Serendip is an independent site partnering with faculty at multiple colleges and universities around the world. Happy exploring!

You are here

Segregation Model

Kathy Maffei's picture
Projects: 
Yes, yes, I agree with #1, and I understand that simple models may help define certain aspects, as opposed to the whole of a complex issue. I can see that and the utility of using them. But there are still a couple of observations about this model that I’d like to share: 1) My first interpretation of 50% preference was that it would imply no particular preference (which is why segregated results were so surprising when using 50% or less preference for difference), but this assumption isn’t true. After all, the test is still biased to a particular preference, no matter how low the %. It appears that the continuum of preferences would start at one end with 100% preference for similarity, progress on down to 0% preference for similarity, then to 0% preference for difference, on up through to 100% preference for difference at the other end of the scale. So, it seems to me that 0% of either would be truly no preference, and that there really is no surprise at seeing segregation resulting from a low similarity preference setting. 2) I thought that even if we only consider two colors of people, we should still probably consider the fact that in real life there isn’t often an equal number of each. It seems as though it would be useful to be able to adjust the proportion of color quantities (and so there’s an enhancement suggestion). 3) What came to mind from my sociology reading was that many blacks also rate black neighborhoods unfavorably and prefer white neighborhoods (which tend to be economically better off with more resources). In other words, in the real-life example of white & black, one color preferred similar and one preferred different. And so there’s another enhancement possibility, but I think this one might be less useful because in the simple model world, I suspect this would cause constant movement (as long as preference strengths were high enough) since one color’s goals would be counter to the other’s. While in the real world, economic and socio-political barriers help prevent minorities from moving to resource-rich (predominantly white) neighborhoods. Interesting observation: I initialized the model with segregated distribution and set the preference to difference and the strength of that preference to 77%; I wanted to see how easy it was to undo segregation when agents were motivated to do so. But it ran on and on without ever being able to get lower than 45% similar neighbors. I tried again and found that with 60% preference strength it was attainable, and the % similar went down to 18. What resolved was a screen of colored lines – short parallel lines of green and red in clusters. I repeated both with the same results. I’m guessing that the higher preference scuttles the success by prompting agents to move on before diversity can be achieved in any area, so the breaking point must lie between 60 and 77%.

Comments

BenKoski's picture

I wholeheartedly agree with you--though the segregation model does do a good job of keeping things simple, it almost oversimplifies the problem by leaving out some parameters central to the question of segregation. I disagree with you, however, on your third point: I don't think minorities desire to live in white neighborhoods because they "prefer different" or because they want to live in a white neighborhood, but because minority neighborhoods are classically the most distressed and disinvested (p. 9). I think it's a bit of a stretch to model segregation based the assumption that minorities "prefer different", unless you mean that minorities prefer to live in a different neighborhood. I think this point raises an intriguing possibility for the model: what if the model integrated an "environment" grid behind the turtles that affected their action? Such a grid could incorporate land values (which are bid up by turtles trying to move away from other groups) and a neighborhood investment factor (another constant that could be used to tune a turtle's propensity to move), among other things. I suppose my basic criticism of the model is that there is no way to control a given turtle's probability of moving--though I recognize that we're just modeling simple segregation, I think it's a bit much to assume that all turtles are equally likely to move in all situations. A p value associated with each turtle could be an interesting enhancement, I think. This could also be tuned to consider the opportunity cost of moving--I would venture to guess that the model would look a little bit different if a turtle could only move once every x turns, depending on the sort of move that it was trying to make. So...perhaps I am getting ahead of myself in terms of complexity, but I am excited by the possibilities of this model.
Kathy Maffei's picture

Ack! I hope my attempts at brevity didn't cost clarity - I definitely wasn't trying to say that minorities prefer to live with whites. (On the contrary there is plenty of literature devoted to the discomforts, difficulties, and disadvantages of being in the minority) I was trying to say that I've read that minorities tend to prefer predominantly-white neighborhoods because the neighborhoods are more desirable - as you point out, it's an economics / resource thing. My point was that modeling both colors with the same preference (of sameness or difference) was far from the reality of racism and economics. A (grossly) oversimplified model of this issue might have one color prefer sameness while the other prefers difference, since one color's neighborhoods are where most of the political & economic resources reside. But I suggested this alteration wouldn't be terribly useful because it's such an oversimplification. But your point is well-taken that a much more accurate model of this issue would account for environment (different kinds of neighborhoods) and purchasing power. You had some really interesting suggestions as to how that might be done. It would be quite a challenge, and I'd be interested in seeing what you come up with!