This paper aims to analyze causality relationship between income and health, seeking to control the potential differences of this relation over the Brazilian territory. Three tests for Granger causality in panel data, proposed respectively by Holtz-Eakin et al. (1988), Granger and Huang (1997) and Hurlin (2005, 2007), are applied to a database of Brazilian states over the period 1981-2007. The main results show that conclusions can be misleading when they are based on the causality tests with homogeneous structure parameters. And so, the test proposed by Hurlin (2005, 2007), which controls the different types of heterogeneity, indicates that in Brazil the evidence is clearer for causality in direction from health to income.
income; health; Granger causality; panel data
