Abstract
In this article we examine the effects of self-reported sexual orientation on substance abuse. Using data on a random sample of 6,713 individuals in Washington State, this study examines causes and correlates of substance use by sexual minorities, an at-risk and treatment underserved population. Logistic regression results indicate homosexual orientation is a significant positive predictor of past year marijuana use, past year hard drug use, past year binge drinking, and lifetime alcohol addiction. Bisexual orientation is a significant predictor of past year marijuana use, past year hard drug use, and past year binge drinking. Potential causal mechanisms for these elevated patterns of substance use are discussed.
Notes
1The current article, due to data constraints specified below, focuses on LGB along with heterosexuals. Our data do not enable us to examine individuals identifying as transgendered or queer or questioning (respectively, the “T” and “Q” in LGBTQ) or other sexual minority identities. Although LGBTQ populations are often combined and presented as a single entity for either research or advocacy purposes, these are distinct groups with distinct concerns (IOM Citation2011). Furthermore, it is important to acknowledge that the experiences of these sexual minority populations “are not uniform and are shaped by factors of race, ethnicity, socioeconomic status, geographical location, and age, any of which can have an effect on health-related concerns and needs” (IOM 2011:1).
2As our data contain measures of past 30 day marijuana and hard drug use, but only past year binge drinking, we chose to utilize past year measures both for uniformity and for statistical power. Regressions of past 30 day marijuana and hard drug use were run and were substantively identical to past year results in both magnitude and direction of all coefficients, though the rarity of hard drug use, combined with the relatively small number of gay and bisexual respondents, reduces statistical significance to above .05 but below .10 in that case only.
3Use of Hc1 robust standard errors did not significantly alter standard error values.
4The ordinal measures for education and emotional health sufficiently conform to linear interval variable behavior and therefore do not violate assumptions of logistic regression.
*p < .05, **p < .01, ***p < .001.
*p < .05, **p < .01, ***p < .001.