Online Appendices
“When Does Regulation Distort Costs? Lessons from Fuel Procurement
in U.S. Electricity Generation”
Steve Cicala
A Data Appendix
Data on Divestitures
Data on divestitures are compiled from the“Electric Utility Plants Sold/Transferred and Reclassified
as Non-utility Plants” Tables across various years of the March Issue of EIA’s “Electric Power
Monthly” report. It is also possible to identify the month of divestiture prior to 2002 because plants
cease reporting fuel costs at that time. A third source of divestiture date is a change in regulatory
status reported on Form EIA-906, “Power Plant Report.” In the relatively uncommon case that
these dates disagree, I rely first on the cost data (a signal of operational changes at the plant), then
the sale data, and finally the “Power Plant Report” data.
Table A.1 breaks down this history of coal-fired plant divestitures by state. Divestiture of utility-
owned plants in a state was usually complete following passage of passing restructuring laws. Not all
states that restructured have coal-fired plants to use in this study. Although California restructured
its electricity markets, its IOUs did not own any coal-fired capacity. Washington, DC was also
restructured but its two coal-fired plants are used sufficiently little to avoid fuel delivery reporting
requirements. All New England states except Vermont restructured their electricity markets, but
Maine and Rhode Island do not have coal-fired generating assets. New Hampshire did not require
divestiture of the two coal-fired plants owned by Public Service of New Hampshire and these plants
continue to report costs after the introduction of retail competition.
There have also been a number of divestitures in states that remain otherwise rate-regulated.
The plants divested in Indiana, and Virginia were owned by IOUs based in restructured states,
and were forced to sell for this reason. Montana has suspended restructuring but Montana Power
Company assets were divested in 2000 after its failed telecom investments during the dot-com bust
led the company in to bankruptcy. The Centralia station in Washington state was sold amidst
conflict among the plant’s eight co-owners.
Divestiture status in Ohio and Texas varies by utility service area. The only IOU plants in Texas
that remain to be divested belong to Southwestern Electric Power Company, which is connected to
a separate grid from the rest of the state. The lack of markets available in this service area has
delayed divestiture. In Ohio, two Duquesne Light Co. coal-fired plants were divested in 2000 as
part of Pennsylvania’s restructuring program. Although Ohio implemented retail choice in 2000,
FirstEnergy’s plants in Ohio would not be divested until 2005. Plans to divest of the remaining
IOU plants in Ohio have been tied up between the Public Utilities Commission of Ohio (PUCO)
and the courts since that time. The owners of these plants remain rate-regulated and require PUCO
approval to change electricity prices.
Coal Prices
This study uses detailed data on coal deliveries to power plants from the Energy Information
Administration (Forms EIA-423, “Monthly Report of Cost and Quality of Fuels for Electric Plants,”
1
and EIA-923, “Power Plant Operations Report”) and Federal Energy Regulatory Commission (Form
FERC-423, “Monthly Report of Cost and Quality of Fuels for Electric Plants”). This is shipment-
level data, reported monthly for nearly all of the coal burned for the production of electricity in the
United States (all facilities with a combined capacity greater than 50MW are required to report).
1
The data record the county or mine of origin, whether purchased on the spot market or long-term
contract, characteristics of the coal (heat, sulfur and ash content), rank (bituminous, etc.), and
the price per million British thermal units (MMBTU). Although data on prices are redacted from
public release for non-utilities, restricted-access data on prices were made available for this study
under a non-disclosure agreement with EIA.
As described in the text, deregulated plants were not required to report fuel prices to the EIA
until 2002. This means that IOU plants that were divested ceased reporting from the time of
divestiture until 2002. There is no gap in reporting for the limited set of plants that have been
divested since 2002. An exception to this rule is for the six FirstEnergy plants in Ohio that stopped
reporting once retail competition began in June of 2000, but did not resume reporting until actual
divestiture at the end of 2005. All results are robust to the exclusion of these plants.
Coal delivered to combined heat and power plants (4% of reported coal deliveries after 2002)
is not included in any of the analysis. These are plants that also sell steam, either for heating
or industrial processes. One reason is practical: 36 of 49 coal-fired co-generation plants were not
required to report until 2002, so they lack data in the pre-divestiture baseline period. The second
is that it is unclear how to categorize the regulatory structure these plants face: a plant owned
by an IOU may be free to privately contract for steam to nearby industrial plants. In addition,
four small facilities (typically produced <50MWh/month) that were divested, but never reported
post-divestiture are also dropped. They are the Hickling and Jennison plants in NY, Grand Tower
in IL, and Edgewater in OH.
Figure A.1a shows the total heat content of coal deliveries reported to FERC/EIA from 1990-
2009. The vertical lines represent the points at which divestitures begin in 1998, and when reporting
for divested plants resumes in January 2002. There is clearly a substantial amount of non-reporting
induced by divestiture. Aside from this dip, there is a 15-25% increase in coal delivered over this
20 year period.
2
It is important to note that nearly all of this came from an increase in production
at existing facilities, not entry of new plants.
Another feature of Figure A.1a worthy of note is the expansion of sub-bituminous coal, both
in levels and as a share of coal consumed for electric power. The Clean Air Act of 1990 created a
cap-and-trade program to reduce sulfur emissions from electricity generating and large industrial
units. Putting a price on sulfur increased the relative value of low-sulfur sub-bituminous coal (95%
of sub-bituminous coal mined in the United States in 2009 was from the Powder River Basin [PRB]
in Wyoming). Switching to PRB coal provided an alternative to building capital-intensive scrubbers
to reduce sulfur emissions. Technological improvements as demand for PRB coal expanded further
reduced the price of extraction, making PRB coal a potentially economical choice regardless of
environmental compliance considerations. Shipments of PRB coal more than doubled over the
twenty year period of study, accounting for about 40% of the coal heat delivered in 2009.
