It has long been recognized that early adversity represents a strong risk factor for the development of later psychopathology. However, proving a causal link between adverse exposures and mental disorder is often difficult because researchers cannot ethically randomize children to experiences that might cause them harm.
Traditionally, researchers have attempted to circumvent this problem by approaching causal inference through a series of increasingly rigorous observational designs. Longitudinal studies, for example, have shown that early adversity predicts multiple psychiatric disorders not only cross-sectionally but prospectively, suggesting that it is actual adversity—rather than simply the memory of adversity—that increases risk. Many studies of adversity also take steps to reduce the possibility of confounding by a third variable, either through introducing statistical controls, or by using propensity score–matching techniques in an attempt to “balance” covariates across exposed and nonexposed groups. However, the primary limitation of these approaches is that they account for only observable, measurable factors, which means that confounding by unobserved or unmeasured factors is still possible.