Earlier this month, I showed how CIS’s alltoowidelycited estimate for the scale of birth tourism was unreliable and egregiously exaggerated how many foreign-born mothers are birth tourists. After I replied to Steven Camarota’s defense of his estimate, Camarota conceded one of two major issues I raised with his original estimate.

Rather than revising his estimate with the data we had been using, Camarota found newly released CDC data that makes for much more reliable analysis of births in the US. Camarota’s analysis of the new data concluded that he needed to revise his original estimate down by 20-40 percent to “20,000 to 26,000 possible birth tourists a year” rather than the 50 percent revision the old data would have required.

However, a closer analysis of this new data reveals that CIS’s revised number of birth tourists double-counts about 10,000 mothers. Adjusting for this leaves an estimate of birth tourists statistically indistinguishable from zero.

CIS’s revised estimate still includes tens of thousands of long-term residents who cannot possibly be birth tourists

Before discussing the new data, I’ll note that CIS’s revised estimates still include large numbers of long-term residents. As I covered in my previous posts, the last step in Camarota’s method takes a raw figure (of new foreign-born mothers found in CDC data but not the ACS) that includes citizens, and adjusts it down by 35 percent to try to exclude citizens from the final birth tourism estimate. However, we also can safely say that the raw figure includes—in addition to citizens—green card holders, long-term guest workers, and other long-term residents. Camarota was the first to bring up this issue, but he makes no adjustment to try to exclude them as he did for citizens.

The problem is not any asymmetric lack of data. There is no more data about the citizen share of the raw figure than the long-term resident share. Recall, the 35 percent share Camarota uses comes from an implicit assumption he made: specifically, that new foreign-born mothers in the CDC data but not in the ACS have the same characteristics (e.g., likelihood of being a citizen) as those who are in the ACS. I suggested using the same assumption for long-term residents as citizens. With the new range of CDC figures Camarota offers in his latest piece, this would yield estimates of  2,100-2,700 (slightly higher than our previous 1,800 figure(. However, even this may overstate the number of birth tourists as we’ll see below.

Camarota maintains that the same assumption for long-term residents as citizens has “seeming implausibility.” Is it really implausible to believe that guest workers and immigrants are significantly more likely to go abroad than citizens? It’s surely more plausible than Camarota’s preferred assumption that the share of long-term residents among new mothers who leave the country is zero.

The new CDC data show that CIS’s revised estimates double-counts over 10,000 new mothers

I’ll let Camarota explain the benefits of the new data:

The new Natality for 2016 – 2018 (expanded) data at the CDC website now allows users to generate tables that show exactly how many births there were to mothers who themselves were born in outlying territories. These individuals can now be easily subtracted from the old CDC definition of the foreign-born, making for more accurate estimates. Even better, it is now possible to calculate the number of births to foreign-born mothers by month. This also allows us to more accurately match birth figures from the ACS.

What’s more, the new data allow us to determine separate numbers for births and mothers. As I mentioned last week, one of Camarota’s unspoken simplifying assumptions is that a woman can only give birth to one baby in a year—in other words, that there is no such thing as a twin or other multiple births (or, for that matter, singleton siblings born within a year). The motivating reason behind the assumption is understandable as it allows an easy comparison between ACS data on new mothers and the old CDC data on births.

But with the new data, this assumption is entirely unnecessary. The new data include which births were singletons and which were twins, triplets, quadruplets, etc. This allows us to derive the number of new mothers who were born outside the United States. The new data reveal that the assumption of unique mothers for every birth causes an overestimate of the number of mothers by about 1.5 percent.

Table 1. Estimating Foreign-Born Mothers from CDC Data


Second half of 2016First half of 2017Total
Births to mothers born outside 50 states and DC471,984429,313901,297
    Singletons458,052416,655874,707
    Twins13,50012,26625,766
    Triplets400369769
    Quadruplets321143
    Quintuplets1212
Derived mothers464,943422,916887,859




Births to mothers born in outlying U.S. territories12,36011,52123,881
    Singletons12,00111,11923,120
    Twins356398754
    Triplets347
Derived mothers born in outlying U.S. territories
12,18011,31923,499




Births to mothers with unreported birthplace4,2003,6727,872
    Singletons4,0563,5037,559
    Twins144163307
    Triplets66
Derived mothers with unreported birthplace
4,1283,5917,719




Derived mothers born outside US and territories
452,763411,597864,360
Estimated mothers born outside US and territories* 453,713412,423866,135

Note: Estimation follows the apportionment recommended by Camarota

Comparing the results in the table above to CIS’s figures shows over 13,000 fewer new mothers born abroad. Without the double count, CIS’s revised estimate would have been around 12,000 rather than 20,000.

Table 2. Birth tourism estimates


CDC-derivedACS-derivedBaseline differenceReplicating CIS’s estimateExcluding citizens and long-term residents
Derived foreign-born mothers864,360845,944
(10,962)
18,416
(10,962)
11,621
(6,917)
1,244
(741)
Estimated foreign-born mothers 866,135845,944
(10,962)
20,191
(10,962)
12,742
(6,917)
1,364
(741)

Note: none of these estimates is statistically significant

Further, the 12,000-13,000 figure proves not to be significant. In the ACS, we find 845,944 new mothers born outside the U.S. with a standard error of just under 11,000. My new ACS derived estimate is about fifty higher than CIS’s. This is because I use the birthplace question where Camarota still uses the citizenship question. In any case, the difference is trivial and does not change the significance. However, the new CDC number (still a slight overestimate since it doesn’t account for mothers giving birth to singletons twice within a year), falls within the ACS’s margin of error at the 95% confidence interval.

In short, using the new CDC data with CIS’s method leaves us with an estimate of the number of birth tourists that is not statistically different from zero.

The new rules for pregnant women appear less justified than ever.