Data and method Summary and conclusion
5 Results

The only covariates that turn out to have a significant influence on the dependent variable in all equations are the mother’s age at first birth and the interbirth interval, both with the expected negative sign of the regression coefficient. Since we are primarily concerned about the effects of the sex composition, we do not report the estimated coefficients of independent variables other than for the sex-combination variables in both models, which are displayed in [Table 2].

The findings of Model 1 show that in every third analyzed country there is no gender preference at all. These countries are Finland, France, western Germany, Norway, Poland, and Portugal. In the other countries (Austria, Belgium, Czech Republic, eastern Germany, Hungary, Italy, Latvia, Lithuania, Slovenia, Spain, Sweden, and Switzerland), a couple is significantly more likely to progress -or express an intention to progress- to parity three, if their previous two children are of the same sex, than in the case of a mixed sex combination. This points to a preference for a sex-mix. For Belgium and eastern Germany, however, the coefficient of the sex-composition variable is significant on the 10%-level only.

The subsequent analysis (Model 2) produces more detailed results, which basically confirm the findings of our first model. In most cases with a significant sex-composition variable in Model 1, the coefficients for the sex-combination dummies used in Model 2 are either both significant (Austria, Hungary, Italy, Latvia, Slovenia, Spain, and Switzerland), or -if only one turned out to be significant- are not significantly different from each other (Belgium, eastern Germany, and Sweden). All this supports the findings from older studies that (if there is any gender preference at all) couples prefer to have at least one child of each sex.

In contrast to Model 1, however, France turns out have a significant and positive effect on the girl-girl dummy, which would indicate a preference for boys. The coefficients of the sex-composition dummies do not differ significantly, though. Therefore we keep our classification of France as a country with no gender preferences. Also inconsistent with Model 1 is the Portuguese girl preference. Here we even find a significant difference between the coefficients of boy-boy and girl-girl. A highly significant girl preference is also found in the Czech Republic, and in Lithuania. The latter three countries are the only ones with significant differences between the sex-combination dummies (on either the 5%- or 10%-level of significance).

Taking together the results of our two models, we find a geographical pattern of gender preferences, which -contrary to our presuppositions- does not suggest a particular regional grouping [see Figure 1]. A hypothesis along the lines that more traditional societies tend to prefer boys (Southern Europe), while more progressive societies tend to prefer girls or a sex-mix (Northern Europe), cannot be supported by our findings. Although a particular regional pattern with regard to fertility levels can be found in Europe [7], which is influenced by different socioeconomic conditions and family policies, the pattern of gender preferences we find in our paper is unlikely to be caused by differences in these factors. While socioeconomic conditions and family policies are important determinants of the fertility level, their effect on childbearing is usually gender neutral. A similar argument holds for many other socioeconomic incentives that affect the overall desire for children. Moreover, the socioeconomic incentives discussed in [Section 2], which can lead parents to desire offspring of a particular sex, have diminished in contemporary European societies. We therefore suppose that gender preferences may vary because of differences in cultural and social institutions across European countries.

Unfortunately, the social and cultural institutions that may lead parents to prefer different sex-compositions of their children cannot be analyzed in greater detail with the data available in the FFS. The investigation of these factors, however, deserves future research efforts, and necessitates detailed, possibly qualitative, and country-specific studies.

In addition to the analysis above, we investigated, whether the sex of the first child had any influence on the parent’s decision -or intention- to have a second child. It turned out that only in Portugal (on the 5%-level) and in eastern as well as western Germany (on the 10%-level), there is a significant effect of this variable. Consistent with our above findings, girl preference is evident in Portugal. Interestingly, there is again a clear difference between eastern and western Germany: while in western Germany having a boy as first born decreases the parent’s desire to progress to parity two, the reverse effect is true for eastern Germany [Note 4].

Data and method Summary and conclusion

logo70.gif (2450 bytes)

Gender Preferences for Children in Europe: Empirical Results from 17 FFS Countries
Karsten Hank and Hans-Peter Kohler
© 2000 Max-Planck-Gesellschaft ISSN 1435-9871