Family income and self-rated health in Canada: Using fixed effects models to control for unobserved confounders and investigate causal temporality.

The objective of this study was to investigate whether the association between family income and adult self-rated health in Canada is causal in nature. The data came from the 2012, 2014 and 2016 waves of the Longitudinal and International Study of Adults linked to current and historical family income data from the Canada Revenue Agency. We used fixed effects models to describe associations between changes in self-rated health between 2012, 2014 and 2016 and changes in equivalized family income between (i) 2011, 2013 and 2015, (ii) 2008, 2009 and 2010, (iii) 2005, 2006 and 2007, (iv) 2002, 2003 and 2004 and (v) 1999, 2000 and 2001. We identified weak negative associations between family income and self-rated health operative over fourteen years or so for both women and men. These associations may be causal in nature. In addition, the implementation of models where changes in income preceded changes in self-rated health suggests that the associations reflect the causal effect of family income on self-rated health rather than the converse. These results make a contribution to the Canadian literature where nearly all previous research on associations between income and self-rated health is cross-sectional in nature and therefore incapable of establishing causal directionality.