Yingying Lu, my post-doc on our ARC project, and myself have a new working paper on our research on the effects of assumptions about substitutability on the estimated costs of climate change mitigation policy. Some of the results are in line with our expectations and some are quite surprising...
I originally proposed this project because I was surprised that there could be such different views on the costs of stopping climate change. The mainstream economic community working on these issues usually finds that the costs of even quite strong action are in the neighborhood of lowering GDP by 1-4% below what it would be under business as usual (BAU). As GDP is expected to continue to grow strongly in such models, this seems to be a quite trivial cost to avoid disaster. It implies that doubling today's level of GDP will be delayed by just 1 to 2 years. Tavoni and Tol argued that these figures ignore those models which failed to be able to simulate the stronger policy scenarios. But even when they compensate for that bias they estimate that the net present value of the reduction in GDP is about 8% of BAU GDP. On the other hand, Tim Jackson argued that we need to stop economic growth in order to have any chance of dealing with climate change. This does not seem to be an uncommon view among natural scientists, environmentalists, and also many climate skeptics. Roger Pielke argues that such unprecedented decarbonization is "all but impossible". Again, the implication is then that growth must be stopped in order to reduce emissions.
So, I wondered whether mainstream climate models are somehow missing something. Specifically, are they assuming that it is easier to reduce fossil fuel use than it actually is. If the economy was less flexible - if the parameters known as elasticities of substitution were smaller - it would presumably be harder to reduce fossil use. Very little research has been published on the sensitivity of climate policy costs estimated by mainstream computable general equilibrium (CGE) models to changes in the elasticities of substitution. And what there is is not really designed to answer this question.
Our research uses McKibbin and Wilcoxen's G-Cubed model. We ran the model under BAU and four policy scenarios ranging from a 20% global cut in emissions by 2030 relative to 2010 to a 20% increase, which matches the RCP scenarios quite well.
We perturbed most of the elasticities of substitution in production and consumption (but not those between domestic and foreign goods and services) by increasing them by 50% and reducing them by 50%. We also tried some other parameter sets, including setting all elasticities to 0.5; setting all elasticities of substitution between capital, labor, energy, and materials to 0.5 and all those between fuels to one; setting all elasticities to 0.1; and setting all elasticities to 2.
Not surprisingly, as we reduce the elasticities, the cost of abating a tonne of carbon increases and vice versa. What is surprising, is the extent to which the BAU emissions path is changed. BAU emissions are reduced in the less flexible economies relative to emissions in the default model. This effect is so strong that usually the total cost of reducing emissions increases with increasing flexibility and vice versa. In fact, in our most extreme low flexibility scenario, emissions grow so slowly that the more moderate policy scenarios are not binding. Economic growth is in fact halted and so there is a much reduced climate problem to deal with. This seems to be an example of the de La Grandville hypothesis that, the greater the elasticity of substitution, the faster the rate of economic growth.
Yingying and I debated whether the growth effect is real or an artefact of our modelling. Jorgenson et al.'s study avoided the issue by looking at policy scenarios that are based on percentage reductions in emissions relative to business as usual. Babonneau et al. adjust the rates of technical change so that the BAU scenario reproduces the expected rate of economic growth in the European Commission's World Energy Technology Outlook. We believe that this is likely to be a real effect. On the other hand, G-Cubed assumes that the rate of technological is exogenous, whereas the rate would also likely vary with the elasticities of substitution. Additionally, the baseline levels of output and prices at the start of our simulation are based on the real world level of these variables which would also differ if the economy was very different. Therefore, our results are not a reliable indication of the relative performance of more and less flexible economies in the real world
So, what is the bottom line?
1. Because a less flexible economy has higher abatement costs per tonne of carbon but less emissions growth, if what we care about is the total costs of climate policy then it is not so important to get good estimates of elasticities of substitution. If we care about average and marginal costs of abatement, then these parameters are critical. We again find that the distinction between marginal and total costs of abatement is important.
2. Though stopping growth reduces the climate change problem, the reverse isn't true. We cannot find a model economy where the costs of climate mitigation are so high that such a policy would result in stopping economic growth or that mitigation cannot be achieved without stopping growth. Certainly, assuming that the economy is a lot less flexible than it is cannot generate high total costs. In fact the reverse is true.
