Saving/Consumption Behaviour of Russian Households: APC Analysis

Aim: The existing research base on the topic of consumer and saving behavior in households emphasises the discrepancies between the traits inherent to younger and older individuals arising from either cohort or age effects. This paper attempts to determine whether there exists significant cohort effect in Russian household consumption behavior and aims to disentangle age, period and cohort effects to provide better understanding of consumption profiles. Methods: Data was gathered from annual nationally representative panel survey from 1996 to 2018. Data included 29,950 individuals aged 14–102. Cohorts were defined as 19 groups within 5-year birth interval which included individuals born 1910-2004. Generations were defined as 6 groups the length of which based on the historical transitions in Russia. The data is aggregated into cohorts and generations to result in 386 and 117 observations respectively. The estimated mixed, random, and fixed effects model allow to overcome identification problem and test the significance of the introduced controls. Random effects coefficients of cohorts were estimated, and the variable of age together with age-squared were introduced into equations. Period is accounted for as data is structured in concordance with time. Explanatory variables were estimated on household and household head level. Controls were higher education, marital status, gender, children in households, employment, smoking habit, and urban status. Results: The regression results indicate significant and positive contribution of almost all controls, except for marital status, which is negatively related to consumption. Observing random effects coefficients of cohorts and variance in dependent variable which can be explained by cohorts makes clear that most of the cohorts significantly contribute to logarithm of non-durable consumption. Tukey test is provided to conduct direct comparison of logarithm of consumption between generations and cohorts, which ends up being significant in most of the cases.