When is some number really better than no number? On the optimal choice between non-market valuation methods

Decision-makers have a wide variety of competing and complementary methods for non-market valuation, but there is little formal advice on the choice of method. I offer a formal approach, using a loss function (the mean square error) to compare contingent valuation, Citizens'Jury and methods where by intention only a portion of total value is estimated, when a) preferences vary across the population and b) methods are more or less susceptible to framing effects. Illustrative simulations suggest con-ditions under which the Citizens'Jury may dominate contingent valuation when framing effects are significant.