Testing the cross-section implications of Friedman's permanent income hypothesis