Benefit-cost analysis for the ETS
Last month I gave a presentation at the 4th International Conference on Climate Change in Chicago. This is an overview of what I said.
Over the last few years there has been a vigorous debate about climate change and the Australian government’s proposed climate policy — an emissions trading system (ETS). A handful of studies have been done to assess the impact of an ETS, the most famous being the Garnaut Report and the Treasury modeling. Surprisingly, none of these studies have included the results of a benefit-cost analysis.
Good policy requires more than good intentions and spending lots of money. Critically, good policy needs to have more benefits than costs. This may seem like a simple and self-evident statement, but it is one that is often ignored in politics. Before we embrace an ETS as a way to “do something” we need to first carefully assess the consequences of the policy and check whether the benefits exceed the costs.
To do a benefit-cost analysis on the ETS it is necessary to estimate several key variables. First, we need to know the costs of anthropogenic global warming (AGW). Second, we need to know how much of this cost can be prevented by an ETS. Third, we need to know the costs of the ETS. And finally, we need to determine a discount rate to use, to bring all of the benefits and costs back into present value so we can make a proper comparison. Once we have all of these variables, we divide the benefits by the costs. A number above one is a theoretical “pass” while below one is a “fail”. (Though normally we would want a number significantly above one before going ahead with a policy.)
** The costs of AGW
Despite thousands of peer-reviewed papers having been written about climate science, surprisingly there has been only 14 complete peer-reviewed studies of the costs of climate change. Of these, only 10 were based of the IPCC mid-point projection for future warming (3 degrees warming by 2100). While the results are reported in terms of money, the studies attempted to include non-financial costs such as deaths and environmental changes.
The average conclusion of the peer-reviewed studies was a cost of 0.9% of GDP, with the highest individual estimate being 2.5% of GDP by 2100. (Some of the studies suggested a zero cost, or even a small benefit.)
There have been some high-profile non-peer reviewed estimates of AGW costs, the most famous of which was the Stern Review done in the United Kingdom. Using a range of more pessimistic assumptions (which have been widely criticised elsewhere in the literature), the Stern Review estimated a cost of AGW at about 2.9% of GDP by 2100.
Of course, if we have more warming, then the costs will be higher. In one of the peer-reviewed studies, Nordhaus looks at what will happen with more warming, and finds that the costs could be as high as 5% of GDP by 2100.
Given this context, it is surprising that the Garnaut Report found a cost of about 10% of GNP… some three times higher than the Stern Review. (The costs were 8% of GDP, but GNP is a better measure for human welfare.)
Years ago Ross Garnaut commented on some economic modeling that it “didn’t pass the laugh test” because the conclusions were so far from expectations. It is tempting to say that the same may apply in this instance. But before dismissing the Garnaut numbers, it is informative to look at how they were made up. The first thing to notice is that Garnaut did not use the mid-point IPCC projection of 3 degrees of warming, but instead decided to double-guess the science and factor in an expected warming of 5.1 degrees by 2100. From this point, the 10% cost is made up of four main elements:
1. Garnaut assumed that Australia must continue to produce the same amount of food that we currently produce. One consequence of warming was a less productive agricultural sector, and so by assuming no change in output, Garnaut assumed that we would take resources out of efficient areas of the economy and put them into inefficient areas of the economy. The consequence of this was a cost of about 1.5% of GDP. A more realistic assumption is that resources will leave inefficient sectors and go to efficient sectors. If this assumption was used, the cost estimate would be much lower.
2. The largest cost was infrastructure. It makes sense that this should be a major cost of AGW, as a changing environment may speed up depreciation of some infrastructure and require updated and modified versions to be built. Garnaut gave the examples of desalination plants, damaged roads, and new and larger ports. However, the cost estimate of 4% of GDP is hard to believe. To put this in context, infrastructure (including ports, rail, roads, water, electricity etc) currently takes up about 4.4% of GDP. For Garnaut to be correct, infrastructure productivity would need to more than half. Given that we have another 90 years to adjust, that seems like a significant over-estimate. It should be noted that the Garnaut infrastructure number did not come from direct estimates of required infrastructure spending, but instead were dependent on industry estimates for a range of productivity shocks, which were then fed into a general equilibrium model.
3. Another cost identified by Garnaut was a change in the value of our trade, creating a cost of about 3% of GDP (the same as the entire Stern Review). This is ironic because a major element of this cost was that Australia would not be able to sell as much coal on the world market. If the other costs of AGW are lower than expected, then this cost will also be lower than expected.
