Optimal Compliance with Environmental Policy: How Large are the
Cost Savings from Emissions Trading?∗
Hei Sing (Ron) Chan†
University of Maryland
June 2, 2013
Extended Abstract
Recently EPA has been designing a new emissions trading scheme on electric utilities that has faced lots of
legal disputes (due to the nature of trading across state borders), which delays the implementation and lowers
the effectiveness of the environmental policy. While economists have long argued that tradable permits or
emissions tax are more economically efficient as firms with higher pollution abatement costs can buy permits
from lower cost firms to lower cost for both parties, it would be empirically important to understand how
much the cost savings are, compared to more traditional form of regulation like emissions standard. If the
emissions trading does not generate too much savings, it is not a terrible idea to use emissions standard
which is easier to implement.
The objective of this paper is to quantify the cost savings from a market-based instrument compared to a
command-and-control instrument. I use ex-post data in the first three years in the Phase II of the Acid Rain
Program (2000–2002) to help identity their best choice of coal as well as a scrubber installation decision,
which would remove most of the emissions. Cost heterogeneity arises primarily because of geographical
difference as some generating units are located closer to the source of low sulfur Western coal. In this ex-post
approach, as compared to ex-ante approaches that use pre-program data (Carlson et al., 2000), I estimate a
random coefficient logit model, based on their observed coal procurement and scrubber installation behavior,
to identify their optimal compliance strategy and recover parameters associated with the compliance cost
function. Other than the benefits of using ex-post actual program data, I also consider a wider range of
choices that they can make (compared to Keohane (2006)) and this is the first ex-post study of the full
participation Acid Rain Program.
In my model each generating unit picks a compliance strategy to minimize cost. Compliance cost depends
on the emissions, ash content of coal, cost of operating a scrubber and additional cost of using low sulfur
Wyoming coal which also depends on the unit age. I also allow the compliance cost to be affected by state
∗I
would like to thank Maureen Cropper, Andrew Sweeting, Rob Williams for their motivation and suggestions towards
the early development of this paper. This paper also benefits from discussions from Daisy Dai, Ginger Jin, Arik Levinson,
Nuno Limão and Tara Kaul. I would like to thank Jacob Bournazian of the U.S. Energy Information Administration (EIA)
in assisting and making the proprietary cost data available for this research. This paper in this publicly available form is
ensured that no proprietary information can be identified and leaked out. The remaining errors are of my own.
† Department of Economics, 3114 Tydings Hall, University of Maryland, College Park, MD 20742.
Contact:
chan@econ.umd.edu or ronhschan0@gmail.com.
1
policy and other locational preferences such as bias towards capital intensive investment, electricity market
deregulation, credits for using in-state coal and minemouth status of the unit. These observable characteristics
can affect scrubbing decision (as rate-of-return regulation favors scrubbing) as well as the choice of coal
(as documented in Cicala (2012) and Chan et al. (2013)). I assume the unobserved component cost to be
generalized extreme value distributed to estimate my model using the family of multinomial logit.
The most basic form of multinomial logit model assumes that the unobserved term is independently distributed across alternatives. If the unobserved component represents the cost not observed by the econometrician, one would argue that the unobserved cost may be highly correlated among choices, given that use of
Illinois coal and use of Illinois coal plus scrubbers are considered two distinct categories. As some of these
alternatives may be highly correlated, I estimate my model in a more relaxed generalized nested logit model
(GNL), proposed by Wen and Koppelman (2001), which allows for correlation across alternatives. I estimate
a model where the coefficients are allowed to be random as in the mixed logit, to capture some of the unobserved policy that affects scrubbing and fuel purchase decisions in particular. The random coefficient is
drawn to allow correlation across generating units within the same plant.
I found a huge heterogeneous effect of scrubber cost on the compliance choice based on both observed and
unobserved components (from the random coefficient. Observable effects have expected signs that significantly influence the compliance choice decision. I also found that low sulfur coal in Wyoming represents a
54-cent premium per heat input and it increases with the age of the generating unit. Based on my estimated
model, I simulate what would happen to the actual compliance cost when we have a uniform emission standard that achieves similar emissions, compared to the current market-based instrument. Uniform emission
standard is modelled where some compliance choices are ruled out because one cannot comply with them.
Conditional distribution of errors and random coefficients are used to capture potential permanent effects. I
found that the cost savings are around 105-230 million 1995 US dollars per year. This number is considerably smaller than estimates from earlier literature. I postulate that two effects may lead to the difference in
our estimates: (1) lower transportation cost induces less cost heterogeneity across generating units, and (2)
technological improvement in scrubbing technology also lowers the marginal abatement cost curves.
References
Carlson, Curtis, Dallas Burtraw, Maureen L. Cropper, and Karen Palmer. 2000. “Sulfur Dioxide Control by
Electric Utilities: What Are the Gains from Trade?” Journal of Political Economy, 108(6): 1292–1326.
Chan, Hei Sing (Ron), Harrison Fell, Ian A. Lange, and Shanjun Li. 2013. “Efficiency and Environmental
Impacts of Electricity Restructuring on Coal-Fired Power Plants.” CESifo Group Munich CESifo Working
Paper Series 4160.
Cicala, Steve. 2012. “When Does Regulation Distort Costs? Lessons from Fuel Procurement in U.S. Electricity
Generation.” Harvard University mimeo.
Keohane, Nathaniel O. 2006. “Cost Savings from Allowance Trading in the 1990 Clean Air Act: Estimates
from a choice-based model.” Moving to Markets in Environmental Regulation: Lessons from Twenty Years of Experience, Chapter 8. New York:Oxford University Press. edited by Kolstad, Charles D. and Freeman, Jody.
Wen, Chieh-Hua, and Frank S. Koppelman. 2001. “The Generalized Nested Logit Model.” Transportation
Research Part B, 35: 627–641.
2