Childbearing Preferences in the Face of Work, Family, and Financial Constraints

Download data and study materials from OSF

Principal investigators:

Julia Behrman

Northwestern University

Email: julia.behrman@northwestern.edu

Homepage: https://sociology.northwestern.edu/people/faculty/core/julia-behrman.html

Emily A. Marshall

Franklin and Marshall College

Email: emily.marshall@fandm.edu

Homepage: https://www.fandm.edu/directory/emily-marshall.html

Florian Keusch

University of Mannheim

Email: f.keusch@uni-mannheim.de

Homepage: https://floriankeusch.weebly.com/


Sample size: 1794

Field period: 04/12/2023-11/01/2023

Abstract
US ideal family size has remained relatively stable in recent years, reflecting widespread preference for two-child families, we know very little about the strength of these preferences among young adults. To examine the relative strength of preferences for family size, compared to other attributes of family life, in this experiment we conduct a forced-choice online conjoint survey experiment on a nationally representative sample of 1,785 American women ages 18-35. We find that when family size is included as one of six attributes in a family scenario, preference for scenarios with two children is not significantly different from preference for scenarios with zero, one, or three-child families, net of other attributes—although four-child scenarios are significantly less preferred than two-child scenarios. Evaluation of the relative magnitude of different attribute effect sizes shows that preference for two-child scenarios is comparatively less important than preference for many of the other attributes. Our findings suggest that even if ideal family size remains at or above two children in standard survey research, preferences for two-child families are surprisingly weak in hypothetical scenarios that account for competing family demands. In additional analyses we explore whether there is heterogeneity in our core findings by partnership status and economic stability.
Hypotheses

1) We hypothesize that women will prefer families with two children (compared to zero, one, three, or four children), reflecting the endurance of the two child norm.

2) We hypothesize that the Average Marginal Component Effects (AMCEs) of zero, one, three, or four children (compared to the reference category of two children) will be significantly larger in absolute magnitude than the AMCEs of other attributes (relative to their reference categories) in predicting preferred family scenario, thus reflecting the importance of childbearing in American conceptions of family.

3) We hypothesize that for each for each category of childbearing (i.e., one child, two children, etc.) the predicted probability of being the preferred family scenario will be significantly higher for scenarios where respondents are married compared to scenarios where respondents are unmarried or cohabiting, thus reflecting the enduring symbolic importance of marriage for childbearing in the U.S.

4) We hypothesize that for each category of childbearing (i.e., one child, two children, etc.) the predicted probability of being the preferred family scenario will be significantly higher for family scenarios where respondents are highly financially stable compared to moderately stable or financially struggling, thus reflecting the importance of financial stability for childbearing.

Experimental Manipulations
We conduct a forced-choice paired conjoint online survey experiment. In our experiment, respondents view two family scenarios, each comprising six categories of randomly varied attributes: number of children, division of care and housework, time intensity of career, family policy, financial stability, and partnership status. These attributes were chosen to represent the competing work, family, and financial constraints and opportunities that make up family life. Respondents are asked to choose which of the two scenarios they would personally prefer.
Outcomes
The dependent variable is a binary indicator of preferred family scenario. For a given pair of family scenarios (our unit of analysis), the scenario the respondent prefers is coded as one and the family scenario the respondent does not prefer is coded as zero.
Summary of Results

1) We find support for preferring two over four children: The probability of preferring family scenarios with four children was 10 percentage points lower than the probability of preferring scenarios with two children, net of other attributes (p<0.001). However, there is no evidence that family scenarios with two children were significantly preferred over those with zero, one, or three children.

2) Our study does not find evidence that preference for two children is more important to young adult women than preference for other components of family life, and we find evidence that it is less important in some cases. More specifically, the coefficients for zero, one, three, and four children (relative to two) have significantly smaller effects on the probability of a scenario being preferred compared to the coefficient for financially struggling (relative to having plenty extra money). The coefficients for zero, one, and three children (relative to two) are also significantly smaller in magnitude than the coefficients for single and cohabiting (relative to married), 60-hour work week (relative to 40-hour), having just enough money (compared to plenty extra), and having no family policy (compared to generous family policy). The coefficient for four children (relative to two) does not significantly differ in magnitude from any of these coefficients. There are a few other notable instances where there are not significant differences in coefficient magnitudes. For example, the magnitudes of childbearing coefficients do not significantly differ from the magnitude of the coefficient for the respondent specializing in housework and care work (relative to sharing equally). Of the childbearing coefficients, only the coefficient for four children (relative to two) has a larger effect (in absolute terms) than the coefficient for the respondent’s partner specializing in housework and care work (relative to sharing equally); the other childbearing coefficients do not significantly differ in magnitude.

3) We find no evidence of preference for partnership status (single vs. married vs. cohabiting) in respondents’ evaluations of scenarios without children. Yet, there is strong preference for marriage (relative to cohabitation or being single) in scenarios with children, regardless of the number of children.

4) We find respondents are not significantly more likely to prefer having children (or prefer more children) in scenarios of economic stability, and they are not significantly less likely to prefer having children (or prefer fewer children) in scenarios of economic instability.

References

Behrman, J.A., Marshall, E., and Keusch, F. 2024. A Conjoint Survey Experiment on Childbearing Preferences in the United States (WP-24-08)

Behrman, J.A., Marshall, E., and Keusch, F. An Experimental Approach to Assessing Young Women's Childbearing Preferences: A Research Note on the United States. Forthcoming at Demography.

Marshall, E., Behrman, J.A., and Keusch, F. The Interaction of Preferences for Marriage and Childbearing: An Experiment Among U.S. Women. Working Paper and Forthcoming Conference Presentation at the Population Association of America Annual Meeting.

Behrman, J.A., Marshall, E., and Keusch, F. Economic Instability and Childbearing: An Experimental Approach. Working Paper.