Saturday, August 04, 2007

A Devastating Influence?

".... immigration and ethnic diversity have a devastating short- and medium-term influence on the social capital, fabric of associations, trust, and neighborliness that create and sustain communities." That's the finding of Robert Putnam's just-published research, according to a story by John Leo in City Journal. The argument is a familiar one--for example, Pat Buchanan has been saying the same thing for years--but there are two things that make Putnam's work stand out: first, he's not just offering an opinion, but a conclusion based on an analysis of data from a large number of communities, and second, it's definitely not the result that he wanted to find. He supports immigration and ethnic diversity, is clearly troubled by his findings, and spends the last few pages of the article suggesting policies that might reduce the "tradeoff between diversity and community." Leo even suggests that Putnam sat on the results for five years because of his discomfort with them, although Putnam disputes this (see his comment on Leo's story at the link above).

Social science research often gets distorted and oversimplified when it's reported in the popular press, but that's not the case here. Here's a quote from Putnam's own article: "inhabitants of diverse communities tend to withdraw from collective life, to distrust their neighbors . . . to withdraw even from close friends, to expect the worst from the community and its leaders, to volunteer less, give less to charity and work on community projects less often, to register to vote less . . . and to huddle unhappily in front of the television." It's not a pretty picture. After reading the article, I'm pretty convinced that ethnic diversity really does reduce the sense of community and trust (what social scientists often call "social capital"). The basic question with any statistical analysis is "have you considered X?," where X is some other variable that might be the real cause of whatever you're trying to predict. For example, cities tend to be more ethnically diverse than rural areas and small towns, so maybe the real story is that cities have less social capital. Putnam considered that, and a lot of other things. Of course, you can never consider everything, and maybe someone will come up with some other "X" that is the real culprit, but until that happens I think that his conclusion should be accepted.

However, I'm not too concerned. While I agree that ethnic diversity appears to reduce social capital, the effect really isn't very big. At first glance, it seems substantial: for example, in the most ethnically diverse communities in his sample (e. g., San Francisco) about 30% of people say that they trust their neighbors "a lot"; in the least diverse (e. g., the state of New Hampshire), it's 55-60%. However, this is without taking account of all the other differences between more and less diverse communities--the "X" factors. The standard way to take account of them is with a statistical technique called "multiple regression." Putnam presents the results from a multiple regression predicting trust in neighbors, which he says are typical of those found for other variables related to social capital. The regression results indicate that ethnic diversity has an effect, but not a very big one. For example, suppose you're comparing two otherwise similar people, one of whom lives in a highly diverse neighborhood, the kind that's typical of San Francisco, the other in a very homogeneous one, the kind that's typical of New Hampshire. The predicted difference in the variable measuring trust in neighbors is .10. Trust isn't a tangible quality like height or weight, so the exact number isn't meaningful, but we can use it for comparisons. For example, suppose you compare two otherwise similar people, one a college graduate, the other a high school graduate who didn't attend college. The college graduate will have a higher predicted value, by about .16. Comparing a 60-year-old to a 20-year-old, the 60 year old will have a higher predicted value, by about .40. Other factors that have big effects on trust of neighbors include home ownership, one's own ethnicity (blacks and Latinos trust their neighbors less), and one's financial situation (people who make more and are more satisfied with their financial situation trust them more).

So his description of the effects of ethnic diversity isn't wrong, but it needs a lot of qualifiers--"somewhat," "a little," that kind of thing. In a few cases it might be possible to give exact numbers--for example, how many more minutes per day do inhabitants of diverse communities spend huddled in front of the television? I haven't analyzed the data myself, and Putnam's article doesn't provide the information needed to calculate the figure, but I'm guessing it's not a big number. Another way to look at it is that, according to Putnam's regression results, the effect of ethnic diversity is about as large as the effect of residential mobility--as you might expect, people living in neighborhoods with more turnover don't trust their neighbors as much. However, few people would say that residential mobility has a "devastating" impact on communities and that something needs to be done to stop it.

So the overall point is one that's routinely taught in statistics classes--that the size of an effect is important--a "highly significant" effect isn't necessarily a "big" effect. The apparent effect of ethnic diversity isn't trivial, but it's not big enough so that it should make much difference in debates over immigration. I was going to say more, but one of the features I like about Brad's blog is that he doesn't make his posts too long, so I'll emulate him and save it for next time.--David Weakliem


Anonymous said...

One would think that Robert Putnam is aware of the difference between statistical and substantive significance. Is he simply trumping up these effects to convince reviewers it's worthy of publication?
I would expect that ethnic diversity would have a stronger, negative effect on trust and community involvement for people who were already not likely to be active in their communities. What do you think?

David Weakliem said...

No, I don't think that was the reason. I'll offer my speculations in a post later today or tomorrow.

I could see how ethnic diversity might have more effect at lower levels of trust and involvement, but I could also see how it could apply pretty uniformly, so I'll just say take the easy way out and say that more research is needed.

Scott Kemp said...

Is the reason that Putnam's conclusions seem so significant that he expected different conclusions? From your post it sounds like he expected a positive correlation between diversity and community, and got a negative correlation. That is more significant to him than it would have been to, say, Pat Buchanan - who was probably not surprised by the result.