Estimating the Value of a Statistical Life: The Importance of Omitted Variables and Publication Bias