I appreciate that the New York Times piece discussing the Fish v. Kobach verdict no longer misleads people quite as dramatically as some earlier coverage concerning my methodology. That said, I am disappointed that it continues to express confusion about and ridiculously magnify a tiny aspect of my report and testimony that I spent a long time discussing with the reporter.
“People with the last name Lopez, for example, were counted both ways. In perhaps the trial’s most startling moment, Mr. Richman said he would have counted as foreign the name Carlos Murguia — which happened to be the name of a judge on the Kansas Federal District Court.
Over the phone with me, Mr. Richman called the moment “ridiculous” and the fallout from it “painful.”
I am quoted in the Times story in just one word — an assertion that the effort by the ACLU attorneys and a previous Times story and in other accounts to make anything of this coding was “ridiculous.” One word taken out of context was hardly sufficient to characterize my argument though.
As part of my analysis of the KS survey of individuals on the suspense list I weighted the respondents by a number of demographic variables including a coding of the respondent names. The goal of this weighting, along with weighting by gender, age, and first three digits of zip code among other characteristics, was to investigate whether the respondents to this survey were representative of the broader population on the suspense list. The name coding was based on whether at least two of the respondents names appeared to come from the same national origin. The intention was for someone with the name of Vladimir Putin to show up as having a foreign name, but someone named Vincent Putin not to.
The story points out that individuals with the last name of Lopez were counted both ways, as they should have been if the coding was applied consistently (and it was — inter-rater reliability was 95 percent). Therefore, this alleged inconsistency primarily reflects correct application of the methodology. And as such it is ridiculous to make anything of it.
This is not and never was, a coding intended to identify specific individuals as citizens, immigrants, or anything else. And contrary to some misreporting in earlier stories, it was never used to identify whether anyone was a citizen or a non-citizen. My goal was to create a measure correlated with demographic characteristics and as such useful for partially correcting a bias if it existed in the degree to which individuals with different characteristics who were on the suspense list responded to the survey. If the coding is applied consistently, then it can be used to weight responses. It was applied consistently. Coding based on names is a common practice. Two of the four papers I recently discussed for a panel on voting issues at the recent State Politics and Policy Conference relied on some form of coding based on names in voter files to estimate demographic characteristics. Attacking use of a coding based on names to ensure that people in different name categories are appropriately represented in a survey is ridiculous.
It was also ridiculous that the ACLU and various reporters following their lead made anything of the fact that a correct application of my coding would identify someone named Carlos Murquia as having a foreign name. Of course it would. Like measures for age, zip code, gender, etc., the purpose here was to weight survey responses to assure that the sample was representative.
Finally, it is ridiculous to make anything of a coding that ultimately made little difference in the results. As it turns out, the sample was already largely representative. Weighting on all of the characteristics (age, gender, zip code, name, etc) shifted the estimated number of non-citizens in the sample from 7 to 9, a change of only two.