364 – What is segregation?

Posted on April 8, 2016 by


“Identity politics is back,” opines a comment to 362. I admit to having difficulties with the concept of “identity.” To put my queasiness into perspective, let me take a roundabout way and reflect on “segregation” instead.

The article: This simple experiment shows how easy it is for society to become segregated[1] which I shall discuss presently, allows me to visit the issue in a somewhat structured way.

What is segregation?

The two authors of the article do not define segregation:

But that’s not how society works. In reality, segregation is widespread: in residential neighborhoods, at the workplace, in schools, even online.

If one posits “segregation,” one should indicate what “non-segregation” looks like. The text is unclear on this point. From the experiment the two authors engage in, one might infer that for them “segregation” is any non-Brownian (or random) array of people scattered across a homogenous space.

From the asserted wide presence of segregation, the authors draw the policy conclusion:

And segregation is not a good thing: people who are physically separated are unlikely to exchange ideas, share resources, or resolve problems. Segregation worsens inequality and conflict.

The article introduces this map so as to prove its premise that segregation is rampant:


Racial segregation in New York City. Image Copyright, 2013, Weldon Cooper Center for Public Service, Rector, and Visitors of the University of Virginia (Dustin A. Cable, creator)

Segregation in New York is not a matter of law. We are talking here about de facto segregation. What might be its origins?

Segregation as optical or definitional artifice

Aggregation rules and coarse grain may help explain what appears on the map as segregation. If most resident couples were mixed, e.g., would this show up on the map or would the aggregation rules and color dominance overwhelm the actual distribution? Just a question.

A further point is definitional. The map highlights four “races” and disallows crossbreeds. Assigning people of mixed parentage to either race is far from objective.[2] Even self-declaration (e.g. in New Zealand) is a dubious criterion when it codes to affirmative action. On this basis, segregation might appear stronger than in reality.

The map highlights the place of residence – where people sleep. Is this the relevant criterion nowadays? How would a map based on “place of work” or “place of social interaction” look? In the past, the railway tracks separated two social worlds, and one race alone moved, just to work. Is this still the case? If it is good to “exchange ideas, share resources, or resolve problems,” does it have to be all the time and by residence rather than by presence?

Agglomerations as fractal constructs

Agglomerations follow fractal rules.[3]


London: 1786 to 2010. http://dx.doi.org/10.1103/PhysRevE.92.062130 

The fractal character of agglomerations is very old and has e.g. underlain ancient African cities.[4] It may shape all sorts of settlement patterns, which change over time, yet leave a historical trace. For our purposes, the insight indicates that what appears as segregation may at least in part reflect settlement patterns old and new who have little to do with de facto segregation.

A “simple” experiment?

To explain the perceived pattern of segregation, the authors claim: “the Nobel Prize-winning economist Thomas Schelling proposed a model. Schelling imagined a world where two types of individuals (we’ll make them blue and yellow) are randomly located on a flat square world. In Schelling’s model, individuals prefer to have some similar neighbors, but they do not discriminate against different neighbors – in short, they are tolerant. If individuals are unhappy with their neighborhood, they can freely move to a neighborhood with a preferable composition. In the example below, the yellow individual is unhappy about her assigned location because she does not have enough yellow neighbors, so she decides to move to a new neighborhood. But when she moves, the composition of both her old and new neighborhoods change. As a result, an old yellow neighbor and a new blue neighbor also decide to move. This causes a domino effect that leads neighborhoods to separate into yellow and blue ghettos. In the end, although no single individual prefers it, everyone ends up in segregated neighborhoods.”


Domino effects in the Schelling model of segregation.

The first assumption – that people are located randomly on a flat square world – is dodgy. The second assumption is even worse: that people are either yellow or blue. That is, they have but one and invariant identity.

There is no basis for this assumption – it is a requirement of the mathematical instrument used. People have many identities, which they continuously fore- or background depending on the situation.[5] It is this ability to switch that gives the illusion of only having one – it is a form of availability bias.

The article’s authors “improve” on the Schelling model by going beyond tolerance – which to them is akin to indifference – and explicitly aiming for diversity. Unsurprisingly, the authors get what they want – integration.

They conclude: “Segregation is not unavoidable, but there is a need to continue educating people about the benefits of diversity and to continue devising policies and incentives that prevent or ease segregation.”

When assumptions are questionable

Strip away the sociological context and the “simple experiment” proves what popular wisdom knew all along: “birds of a feather flock together.” It looks as if the models just explain murmuration (which employs similar rules):


I would shy away, however, from suggesting any policy based on such simple models.

[1]           http://bit.ly/1RZ5qce

[2]           Traditionally, in the US one “drop” of black blood made a person “black” – unless the person could pass off as white. As long as slavery existed, “blackness” coded to the slave owner’s property: race was defined to maximize it. While slavery as such was eliminated, the census and the birth certificate failed to recognize the new reality.

[3]           Michael BATTY – Paul LONGLEY (1994): Fractal cities: a geometry of form and function.  Academic Press, San Diego. Fractal patterns repeat themselves over different length scales: If you zoom in on one small portion of a fractal, you’ll find a motif nearly identical to the pattern as a whole. Over time, the fractal network of streets may change from being multifractal, scaling spatially according to multiple rules, to mono-fractal, with a single scaling rule (e. London).

[4]           Ron EGLASH (2005): African fractals. Modern computing and indigenous design. Rutgers University Press, New Brunswick.

[5]           Amartya SEN (2015): Identity and violence. The illusion of destiny. Penguin, New York.

Posted in: Uncategorized