and of alternative regulatory strategies. Fluorocarbons are widely believed to threaten the stratospheric ozone layer and to influence climate. In fact, all the pertinent effects are highly uncertain, and could well be entirely different from what currently popular hypotheses suggest. The eventual ozone-depleting effect of fluorocarbons has been reduced approximately to zero in recent trials of the going models of stratospheric chemistry, due to new experimental data on the pertinent chemical reactions. The climatic effects could imply either net benefits or costs. Subject to these wide uncertainties, the recent data revisions lead my assessment of benefits of costs to a net benefit expected from continued unregulated emissions of fluorocarbons, due mainly to energy-related benefits of climatic warming. That is, it suggests that regulation would do more harm than good. However, there is a nontrivial possibility of extreme damage, and we might wish to insure against this threat by restricting emissions. Analysis of the buildup of risk over time suggests that the optimum time for a decision on regulation will be ten years or so in the future. In the meanwhile, further research will have narrowed down the uncertainties. At present they are so wide that a wrong decision either way could be very costly to the U.S. economy. The worldwide character of the problem adds both to the uncertainties and to the

There has been a great deal of discussion about whether the marginal utility of income rises with income. Most notably, Milton Friedman and Leonard J. Savage argued in their classic paper that the willingness to gamble implies that the marginal utility of income is rising over a range. The discussions of rising marginal utility and of Friedman-Savage gambles have proceeded independently of the literature on time preference, although the issues are in fact related logically. Drawing on this logical relationship we shall show 1) that at least when intertemporal utility is separable, the stable levels of consumption that are usually observed imply that the marginal utility of income decreases as income rises. 2) Even if the marginal utility of income does increase, Friedman-Savage gambles normally will not maximize utility; saving and dissaving can attain the levels of consumption which generate the most utility per dollar of income at a lower cost than gambles unless imperfections in the capital market are severe. 3) Even when the utility function is not temporally separable, repeated gambling cannot be a rational way of dealing with a rising marginal utility of income. We therefore conclude that observed gambling is seldom if ever explained by the logic set out in Friedman and Savage's seminal paper. In view particularly of the stability of consumption levels and the lack of FriedmanSavage gambles, we conclude that marginal utility of income does not rise with income. I. A Conceptual Framework