Survey protocol and income effects in the contingent valuation of public goods: A meta-analysis