first into lawsuits. Leftists, well, they’ve got quite a lot more to learn.