Evidence of a Fraternal Birth Order Effect on Male and Female Same-Sex Marriage in the Dutch Population: A Reply to Blanchard and Semenyna, Gómez Jiménez & Vasey

Click to increase image sizeClick to decrease image size

Acknowledgments

This research was partially supported by the Australian Research Council Centre of Excellence for Children and Families over the Life Course (project number CE140100027) and an Australian Research Council Discovery Early Career Researcher Award for a project titled ‘Sexual Orientation and Life Chances in Contemporary Australia’ (2017–2020). Authors declare no competing interests. All authors contributed to the study equally. The data used in these analyses were provided by the Dutch national statistical office, Statistics Netherlands. Inquiries regarding the data access should be addressed to [email protected].

Disclosure Statement

No potential conflict of interest was reported by the authors.

Notes

¹ Of note, ours are sex ratios at the point of observation. These may be slightly smaller than the sex ratios at birth due to higher mortality rates among men up to the point of observation.

² While this pattern may seem counterintuitive, it is also a consequence of endogenous stopping rules. For illustrative purposes, let us assume that all parents follow a strict stopping rule of having at least one son and one daughter (i.e., they will keep having children until the sex composition of their children aligns with the stopping rule). If we observe an individual who is the youngest child and whose older siblings are all brothers, then this individual must be female. In contrast, if we observe an individual who is the second-youngest child and whose older siblings are all brothers, then this individual must be male. In practice, compliance with stopping rules is far from strict. Yet this process is why youngest children whose older siblings are all brothers are more likely to be female, whereas second-youngest children whose older siblings are all brothers are more likely to be male.

³ To formally test whether the model coefficients corresponding to men and women were statistically different to each other we used the same interaction-based procedure applied in Ablaza et al. (2021).

⁴ This reasoning, however, would not apply if the stopping rules themselves had a direct effect on the probability of same-sex union entry (e.g., children whose parents have a diversity preference being more likely to enter a same-sex union). We consider this to be an unlikely situation, both because there is little theoretical ground to expect such an effect and because our coefficients do not change when we exclude youngest children from the estimation sample.

⁵ In the specific case of Blanchard and Lippa’s (2007) and Blanchard (2021a), we believe that the decision to exclude youngest siblings from the estimation sample was justified. Yet the underlying problem with their data was likely sampling biases unrelated to endogenous stopping rules. The latter becomes apparent when comparing the sibship characteristics of youngest siblings in our data (see Figure 2 below) to those in Blanchard’s data (see Figure 1 in Blanchard, 2021a). The comparison suggests that the online survey used by Blanchard and Lippa (2007) and Blanchard (2021a) oversampled homosexual participants with older sisters, which is a likely reason why their full-sample models failed to yield statistically significant estimates of the FBOE.

⁶ Of note, Blanchard and Lippa’s (2021) extension of the Khovanova (2020) method yielded significant FBOE estimates in the sample used by Blanchard and Lippa (2007) and Blanchard (2021a). This led the authors to speculate that the new procedure may bypass potential biases caused by endogenous stopping rules. While we find much to like about this procedure, we cannot think of a sound methodological argument that would support its superior performance under endogenous stopping rules. Rather, we believe that the performance of this procedure will depend heavily on the nature of sampling biases present in the data. Overall, we expect our modeling approach to perform better, because it operates with an unrestricted estimation sample that should be less vulnerable to small-sample biases.

⁷ As explained in our initial piece, the deficiencies of many earlier FBOE and FFE studies also encompassed a small sample size. For instance, a third of the studies reviewed in Blanchard (2018) included less than 100 homosexual participants. The larger size of our data enables us to detect associations that would be difficult to identify in such samples.

Additional information

Funding

This work was supported by the ARC Centre of Excellence for Children and Families over the Life Course [CE140100027].