When should the distant future not be discounted at increasing discount rates?

A number of governments have already adopted the policy of applying Declining Discount Rates (DDRs) to long lived projects, a move that will significantly affect public sector investment decisions. This paper argues that such policy is misguided, and revisits the discussion that led to it. A 2009 paper by Christian Gollier and Martin L. Weitzman is widely regarded as having solved the Weitzman-Gollier Puzzle, which is that the definition of expected present value (EPV) proposed by Weitzman’s in 1998 is inconsistent with the calculation of expected future values (EFV) when market interest rates are stochastic but perfectly auto-correlated. The inconsistency is actually due to the fact that Weitzman’s EPV formulation is incorrect. When it is replaced by the correct formulation, the puzzle disappears, and risk neutral certainty equivalent rates (CERs) turn out to be growing, rather than declining under the assumptions of Weitzman’s model. This removes the justification for the use of DDRs. This paper shows that Gollier and Weizmann (2009) fail to resolve the puzzle. Adding risk aversion to Weitzman’s 1998 model to derive risk adjusted CERs cannot resolve the inconsistency between alternative methods of computing expected monetary yields, because investors’ risk aversion only affects their own valuations, not market yields. If monetary CERs increase, the underlying efficiency of investment projects must generally match the growing monetary CERs of capital markets for them to be worth investing in, even for risk averse investors. The distant future should only not be discounted at increasing discount rates if Weitzman’s 1998 assumption of perfectly auto-correlated interest rates fails to hold sufficiently.