Possible Insight

In my previous two posts (1, 2), I showed how climate models forecast too much warming and that we should actually expect roughly 1.5 to 2 deg C of warming from 1900 to 2100 due to CO2 emissions (0.6 to 1.1 deg C more than today).

Now I’ll turn to what we should do about it. From what I’ve researched, this warming could still result in enough harm to justify action. I admit that it’s somewhat of a judgement call, depending on your views about the competence of government programs. But we can quickly adjust my preferred policy if we encounter greater than expected warming. This adaptability pushes me over the edge towards taking action.

However, in terms of social fairness, any policy response must balance competing goals. On the one hand, burning fossil fuels emits CO2, which causes warming, which has harmful effects. On the other hand, burning fossil fuels produces energy cheaply, which allows us to produce food, shelter, and medicine cheaply, which has beneficial effects. If other energy sources were as cost effective in all cases, we would already be using them exclusively. So what we want is for people to use fossil fuels only in cases where the benefits exceed the harms relative to other fuels. In my opinion, any other goal most likely suffers from either wishful thinking or unreasonable ideology.

This balancing act has been studied by economists for about a hundred years. Economic consequences to third parties are known as “externalities”. (For a high level overview, I recommend this article, this one, and this video) . When bargaining costs are high, the standard recommendation is a “Pigouvian Tax”, named for Arthur Pigou, who introduced the concept in his 1920 The Economics of Welfare.

Pigou argued, and economists have since demonstrated with microeconomic models and some real world successes, that the optimal way to address such an externality is with a tax (or subsidy) set to the social cost (or benefit). This 2013 paper from economists at MIT and the Federal Trade Commission reviews some of empirical results for taxes on transportation fuels. For emissions fundamental to the combustion process like CO2, a tax is reasonably effective at reducing those emissions. This 2018 paper from economists at the U of IL and Harvard shows that electricity demand does indeed decrease in response to price increases in the range we’re talking about.

What is the social cost of emitting CO2? In 2016, the Obama administration released an interagency report estimating the “social cost of carbon” in 2020 at $62 per metric ton in 2007 dollars given a 2.5% social discount rate (the lowest discount rate tested, which results in the highest social cost). 2007 dollars are forecast to be worth $1.26 2020 dollars. So that would be $78 per MTon in nominal dollars for 2020. This translates into 69 cents per gallon for gasoline, 3.1 cents per kwh for electricity from natural gas, and 7.8 cents per kwh for electricity from coal.

That report used climate sensitivities my last post showed are too high. Dayaratna et al 2017 rerun the models with an empirical probability distribution on sensitivity that ranges from 1.25 to 2.45 deg C. The highest estimated social cost of carbon in 2020 was $29 per MTon in 2007 dollars at a 2.5% social discount rate. That would be $36 per MTon in 2020 dollars. This translates into 32 cents per gallon for gasoline, 1.4 cents per kwh for electricity from natural gas, and 2.8 cents per kwh for electricity from coal.

(Average gas prices in the US for Aug 2019 were $2.59 to $3.41 depending on grade and formulation. Average electricity prices in the US for Jun 2019 ranged from 9.9 cents/kwh in Washington to 32 cents/kwh in Hawaii with an average of 13 cents/kwh.)

Importantly, to work as advertised and deliver the greatest warming reduction at the lowest cost, this tax (with some refinements I will discuss) must replace all other policies targeted at warming. Otherwise, we waste resources.

Moreover, even if someone didn’t like the tax concept, we know the following policies are not cost effective under any circumstances: dedicated battery electric vehicle subsidies, biodiesel, energy efficiency programs targeted at China, most renewable portfolio standards, most low carbon fuel standards, most solar PV subsidies, and some types of wind subsidies.

The rest of this post discusses why a Pigouvian tax is superior to command-and-control solutions, describes the optimal policy, and examines some important subtleties. As usual, I end with some final thoughts.

Why Taxes Are More Effective

To demonstrate the benefits of a carbon tax, let’s start by analyzing the cost-benefit of a popular command-and-control measure: subsidies for electric cars. Consider a Tesla Model S.

