Uncertain technical change, Bayesian learning and the Green Paradox

The phenomenon of the ‘Green Paradox’ has been widely discussed in the climate economics literature for the last 15 years. The term refers to a situation in which a well-intended climate policy leads to adverse results, such as a rise in greenhouse gas emissions. The emergence of the paradox is usually attributed to the exhaustibility of fossil fuels: firms extracting these resources seek to equalise the present value of resource rents in each time period, so, anticipating future tax increases, they increase current production. On the other hand, if the productivity growth in the green sectors is driven by learning-by-doing, the Green Paradox does not emerge. In this paper, I show that the Green Paradox may arise as a consequence of an ex ante optimal policy in the absence of exhaustible resources if technological change in the clean sector is subject to uncertainty. That is, if the true speed of technological progress is an unknown parameter, economic agents have to form their expectations regarding future technological development with the help of Bayes’ rule and make their decisions accordingly. If the market expects (a priori) the demand for dirty capital to shrink more rapidly due to technological progress than policymakers do, the latter must cut carbon taxes or even subsidize investment in dirty capital to avoid underinvestment in this type of capital and, consequently, the underproduction of energy in the present. If the flow of subsidies is persistent enough, CO2 emissions may rise.