A Critical Introduction to Instrumental Variables for Sibship Size Based on Twin Births

Abstract

Research on how children are affected by their sibship size, i.e., their number of siblings, faces a serious hurdle because sibship size is endogenous in the model. The current “gold standard” method to solve this is to use (parity-specific) twin births as instrumental variables. The purpose of this chapter is twofold. Firstly, it introduces instrumental variable methods in general. Secondly, it argues that we need to update our interpretation of instrumental variables for sibship size based on twin births. The essence of my new interpretation is that only some twin births lead to an exogenous increase of the number of children. This seemingly minor update of the interpretation highlights some previously overlooked necessary assumptions. Most importantly it shows how the method is intrinsically linked to fertility preferences and how assumptions regarding unintended pregnancies are crucial for identification. We need to assume that families have a fixed desired number of children. We also need to assume either that there are no unintended pregnancies or that parents who have an “unwanted” child through single births are no different from parents who have an “unwanted” child as the result of a twin birth. My introduction therefore contributes to the mounting critique of using twin births as instrumental variables for sibship size. We need to re-evaluate this current “gold standard” method and possibly reinterpret previous results using it.