A Research Note on the Possible Bias of Within-Estimators for Binned Variables

In this research note, I discuss the existence of a possible attenuation bias in research involving within-estimation (e.g., such as in sibling comparison designs) and the use of binned variables—i.e., categorical versions of continuous variables broken down into intervals. I draw on a specific example of recent research on the effects of parental age at birth and birth spacing, although the problem I address is more general in nature. I explain the nature of the bias, show how and under what conditions it arises, and provide simple recommendations on how the bias can be either avoided altogether and/or accounted for in the inference. This paper builds on research supported by the German Research Foundation (DFG), grants LE 3253/4-1 and YA 778/1-1.