130 likes | 214 Views
Diversity in Core Networks. Ana Prata and Sadie Pendaz University of Minnesota. Networks and Diversity.
E N D
Diversity in Core Networks Ana Prata and Sadie Pendaz University of Minnesota
Networks and Diversity • Using American Mosaic Project national phone survey data, we examine diversity within the close ties in people's personal networks. While most studies of network diversity theoretically adopt a structural perspective, making network diversity a direct outcome of the differences in people's social structural locations, there has been a lack of empirical testing of this connection (Moore 1990), which our research serves to address.
Previous Research on Core Personal Networks • We capitalize on a unique subset of questions in the survey that asked people to identify up to five persons with whom they had discussed important matters in the last six months. Our data replicates in part the 1984 and the 2006 research data collected in the General Social Survey of 1985 and 2004, regarding the confidants with whom Americans discuss important matters. The 2006 research (McPherson, Smith-Lovin, Barshears 2006) showed that the core discussion networks of Americans have changed greatly in the last two decades.
Comparisons with the GSS • A comparison between the 1985 and 2004 GSS data demonstrates that the number of people with no confidants has nearly tripled, the mean network size has decreased to just one confident, and non-kin ties have had the greater decrease, supporting the body of work on the diminishing social capital in the fabric of American life (Putnam 1995).
Data Transformations • The AMP dataset has an N of 2,081. We expanded the dataset by the number of persons a respondent’s core network (people could name from 0 to 5 persons). The transformed data set has an N of 6,490.
Creating Match Variables • Respondents were asked: • With how many people have you discussed important matters in the last six months? • Where do you know this person from (work, family, church, something else/just a friend, etc.)? • What is this persons’ race? • What is this persons’ religion? • Is this person more liberal or conservative than you, or about the same?
More on Match Variables • We made dummy variables that identify whether or not respondent matched their friend on three dimensions: • Race • Religion • Political Ideology
Results on Descriptives Number of persons in core network (N=2,081)
Results on Descriptives Number of persons in core network (N=6,490) Number of persons Freq. Percent Cum. ------------+----------------------------------- 0 | 142 2.19 2.19 1 | 270 4.16 6.35 2 | 750 11.56 17.90 3 | 1,266 19.51 37.41 4 | 1,192 18.37 55.78 5 | 2,870 44.22 100.00 ------------+----------------------------------- Total | 6,490 100.00
Homogeneity Findings • Approx 77% of named friends are persons who have the same race as the respondent. There is a jump in racial heterogeneity between the 4th (77%) and 5th (74%) named friend. • Approx 50% of named friends are persons who have the same religion as the respondent. However, religious heterogeneity progressively increases as we move from the first named friend (53%) to the last named friend (46%).
More on Homogeneity Findings • Approx 51% of named friends are persons who have the same political ideology as the respondent. However, ideological heterogeneity progressively increases as we move from the first named friend (55%) to the last named friend (49%).
What’s Next? • We will determine the characteristics (age, education, race, gender, geographical location, etc) of two main groups of interest: • Respondents with no friends in core network. • Respondents with more heterogeneous networks. • What combination of characteristics make more heterogeneous networks.
Future Regression Models • We will use measures of network size and network composition in regression models with our independent variables: age, gender, race, religion and political orientation. Because past research has suggested that social structural variables interact with one another in the formation of close ties and personal networks, we also use interaction models in our testing (Brass 1985).