- “What We Know and Don’t Know About Climate Change, and Implications for Policy,” by Robert Pindyck. June 2020. SSRN #3614104.
The author who has most informed my thinking about climate change is Robert Pindyck, professor of economics at MIT’s Sloan School of Management. In this paper, he explains clearly how little we know, why we know so little, and how that lack of knowledge matters for policy.
He first looks at the science of climate change and projections of the likely effect of carbon emissions on future temperatures. He reviews the 140 studies that have been published since 1970 on “climate sensitivity” — the increase in the global average temperature that would result from a doubling of atmospheric carbon concentration. Most of the studies (115 of the 131) have “best estimates” of this increase that range from 1.5° to 4.5°C. That is a wide range, and if we include the outlier 16 studies’ “best estimates,” that range expands to between 0.5° and 8°C. The uncertainty in the estimates is increasing slightly over time: the standard deviation in post‐2010 studies is 1.13 as compared to 1.03 in pre‐2010 studies.
Why is there such uncertainty? The short answer is feedback loops: changes in the underlying physical processes arising from initial temperature increases created by increases in carbon concentration. We do not know if feedback is normally distributed nor do we know its mean and standard deviation. An important article in Science in 2007 (“Call Off the Quest,” by Myles R. Allen and David J. Frame, 318: 582–583) argued uncertainty about climate sensitivity is in the realm of the “unknowable” and that the uncertainty will remain for decades.
What economic damages result from temperature increases? We have some sense of how higher temperatures might affect agriculture. But those estimates are from short‐term changes in weather, not long‐term changes in climate. The latter will occur slowly and we will adapt. Even to the extent we don’t, losses of agricultural output in some regions of the world (near the Equator) might be offset by increased output in other regions (northern Canada and Russia). And agriculture is only 1%–2% of gross domestic product for industrialized countries and 3%–20% for developing countries. For the rest of the economy, economic activity is not related to temperature.
For Pindyck, the key for policy is the possibility of a catastrophic loss of GDP in the future. How much should we pay currently for carbon abatement to avoid catastrophe in the future? In Pindyck’s formulation, future generations may deeply regret irreversible environmental damage. But they also could find such preservation less valuable than we currently expect, in which case they would regret the irreversible expenditure that we made on preservation. He writes:
Should we hold back on emissions abatement because of the sunk cost, or should we accelerate abatement because of the irreversible environmental damage caused by emissions? And by how much should we hold back or accelerate? Sorry, but I can’t answer these questions. Why not? Because we simply don’t know enough about the climate system and about the impact of varying amounts of climate change.
Ironically, the lack of knowledge that makes “climate insurance” valuable prevents us from determining exactly how large that value is.
COVID-19 and Non‐ Pharmaceutical Intervention
- “Four Stylized Facts about COVID-19,” by Andrew Atkeson, Karen Kopecky, and Tao Zha. August 2020. NBER #27719.
- “Epidemiological and Economic Effects of Lockdown,” by Alexander Arnon, John Ricco, and Kent Smetters. September 2020. www.brookings.edu/wp-content/uploads/2020/09/Arnon-et-al-conference-dra…
The COVID-19 pandemic has induced people to reduce their interaction with others in order to reduce the risk of infection and has induced governments to enact policies that mandate reductions in interaction through the closure of businesses and large events. The economic recession resulting from reduced interaction has led to political dispute over the relative contributions of voluntary and mandated social distancing.
The first of these papers uses patterns in the fatality data to make inferences about the relative roles of voluntary behavior and policy. It examines the COVID-19 fatality data as of July 22, 2020 across the 23 countries and 25 U.S. states that have experienced at least 1,000 cumulative deaths. Across this diverse set of places (Argentina, Belgium, Brazil, Canada, Chile, France, Germany, India, Iran, Ireland, Italy, Japan, Mexico, Netherlands, Panama, Peru, Portugal, Russia, Spain, Sweden, Switzerland, Denmark, Turkey, and the United Kingdom), once the cumulative announced deaths reach 25, the growth rate of daily deaths from COVID-19 fell rapidly everywhere within 30 days and then remained at zero or below. The variance in the growth of deaths across countries and states fell within 20 days of cumulative deaths reaching 25 and has remained low relative to its initial level.
The authors claim that though the variance in policy across this diverse set of countries and states was large, the patterns in the fatality data were similar, and therefore peoples’ voluntary reactions, rather than policy, largely must explain the evolution of the pandemic. Put differently, despite the large variation in policy around the world, the actual change in behavior across countries was quite similar, resulting in a pandemic that is neither exponentially growing nor extinguished.
