From Blanchflower and Oswald 2006:
Instead, using samples of individual workers, [Blanchflower and Oswald 1995 and 1996] documents the existence of a logarithmic curve – what physicists would call a power law – linking the level of the wage to the unemployment rate in the local area. Their book’s conclusion is that, in sixteen nations, including the United States, the data are well described by a wage curve with an unemployment elasticity of approximately –0.1.
That is, if an area X has double the unemployment of another area Y it would have (approximately) 10% lower wages. The elasticity figure may be a slight overestimate due to the self-censoring involved in publication (no effect results don’t get published). As the authors note, no good theoretical reason for the existence of this empirical regularity has yet been provided.