Government vastly underestimates potential health and environment damages from climate change

According to a new (peer reviewed[i]) analysis , the economic, social, and environmental damages to our children and grandchildren from climate change could be far larger than the value currently used by the US government to evaluate climate policy. Discussed more below, now that the Environmental Protection Agency (EPA) must set carbon pollution standards for cars, power plants, and other sources, the value placed on climate damages will be crucial in any evaluation of the benefits and costs of proposed standards.

The authors find that if some of the more dangerous projections from recent climate research come true, damages could exceed the official figure by a factor of 42, or $21 per ton of CO2 emitted versus $893. The difference between the two is equivalent to $0.21 per gallon of gasoline versus $9.00.

The authors were conservative. The $893/ton estimate may still understate potential damages from climate change: it was not the maximum value among the authors’ estimates. It was the “95th percentile” value, which means that 5% of the estimates under dangerous climate change were higher, but not reported. Equally scary, many of the most important damages were not taken into account because they cannot be monetized. To name just a few: damages resulting from socio-political conflicts over land and water resources, from socio-political conflicts from potentially millions of climate refugees, from the loss of all coral reefs, from mass species extinction, or from pain and suffering over loved ones who die or get sick (e.g. from extreme weather events, or disease and illnesses such as malaria, westnile virus, and asthma). The easier things to measure comprised things like changes in heating and cooling costs, and property losses due to sea level rise.

Also not factored into the equation is the tremendous injustice to the poor associated with climate change: some of the world’s poorest regions (and populations within countries) will bear the worst damages (e.g. Africa), while having contributed the least to the problem. Ironically, since many of these areas did not share in the wealth generated from fossil fuel combustion, they have the least ability to deal with impacts. As discussed below, weighing damages by income can increase damage estimates by an order of magnitude (see just before Table 1 below).

Still, $893/ton of CO2 is orders of magnitude higher than estimated costs for even the most aggressive emissions targets that have been proposed, all of which would pass a cost-benefit test with flying colors. For example, the most recent proposed bill in the US Senate, the American Power Act, called for a 17% reduction in CO2 emissions below 2005 levels by 2020 and 83% by 2050. The EPA estimated the cost of this at roughly $21/ton; their worse-case scenario projected $46/ton.

These numbers matter, and will gain increasing traction in the coming years. As the Supreme Court has affirmed, the Clean Air Act requires EPA to regulate CO2 and other global warming pollution now that its scientists have determined they endanger public health. As part of the regulatory review process, the agency compares damages to mitigation costs; estimated damages from climate change will play a prominent role. Indeed, EPA has already used damage estimates in several rulemakings.

Scenarios examined

The authors re-calculated the government’s estimates using one of the three models used by EPA—the one that produced values in the middle of the other two, called DICE. They looked at four scenarios, changing a few key assumptions to reflect recent scientific research. Much of the data EPA used relied upon out of date knowledge, based largely upon the United Nation’s International Panel on Climate Change (IPCC) 2007 Assessment Report. Since then, scientists’ projections have become more ominous by the day, and impacts are happening much faster than originally predicted.

Accordingly, the scenarios examined used more pessimistic climate change assumptions:

• Scenario 1: The default scenario to which the others are compared. Uses EPA’s assumed damages for a given temperature increase, and EPA’s assumed rate of change in damages the higher the temperature increase.
• Scenario 2: Damages assumed to increase by a factor of four for a given temperature increase, corresponding to a 2010 publication by Michael Hanemann[ii] (University of California—Berkeley); no alteration to the rate-of-change assumption.
• Scenario 3: Damages assumed to be increasingly larger the higher the temperature increase, corresponding to a 2009 publication by Martin Weitzman[iii] (Harvard University). The default model assumes temperature increases have to reach 34 degrees F before damages reach only 50% of world GDP. The new “damage function” (the mathematical formula that converts temperature changes into monetized damages) lowers this threshold to 11 degrees, and for 22 degrees stipulates a 99% loss. Put another way, a 22 degree increase would mean that any output produced in the economy would be canceled out by climate damages. Importantly, this puts a cap on total impacts. It also assigns an equal value to damages incurred by poor areas as those to rich areas. This is because GDP reflects only losses that can be monetized (the market value of goods and services produced in the economy), [iv] and places no value on equity. As discussed above, in that sense the authors’ estimates were conservative.
• Scenario 4: Scenarios 2 and 3 combined.

