magazine issue 6 2015 / Issue 26

To affinity and beyond! How our preference to be among similar people interacts with our social ecology

• written by Joanna Schug & Philip S. Lu
• edited by Laysee Ong

This might sound like a good thing, but a lot of other research suggests that there are unfortunate societal level consequences to people getting what they want in societies high in relational mobility. For example, people who are highly desirable will have no trouble meeting and forming relationships with others who are highly desirable. While this might be good for them, this means that people who are a little less desirable might have a much harder time with relationships in high-mobility societies. Thus relationships have the potential to be much more stratified socially, with those who are more highly desirable winning out.

One example of this social stratification can be seen in educational homogamy. Since the 1960s, the tendency for people to marry others with the same level of education has been increasing in the US (Mare 1991, Schwartz & Mare 2005). Now more than ever, people are marrying and having children with those of similar educational backgrounds.

Now you might be wondering whether this increase in educational homogamy is due to an actual increase in the preference for similarity in educational attainment, or due to some other factor, such as the fact that more women are attending college these days. The answer appears to be a bit of both. One commonly accepted explanation is that because women’s access to education and opportunities in the workforce have improved, the education and economic status of women is now a relevant factor for men to consider in selecting their mates (Oppenheimer 1988, Sweeney 2002), compared to when it was still relatively rare for women to go to college or join the workforce.

Obviously, educational attainment has a big impact on occupational status and income. What this means is that now more than ever, couples are either very well-off (both earning a high income) or not at all (both earning low incomes). The uncomfortable result of this is that socioeconomic inequality among households for our generation is much higher than it was when households had spouses of mixed educational levels. In fact, economists have shown that rising income inequality in the US can be explained by people marrying people with similar levels of educational attainment (Greenwood, Guner, Kocharkov, & Santos, 2014). If anything, it appears that if it weren’t for people marrying people with similar levels of education, economic inequality is the United States would be falling instead of rising!

Preferences for similarity and residential segregation

The same preference for similarity can also have implications for segregation and fragmentation of the neighborhoods in which we live. Social scientists have been studying segregation of neighborhoods, based on factors such as ethnicity, race, and income for decades (e.g., Massey & Denton, 1988).  How did cities in the US get so segregated? In the case of ethnic and racial segregation, one common explanation is ‘white flight’ or tipping (Grodzins, 1957). When “undesired” minorities move into a predominantly white neighborhood, the theory is that whites tend to pack up and leave.

But what if our seemingly benign preference for similar others is once again the culprit here? Economist and Nobel laureate Thomas Schelling (1971) demonstrated through a simple agent-based simulation model that this indeed may be the case. Schelling created a grid-like structure where the cells represented housing opportunities, and agents of different types would occupy these cells with the opportunity to relocate. The ‘free-will’ of these simulated agents worked along a threshold mechanic. At each step of the simulation, the agent would take a look at the other agents who “lived” in the adjacent cells. If enough of these agents were of a different type, the agent would more to a new neighborhood, or in other words, they would relocate to a random empty cell that was surrounded by more similar agents.

Schelling found that when agents preferred cells that were surrounded by at least one-half of similar agents, striking patterns of segregation would occur. In fact, eventually the system would achieve close to the mathematically maximum amount of segregation possible given the parameters. Schelling also found that uneven populations would enhance the segregation. When one population outnumbered the other 3:1 or 4:1, the minority population would very quickly form a large homogenous cluster. This was because the low initial numbers of cells containing similar others caused the minority to relocate aggressively in the early stages of the simulation.

More recently, researchers have tried to update Schelling’s original model to accommodate the more complex nature of real neighborhoods. The fashionable thing to do now is use continuous preference functions, where the probability of moving isn’t black or white, but comes in shades of grey. Bruch and Mare (2006; 2009) argue that these functions generate less segregation compared to Schelling’s threshold functions. However, Van de Rijt, Siegel, and Macy (2009), found a disconcerting result: when simulated agents were sensitive to how the composition of their neighborhoods were changing in terms of the number of agents similar to themselves, the simulation could become very segregated once again, even with built-in parameters for diversity and tolerance.