Measurements of attitudes towards gender roles in cross-cultural surveys: Internal and external validity

Measuring attitudes towards gender roles in comparative researches is a difficult task, previous research pointing out the difficulties of finding a cross-national invariant indicator. This paper compares two scales one used by the International Social Survey Program 2002 and the other by World Values Survey 2005, testing for both internal and extern validity of each, in order to see which of them is a more accurate instrument for measuring attitudes towards gender roles. We have employed Confirmatory Factor Analysis aiming to find an invariant measure for attitudes towards gender roles across countries. We have test for external validity using hierarchical multi-level models, which allow testing expected effects of theoretically relevant variables.

For the items used in ISSP 2002 the Confirmatory Factor Analysis confirmed the existence of two different factors, one tapping attitudes towards the traditional division of gender roles: male breadwinner - female homemaker, while the other indicating attitudes towards male involvement in domestic work and childcare. The measurement model is partially invariant in most of the countries included in the survey. The external validation pointed out that the first indicator is a robust estimator of attitudes towards traditional gender roles, but does not produce similar results for the second indicator.

World Values Survey combines one item that refers to private gender roles with items measuring attitudes towards equal involvement of men and women in public sphere (in education, employment and politics). The confirmatory factor analysis indicates the existence of one factor, tapping only attitudes towards equal involvement in public sphere. The external validation shows that this factor has weaker connections with variables measuring family context (like marital status or parenting), but it is well connected with indicators tapping political participation or employment.