Child effects in socialization research: Some conceptual and data analysis issues

This paper examines issues associated with the conceptualization of child effects in socialization research and outlines possible data analysis strategies that may be used with contemporaneous correlational data from parents and children. The paper begins with an outline of some recognized prohlems in the studv of child effects. The role of cognitive factors, which complicate efforts to disentangle child effects, is emphasized, in an attempt to clarify further the meaning of child effect, three different types of effect are outlined and then a distinction is made between moderator and mediator variables. In general, a moderator variable specifies when certain outcomes will hold (e.g. for one type of child but not another), whereas a mediator variable indicates how or why the outcome occurs. The distinction between moderator‐mediator variables provides the main focus of the paper. It assists in the conceptualization of child effects, as well as having direct implications for data analysis strategies. The relevance of the distinction is illustrated via data from a study of family relationships. The use of correlations, multiple regressions with interaction terms, and latent variable path analysis are examined as ways of investigating moderator and then mediator variables. Conclusions are reached about conceptual and data analysis difficulties associated with attempts to determine child effects.