The more the merrier? Perhaps so!
In a new publication in Social Indicators Research, Thijs van den Broek and Marco Tosi analysed data from the Generations & Gender Programme (GGP) of older parents from eight Eastern-European countries. They were interested in determining if having multiple children has a causal protective effect against later-life loneliness among women in this region. Earlier research has already shown that parents of many children tend to be less lonely than their counterparts with fewer children, but this does not necessarily mean that high fertility protects against loneliness. For instance, people with a tendency to be lonely may tend to remain unmarried and have lower fertility. Additionally, certain background characteristics may increase the likelihood of low fertility, as well as feelings of loneliness.
To address these issues the authors needed to identify two groups of parents with similar background characteristics, but different fertility patterns. The groups were identified through the sex of the first two children, distinguishing between parents whose first two children were of the same sex and parents with opposite sex children. On the one hand, it is randomly assigned by nature whether the sexes of the first and the second child are similar. Therefore, the two groups of parents are bound to have identical background characteristics. On the other, a preference for mixed-sex offspring makes parents more inclined to desire an additional child when they have either two boys or two girls. This was indicated by the fact that they were more likely to have a third child when the first two children were either both girls or both boys, as opposed to a girl and a boy. Consequently, parents whose first two children were of the same sex had higher completed fertility. By using this strategy, known as an instrumental variable approach, the authors show that for women, additional children had indeed a causal protective effect against loneliness. For men, a similar association was also found, but no causality could be inferred.