the hypothesized associations in our analyses. This indicates that our findings may actually underestimate the self-medication/dependence associations. Yet reporting biases may work the other way, in that individuals who drink heavily may want to use self-medication as a way of rationalizing their drinking, even if they don’t have mood symptoms. We were not able to assess whether individuals without affective symptoms or depression also report self-medication drinking, as the questions concerning self-medication were only asked of individuals with mood and anxiety symptoms. Third, although we were able to hold constant a large number and range of confounding characteristics, the potential for residual confounding remains. Individual-level characteristics such as coping skills, neighborhood-level items such as poverty and availability of liquor stores, as well as medical community-level factors such as access to mental health care or substance disorder treatment may explain some of the associations found in our current study. Although we were not able to hold constant all potentially explanatory characteristics, we were able to utilize propensity score methodology to assess the causal inferences of these associations. The use of this methodology reduces confounding due to observed characteristics and any unobserved characteristics associated with the observed ones. However, non-causal explanations for