I originally proposed this project because I was surprised that there could be such different views on the costs of stopping climate change. The mainstream economic community working on these issues usually finds that the costs of even quite strong action are in the neighborhood of lowering GDP by 1-4% below what it would be under business as usual (BAU). As GDP is expected to continue to grow strongly in such models, this seems to be a quite trivial cost to avoid disaster. It implies that doubling today's level of GDP will be delayed by just 1 to 2 years. Tavoni and Tol argued that these figures ignore those models which failed to be able to simulate the stronger policy scenarios. But even when they compensate for that bias they estimate that the net present value of the reduction in GDP is about 8% of BAU GDP. On the other hand, Tim Jackson argued that we need to stop economic growth in order to have any chance of dealing with climate change. This does not seem to be an uncommon view among natural scientists, environmentalists, and also many climate skeptics. Roger Pielke argues that such unprecedented decarbonization is "all but impossible". Again, the implication is then that growth must be stopped in order to reduce emissions.
So, I wondered whether mainstream climate models are somehow missing something. Specifically, are they assuming that it is easier to reduce fossil fuel use than it actually is. If the economy was less flexible - if the parameters known as elasticities of substitution were smaller - it would presumably be harder to reduce fossil use. Very little research has been published on the sensitivity of climate policy costs estimated by mainstream computable general equilibrium (CGE) models to changes in the elasticities of substitution. And what there is is not really designed to answer this question.
Our research uses McKibbin and Wilcoxen's G-Cubed model. We ran the model under BAU and four policy scenarios ranging from a 20% global cut in emissions by 2030 relative to 2010 to a 20% increase, which matches the RCP scenarios quite well.
We perturbed most of the elasticities of substitution in production and consumption (but not those between domestic and foreign goods and services) by increasing them by 50% and reducing them by 50%. We also tried some other parameter sets, including setting all elasticities to 0.5; setting all elasticities of substitution between capital, labor, energy, and materials to 0.5 and all those between fuels to one; setting all elasticities to 0.1; and setting all elasticities to 2.
Not surprisingly, as we reduce the elasticities, the cost of abating a tonne of carbon increases and vice versa. What is surprising, is the extent to which the BAU emissions path is changed. BAU emissions are reduced in the less flexible economies relative to emissions in the default model. This effect is so strong that usually the total cost of reducing emissions increases with increasing flexibility and vice versa. In fact, in our most extreme low flexibility scenario, emissions grow so slowly that the more moderate policy scenarios are not binding. Economic growth is in fact halted and so there is a much reduced climate problem to deal with. This seems to be an example of the de La Grandville hypothesis that, the greater the elasticity of substitution, the faster the rate of economic growth.
Yingying and I debated whether the growth effect is real or an artefact of our modelling. Jorgenson et al.'s study avoided the issue by looking at policy scenarios that are based on percentage reductions in emissions relative to business as usual. Babonneau et al. adjust the rates of technical change so that the BAU scenario reproduces the expected rate of economic growth in the European Commission's World Energy Technology Outlook. We believe that this is likely to be a real effect. On the other hand, G-Cubed assumes that the rate of technological is exogenous, whereas the rate would also likely vary with the elasticities of substitution. Additionally, the baseline levels of output and prices at the start of our simulation are based on the real world level of these variables which would also differ if the economy was very different. Therefore, our results are not a reliable indication of the relative performance of more and less flexible economies in the real world
So, what is the bottom line?
1. Because a less flexible economy has higher abatement costs per tonne of carbon but less emissions growth, if what we care about is the total costs of climate policy then it is not so important to get good estimates of elasticities of substitution. If we care about average and marginal costs of abatement, then these parameters are critical. We again find that the distinction between marginal and total costs of abatement is important.
2. Though stopping growth reduces the climate change problem, the reverse isn't true. We cannot find a model economy where the costs of climate mitigation are so high that such a policy would result in stopping economic growth or that mitigation cannot be achieved without stopping growth. Certainly, assuming that the economy is a lot less flexible than it is cannot generate high total costs. In fact the reverse is true.
No comments:
Post a Comment