• The best Model S gets 32 kwh/100 miles. That’s .32 kwh/mile.
• On average, generating a kwh of electricity in the US produces .459 kg of CO2.
• At an SCC of $36/Mton in 2020, that yields a social cost of $0.0053 per mile.
• Compare this to the average gasoline car that gets 24.7 mpg. That’s .040 gallons per mile.
• A gallon of gasoline produces 8.9 kg of CO2 when burned.
• At an SCC of $36/Mton of CO2, that’s yields a social cost of $.013 per mile.
• The difference is .0077 per mile.
• Consumer Reports seems to think 200K miles is a very good outcome for vehicle life.
• That means a Model S could save roughly $1,540 in SCC if you drove it into the ground.

Several obvious problems:

• The actual Federal Tax Credit for electric cars is $7500, though Tesla hit their limit. So we were paying Model S owners almost $6000 more than they deserved—money we could have spent on more efficient emissions reduction policies or some other worthy cause. Dividing the the subsidy by the amount of CO2 emissions saved yields an “abatement cost” of $175/MTon–about 3X the Obama administration’s estimate of the social benefit!
• The same subsidy applies to every electric car, regardless of efficiency. The best Tesla Model 3 uses only 26 kwh/100 miles, justifying a subsidy of roughly $1740. We should not be incentivizing the Model S and Model 3 the same amount.
• The same subsidy applies regardless of the ultimate charging fuel. The US Energy Information Administration calculates the regional breakdown of CO2 emitted due to electric generation. In Washington, where there is a lot of hydroelectric, the emission rate is only .095 kg CO2/kwh. In Texas, it’s .530, more than 5X as much! It’s incredibly wasteful to incentivize the purchase of electric cars in Texas if there are people in Washington who want them. And if the car’s owner has his own solar panels, he’s potentially saving the full $2600 social cost of driving a gasoline car (ignoring separate rooftop solar subsidies).
• These subsidies completely ignore alternative arrangements that avoid driving altogether. What if someone faces the choice between taking a job in a big city 50 miles away versus one in the local town that pays $5000/year less? An electric car subsidy does nothing to compensate for the environmental benefit of taking the job close to home. Or what if someone can arrange to bicycle to work by buying a nice commuter bike and associated gear? Where’s the incentive for that?

There are similar problems with other popular command-and-control measures. For example, a rooftop solar installation in Southern California will produce about 25% more electricity than the same one in New England, but the federal subsidy program pays the same for both. This is another incredible waste. We should not be subsidizing rooftop solar in New England until all good opportunities in the Sun Belt are exhausted.

The basic idea behind an electric car or rooftop solar subsidy is to make those options more attractive because of their environmental benefit. A reasonable idea in theory. However, because it’s impossible for the government to maintain a continuously up to date list of all the potential CO2-avoiding measures with dollar estimates of their effectiveness in every different situation, this approach wastes resources. It also makes a very attractive target for special interest lobbying.

A carbon tax achieves the exact same goal far more efficiently. By increasing the price of every activity in proportion to the harm it’s CO2 emissions cause, the relative price of every alternative activity automatically adjusts to the correct level. If a particular electric car reduces CO2 emissions per mile a certain amount in a particular location, the tax ensures the relative price of driving a mile in an electric vs gasoline car accurately reflects that advantage. But it also ensures that avoiding a mile of driving altogether saves the correct amount of money.

What a “Carbon Tax” Looks Like

So if a carbon tax is better, what would it actually look like?

First, it wouldn’t be just a carbon tax. It would be a climate change tax and subsidy regime Why? Because CO2 doesn’t directly cause the harm we’re concerned about. Warming causes the harm. So we have to generally deter activities that produce warming and encourage activities that produce cooling. This targeting would have to include at least the following components:

