« Settling | Main | China fact of the day »
Book forum: Tim Harford's chapter six on Schelling's segregation model
Tim Harford has the best exposition of Tom Schelling's segregation model I have read. Maybe no one prefers segregation, but if you mind being a minority in a neighborhood an invisible hand process can lead to segregated outcomes. Individuals will move closer to their compatriots, giving rise to an overall separation of groups. This paper has some good models and fills out the main conditions behind the result.
But is it true? Schelling would be the first to admit he created only a partial model. Human genetics show more and more out-breeding over time. Those first cousins just don't cut it any more. No, the earth isn't flat but outmigration is increasing and many more people are choosing to live as minorities in foreign lands, most of all in the EU. I live in Northern Virginia, one of the most successfully integrated regions of the United States, whether it be along lines of race, religion, or nationality. Latino arrivals are concentrated in the American Southwest but over time they are spreading out to many other states. What is the segregation model missing?
Gains from trade, in a nutshell. If I'm the first Mexican to arrive in North Carolina, yes maybe I feel lonely. But I also can fill some empty economic niches and overall it may beat East L.A. Other immigrants will follow, but if too many come some of them will move on to South Carolina. And so on.
High levels of inequality often bring more integration, at least in terms of spatial proximity. Even with high rents there is a large community of Latinos living just outside of Aspen, Colorado. Guess why. They don't live right next to the very rich but they do live among non-Latinos. And the greater availability of cheap services is one reason I prefer life in the United States to Western Europe. Cheap shipping of goods means I still can get French cheese and German books.
It is harder to ship services. The more we become a service economy, the more you have to live near the people you trade with.
So what's the problem in Newark, NJ or for that matter Northeast Washington? Schelling's model seems to work better there perhaps because of high unemployment and fewer services. That said, both areas have seen considerable Latino integration over the last twenty years, as well as outmigration to the suburbs.
Thus the more general model starts with the idea of gains from trade and then asks when those gains won't be especially strong, or when they won't require much physical proximity. Note that Schelling's original paper, published in 1971, very much represents a 1960s perspective on its topic.
Addendum: Tim Harford also discusses urban crime and its control; here's a good new paper on that topic.
Posted by Tyler Cowen on February 11, 2008 at 06:11 AM in Books | Permalink
Comments
Please note that the comments section on this post is for a discussion of segregation and integration and not for a treatment immigration policy per se. You get plenty of other opportunities to debate the latter.
Posted by: Tyler Cowen at Feb 11, 2008 7:32:57 AM
Are government actions also in part responsible for segregation? After listening to Venkatesh being interviewed by Brian Lamb I couldn't help thinking that those huge housing projects in the south side of Chicago were preventing blacks from moving where the jobs actually were and thus trapped them in a de facto state sponsored segregation model.
Posted by: Unit at Feb 11, 2008 8:45:07 AM
Integration may reflect gains from trade but the most parsimonious model would start with entropy as the baseline factor. Random shuffling alone will lead to mixing of groups. If transportation and relocation costs fall (due to technology, more flexible institutions, or more accepting culture), then groups which started out segregated would be expected to be more and more intermingled over time.
Posted by: A student of economics at Feb 11, 2008 9:14:44 AM
Thanks, Tyler.
In the light of your comments it is not that surprising to find that racial segregation is falling (but slowly) and the fall is driven by the fact that fast-growing economies are showing much more racial mixing.
A link to a good explainer from Brookings is here (pdf): http://www.brookings.edu/es/urban/census/glaeser.pdf
Jacob Vigdor is the man with the data:
http://trinity.aas.duke.edu/~jvigdor/segregation/index.html
Posted by: Tim Harford at Feb 11, 2008 12:25:40 PM
Is the Schelling Model limited to race? What if we apply it to tastes in general? I'd imagine that in more sophisticated societies, skin color is only one of the many factors in picking peers/neighbors.
Posted by: Biomed Tim at Feb 11, 2008 12:43:41 PM
I think there are three main factors involved in the persistence/break-down of desegregation
1) Segregation is on the extreme local level (particular bars and streets will be predominantly a particular race group, but the preference breaks down over larger geographic areas like whole states)
2) Preferences for segregation are very personal and very wildly between individuals. That first Mexican that moves to North Carolina has a low preference for segregation, especially compared to the potential gains to move. And his presence lowers the threshold for the next potential migrant.
