That tax haven policies contribute to favorable economic growth in tax haven countries is commonly accepted. Empirical investigations, however, do not substantiate this assertion and are subject to endogeneity bias. Using a sample of 155 countries from 1982 to 2003, we find that the standard tax haven variable is endogenous to the error term in a typical growth regression. We offer land area measures as valid instruments for tax haven status. Results based on two-stage least squares estimation with heteroskedastic standard errors and controls for initial conditions provide support for the claim that tax havens "flourish" compared with non-tax haven countries even when accounting for the self-selection of tax haven status.