• Tax on portable fuels like gasoline, diesel, jet fuel, LNG, and propane. The tax would be based on the CO2 emissions from burning a unit of fuel.
• Tax on heating/cooking fuels like natural gas and heating oil. The tax would be based on the CO2 emissions from burning a unit of fuel.
• Tax on electricity generated from fossil fuels or a tax on utility-scale fossil fuels themselves. The tax would be based on the CO2 emissions from either producing a unit of electricity or burning a unit of fuel.
• Subsidy for reducing CO2 in the atmosphere through such activities as planting trees, using CO2 scrubbers, or deploying carbon capture and sequestration. The subsidy would be based on how much CO2 is removed per unit of activity.
• Tax on other warming activities such as the emission of methane. The EPA lists greenhouse gases and their relative contributions here. In addition to covering all these gases, a tax might also be necessary on any large scale activity that produced warming through other channels such as urbanization, which causes the absorption of more sunlight by roads and buildings. The tax would be based on how much warming is caused by a unit of the activity.
• Subsidy for reversing the other warming activities such as capturing and sequestering methane or increasing the reflectivity of urban landscapes. The subsidy would be based on the amount of warming that would be avoided by a unit of the activity.
• Subsidy for “geoengineering” measures that directly produce cooling. The subsidy would be based on the amount of cooling produced by a unit of the activity, but netting out any other adverse environmental impacts the technology causes.

Note that this regime would require some sort of mechanism for measuring the warming and cooling contribution of each activity, then determining and administering the corresponding tax or subsidy. New technologies and approaches would create a constantly changing landscape.

This regime would also require some sort of mechanism for adjusting taxes and subsidies at periodic intervals. New information, technologies, and behaviors will alter the forecast of warming damage, so the system must respond appropriately. For example, we might learn that the climate sensitivity to CO2 or methane is higher or lower than expected. Or the forecast amount of warming damage could be higher or lower based on new demographics or technologies. Very cheap nuclear or very cheap geoengineering could reduce either the expected future demand for fossil fuels or the cost of counteracting warming to very low levels. Even more subtly, as people’s behaviors adapts to new incentives and new technologies, interactions among all these effects could alter the forecast. So we would have to adjust the regime accordingly.

Finally, this regime only works if it replaces all other climate change policies. Because it adjusts the relative prices of all climate-change-related activities to their appropriate levels, other interventions don’t do any additional good. In fact, they waste resources and end up being pure transfers to whatever special-interest industry produces compliance with the policy–at everyone else’s expense. This requirement even applies to emotionally appealing regulations like green building standards and mass transit subsidies, to the extent they are motivated by climate change concerns.

Backup Position: Abatement Costs

Now, some people may not like the climate change tax and subsidy regime. It may seem too complicated. It may not be “visible” enough. I personally don’t feel these are valid concerns. Any effective policy is likely to be complicated because our society and the climate are both complicated. And “visibility” doesn’t seem like a feature we should force the public to pay for. However, even if we accept these objections, we still have to respect the reality of abatement costs.

As I showed for a Tesla Model S’s electric vehicle subsidy, we can calculate the “abatement cost” of a policy–how much it costs to either prevent or eliminate a metric ton of CO2 in the atmosphere. For non-carbon heating or cooling activities, we can calculate the cost or benefit in terms of the equivalent CO2 effect (we could also convert all measures into a warming-based metric).

• If the abatement cost is more than the worst case estimate of the social cost of carbon, we should not pursue that policy under any circumstances. It does more harm than good under all reasonable assumptions. One might even say it is “unscientific”.
• If the abatement cost is between the best estimate and the worst case estimate, we should only pursue the policy after all better options are exhausted and we have done a more thorough risk analysis.
• For all policies with abatement costs lower than the best estimate, we should pursue them in order of lowest to highest abatement cost, recalculating as we observe the effects of the lower cost policies.

The worst case (95th percentile) social cost estimate from the aforementioned Obama Administration report was $123 per MTon. Using a more accurate, consensus based estimate of the climate sensitivity from Dayaratna et al 2017, the worst case is $76. Recall that the best estimate from the Obama Administration at a 2.5% discount rate was $62 and the best estimate from Dayaratna et al 2017 was $29. All these are 2007 dollars.

