25.01.2015 Views

CONNECTIONS - INSNA

CONNECTIONS - INSNA

CONNECTIONS - INSNA

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

<strong>CONNECTIONS</strong><br />

Configurations of Homophily<br />

as the configuration having membership in the<br />

outcome set (Y). In the present study, the<br />

outcome is whether or not the configuration of<br />

homophily occurred more often than expected by<br />

chance. Configurations whose frequency exceeds<br />

chance are in the outcome set; those that equal or<br />

fall below chance rates are outside of the outcome<br />

set. Only those elements of P(X) that have<br />

membership in the outcome set are members of Y.<br />

The outcome is recorded for each configuration,<br />

resulting in a truth table that serves as the basis of<br />

analysis (i.e., Table 1).<br />

Arriving at a QCA solution takes place through a<br />

systematic comparison of configurations that<br />

exhibit the outcome. If two configurations with<br />

the outcome differ on only one condition then that<br />

condition is considered irrelevant and can be<br />

dropped to produce a simpler expression. For<br />

example, given two configurations “Age and Race<br />

Homophily” and “Age Homophily” that both<br />

exhibit the outcome (i.e., occur more often than<br />

expected by chance), similarity on race is<br />

irrelevant to the outcome. Dyads existed more<br />

often than by chance if they were similar on age,<br />

regardless of their similarity on race. Thus, the<br />

simpler expression “Age Homophily” is an<br />

implicant of “Age and Race Homophily” and<br />

“Age Homophily” because both configurations<br />

are contained within the set “Age Homophily.”<br />

Comparisons between implicants can also be<br />

made and conditions that are irrelevant can be<br />

dropped. An implicant that cannot be simplified<br />

through comparison with another implicant is a<br />

prime implicant. The set of prime implicants for<br />

the configurations with the outcome will contain<br />

as a subset only those configurations that exhibit<br />

the outcome. Thus, the set of prime implicants for<br />

Y provides a simplified account of the<br />

combinations of conditions where the outcome<br />

exists. A QCA solution contains the combinations<br />

of conditions that are associated with the outcome.<br />

The Quine-McCluskey algorithm provides a<br />

systematic means of comparing configurations<br />

and implicants to reduce a truth table and is<br />

incorporated within the fs/QCA software used for<br />

this research (Ragin, Drass, & Davey, 2003).<br />

The following analysis uses QCA procedures to<br />

simplify the configurations of homophily that<br />

exceed the baseline and produce a solution that<br />

identifies the basis of inbreeding tendencies.<br />

Using fs/QCA software (Ragin et al., 2003), a<br />

crisp set analysis was performed to determine the<br />

types of configurations with more dyads than<br />

expected by chance. In the QCA models that<br />

follow, the outcome is present when the frequency<br />

of dyads in a given configuration exceeds the<br />

baseline. To simplify presentation, solutions are<br />

described as leading to the outcome or not, with<br />

solutions leading to the outcome indicating more<br />

dyads than expected by chance.<br />

RESULTS<br />

The results for models predicting relations of any<br />

type are presented in Table 2. The QCA solution<br />

can be interpreted as follows. The first column<br />

includes the set of solution terms from which all<br />

solutions are drawn. In each term, characteristics<br />

are connected by logical “and” (represented by *).<br />

Characteristics in uppercase letters refer to<br />

similarity on a dimension while lowercase letters<br />

refer to dissimilarity. The columns to the right<br />

indicate which terms are contained in the solution<br />

for each type of relationship (designated with a<br />

“•”). The terms for each solution are connected<br />

by logical “or.” Thus, the first solution (column<br />

2) indicates that more dyads than expected by<br />

chance existed when ego and alter had similar:<br />

sex and race and religion, or<br />

age and race and religion, or<br />

race and education and religion, or<br />

age and race and education, or<br />

age and sex and education and religion<br />

The strong effect of race homophily is evident in<br />

the solution. All terms in the solution term,<br />

include similarity on race except for the final<br />

which includes similarity on every dimension<br />

except for race. Religion is also quite strong in<br />

that four of the terms include similarity on<br />

religion. However, the number of dyads with just<br />

similarity on race and religion did not exceed<br />

chance. It was necessary to combine similarity on<br />

race and religion with at least one other dimension<br />

in order to produce inbreeding homophily.<br />

This solution can be better understood by<br />

examining the lattice in Figure 1. Because all<br />

30

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!