Identifying Interaction Effects Using Random Fertility Shocks

BACKGROUNDFertility interaction effects in social networks increasingly attracts the interest of demographers. While these theories propose a causal mechanism, they are rarely put to test in a plausibly causal statistical design.OBJECTIVEWe aim to differentiate network effects from selection by using an instrumental variable (IV) approach to achieve exogenous variation in fertility. We use interaction effects between siblings as an empirical example.METHODSWe draw data from Norwegian administrative registers (N similar to 170,000 men and women). We use twin births and children's sex composition as random fertility shocks (IVs), generating exogenous variation in third births.RESULTSIn our full study sample, we find no significant effects on ego's fertility of random shocks to the propensity to have three children. Subgroup analysis by sex and parity indicates positive effects for firstborn women when the sibling's birth is intended (as captured by the same-sex instrument). We find no evidence that similarity strengthens interaction effects.CONCLUSIONSThe study contributes to the literature by testing long-standing hypotheses of fertility interaction effects in a plausibly causal design. With the exception of firstborn women, we find no evidence of interaction effects of a sibling's third birth.