4. The final cost was for left-over issues, such as potentially higher defence spending and potentially lower tourism due to the loss of the Great Barrier Reef. There was no methodology used for these costs, but simply a guestimate of perhaps 1-2% of GDP (higher than the average peer-reviewed estimate for the total costs).
(The Garnaut review also looked into the costs of more severe cyclones and more heat-related deaths, but the costs from these were relatively low.)
If we were to make some rough adjustments to reflect more reasonable assumptions, then the infrastructure costs might be 2-3% of GDP, trading costs might be 1-2% and other costs may be another 1%… giving us a total cost of about 5% of GDP. This is about the same as Nordhaus found when he used a higher warming scenario. It might be reasonable then to conclude a range of costs from 0-5% of GDP (noting that peer-reviewed estimates are about 1% of GDP). But for the sake of the current benefit-cost analysis, we will extend the range to include the pessimistic Garnaut estimate of 10%.
** Benefits of mitigation
The benefits of mitigation are the AGW costs prevented. There are two reasons why this may be less than the full costs of AGW.
First, it will not be possible to prevent all AGW and so some of the costs will be felt irrespective of the policy. Bjorn Lomborg explains that if every country met their Kyoto obligations up to 2012, that would simply delay global warming by one week. More recently, Pat Michaels and colleagues have shown that if the entire world signed up to the American “cap and trade” legislation it would decrease world temperatures in 2100 by about 0.2 degrees. Clearly, the benefits of mitigation are only going to be some percentage of the potential costs of AGW.
On this issue, Garnaut is very optimistic. He assumes that with a binding and coordinated world agreement it may be possible to prevent some 80% of the costs of AGW.
The second concern when looking at the benefits of policy is that Australia does not get to set world policy. Our emissions make up about 1.4% of global emissions and that number will continue to decrease into the future, so the benefit of Australia acting alone would be effectively zero. As Garnaut notes, this means “the efficient amount of mitigation is easy to work out — it would be zero”.
However, Garnaut also points out that Australia does have some influence on the world stage. If Australia is able to go forward with an effective policy, then other countries may copy our approach, making it more likely for us to achieve a binding and comprehensive global treaty. While this is possible, it is not guaranteed. Therefore, we need to estimate the likelihood that Australia will change the decisions made by other countries around the world — primarily China, America and India. While our true influence is probably between zero and 1%, we could charitably suggest that Australia’s actions have a 10% influence on the rest of the world.
(Note that our pre-emptive action may actually create a negative impact on global negotiations. It may make more sense for us to have the bargaining position that we will act only if other countries act. This will then increase the incentive on other countries to act.)
A generous range of assumptions would be that Australia’s actions have a 1-10% impact on the rest of the world, and a binding and comprehensive treaty will prevent 50-80% of the costs of AGW.
** Costs of an ETS
On the other side of the ledger are the costs of an ETS. Before Garnaut there were a range of Australian estimates for ETS costs (including ABARE, Concept & Allens consulting). The mid-point of these studies was a cost of about 6% of GDP. (It is worth noting the different time frame being used — ABARE & Allens use 2050, Concept used 2030, while the above AGW costs were for 2100.)
Regarding the costs of an ETS, the Garnaut report is relatively optimistic, suggesting costs of about 4.5% of GDP by 2050. While lower than the previous average, this was well within the ranges of the previous studies.
The most comprehensive study of the costs of an ETS was done by the Commonwealth Treasury, which estimated a cost of 5.1 to 6.7% of GDP by 2050. This study has been criticised by a range of economists for assuming a world agreement (which would lower the costs), for assuming the removal of other costly policies, for assuming no distortions in the compensation payments, for assuming that carbon capture & storage becomes common after 2020, and for failing to fully estimate the long-run inflation and employment consequences. While there is some validity to these complaints, modeling is always an imperfect process and the Treasury numbers are probably a relatively good estimate of the potential costs we could face from the proposed ETS.
It should be noted that the proposed ETS does not meet the standards required by the Garnaut review. Instead, it could be seen as a first step. Consequently, the costs of a full binding and comprehensive global agreement (as outlined by Garnaut) may be higher than the Treasury estimate.
For the sake of this analysis we will use the most optimistic Garnaut number and assume a cost of only 4.5% of GDP by 2050.
** Discount rate
To compare benefits and costs it is necessary to put all of the estimates into a common value. This is done by discounting future benefits & costs by a discount rate so that we can compare “present value” benefits with “present value” costs.
