search

AaronGullickson / panethnicity_intermar

Data for "Patterns of Panethnic Intermarriage in the United States, 1980-2018" forthcoming in Demography
MIT License
0 stars 0 forks source link

Test of American Indian/Asian Indian affinity #18

Closed AaronGullickson closed 3 years ago

AaronGullickson commented 3 years ago

I should re-do the ACS model to the relationship between American Indians and South Asian groups separately. If the issue is still present in ACS time period, we should see substantially higher odds of Asian Indian/American Indian intermarriage than other groups' intermarriage with American Indians.

AaronGullickson commented 3 years ago

I found a few articles on this topic, two of which are by Carolyn Liebler. So it appears that this is a recognized issue. Carolyn uses ancestry reports to filter out at least some of the Asian Indians from the American Indian category. To do this, I would have to DL new data as I did not use the ancestry variables.

AaronGullickson commented 3 years ago

I am trying out a specific Asian Indian/AIAN dummy. This can work in the ACS data but it is perfectly collinear with South Asian/AIAN in the 1980 basis data because the only South Asian group is Asian Indians in 1980. So, I will not add this as a general control in the models but I will test a separate model on the full data, to see if Asian Indians stand out from Pakistanis, Sri Lankans, and Bangladeshis here.

This would address the issue of Asian Indians labeling their US born children "American Indian" which seems to be suggested by the previous research. It would not address the issue of people intending to check American Indian checking Asian Indian instead. I may want to ask Carolyn about this.

AaronGullickson commented 3 years ago

Ok, the model fitting results even in the fullest model are producing non-convergent estimates. This is probably because the AIAN and non-Asian Indian South Asian samples are just too small. The point estimates are highly suggestive that the affinity seen here is purely with Asian Indians (the point estimate of the log odds ratio is -1.02 for other groups with AIAN), but I do not know if I can trust it. So I think I will have to rely upon other work here to make the case.