Income and (un)healthy consumption: Estimating the effect of income on smoking, drinking and sports by exploiting policy changes in the income-dependent combination-reduction (IACK)

This thesis evaluates the effect of income on smoking, drinking and sports by exploiting policy changes in the income dependent combination reduction ("inkomensafhankelijke combinatiekorting", IACK), a Dutch type of EITC. It thereby adds to the existing, small body of literature on the effect of income on health behaviour with a novel approach. As far as I know, this is the first time that the IACK has been used in this type of analysis. Moreover, it is a first empirical test of a model which could potentially explain the socioeconomic status health gradient (SES health gradient). The existence of the SES health gradient is undisputed, but the causal mechanisms are difficult to uncover (Smith, 1999). Galama and van Kippersluis (2017) propose a model that aims to explain this health gradient. This paper test one of their predictions, namely that wealthier individuals engage more in healthy and moderately unhealthy consumption than less wealthy individuals, but engage less in severely unhealthy consumption. The research question of this paper is as follows: 'To what extent does income influence the level of (un)healthy consumption by individuals?' The research question is tested with panel data from the LISS Panel, covering 7 waves in the period 2008-2015 and consisting of 28,838 observations for 8,657 individuals. A 3-step analysis is used. First, OLS models with and without controls for age, gender, marital status and changes in health are tested. Secondly, FE models with individual and year fixed effects are estimated. Thirdly, policy changes in the IACK are exploited as a source of external variation in income. This paper finds some evidence in support for the Galama van Kippersluis' (2017) prediction. The coefficient of income on the external margin of alcohol consumption is significant in the FE model. The average marginal effect of a 10% income increase is a 5.88 percentage point increase on the likelihood of alcohol consumption. However, this effect seems to be mainly driven by 16 to 24 year old, which cannot be explained by income. For this group the average marginal effect on a 10% income increase is a 20 percentage point increase in the probability of drinking. It is difficult to say whether this effect is causal, as the main group that drives the effect is not affected by the IACK policy. Moreover, for other types of unhealthy and healthy consumption no evidence for an effect of income is found. The main limitations of this thesis are that it cannot succesfully use the IACK to exploit external variation in income, and therefore cannot make claims on causality. Moreover, it focuses on three types of unhealthy and healthy behaviour. Whereas its findings suggest that these different types cannot be generalized. Further research should focus on expanding the range of behaviours, isolating external variation and empirically testing other predictions of Galama and van Kippersluis (2017).