This paper from Harvard and a Yale economists estimates the abatement cost for a wide variety of measures (in 2017 dollars, equivalent to $1.18 2007 dollars) . This recent working paper from the University of Chicago calculates abatement cost for renewable energy mandates (in 2018 dollars, equivalent to $1.21 2007 dollars). This US Forest service study reviews the literature on afforestation (planting trees to capture carbon) and calculates the supply curve (in 1997 dollars, equivalent to $0.74 2007 dollars). From these sources, we can categorize various policy measures:

1. Definitely justified with the consensus ECS. Behavioral energy efficiency. Some methane flaring regulation. The aspects of the Obama Administration’s Clean Power Plan aimed at increasing fossil fuel plant efficiency and substituting natural gas generation for coal generation. Certain types of wind subsidies. Afforestation up to 125 million tons of carbon per year.
2. Potentially justified, depending on ECS assumptions, risk tolerance, and program specifics. Reducing federal coal leasing. National clean energy standards. Gasoline tax. CAFE standards. Agricultural emissions policies. Soil management. Livestock management More methane flaring regulation. More types of wind subsidies. Afforestation between 125 and 345 million tons of carbon per year.
3. Not scientifically justified with the consensus ECS. Concentrated solar power programs targeted at China and India. Some types of wind subsidies. All renewable portfolio standards. All low carbon fuel standards. All solar PV subsidies. Afforestation beyond 345 million tons of carbon per year.
4. Not scientifically justified, even with an ECS that is too high. Dedicated battery electric vehicle subsidies. Biodiesel. Energy efficiency programs targeted at China. Most renewable portfolio standards. Most low carbon fuel standards. Most solar PV subsidies. The remaining types of wind subsidies. Afforestation beyond 480 million tons of carbon per year.

If we want to be both scientific and fair, we should stop pursuing all climate change policies in categories (3) and (4). Furthermore, we should unwind any policies in category (4).

Important Economic Subtleties

In addition to the pure externality calculations, there are several other economic considerations worth noting about a climate change tax and subsidy regime

• Revenue Neutrality. Climate change incentives are not an excuse to raise more (or less) government revenue. The proper amount of government spending is an independent issue. To avoid getting climate change policy bogged down in that debate, all sides should accept revenue neutrality as a key element. Any net revenue raised by this policy should be offset by reductions elsewhere, ideally reductions in the most economically inefficient taxes. The nice thing about taxes on and subsidies for externalities is they theoretically improve economic efficiency. Similarly, if we somehow ended up with net subsidies, we should increase the most economically efficient taxes to pay for them.
• Distributional Effects. Because energy comprises a larger share of poor people’s budgets, increases in energy costs disproportionately affect them. Note that this issue applies to other climate change policies as well. Renewable portfolio standards also increase retail energy prices. And the poor are less able to afford the investment necessary to take advantage of solar PV and electric vehicle subsidies. However, by standard economic reasoning, just because climate change policies make the poor relatively worse off does not mean that climate change policies should be used to offset this effect. The standard economic reasoning on the efficiency of assistance to the poor still applies. Cash transfers and refundable tax credits are the most effective programs. Trying to offset an inefficient climate policy like renewable portfolio standards with an inefficient assistance policy like a green jobs program is a vain attempt at using two wrongs to make a right.
• Political Feasibility. Some might argue that the tax and subsidy regime could never get legislated. First, how’s climate change legislation going so far? The only policies we seem to get passed are the ones that benefit special interests and actually do more harm than good. Second, you have to offer something to get something. I imagine that offering to dismantle all the other climate change policies and reduce certain inefficient taxes might generate some support. But it has to be a sincere offer. At the end of the day, if people think other policies are more politically feasible, let’s quantify that. If the gain in political feasibility is more than the loss of efficiency, that’s a reasonable argument.

Final Thoughts

• The goal of climate change policy is to balance the environmental harms of warming caused by fossil fuels against their other benefits.
• If you value “consensus”, a Pigouvian warming tax and subsidy regime is the closest policy you’ll find to a consensus choice among economists that study externalities.
• If you value science, many popular climate change policies do not meet a scientific test of cost effectiveness even using analyses of worst-case scenarios from the Obama administration. These include dedicated battery electric vehicle subsidies, biodiesel, certain types of wind subsidies, most renewable portfolio standards, most low carbon fuel standards, and most solar PV subsidies.
• Any climate change policy must adapt to new information and conditions.
• There is room for lots of debate about the implementation details of a climate change tax and subsidy regime, as well as the cost effectiveness of many specific measures.