Are these assumptions plausible? Yes, if you consider the world today after an increase of only 1.4 degrees F. Extreme weather events (e.g. record heat waves, floods, droughts, hurricanes) and numerous other environmental changes associated with climate change, are accelerating. Of equal if not more concern is that fact that many impacts have not yet been felt from the 1.4 degree increase, due to long lags in the system. One example is declining arctic summer sea ice cover, 30% of which has vanished over the last four decades. As it continues to melt, more of the earth’s surface will be dark, resulting in less heat being reflected back into space. Or consider melting permafrost on the earth’s surface. As this progresses, carbon stored in the earth is being released, increasing greenhouse gases further.

Results

For each scenario analyzed, the authors presented estimates using two different “discount rates,” 1.5% and 3% (discussed further below under “Explanation of the discount rate”). They also presented the average and 95th percentiles (see third paragraph above for explanation). As you’ll see shortly, these specifications make a huge difference.

Table 1 below summarizes the results. I supplement them with another analysis (Hope and Johnson, 2010[v]), where we calculated the default scenario at the 99th percentiles using a 1% discount rate—these are highlighted in purple. Notably, government guidelines stipulate that discount rates of 1% to 3% are appropriate for costs and benefits spanning multiple generations. Given this, it is curious that the lowest rate used by EPA was 2.5%. In this respect, the authors were again being conservative, by stopping at 1.5%. (See “Explanation of the discount rate” section below for discussion of what discounting is and why a range is considered appropriate). In addition, the authors weigh damages to poor countries equally as to those to rich countries. Similar to other research in the literature, Hope and Johnson (2010) show that “equity weighting,” a technique that weighs damages more highly for poorer people, results in much higher damage estimates. For one of the models used by EPA (FUND), equity weighting increased damages by as much as 10 fold.

The report’s highest estimate, $893, is highlighted in yellow. For any point on the distribution in the EPA scenario, damages approximately double when going from a 1.5% to 1%, and more than double when going from the average to the 95th percentile. The average increase when going from the 95th percentile to the 99th percentile, for any given discount rate, is about 20%.

In Table 2, I apply these rates of increase to estimates in Scenarios 2, 3 and 4 (highlighted in blue), to get a crude idea of how much higher their estimates might have been at the 1% discount rate and 99th percentiles. This is of course very inexact, as the actual percent increases would differ somewhat if they were actually calculated, but still provides a reasonable approximation. With a discount rate of 1%, and measured at the 99th percentile, damages could exceed $2,000 per ton of CO2 emitted in Scenario 4. Again, this makes no adjustment to weigh damages to the poor more heavily, which can radically increase damages, and excludes non-monetizable damages.

But even if you stick with Scenario 1’s optimistic climate change assumptions, using the lower end of the government’s recommended discount rate for intergenerational discounting (1%) gives an average damage estimate of $229 per ton of CO2. The 95th and 99th percentiles give $527 and $643, respectively. At worse, EPA’s $21 central estimate for damages (the average over the three models used, using a 3% discount rate (not in Tables 1 and 2)) is equal to the estimated cost of reducing emissions.

Explanation of the “discount rate”

There is a range of opinions as to the correct discount rate to use, based upon the concept of returns on investment.

Taking energy efficiency as an example, suppose I am trying to decide whether to make a $1,000 investment to improve my house insulation. I might want to make a return (through reduced energy bills) at least as good as one I can get from investing in alternative assets. For example, assume other assets give an annual return of 3%. To give up $1000 now to improve insulation, I might require at least $30 in savings per year—what I could have earned elsewhere (I may require less, if I also get personal satisfaction from reducing pollution).

This framework works well for many types of investment decisions, but falls apart with climate change. One reason is that the decision to invest in emissions reduction is not as simple a trade-off as the one between energy efficiency savings and market returns. Unlike these two investments, climate mitigation does not have a positive expected return between the two alternatives. Nor are the returns even remotely predictable. Instead, climate mitigation is a highly uncertain investment, one outcome of which could be preventing enormously negative returns resulting from catastrophic, or even highly damaging, climate change. Another possible outcome is getting a slight net gain or loss relative to what you could have earned in the market. For instance, mitigation might prevent a 3% loss in GDP at an opportunity cost of a 2% market return (a slight net gain of 1%), or vice versa (a slight net loss of 1%).

