So far, the results emphasized empirical regularities and avoided statements about causality. One issue is that consumption might be endogenous due to joint determination with SWB. This section uses instrumental variables (IVs) for consumption in order to provide causal inference. At the same time, I test for endogeneity to see whether IV estimation is even necessary. A good IV has two properties: validity and relevance. Validity means that the instrument is uncorrelated with the error term and not part of the original equation. Relevance means that the instrument is (sufficiently) correlated with the endogenous variable, in order to explain its variation.