Homogamy in Gender Role Attitudes Among Young Couples - Evidence from Germany

Ansgar Hudde, University of Bamberg / Bamberg Graduate School of Social Sciences (BAGSS)
Henriette Engelhardt, University of Bamberg

How similar are young romantic partners in their gender role attitudes? We challenge the idea that there is strong assortative mating on gender role attitudes: we give theoretical reasons for why people might choose a partner with different gender role attitudes, and test the idea empirically. To do so, we compare real couples to three types of ‘synthetic’ couples, couples that are re-matched in different ways: (1) re-matched to create maximum similarity in attitudes, (2) re-matched to create maximum similarity on other characteristics, education, age, and religiosity, and (3) randomly matched couples. We use pairfam-data from around 300 young (duration of relationship maximum three years) German couples. Results show that similarity in attitudes of real couples is much closer to randomly matched couples than to couples with maximum similarity in attitudes. We discuss potential consequences for couple stability and family formation.

Why People Form Relationships with Partners That Hold Dissimilar Gender Role Attitudes

There are (at least) four potential reasons.

(1) Lack of information on partner’s attitudes.

(2) False consensus effect/ bias.

(3) Low initially perceived importance of gender role attitudes.

(4) Overestimation of convergence over time.

Data and Sample

Data. We analyse data from the first eight waves (2008/09-2015/16) of the German Family Panel (pairfam), release 8.0 (Brüderl et al., 2017; Huinink et al., 2011).

Case selection. We study the couples in which the anchor is in the cohorts 1981-83 and 1971-73, anchor persons that are in their mid-to late twenties and thirties at wave one (n=8,064). Our level of analysis is the couple. Couples enter the sample if (a) the anchor is in an opposite-sex relationship in wave one (n=6,048), (b) that partner also participated in the first wave of the survey (n= 3,375), (c) duration of relationship for maximum three years (n= 435), and (d) have no common children at wave 1 (n=371).

Analytical Strategy and Measures

Gender role attitudes. The following Likert-scaled items measure female and male gender and family role attitudes. Answer categories range from 1 (agree completely) to 5 (disagree completely).

  • women: family > career: Women should be more concerned about their family than about their career.
  • child <6 suffers if mother works: A child under age six will suffer from having a working mother.
  • housework: female involvement = male involvement: Men should participate in housework to the same extent as women.
  • child suffers if father focus work: Children often suffer because their fathers spend too much time at work.


Real couples. These simply are anchor-person matched to their actual partners.

Randomly matched couples. We divide the real couples into two data-sets, one that consists of men, the other of women. We then re-match couples based on a random variable. To achieve reliable estimates of similarity of randomly matched couples, we perform this step 10.000 times and compute the average value.

Matched to achieve maximum similarity in gender role attitudes. As for randomly matched couples, we start with two data-set, one containing women, the other men. We run an iterative process. First the intuitive description of the procedure: think of a speed-dating scenario, with nine women and men. Women sit on ‘the inside’ (blue dots with numbers), men sit ‘on the outside’ (orange dots with letters). Unlike regular speed-dating, in this case there is more than one round. In the first round, women and men are assorted randomly, and everyone is looking for a perfect fit. If a man and woman, that sit opposite of each other, gave identical answers to all items, they are a perfect match and leave the dating round (3c and 7g).

These two couples (3c and 7g) leave the round, and the remaining men rotate by one seat.

Those that have not found a partner yet, enter the next round and lower their standards: they now accept a partner that is one-Likert scale-point away from one’s one attitudes. h9 and b4 are a match now, and they leave the round. Rotation goes on until all possible combinations met.

Some women and men remain, as there is no perfect partner for them. They enter the second round. The second round proceeds in the same way as the first round, with the exception that people now ‘accept’ a partner that has slightly different attitudes (one Likert-point away on one item). After this round, there is a third, fourth etc. round. In each new round, the remainders lower their standards for similarity in attitude by one point – until all couples are matched.

We then calculate average similarity in attitudes among these couples. Note that this matching procedure is not necessarily the ‘best’ possible match; it is an approximation. To find the best possible match, we would need to compare 300! (=300*299*298*297…) couples, which quite impossible. Probably our matching procedure is a sufficiently good approximation though. Also, it is a Pareto-efficient matching: we could not give one person a more-suitable partner without giving another person a less suitable one.

Matched to achieve maximum similarity in education, age, religiosity. This follows the same rationale as matching to achieve similarity in gender role attitudes.

  1. Preliminary Results

Preliminary results show that similarity in gender role attitudes among real couples is much closer to similarity in random couples, than in couples that are matched to maximize similarity in attitudes. These results suggest there is no strong assortative mating concerning gender role attitudes among young German couples.

  1. Further Analyses

We further show whether couples experience attitude alignment over time. We find that there is statistically significant alignment (fixed-effects panel regression); however alignment is low in substantial terms.

Presented in Session 1097: Families and Households