The second paper examines daily counts of COVID-19 cases and deaths in the United States as well as mobile phone data to estimate population contact rates (that is, how often a person encounters another person closely enough and long enough for COVID infection to occur) and employment rates. Like the first paper, the second concludes that almost all of the reduction in contact was voluntary rather than the result of policy. State and local non‐pharmaceutical interventions (NPIs) explain only 7% of the reduction in the contact rate by mid‐April, when it reached its lowest point.
The second paper goes further, however, and estimates the incremental mortality reduction from NPIs and the relative efficacy of business closures (which restrict firms) versus stay‐at‐home orders (which restrict individuals). The authors conclude that, on average, NPIs reduce employment (15%) much more than they reduce social contact (7%). Business closure orders performed particularly poorly from a benefit–cost perspective, accounting for 48% of the decline in employment as compared to a 22% decline in the contact rate. On the other hand, stay‐at‐home orders accounted for 30% of the decline in employment but 50% of the decline in the contact rate. Thus, the NPI of first resort should be stay‐at‐home orders rather than closing businesses.
Through the end of May, NPIs lowered confirmed COVID-19 deaths by more than 33,000 and reduced employment by an average of 3 million. Using the current conventional value of a statistical life (VSL) of $10 million, the benefits of saving 33,000 lives would be $330 billion.
An overestimate of the costs of reduced employment divides U.S. gross domestic product ($21 trillion) by total employment (165 million) to value each job at $127,000. For a better estimate, let’s use $100,000 per job; thus, 3 million lost jobs would entail $300 billion in costs. The NPIs would pass a benefit–cost test in the aggregate if, indeed, they have benefits of $330 billion and costs of $300 billion.
In a recent Cato paper titled “Balancing the Tradeoffs between Liberties and Lives,” Jeff Miron and I concluded that the appropriate VSL to use in the pandemic context was approximately $5 million rather than the conventional $10 million because the infection fatality rate for COVID is 100 times the employment fatality risks used to estimate VSL. That would lower the benefits of 33,000 saved lives to $165 billion. Those benefits would be greater than the costs of reduced employment only if the compensation of the average lost job was less than $55,000.
Truck Energy Efficiency Standards
- “Mode Choice, Energy, Emissions and the Rebound Effect in U.S. Freight Transportation,” by James B. Bushnell and Jonathan E. Hughes. October 2020. SSRN #3689848.
Energy efficiency regulations are more popular with voters than Pigouvian taxes on energy use because the costs of the former are much less visible than the latter. But a perverse effect of energy efficiency regulations is that they decrease the price of energy use slightly because of the increased efficiency, resulting in slightly more energy use that offsets some of the efficiency gains. This increased use of energy is called the “rebound” effect. In contrast, higher energy prices from taxation discourage more use even if efficiency improvements occur.
In August 2016, the U.S Environmental Protection Agency released new efficiency standards for heavy‐duty trucks (think tractor‐trailer trucks, though this class includes very large pickups) produced through the 2027 model year. The EPA predicts the new standards will improve new truck tractor fuel efficiency 11%–14% by 2021 and 19%–25% by 2027.
Trucks are not the only method of shipping freight. Truck shipments are approximately 46% of total ton‐miles while rail accounts for approximately 48% of ton‐miles. Rail is slower but much more fuel efficient. Rail can move 1 ton of freight approximately 450 miles on a gallon of fuel while trucks move approximately 70 ton‐miles per gallon.
Intuitively, if truck fuel economy improves, marginal shippers who previously paid lower rail rates but higher inventory costs (because they stored more stuff in inventory because of slower rail shipping times) may now shift to truck because the price of speed has decreased. Those shippers would get both more speed and lower inventory costs. But because rail is four times more fuel efficient than truck, the shift to trucks would increase fuel consumption. The authors label this the “cross‐rebound” effect.
The authors use 2012 data on a random sample of freight shipments in the United States to estimate the relationship between fuel costs and the use of rail or truck to ship. They then use those estimates to run a simulation in which fuel costs were reduced by 5% for trucks because of increased efficiency. They use 5% rather than the actual efficiency increase for new trucks to account for the gradual replacement of the truck fleet over time.
In the simulation, truck fuel‐economy efficiency regulation shifts approximately 16 billion ton‐miles, or approximately 1.3%, of business‐as‐usual rail freight to trucks. This shift reduces the effectiveness of truck fuel‐economy regulations. Without mode shifts, a 5% reduction in truck fuel intensity lowers total freight‐sector fuel consumption by approximately 4%. However, modal substitution from rail to truck dampens that effect, reducing fuel consumption by 3.3%. This implies a “cross‐rebound” effect of approximately 18%.