Last week, I showed how CIS’s widely-circulated estimate for the number of birth tourists was not a reliable figure and instead severely overstated the scale of birth tourism. This morning, CIS’s Steven Camarota published a response that defended his estimate. Unfortunately, the explanation proves unconvincing and reveals still deeper problems with his method.
Misunderstanding the data
Camarota says that it “seems [I] mistakenly used the year-of-arrival question in the [American Community Survey (ACS)] to identify new foreign-born mothers, not realizing that many U.S.-born women also give a year of arrival in the ACS.”
Both claims are incorrect. Only the foreign-born are included in the year-of-arrival question. In any case, I did not use the year-of-arrival question, but instead used the birthplace question, which is the obvious way to tell if someone is foreign-born.
In my original post, I reported that I found it impossible to replicate Camarota’s number, but that “this is not necessarily a problem if CIS can provide more information.” Camarota’s reply gave me the information I had hoped for, allowing me to replicate his number (and revealing what I had identified as a subtraction error was truly a transposition error). Specifically, we learn that Camarota attempted to identify the foreign-born using the ACS question on citizenship status, rather than birthplace.
Camarota is mistaken to use citizenship status rather than birthplace to identify foreigh-born status, because although similar, they are far from equivalent. This distinction is especially important when we are comparing ACS data to CDC data, which explicitly identifies the relevant category as “mother’s place of birth.”
In fact, the CDC only distinguishes between mothers “born in the 50 states and the District of Columbia” and mothers “born elsewhere.” This means that when Camarota excludes tens of thousands of recent mothers from Puerto Rico, Guam, and other U.S. territories from the ACS number because they are U.S. citizens, he is excluding mothers from his ACS number who are included in the CDC number. The effect is to inflate his estimate by tens of thousands.
In addition, Camarota accuses me of “assuming that new foreign-born mothers not in the ACS, but found in the CDC data, have the same characteristics as those who are included in ACS data.” He adds that I do “not provide any reason why [I think] this is the case.”
As it happens, I don’t think this is the case. However, the assumption is not my own—I freely borrowed this assumption from Camarota, who made use of it when he adjusted the difference he found between the CDC number and his ACS number downward by the citizen share from those in the ACS data. (I’ll add that he only used this assumption in the 2017 estimate and not in his 2012 estimate, which makes his comparison of the two estimates questionable).
I never intended my piece to be an endorsement of Camarota’s method for estimating the scale of birth tourism or the assumptions his method requires. I left unchallenged a number of questionable assumptions underlying CIS’s method: not only that the ACS characteristics describe those of the “missing” mothers, but also that that a mother can only give birth to one baby within one year, that the ACS does not systematically undercount immigrants, that the ACS includes tourists, etc.
I only intended to show that if we adopted Camarota’s assumptions, reasonable or not, then we should find a much lower number than he reported. I leave it to the reader to determine whether I succeeded.