Support #869

Unexpectedly strong gender homophily in Understanding Society compared with the BHPS

Added by Till Hoffmann about 6 years ago. Updated 3 months ago.

Survey design
Start date:
% Done:




I am interested in homophily effects that can be derived from the social network questions in waves B, D, F, H, J, L, N, P, R of the BHPS and waves C and F of Understanding Society. Homophily with respect to occupational status is relatively constant across all waves, and age homophily only changes slightly. However, homophily with respect to gender increases sharply from the last wave of the BHPS (R) to the first wave of Understanding Society (C). Do you happen to have any ideas whether a data error (or misuse on my part) might cause such an effect?

Details and code for reproducing the observations

For each wave of the survey I compute the contingency table of the sex of the ego (respondent) and alter (nominated friend) and apply the Fisher exact test ( to compute the odds ratio of association and test for statistical significance.

To account for effects of the survey design I use the weight

for the BHPS and
for Understanding Society (the results are similar when omitting the survey weights). The results below include survey weights but the contingency tables have been rounded to the nearest integer. The code to reproduce the results can be found here:

Thank you very much for your help!

BHPS wave: b
Contingency table
[[ 8862  3011]
 [ 2202 11176]]
Odds ratio: 14.947, p-value: 0.000000

BHPS wave: d
Contingency table
[[ 8435  2817]
 [ 2053 11044]]
Odds ratio: 16.116, p-value: 0.000000

BHPS wave: f
Contingency table
[[ 8500  3253]
 [ 2195 11149]]
Odds ratio: 13.277, p-value: 0.000000

BHPS wave: h
Contingency table
[[ 8328  3034]
 [ 1980 10994]]
Odds ratio: 15.249, p-value: 0.000000

BHPS wave: j
Contingency table
[[ 7936  2935]
 [ 1941 10541]]
Odds ratio: 14.697, p-value: 0.000000

BHPS wave: l
Contingency table
[[ 7625  2791]
 [ 1888 10157]]
Odds ratio: 14.696, p-value: 0.000000

BHPS wave: n
Contingency table
[[7320 2663]
 [1827 9780]]
Odds ratio: 14.713, p-value: 0.000000

BHPS wave: p
Contingency table
[[7099 2597]
 [1714 9436]]
Odds ratio: 15.052, p-value: 0.000000

BHPS wave: r
Contingency table
[[ 7682  2910]
 [ 1964 10312]]
Odds ratio: 13.865, p-value: 0.000000

Understanding Society wave: c
Contingency table
[[36818  9400]
 [ 6777 44476]]
Odds ratio: 25.710, p-value: 0.000000

Understanding Society wave: f
Contingency table
[[30372  7348]
 [ 5004 36279]]
Odds ratio: 29.972, p-value: 0.000000

Updated by Stephanie Auty about 6 years ago

  • Category set to Data analysis
  • Status changed from New to In Progress
  • Assignee set to Stephanie Auty
  • % Done changed from 0 to 10
  • Private changed from Yes to No

Many thanks for your enquiry. The Understanding Society team is looking into it and we will get back to you as soon as we can.

Best wishes,
Stephanie Auty - Understanding Society User Support Office


Updated by Stephanie Auty about 6 years ago

  • Status changed from In Progress to Feedback
  • Assignee changed from Stephanie Auty to Till Hoffmann
  • % Done changed from 10 to 80

Dear Till,

There are several differences in the way these questions were asked from Wave 18 of the BHPS to Wave 3 of Understanding Society.

1. In Understanding Society, these questions are only asked if the respondent says they have more than one close friend. In the BHPS there is no such filtering and so they are not allowed to say they don’t have any close friends. As a result this could lead to someone in the BHPS answering about people they would not count at ‘close’ friends.

2. In the BHPS the question specifies that the friends to whom these questions refer should not live with the respondent but could be a relative. In Understanding Society, there is no such restriction.

3. Also, note that you are using weights which are specific to the self-completion section in Understanding Society, but in BHPS there are no self-completion weights.

4. In Understanding Society, a much higher proportion of the sample are ethnic minorities as compared to the BHPS sample. Even if you use weights which should correct for the over-sample of ethnic minorities in the UKHLS, UKHLS and the BHPS represent different populations (1990 Vs 2009/10). If these populations have different propensities of having same-gendered friends, then there may be differences.

Best wishes,
Stephanie Auty - Understanding Society User Support Officer


Updated by Stephanie Auty almost 6 years ago

  • Status changed from Feedback to Resolved
  • % Done changed from 80 to 100

Updated by Stephanie Auty almost 6 years ago

  • Status changed from Resolved to Closed

Updated by Understanding Society User Support Team 3 months ago

  • Category changed from Data analysis to Survey design

Also available in: Atom PDF