Five years ago Jen’nan Read and I published an article about women’s employment for 12 different ethnic groups in the U.S. In my Gender, Work, and Family seminar the other day we read it, and because of the timing in the course we were talking about developing and framing research questions. We looked at a couple of the images Jen’nan and I used to work through the many patterns we were trying to sort out in the data, and I thought the sequence might be interesting for blog readers, too. (These use data from the 2000 Census, with “employed” meaning any employment in the year 1999.)
The overall employment rates ranged widely, from 65% (Mexican, Asian Indian) to 84% (Filipina). First thing, break the groups out by education:
That helps a little. At the low end of education, employment rates are low. But at the high end, the spread is great — the line isn’t helpful — with high- and low-employment groups all found above 14 years of average education.
Next step was nativity (where they were born) and year of immigration for those born elsewhere. Looking at the high and low groups, there were distinct patterns:
Mexican women are on a diagonal line, showing a temporal progression from recent immigrants having low education and employment rates, back in time to those born in the U.S. (mostly children of earlier immigrants) with higher levels of both. The Asian Indian women are on a vertical line, with high education but low employment rates for recent immigrants, who may have come as spouses of men who had (or were getting) jobs here. The Filipinas, who are mostly labor migrants — that is, they came for their own jobs — are bunched up at high levels of both employment and education regardless of immigration timing.
This must have been the time Jen’nan hit on the title of “One Size Fits All?”, since the stories clearly diverged, and we decided to fit models of employment predictors for each group separately (details are in the paper).
One other image, a nice idea but hard to read, is also sitting in the old folder (click to enlarge):
Here there’s a dot for each ethnic group for each year of immigration. That is, the dot down in the bottom left is Mexican women who immigrated in 2000 (“M0”). The line is a weighted regression line, so it’s the overall relationship between education and employment rates. You see the Mexican women crowded around the lower left; the Filipinas in the top right; and recent Arab, Japanese, Korean and Asian Indian immigrants along the bottom. (The other initials you see are for Vietnamese, Puerto Rican, White, Chinese, Cuban, and Iranian).
When possible, I like to do this kind of descriptive exploration — making progressive slices in the data and looking at the patterns — before building complicated statistical models with a lot of factors at once. If the descriptive image survives the statistical analysis, there’s a chance of not only getting the story right, but being able to explain it.
This is cross-sectional data — that is, what each group is doing in 2000, based on the year they arrived. But it is suggestive of different immigration pathways. I’m not expert on the immigration side, but since I brought it up I should give a few pointers to recent literature on “segmented assimilation,” which focuses on such differences. The idea dates to the early 1990s, maybe this article:
- “The New Second Generation: Segmented Assimilation and Its Variants,” by Alejandro Portes and Min Zhou.
- “Generation and Earnings Patterns Among Chinese, Filipino, and Korean Americans in New York,” by Sookhee Oh abd Pyong G. Min
- “The social context of assimilation: Testing implications of segmented assimilation theory,” by Yu Xie and Emily Greenman
Feel free to add suggestions for additional resources in the comments.