آیا رضایت از زندگی است با مصرف الکل شکل برجسته ای دارد؟ شواهدی از داده های پانل روسیه

There has been a growing interest in the study of the shape of the relationship between alcohol consumption and psychological well-being in recent years. Overall, evidence is however still mixed and debated, the type of measures and methods of analysis having been emphasized as key elements in these studies. This paper contributes to this debate by providing new evidence relying on a large-scale population-based study. We used the Russia Longitudinal Monitoring Survey to build an unbalanced panel of 17,953 individuals providing 97,973 observations throughout 10 rounds. We studied the shape of the relationship between alcohol consumption (defined in grams of pure alcohol consumed in the last 30 days) and life satisfaction (measured by a five-item scale) by running a set of regressions. We successively introduced a large number of control variables (age, gender, marital status, occupation, income, health condition, education, living area, smoking status, and body mass index) and individual fixed effects in order to take both potential confounders and unobserved individual heterogeneity into account. Unadjusted analyses indicated a clear hump-shaped relationship between life satisfaction and alcohol use. The association was inverse J-shaped among men and inverse U-shaped among women. When control variables and individual fixed effects were introduced, the hump-shaped curve became increasingly flattened in all samples. Among women, all specifications (linear, quadratic and based on quartile dummies) turned non-significant. The quadratic specification for alcohol use remained however significant in the full sample and among men. In addition, in these two samples, being a fourth quartile drinker was negatively associated with satisfaction.

There has been a growing interest in the study of the shape of the relationship between alcohol consumption and psychological well-being in recent years. A non-linear relationship between alcohol consumption and symptoms of depression has already been described, alternatively as a J-shaped and a U-shaped relationship (Alati et al., 2005, Lipton, 1994, Rodgers et al., 2000 and Skogen et al., 2009). Overall, evidence is however still mixed and debated. Firstly, some authors have regarded some of these results as a statistical artifact due to study design and interpretation (Taylor & Rehm, 2005). Secondly, important gender differences have been underlined, with a number of studies suggesting a linear relationship among women (Alati et al., 2004, Caldwell et al., 2002 and Zhan et al., 2012). Thirdly, the types of measures used for both alcohol consumption and well-being have been emphasized as being key issues, leading to inconsistent findings, in such studies (El-Guebaly, 2007 and Graham et al., 2007). For all these reasons, it seems important to accumulate new evidence about the shape of the relationship between alcohol consumption and well-being using various samples, various types of measures for the two variables of interest and various statistical methods. In this paper, we wish to contribute to this aim by studying the relationship between alcohol consumption and life satisfaction data using a large-scale population-based study. Existing evidence concerning the relationship between alcohol consumption and positive measures of well-being, such as life satisfaction or happiness scores, is very limited. Restricting ourselves to surveys in the adult population, we found five studies that addressed this issue, often as a secondary objective. They relied on quite heterogeneous measures of well-being: happiness categories (Brenner, 1967), a happiness score (Ventegodt, 1995; cited by Veenhoven, 2003), the Cantril Self Anchoring Striving Scale (Levy, Bell, & Lin, 1980), a life satisfaction score (Koivumaa-Honkanen et al., 2012) and the Personal Well-being Index (Cummins, 2008). One of these studies (Levy et al., 1980) reported a U-shaped relationship: excluding heavy drinkers, a decreasing linear relationship was found between drinking and perceived satisfaction. Heavy drinkers however reported a higher satisfaction than moderate drinkers. A possible explanation, put forward by the authors, is that the levels of satisfaction reported by heavy drinkers might be inappropriate since alcohol abusers would be more susceptible to deny or fail to perceive or report their dissatisfaction. The four other studies tended to confirm the plausibility of a hump-shaped relationship. First, Brenner (1967) indicated that light drinkers had a higher mean happiness score than abstainers and medium/large drinkers. Statistical significance was however not reached (results based on our own calculations using Brenner's data — see Appendix A). Second, Ventegodt (1995) (cited by Veenhoven, 2003) reported that correlations were low and not significant, but an inverse U-pattern seemed plausible: moderate drinkers tended to be happier than abstainers and heavy drinkers. The greatest satisfaction was reached at 3–4 glasses consumed the week prior. Third, Cummins (2008) noticed not only that drinking a small amount of alcohol each day was generally associated with high well-being, but also that important differences appeared when gender and age were taken into account. More precisely, he found that females who never drank and females who drank more than three drinks per session had below normal well-being, while males who drank every day had above normal well-being, with no systematic change in male well-being with the number of drinks consumed. In terms of age, not drinking alcohol disadvantaged well-being for the 36–65 year group, as well as drinking more than three drinks per session for the 46–55 year group. Fourth, Koivumaa-Honkanen et al. (2012) found an inverse J-shaped relationship between a life satisfaction score and alcohol consumption for both men and women in their cross-sectional unadjusted analysis. All these studies relied on unadjusted analysis, i.e. did not control for confounding variables, except Cummins (2008) who used ANOVA analysis with covariates of gender, age and income. Maybe even more importantly, they did not use individual fixed effects, which have been emphasized as crucial in life satisfaction studies (Ferrer-i-Carbonell & Frijters, 2004). We overcame these limitations in our study.