Detecting and correcting for family size differences in the study of sexual orientation and fraternal birth order

The term "fraternal birth order effect" denotes a statistical relation most commonly expressed in one of two ways: Older brothers increase the odds of homosexuality in later born males or, alternatively, homosexual men tend to have more older brothers than do heterosexual men. The demonstrability of this effect depends partly on the adequate matching of the homosexual and heterosexual study groups with respect to mean family size. If the homosexual group has too many siblings, relative to the heterosexual group, the homosexual group will tend to show the expected excess of older brothers but may also show an excess of other sibling-types (most likely older sisters); if the homosexual group has too few siblings, it will tend not to show a difference in number of older brothers but instead may show a deficiency of other sibling-types (most likely younger brothers and younger sisters). In the first part of this article, these consequences are illustrated with deliberately mismatched groups selected from archived data sets. In the second part, two slightly different methods for transforming raw sibling data are presented. These are intended to produce family-size-corrected variables for each of the four original sibling parameters (older brothers, older sisters, younger brothers, and younger sisters). Both versions are shown to render the fraternal birth order effect observable in the deliberately mismatched groups. In the third part of the article, fraternal birth order studies published in the last 5 years were surveyed for failures to find a statistically significant excess of older brothers for the homosexual group. Two such studies were found in the nine examined. In both cases, the collective findings for older sisters, younger brothers, and younger sisters suggested that the mean family size of the homosexual groups was smaller than that of the heterosexual comparison groups. Furthermore, the individual findings for the four classes of siblings resembled those for the present experimentally mismatched groups in which the mean family size of the homosexual group was significantly smaller. This illustrates the necessity of comparing groups on measures of mean family size and removing this confound in some way when those means are markedly different.