1.5 Generation

[AaronGullickson]

Right now, I am coding birthplace without regard to the 1.5 generation among the foreign-born, those that came to the US as children. Its reasonable to think that this population might be more likely to think of the US as their birthplace in terms of finding partners. I can think of a couple ways to handle it:

• Set some age threshold for US arrival and replace actual birthplace with USA if below the threshold. This is easy, but doesn't allow me to see if the effects for the 1.5 generation really are different.
• Include a dummy variable for the 1.5 generation based on an age threshold and interact this with the birthplace endogamy variable. Its a little tricky to think how to do this in the marriage models with both spouses. It gives me more information on 1.5 generation differences, but I am not sure it a correct way to think about the process, because it just assumes greater exogamy rather than a different frame of reference for endogamy.
• consider USA or country of birth to be endogamous.

None of these seem quite right at the moment, so I will have to give this more thought.

[AaronGullickson]

Pan-Ethnicity and Intermarriage

[AaronGullickson]

Ok, I think I might have something that works. I keep the birthplace variable as is.

• Add another set of dummies for 1.5 generation that are as follows:
  • TRUE if one partner 1.5 generation and birthplace endogamous (x1)
  • TRUE if both partners 1.5 generation and birthplace endogamous (x2)

The result for x1 is expected to be negative and will tell me that 1.5 generation is less birthplace endogamous than true first generation. The result for x2 will tell me if there is a tendency for the 1.5 generation to marry other 1.5 generation people from the same country.

[AaronGullickson]

Ok, on reflection. I don't like that method. I think the way I want to do this is set certain age cutoffs for entry into the US. At a certain cutpoint, birthplace will just be re-assigned as USA (e.g. under 6 years of age). I might also add a second cutpoint where the union is considered endogamous both for country of birth and US (e.g. under 16 years of age). I should look at the literature for some of the common definitions of 1.5 generation and then code up a few alternate birthplace_endogamy variables. I can then test them out and decide on the best one based on model fit.

[AaronGullickson]

I put an article by Rumbaut in the Zotero library that distinguishes the 1.25, 1.5, and 1.75 generation. The 1.75 generation is the one that came to the US age 0-5, the 1.5 generation came age 6-12, and the 1.25 generation that arrived aged 13-17. I think I can use this basic framework to create variables that potentially treat each of these categories differently, like so:

Gen	All 2nd	Gradation 1	Gradation 2	Gradation 3	Gradation 4	Gradation 5	Full USA
1.75	Birthplace	USA	USA	USA	USA	USA	USA
1.5	Birthplace	Birthplace	Both	Both	USA	USA	USA
1.25	Birthplace	Birthplace	Birthplace	Both	Both	Birthplace	USA

Gen	Full Flexible	Partial Flex 1	Partial Flex 2
1.75	Both	Both	Both
1.5	Both	Birthplace	Both
1.25	Both	Birthplace	Birthplace

I believe that these 10 possible coding schemes exhaust all the logical cases where a coding for a later generation can't be more "liberal" than an earlier one (out of 27 altogether).

I will build a function in R to calculate birthplace endogamy variables based on all 10 of these criteria. I will then run 10 test models on the ACS data using the simple exogamy codes. I can compare the deviance of models directly since they will all have the same degrees of freedom. This should help me decide on the best coding scheme.

[AaronGullickson]

here are the results!

model        gen1.75      gen1.5       gen1.25       deviance
-----------  -----------  -----------  -----------  ---------
flex2        Both         Both         Birthplace     1989358
flex1        Both         Birthplace   Birthplace     1989831
grad2        USA          Both         Birthplace     1989857
grad1        USA          Birthplace   Birthplace     1990587
all_second   Birthplace   Birthplace   Birthplace     1990728
all_flex     Both         Both         Both           1990871
grad3        USA          Both         Both           1991090
grad5        USA          USA          Birthplace     1991351
grad4        USA          USA          Both           1991842
all_usa      USA          USA          USA            1993244

Seems like partial flex as the winner! This looks pretty good, but I want to run some sanity checks on the coding and do a bit of a write up on the analysis markdown file before closing.

[AaronGullickson]

Commit a7d521f provides a writeup of the section in the analysis document. I still want to do some checks on the coding before closing this issue however.

[AaronGullickson]

I added some checks in commit fd21c13. Everything is working as expected. Closing the issue.