The problem with measuring any sort of relationship in social sciences is that there can be confounding variables that can make us think that a causal relationship exists where it really does not.
In the case of a relationship between unemployment and health, we can think of other factors that can make unemployed people less healthy than employed people. For example, the unemployed generally come from a much lower socioeconomic status than the whole population of the employed. They may well have worse health care and more stress in their lives even before they become unemployed. In other words, it may be their socioeconomic status over their whole lives that has made them less healthy, not the fact that they are unemployed at the moment.