Thread by @KlasfeldReports

Key witness up today in the #2020Census case: Dr. John Abowd, chief scientist at the U.S. Census Bureau. Live-feed ahead.

In case you missed it, video released on Friday. ⬇️

https://twitter.com/KlasfeldReports/status/1061321763529936896

The @ACLU's Dale Ho questions him by going over his CV, moving onto Ross's decision memo.

Ho notes that Ross said the citizenship question was necessary to enforce the Voting Rights Act.

Abowd notes that's true that Ross said that.

Ho notes that "as the chief scientist of the census bureau," Abowd and other census bureau leadership have consistently recommended against a citizenship question.

"Correct," Abowd answers multiple variations of this question.

Q: The memo that you wrote to Sec. Ross, in your opinion, that memo memorialized that ... could be expected to lower the self-response right among households that include non-citizens.
A: Yes, that's correct.

Abowd agrees that's a "large population."

Q: You agree with the conclusions in this memo, isn't that right, Dr. Abowd?
A: Yes, I do.

It as routed to the Secretary of Commerce, and Abowd had a meeting with Ross about it.

Q: "Let's be clear: Secretary Ross had only one meeting the chief scientist of the Census Bureau" before adding the citizenship question.

Abowd agrees.

Abowd agrees that his memo advised against "Alternative B" (i.e. the citizenship question) as "very costly" and "harms the quality of the census counts."

"I support the conclusion that there would be a lower self-response rate and the consequences of that self-response rate, yes," Census Bureau chief scientist John Abowd, on the effects of the citizenship question.

Abowd: "The best estimate we have is 5.8 percentage points," on the differential of adding the citizenship question.

The breakdown:

* 9.8% of households contain non-citizens
* 5.1% estimated lower self-response
* Affecting 630,000 households

Asked about the math of that, Abowd said: ""That's more than a million people, yes."

Q: The 5.8 estimate. That's a conservative estimate, right?

Abowd responds that "conservative" is hard for a statistician to define, but with that caveat, yes.

"It's the best available point estimate in the decline in self-response," Abowd says.

It's as good as the analysis that led up to it can be.

Ho notes that 5.8 estimate is based on responses to the ACS, but the question would be "more prominent" on the decennial census.

Abowd agrees.

Ho pulls up Abowd's white paper for the witness.

Q: In your January memo to Secretary Ross, you concluded that adding a citizenship question would be a sensitive question for Hispanics.
Abowd agrees.

"Increased sensitivity is reflected through the 2017 data, yes," Abowd testified.

Q: You believe that the Census Bureau did provide empirical support for its belief that adding a citizenship question to the census would reduce response rates.

A: Self-response rates. That's correct.

Ho is now grilling Abowd about the lack of RCTs (randomized control testing).

Ho asks whether the political climate around immigration could make the citizenship question affect the count.

A: Yes, it could.

Noting that thee last time the census had a citizenship question was in 1950, Ho asks: "You agree that the macro environment is a little different now."

"Well, I'm not a macroeconomist," Abowd notes, before adding that he believes it is.

Ho reads a focus group report: "Participants in all locations asked about the citizenship question before the moderator asked about it."

They said that they are personally citizens, but they would not answer out of fear, he notes.

Q: People who are afraid of deportation will be an extremely difficult group to count.
A: They would be a very difficult group to count yes.

Ho presses him on whether they'd be "extremely" difficult to count.

Ho asks whether Hispanic focus group members said members of their community care about the census.

Abowd says they did.

Q: Members of this group would be more likely to self-respond if there were not a citizenship question on the census.
A: Yes.

One participant in the Census focus-group testing on citizenship question (CBAM):

"Latinos will not participate out of fear. There’s a hunt for us... It’s like giving the government information: 'Oh, there are more here.'"

We’re back for the afternoon session of the #2020census trial.

Ho asks about whether adding citizenship question will increase costs of census.
Abowd says it will.

By $27.5 million, conservatively, he testifies.

This is a “lower-bound” estimate, Abowd says, because less self-reporting means more NRFU (non-response follow-up).

Another cost will be increased communication campaign expenditures.

Abowd testifies that the citizenship question also will lower the quality of the data.

Q: That’s a bad thing, right?
A: That’s something we try to avoid, yes.

Abowd agrees that Census Bureau found that using administrative records for CVAP data would be a better method and “best meet DOJ’s stated uses.”

He said he communicated that to Ross in February meeting.

Ho asks to play a short selection of John Gore’s video deposition.

DOJ objects. Calls footage unnecessary.

Judge Furman overrules. It rolls.

AG personally directed staff not to meet with the census bureau staff on how to obtain higher quality data.

Asked if this refusal is unusual, Abowd replies: “Yes, it is.”

The Census Bureau believes that noncitizens give citizenship question answers that are inconsistent with the administrative records more than 30 percent of the time, Abowd testifies.

Ho argues that using administrative records to ascertain citizenship data would be more accurate than using self-reported survey data.

Abowd agrees.

“The most accurate data would come from modeling their citizenship status, that’s correct,” Abowd says, contrasting it to soliciting survey data.

Ho notes there has been no cognitive or field testing of the 2020 census, unlike in 2010.

Abowd agrees.

Abowd agrees that a citizenship question could be more sensitive to some respondents than asking for a social security number.

Afternoon recess. Back in 10 minutes.

Census Bureau doc says that the bureau did not feel “bound by” past precedent regarding the citizenship question. Ho presses Abowd to agree that nobody at the Census Bureau wrote that.

“Not to the best of my knowledge,” Abowd replies.

Abowd says that he was not aware about Sec Ross’s conversation with Kris Kobach, or communications relating to Steve Bannon.

Ho plays another video from Gore’s deposition, again over DOJ’s overruled objection.

“My understanding was those conversations were initiated by the Department of Commerce,” Gore said in the footage.

Abowd said he didn’t know that Commerce, not DOJ, initiated.

Abowd agrees that fact “surprised” him.

Q: Among the senior executives at the Census Bureau, everyone you know was also surprised... [by that order of events].

Abowd agrees.

Ho asks Abowd was whether he was under the impression that his worked mattered in convincing Ross about the citizenship question.

“I was under the impression that it mattered within the context of the 2020 census,” Abowd says.

Ho pivots to measures to protect confidentiality of data: “household-level swapping” and “synthetic data noise infusion.”

Cross-examination by the government begins with Abowd talking about his CV.

One immediate observation:

DOJ attorney’s tone is much friendlier than Ho’s, as Abowd is effectively an adverse witness - called by the plaintiffs and employed by a defendant.

Abowd’s testimony against the citizenship question notwithstanding.

Still on Abowd’s CV.

Tl;dr, he’s impeccably credentialed.

They’re still talking up his credentials.

We get it.

Proceedings over for the day. Abowd’s testimony will continue well through tomorrow, with the plaintiffs’ case (perhaps) wrapping up then.

I will have one story for both days of testimony and so perhaps no wrap-up today.

See you tomorrow.