I recently spent a month reading through the vast empirical literature on trust. A simple survey question - usually along the lines of "Can most people be trusted?" - seems to predict many good social outcomes. What most interests me, however, is trust assimilation. Fans of empirical trust research often claim that immigrants retain trust generation after generation. The go-to example: Swedes are highly trusting, and so are Swedish-Americans. If this is generally true, immigration from low-trust countries will durably reduce average trust in high-trust countries that receive them.
One of the most-cited papers on this topic, Uslaner's "Where You Stand Depends Upon Where Your Grandparents Sat" (Public Opinion Quarterly, 2008) dissatisfies me in four ways: (1) despite the title, most of the reported effects of ancestry on trust were quite small; (2) the author's main specification uses dummy variables for national/regional origin, rather than a direct measure of trust in country/region of origin; (3) eight out of the nine countries/regions of origin were in Europe; (4) Uslaner includes multiple control variables that plausibly proxy trust, including whether you're "satisfied with your friendships" and agree that "officials are not interested in the average person," and that it's "not fair to bring a child into the world."
Algan and Cahuc's "Inherited Trust and Growth" (AER, 2010, Section II.B) does a much better job. It uses the World Values Survey to measure trust in "home countries," instead of just national/regional dummies, allowing a direct (though coarse) measure of immigrant assimilation. And it doesn't control for variables that are close proxies for trust. Result: About 45% of trust is persistent - not overwhelming, but hefty. Ultimately, though, Algan and Cahuc still dissatisfies me. Shortcomings: (1) they control for several variables that plausibly capture assimilation, especially education and income; (2) they heavily restrict their sample, (3) other than Africa, India, and Mexico, all of the countries/regions of origin they use are European.
So while I was hoping to outsource trust assimilation to prior researchers, I ultimately decided I needed to crunch the numbers myself. My approach closely mirrors Algan and Cahuc's: I use the General Social Survey to measure individuals' trust within the United States, and the World Values Survey to measure average national trust everywhere else. When multiple years of WVS data were available, I averaged them. When the two data sets' categories didn't perfectly mesh, I did the best with what I had. Example: Since the WVS doesn't include Austria, I gave Austrians the trust rate for Germany. I ended up with this:
Since measuring assimilation is my goal, I include (a) a dummy variable for first-generation immigrants (BORN=0 if foreign-born, =1 if native-born), and (b) separate coefficients for assimilation for first-generation immigrants and their descendants. What do we get?
Using the entire sample:
Now remember: The formula for zero assimilation is TRUST = .00 + 1.00 * ANCESTRAL for foreign- and native-born alike. What the data yield, however, is TRUST = .10 + .70 * ANCESTRAL for foreign-born, and TRUST = .22 + .56 * ANCESTRAL for native-born. The upshot is that low-trust immigrants gain a lot of trust, especially after the first generation. Suppose your parents come from a country where only 10% of people trust others. Your parents' predicted trust in the U.S. is 17%, but yours is 28%, just 9 percentage-points below the U.S. average.
What happens if we distinguish voluntary immigrants from the descendants of slaves? Some prominent cliometricians argue that slavery durably damaged trust in Africa, and it's highly plausible that slavery in America did the same here. Given historic immigration patterns, virtually all native-born U.S. blacks descend from slaves, so I define the dummy variable SLAVE, where SLAVE=1 if (ETHNIC=1 and BORN=1). Then I re-estimate:
Now we see an extremely high rate of trust assimilation for descendants of voluntary immigrants, regardless of their origins: TRUST = .33 + .28 * ANCESTRAL. If your ancestral trust is 10%, your predicted trust in the U.S. is 36%, a mere 1 percentage-point below the U.S. average.
I'm well-aware, of course, that these results aren't good enough for publication. I banged them out in a few hours. But I still think I learned a lot by getting my hands dirty in the data. The most important lesson: While scholarly claims about trust persistence aren't totally wrong, they're heavily exaggerated. Common sense says that descendants of voluntary immigrants readily assimilate to mainstream U.S. culture. And at least for trust, common sense is right.