(For those interested in the mechanics, here's the formula, see just below the conclusion).

If you are risk averse, you’d be thinking of investing in mitigation somewhat similarly to how a parent thinks when investing in life insurance. You’d be willing to pay a lot to insure against a worse case but low probability outcome—your death and its financial impact on your children. On the other hand, if you are risk neutral or love taking risks, you’d view mitigation as worthwhile if expected benefits were equal to or greater than expected costs. You’d be also be assuming monetized impacts are comparable to market returns, and that non-monetizable impacts (e.g. saved lives, avoided violent resource conflicts, preserved ecosystems) either don’t matter or are not that large. 

How does this relate to the discount rate? Insurance investments give negative expected returns—you only “make a profit” in the event of a disaster. For most people, that disaster doesn’t occur, and they get a negative return on insurance investments. Thus, any discount rate higher than 0% implies that you require a positive expected return. In this sense, even a discount rate of 1% might be viewed as high.

Thus, what discount rate you advocate depends critically on the degree to which you view climate mitigation as an insurance investment, versus a problem of maximizing expected returns relative to markets. Separately, to the extent that you view climate mitigation from an insurance framework (discussed in the previous section), the 95th and 99th percentiles are the damage estimates that should be used to guide climate policy: it’s the lower probability but disastrous outcomes that matter.

The formula for discouting is:

Present Value = Future Value/(1 + d.r.)^t, where t is the year in which the benefit or cost occurs, d.r. is the discount rate and ^ is the symbol for raising the term to the power of t. So, as an example, $101 worth of damage in 1 year is counted as $101/(1 + .01)^1 = $100.

Conclusion

The estimated damages from climate change presented here range from $21 per ton of CO2 to more than $2,000.  None of these estimates account for the worst impacts from climate change that could not be monetized, such as those resulting from socio-political conflicts over land and water resources, socio-political conflicts from potentially millions of climate refugees, losses of all coral reefs, mass species extinction, or pain and suffering from public health impacts from increased disease and illnesses such as malaria, westnile virus, and asthma. To name just a few. And they make no adjustments for poorer regions bearing larger burdens (a dollar’s worth of damage is assumed to have the same relative impact regardless of income and wealth).

The $21 estimate ignores worse-case outcomes, relies upon optimistic climate change projections that do not reflect the most recent climate research, and assumes the uppermost value of the government’s recommended discount rate range for intergenerational discounting (see “Explanation of discount rate” section above). As a central estimate, it also assumes people are not risk averse to highly damaging or catastrophic climate change. Even despite all of these flaws, EPA’s central estimate it is not less than its estimated cost of reducing CO2 emissions.

Using EPA’s more optimistic climate projections, at the lower end of the government’s discount rate range (1% from its 1%-3% range), EPA’s own model projects average damages of $229/ton of CO2, and a 5% and 1% chance of damages exceeding $527/ton and $643/ton, respectively. Making adjustments to reflect more recent and pessimistic climate science gives a 1% chance of damages exceeding $2,000.

What left is there to say?

This blog was updated July 27th, 2011 to add the discount rate formula.

[i] The report was peer reviewed by a well published scholar in climate change economics who teaches and does research at the London School of Economics (LSE). Consistently ranked one of the world’s top universities, LSE also comes out way ahead in the social sciences (where economics resides)—ranking 5th, 4th, 3rd and 2nd in the last five years.

[ii] Hanemann, M (2010). Chapter 17, in Climate Science and Policy, eds. Schneider, S.H., and Rosencranz, A., Mastrandea, M.D., and Kuntz-Duriseti, K. Island Press: Washington D.C.

[iii] Weitzman, M.L. (2009). On modeling and interpreting the economics of catastrophic climate change. Review of Economics and Statistics 21:255-287.

[iv] Just about all non-market goods are not monetized. One notable exception is the inclusion of what people are willing to pay to reduce their risk of death. For a discussion of how this is estimated in cost benefit analysis (in the context of conventional pollutants regulated under the Clean Air Act), see http://switchboard.nrdc.org/blogs/ljohnson/neras_flawed_critique_of_epas.html.   

[v] Hope, C. and Johnson, L.T. (2010). Revisiting the SCC estimates developed by the US Government: The effects of intergenerational discounting methods and regional equity weights.*Available upon request.

Comments