Survey Variable: Social Network Homogeneity

Figure 1. See also Table A1.

There are numerous components of social capital: the size of your network, your amount of interaction within it, and the occupational status of your acquaintances all matter. Additionally, the social homogeneity of the network is important, because it is likely to affect the number and range of experiences and perspectives that you encounter. The survey also covered this component of social capital by asking respondents to estimate roughly what percentage of their friends were from the same sex, ethnicity, and religion as themselves. Figure 1 (above, using weighted data) shows the average answers to those questions, as well as the average across them. People tend to have friendship groups that are made up of people from the same sex, religion, and especially ethnicity as them, thought they also have sizeable minorities of friends from different groups. On average, people say that approaching two thirds of their friends are from the same sex (64.2%) and religion (63.2%), and four fifths (81.6%) are from the same ethnicity. In all these cases, the distribution around the mean is quite wide: the standard deviation is roughly twenty percentage points for sex and ethnicity, and more than twenty-six for religion.

Whilst Figure 1 gives us a helpful view of how homogeneous people’s social networks are, there are also some problems with the questions that underpin it. First, it is quite difficult to estimate the proportion of your friends who are from the same group as yourself without listing and categorising them. It is unlikely that survey respondents will go to this effort, so their answers probably represent a best guess that is open to bias, for instance from whether they identify as open and tolerant (which might lead to over-estimating the percentage of friends from the different groups). This fits within wider evidence that people’s perceptions of figures (e.g. the percentage of the population who are ethnic minorities) are often very inaccurate.

The question also takes no account of the groups that respondents are in, which is likely to affect the homogeneity of their networks. For instance, in Great Britain, someone who is Zaroastrian is less likely than someone who is Protestant to go through their lives meeting people from the same religion, and therefore may be less likely to have friendship group comprised largely of people who share their faith. It is still possible, of course, because socialisation often sorts people into groups that are similar to them, but we cannot tell this from the answers to the questions presented above. Finally, when someone indicates that most of their friends come from a different group from themselves, it might still be the case that those friends all come from a single group. In other words, someone might have eight friends who share a single religion that is not their own, or might have eight friends with eight different religions that are also not their own, but we cannot distinguish between the two in the data. Despite these problems with the questions, they still allow us to differentiate between people who mix with more or fewer people who share their characteristics, which might shape the experiences and perspectives that they encounter.

Variable namessc_friends_sex, sc_friends_sexdk,
sc_friends_eth, sc_friends_ethdk,
sc_friends_rel, sc_friends_reldk
Number of cases1,405
Number of categories100-point scale
Categories to code as missing‘Don’t know’ (997), ‘Skipped’ (998), and
‘Not asked’ (999)
Cases to code as missing89 – 65
Recoded variable namessc_friends_sex_c, sc_friends_eth_c,
sc_friends_rel_c, sc_homog_amv
Number of cases1,316 – 1,340
Number of categories100-point scale
New and old categoriesRespondents who skipped (998) or were
not asked (999) each of the original
questions were coded as missing, unless
they had provided answers to subsequent
prompt questions (shown to them if they
skipped the questions). The substantive
(i.e., not ‘Don’t know’ (997)) answers
provided in response to the prompts
were used to replace the missing values
(resulting from initially skipping the
question before being prompted) in the
new variables.

For the average variable, a row average
of the new variables was taken, utilising
whichever of the answers had been
provided in relation to the shared sex,
ethnicity, and religion of friends had
been provided.
Details of the original and recoded social network homogeneity variables.

Published by joegreenwoodhau

Joe Greenwood-Hau is a Lecturer in Politics in the School of Social and Political Science at the University of Edinburgh, where his teaching focuses on Introduction to Political Data Analaysis and he is wrapping up the Capital, Privilege and Political Participation in Britain and Beyond project.

2 thoughts on “Survey Variable: Social Network Homogeneity

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: