Social Network Blog - g10

McPherson et al.’s article, “Birds of Feather: Homophily in Social Networks,” discusses the principle of homophily, which asserts the “contact between similar people occurs at a higher rate than among dissimilar people” (416). In the article, they examine the sources, types of relationships, and varying dimensions of homophily. The article draws heavily on ideas discussed in earlier readings, by Granovetter and Burt, such as cognitive balance theory and transitivity, which both assert the basic concept that “friends of friends tend to be friends.” This also relates to Burt’s criticism of redundant ties and the role of “egocentrism,” which attributes tie formation to cues from shared backgrounds and interests (73). McPherson and his colleges expand on these concepts, identifying baseline homophily, related to demographic variables that create the opportunity for a potential tie, and inbreeding homophily, which is independent from opportunity (choice to form a tie). Not surprisingly, race and ethnicity contributed most significantly to homophily, along with education, sex age and religion.

I was particularly interested in their discussion of “tie dissolution,” which asserts, “nonhomophilous ties are especially likely to be dropped when they are involved in intransitive friendships” (436). This provides an introduction into the process of co-evolution of friendship networks, which are characterized by both the selective formation and dissolution of ties, and is a nice transition into the next article from this week, “Homophily and Assimilation Among Sport-Active Adolescent Substance Users.” In this study, the authors, Pearson et al., “analyse the co-evolution of social networks and substance use behavior of adolescents and address the problem of separating the effects of homophily and assimilation” (47). Their discussion highlights the process of “friend selection” based on homophily v. behavioral assimilation in relation to substance abuse. In this sense, they examined whether students chose friends based on their similar preferences and habits (homophily) or assimilated to the same substance use behavior of their existing friends (assimilation). The strongest homophily and assimilation were associated with alcohol, which found that “drinkers prefer the same drinking behavior of their friends, as well as exercising strong influence on the dimensions of alcohol use” (57). They attribute this finding to the social dimension of alcohol use. How do you think that this reflects substance use later in life, specifically patterns in college?

This study suffers from several laminations. The sample is drawn from 160 students at a single school in West Scotland, this it cannot be generalized to all adolescents. The study is also based on self-report, which is not always reliable when discussing drugs use in a school setting.

The last two readings focus on the estimation of network size, employing very different methodologies and producing two dramatically different estimations. In the first article, “Estimating the Size of Personal Networks,” Killworth et al. attempt to determine the maximum number of individuals known by an informant. In the study, Killworth and his colleges examine four different methods that have been employed to estimate total size of communication networks, by reevaluating data from three separate previous studies, Jacksonville, FL, Orange County, CA, and Mexico City, Mexico. They use four different elicitors, or modules, to determine network size: 1) Approximation of GSS question; 2) Support Network; 3) the RSW instrument; and 4) FT telephone book instrument. Unlike the Christmas card study by Hill and Dunbar, the Kilworth study used four elicitors simultaneously to determine network size and compared the results to determine reliability. The sample used is also considerably larger (i.e. 98 in Jacksonville, 99 in Mexico City) and the methodology is presumably more random, the list is compiled from a telephone book, although participation was voluntary. Kilworth et al. also did an excellent job of citing possible sources of error, such as the problem of “similar names,” size of telephone book, and randomness of names chosen. One of the most obvious limitations of the study is the dependence on the representativeness of the phonebook because not everyone is listed. They conclude with an approximation of 1,526 individuals in the average U.S. informants’ network. Do you think that this is a reasonable estimate? How can one account for the significant variance between the three different samples, particularly Mexico City?

The second article discussing network size, “Social Network Size in Humans,” by Hill and Dunbar, aims to identify “those people an individual considers important and whose relationship they value” (55). The study uses a unique approach, using the exchange of Christmas cards as a measure of network size. Although the study provides adequate justification for this methodology, I found that the study was plagued by bias and sources of error. For example, the participants were solicited through other participants, making it a convenience sample that relied on snowballing, which are two well-known sources of nonrandom sampling error. As a result, it is very likely that the sample in this study is more homophilous that the rest of the population because, as we discussed earlier with McPherson, Burt and Granovetter, friends of friends are usually friends due to transitivity, making overlap and similar network structure more likely. Another more obvious limitation is the use of Christmas cards, which almost ensures homophily based on religion. Not only are none of the participants the same religion, but also people may not send Christmas cards to all of the important people in their network if they have varying religious beliefs (i.e. not sending an X-mas card to a Jewish friend). How could this study be improved?