Carbon taxes delay CO2 emissions, but they don’t reduce them

The intellectual history of the carbon tax is a bit of a tragedy.

I always like to say that economists knew the solution to climate change before climate change existed. Arthur C. Pigou laid out the theory of externalities in 1920, long before global warming became a concern. And Pigou also provided the solution: if there is a negative externality, tax the thing that causes the externality. A carbon tax is the straightforward application of this idea.

It took quite a long time for climate change to become a number one concern in global politics and it took even longer for the carbon tax to become a viable idea in policy circles. There still is a lot of misunderstanding about how such a tax would work and it continues to face a lot of resistance. There is, of course, also disagreement about the appropriate tax rate and what to do with the revenue.

And yet, as many people still fail to realize, most countries already have an implicit carbon tax. Almost all developed countries already tax gasoline. The average gasoline tax in the OECD translates to an implicit carbon tax of 85 euros per ton of CO2 – which is in the ballpark of reasonable estimates of the social cost of carbon. That implies, at least to a first approximation, that the average user of a combustion engine car in the OECD already pays the full social cost of burning gas.

Of course, the problem with the existing implicit carbon taxes is that they are not uniform. The taxes on gas and diesel are much higher than those on coal, natural gas, and kerosine. There is a lot of variation in tax rates between countries. So the existing carbon taxe regime leaves a lot of room for improvement. But enacting an optimal carbon tax wouldn’t require us to change our tax system fundamentally – we would just need to tweak the existing fossil fuel taxes.

And let’s not forget the EU cap-and-trade system which imposes another implicit carbon tax on a broad set of emitters, from electricity generation to steel factories, refineries, and other energy-intensive industries to commercial aviation. The current price of carbon in the certificates market is 51 euros per ton – again, not far from mainstream estimates of the true externality.

Now here comes the tragedy – at the exact time that a carbon tax has become a mainstream idea, economists have started to realize a major flaw with it: A carbon tax may not actually reduce carbon emissions over the long run, but merely change the time-path of emissions.

Readers of this blog have long understood this. Hans-Werner Sinn has understood this even longer. Recently, I have seen indications that more and more economists are becoming aware of it.

Here is a graph from a recent paper by Jose Luis Cruz Alvares and Esteban Rossi-Hansberg:

Source: Cruz Alvares and Rossi-Hansberg (2021)

Under business as usual (represented by the light-green curves above), this model predicts that CO2 emissions would peak around the year 2100 and then start to decline, while temperatures would rise by about 7 degrees relative to pre-industrial levels. With a carbon tax of 50% (brownish curve), the peak of emissions would be shifted in time by a couple of decades and temperatures would rise somewhat less rapidly, but they would end up with the same cumulative emissions and the same long-run temperature increase. A higher tax rate (blue and dark green curves) would simply shift the time path of emissions and temperatures further into the future.

So these model projections show that a carbon tax merely stretch out the cumulative CO2 emissions over time without any long-run impact on the climate. No matter what tax rate is applied, in the long-run we’re going to end up with a rise of about 7 degrees.

This is a devastating result for carbon taxes. (Too devastating for the authors to put in the abstract, where they try to down-play this result as follows: “Carbon taxes delay consumption of fossil fuels and help flatten the temperature curve but are much more effective when an abatement technology is forthcoming.” I think “help flatten the temperature curve” is a little misleading , since it is really just a postponing of the temperature increase.)

The Cruz Alvarez and Rossi-Hansberg paper is notable for a number of other results. For instance, it has bad news on clean energy subsidies:

Clean energy subsidies have only a modest effect on carbon emissions and the corresponding evolution of global temperature since, although they generate substitution towards clean energy, they also lead to a reduction in the price of energy which results in more production and ultimately more energy use. These effects tend to cancel each other out.

Cruz Alvares and Rossi-Hansberg (2021), p 4

And it is the first paper I have seen (though I do not claim to have read every relevant paper) that tries to estimate the welfare cost of climate change taking into account that people will change their behavior when temperatures rise (including, if necessary, leaving their current home and migrate to better climates).

This is their summary of the key findings regarding welfare:

With the quantified model in hand, we can simulate the economy forward over several centuries and evaluate the economic consequences of global warming. This phenomenon is expected to have heteroge- neous effects over space, where the hottest regions in South America, Africa, India and Australia experience welfare losses of 15% and the coldest regions in Alaska, Northern Canada, and Siberia undergo welfare gains as high as 14%. On average, the world is expected to lose 6% in terms of welfare, although the exact number depends on the yearly discount factor.

Cruz Alvares and Rossi-Hansberg (2021), p 4

I have some trouble converting this finding into something tangible. Is a 6% drop in welfare a big deal or not? Imagine a politician saying “Vote for me, and I will make you 6% happier?” Would you vote for the guy? I don’t know. Maybe. I guess the question is: compared to what? Footnote 36 of the paper provides a clue:

Global average welfare losses are calculated as the population weighted average of the relative present discounted value of utility in the baseline case relative to the counterfactual without global warming.

Cruz Alvares and Rossi-Hansberg (2021), p 28

So this is one way to think about it. Taking the authors’ chosen discount factor of 0.964, a 6% reduction in the present value of utility is equivalent to an annual reduction of 0.21% for every year into eternity. For small changes, the change of utility is closely approximated by changes in income. Therefore, a 6% loss in welfare translates into a 0.21% loss in annual income, now and forever. Given that the average income of a world citizen in 2019 (pre-covid) was about 18,000 US dollars (purchasing power adjusted), the annual cost of climate change for the average world citizen is about 40 dollars.

40 bucks isn’t a lot. But there is, of course, a lot of uncertainty and a lot of variation across countries. Helpfully, the authors include this graph showing the regional variation of welfare losses (less helpfully, in the graph, red means welfare gains, blue welfare losses):

Cruz Alvares and Rossi-Hansberg (2021), p 29

In conclusion, I see signs that the academic consensus is slowly shifting away from the carbon tax precisely at a time, when that idea is getting a lot of traction from policy makers. I continue to believe, and I am emboldened by the research discussed above, that the solution to the problem of climate change is to curb the extraction of fossil fuels directly.