3) Preferences are highly dynamic. A person may have a high degree of difference from day to day. The time that an individual decides to move (or not to move) may not even be representative of them over a longer period. Not to mention the huge shifts in norms that have occurred at a much larger scale throughout society.
Posted by: Nate at Feb 11, 2008 1:17:30 PM
Regarding Latinos, despite being somewhat better integrated than
African Americans in major US cities, many do have substantial Latino
neighborhoods, pretty well-defined and segregated. Think East LA.
Posted by: Barkley Rosser at Feb 11, 2008 6:03:48 PM
William Julius Wilson spent four years studying exactly these questions in four neighborhood in the South Side of Chicago for his recent book "There Goes the Neighborhood." His discovery was that Tyler's worldview, which is based on Tyler's experience in upper middle class Northern Virginia, doesn't apply very well at all farther down the social scale.
For details on Wilson's study, see:
http://www.vdare.com/sailer/080203_chicago.htm
Posted by: Steve Sailer at Feb 11, 2008 7:07:50 PM
Much of elite media opinion on immigration and integration is formed by the unusual diversity and quality of immigrants to the Washington DC area. In contrast, in the Sunbelt, such as in an obscure little place called Southern California, our immigration policy has led to enormous regions dominated by Hispanics. Take a look at LAUSD school-by-school demographic statistics if you don't believe me.
www.lausd.net
Posted by: Steve Sailer at Feb 11, 2008 7:14:22 PM
Another reason for the consistent bias on the subject of immigration among DC-NY media elites is that immigration pushes African-Americans out of DC and NY and replaces them with a more amenable immigrant servant class. New York City's African American population, for example, has been declining since 1979.
DC-NY pundits tend to assume that the African-Americans displaced by immigrants are somehow being deported or simply vanishing from our time-space continuum, but in reality they are just moving to less privileged locations. This explains a lot about the current state of Newark and Baltimore.
Posted by: Steve Sailer at Feb 11, 2008 9:12:12 PM
Check out:
John H. McWhorter
Party of Chains
The greatest oppressors of blacks have been Democrats, says Bruce Bartlett.
http://www.city-journal.org/2008/bc0208jm.html
Posted by: jorod at Feb 11, 2008 9:25:32 PM
The preference distribution among individuals for living among one's co-nationals is broad, and given a chance, the tail with zero or negative preference for segregation will desegregate most locales.
Historically, the one widely effective means for creating and maintaining segregation has been ethnic violence and the threat thereof.
Posted by: Cyrus at Feb 11, 2008 10:47:40 PM
Isn't this Chapter 5 you are talking about, not Chapter 6?
Posted by: James K at Feb 12, 2008 12:10:52 AM
I don't see how the gnxp link supports your point, although I do not dispute it. I also agree with A.s.o.e that entropy and transportation have to be taken into account, although the gains-from-trade model also has plausibility.
Posted by: TGGP at Feb 12, 2008 12:44:59 AM
The Pew Center today projected that the Hispanic population, under current immigration rates, will increase from 41 million to 127 million from 2005 to 2050. So, it's pretty hard to imagine how those 86 million additional Hispanics will find enough white people to live near to make Tyler's vision come true. The country will increasingly look like the LA Unified School District, without enough whites to bus around to integrate much of anything.
Tyler, may I ask you a serious question? In all your thinking about immigration, have you ever made a single numerical calculation?
Posted by: Steve Sailer at Feb 12, 2008 4:55:57 AM
My problem with the Schelling model is that it may people stop thinking about segregation. "Oh - it's coz of this dynamic." I haven't read Harford's chapter, so I don't know if he discusses this at all; I heard him interviewed on this subject and it didn't come up.
I'm not anti-model (far from it, I'm a trainee modeller myself.) I just wish more people accepted models as one way to describe reality, and one that shouldn't blind us to important empirical facts. With Schelling, there's a tendency to say, 'the model demonstrates x' and then sit down with a nice cuppa and do no more.
An example of why this is a potential problem: I live in Leeds in the UK. As a member of my department has shown, for a long time some Leeds' estate agents directed specific minority groups to particular areas of the city, and didn't even mention certain properties to them in areas where they believed a minority presence might bring the value down. This, despite the fact that many of the minorities in question could pay cash for properties.
Models *can* produce blindness to this kind of empirical fact. I'm not sure what's to be done about that, except to repeat the mantra: a model is a map, not the territory.
Posted by: Dan at Feb 14, 2008 4:41:41 PM