The fact that people have a time preference is clear from our everyday behaviour. We prefer $100 now compared to $100 next year for a range of reasons. The first reason is that we could invest that $100 and we would have $105 next year. A second reason is that if we have the money now we have more freedom as to when we can spend it (immediately, next month or next year). A third reason people have a “pure” time preference, where we put a slightly higher value on the immediate compared to the future. The truth of our time preference is revealed in our spending and savings decisions. If there was no time preference, then people would save nearly all of their income for nearly all of their life. Instead, most people spend a large percentage of their income when they get it.
A standard discount rate for many benefit-cost analysis is 5%, and perhaps even 10%. However, Garnaut (following Stern) wants to use a much lower discount rate so that we put a much higher value on future events. Garnaut suggests about 2% (he uses 1.35% & 2.65%).
Their reason for preferring a low-discount rate is that they want to factor in the utility of future generations. What this rationale fails to recognise is that our current behaviour already does factor in the welfare of future generations. Parents already arrange their affairs so as to ensure the welfare of their children and grandchildren. Stern and Garnaut might be insisting that parents currently don’t care the “right amount” about their children, though that seems impossible to determine. One simple check for intergenerational equity would be to ensure that future generations are at least we well-off as we are today. As that will be achieved under all modeled scenarios, there is no reason for believing there is an intergenerational problem that requires an artificially low discount rate.
For the sake of this analysis we can look at both a 2% and a 5% scenario.
** The results
No matter which set of assumptions you use, an ETS does not pass a benefit-cost analysis, and is therefore bad public policy.
The most optimistic set of assumptions for an ETS include (1) AGW costs 10% of GDP; (2) Australia influences the world 10% and we can stop 80% of the costs; (3) ETS costs of 4.5% of GDP; and (4) discount rate of 2%. Using these assumptions (and an estimated real growth rate of 2.1%, as used by Garnaut) the benefit/cost number is 0.08.
A more realistic set of assumptions are (1) AGW costs of 0.9% of GST (peer-reviewed); (2) Australia influences the world 1% and we can stop 50% of the costs; (3) ETS costs of 4.5% of GDP (note, this is the same assumption); and (4) a discount rate of 5%. Using these assumptions, the benefit/cost number is 0.00.
Plotting in any range of numbers as outlined in the discussion above gets a result somewhere between 0.00 and 0.08, which is a massive failure. There is literally no way to make this policy pass a reasonable benefit-cost analysis.
One reasonable response to this is to say that we should not act pre-emptively, but that if the world acts then Australia should act. It is possible to test this hypothesis too, by simply assuming a 100% guarantee of a binding and comprehensive world agreement and re-running the numbers.
Under this (super optimistic) assumption of a world agreement, mixed will all of the most optimistic assumptions from before, the benefit/cost number rises to 0.76, which is still well short of a pass. If we drop the discount rate down to their lowest number of 1.35%, the benefit/cost number rises to 0.83, which is still a fail. Once again, there is no way with reasonable assumptions to make this policy pass.
If we assume a world agreement, and put in the above “reasonable” set of other assumptions, the benefit/cost number is 0.03. Plotting in any range of numbers as outlined in the discussion above gets a result somewhere between 0.03 and 0.8, which are all failures.
It seems that the only way to make the ETS look like good policy is to refuse to do a benefit-cost analysis. Or at least, refuse to publish the results. Hidden in the Garnaut review is the comment that an ETS will result in a “loss of present value”, but this is not considered important enough to put in the summary or conclusion, and nobody in the media picked it up.
The first conclusion from all of this is that Australia should not have an ETS. Thankfully, we have managed to avoid the introduction of ETS legislation for the moment. However, the people who support such policies are still determined to promote their policies, and so it is still necessary to subject these ideas to careful scrutiny and insist that we only introduce policies that pass a benefit-cost analysis. The fight against the ETS must continue.
A second conclusion is that it is unlikely that any large and expensive climate policy will pass a benefit-cost analysis. This means that we must shift our thinking either towards low-cost or no-regret climate policies, and away from grand schemes.
This isn’t the right place to discuss potential alternatives, but some that have been mentioned previously include removing perverse subsidies, a tax swap, providing targeted tax cuts, promoting civil society actions, pursuing policies with other tangible benefits, or supporting a private climate insurance market.
But whatever the policy suggestion, all honest people with good intentions should agree that we should only introduce policy if that policy has more benefits than costs.