1
When switching to Form 923 in 2008, the EIA began collecting monthly data from a sample of plants and a
census annually. Monthly data are estimated by EIA from plants that only submitted the annual form. This change
applied more significantly to gas-based generators, as more than 97% of coal deliveries continued monthly reporting.
2
The drop-off in 2009 is the combined effect of the economic downturn and displaced generation due to the fall
in natural gas prices.
2
Plant-Level Data
Data on generator nameplate capacity and vintage come from Form EIA-860, “Annual Electric
Generator Report,” while data on installed abatement equipment are from Form EIA-767, “Annual
Steam-Electric Plant Operation and Design Data” and EIA-923, “Power Plant Operations Report.”
Annual capacity factor is the annual net generation reported on Form EIA-906/759 “Power Plant
Report,” divided by maximum potential output as determined by facility nameplate rating. This
form is also the source for analysis on changes in output at the facility-level. Utility-specific imple-
mentations of Incentive Regulation programs is from Sappington et al. (2001) with updates from
Guerriero (2010). This is linked to the plant-level data by the utility identifiers in the “Power Plant
Report” data.
Data on geographic coordinates of power plants are from the Environmental Protection Agency’s
eGrid database.
Unit-Level Data
Unit-specific characteristics are assembled using the crosswalks between unit components provided
in Form EIA-767, “Annual Steam-Electric Plant Operation and Design Data” available from 1990-
2005. The data on this form were later compiled on Form EIA-923, “Power Plant Operations
Report” after a gap in reporting for 2006.
3
The effects of this gap can be mitigated by the fact
that scrubber installation date is collected, so status in the missing year can be inferred from prior
and subsequent years. Power generating stations have been required to file these forms with EIA
regardless of regulatory status,
4
so this series does not suffer from the intermittent non-reporting
present in the fuel price data. Unit-level generator nameplate capacity and vintage comes from
Form EIA-860, “Annual Electric Generator Report.”
As with the generating facilities themselves, there has also been limited entry and attrition at the
unit level. As a fraction of nameplate capacity, 92% of units reporting in 2009 also reported when
the series began in 1990 (85% of units). These numbers increase to 95% and 93% respectively when
accounting for the expanded coverage among combined heat and power units in 2002. Attrition
was similarly rare, with 96% of capacity and 87% of units reporting in 1990 continuing to report in
2009.
It is worth noting that is that it is not uncommon for facilities to have both scrubbed and
un-scrubbed units operating at the same plant. This can be seen by comparing the number for
any scrubber present at the facility in panel A of Table 1 and the unit-level statistics in panel C.
5
The differences between divested and non-divested units are otherwise similar to those found at the
plant-level, and largely eliminated in the matched sample.
Mine-Level Data
Data on mine labor productivity are from the Mine Safety and Health Administration’s “Quarterly
Mine Employment and Coal Production Report” (MSHA-7000-2). Figure A.2 shows the trends in
production and labor hours over the sample period. The main development over the last twenty
years has been the explosion of production from the Powder River Basin (PRB) in Wyoming. This
3
Plants with a combined nameplate capacity less than 50 MW are not required to report fuel prices (Form EIA-
423/923), while all facilities with a capacity greater than 10 MW are required to report generating unit configurations
and operations (Form EIA-767/923). The discrepancy amounts to an infinitesimal share of production and capacity.
4
Form EIA-767 expanded coverage to a handful of combined heat and power plants in 2002.
5
While scrubbers had only been installed on a small fraction of generating units in 1997, these units were dis-
proportionately large. In 1997 28% of U.S. coal-fired capacity was scrubbed for sulfur emissions. This has grown to
nearly half by 2009.
3
has more than offset the decline of output elsewhere, so that there has been a modest increase
in coal production overall. The shift in output has been accompanied with a sharp decline in
mining employment, which has only rebounded slightly since 2005. The 1990s saw sharp increases
in labor productivity all around: from expanding output faster than employment in the PRB and
by reducing employment faster than output in the East. It requires about seven times less labor to
extract a ton of coal in the PRB.
Wages are calculated by adding up the quarterly hours reported in the MSHA data by FIPS
county and merging this data with the quarterly wage bill in the coal mining sector as reported
in the “Quarterly Census of Employment and Wages” from the Bureau of Labor Statistics.
6
Wage
rates are calculated at the county level by dividing the total county wage bill by total hours.
The thickness of coal seams is from MSHA’s “Mine Dataset,” which contains descriptive data
on all mines under MSHA’s jurisdiction since 1970. To calculate the depth of mine seams, I used a
Perl script to collect the universe of stratigraphic data from the U.S. Geological Survey’s “National
Coal Resources Data System.” The combined USTRAT and COALQUAL databases consist of
over 200,000 geo-coded core samples taken by federal and state geologists in order to map U.S.
coal deposits. Among the many parameters collected from these core samples is the depth of coal
deposits. I use these points to create a surface of estimated seam depth using a spline to interpolate
between points using the geoprocessing toolkit of ArcGIS 10.0. I then intersect the coordinates of
mines with this surface to estimate the depth of coal deposits at each mining site.
The EIA only began collecting source mine identifiers (MSHA ID) on the fuel delivery data in
2008. From 1990-2001, I link deliveries to the name of the supplier listed in EIA’s Coal Transporta-
tion Rate Database (CTRD) based on facility, county of coal origin, and the characteristics of the
coal reported in both the CTRD and EIA-423 data. The name of the supplier is explicitly listed in
the EIA-423 data beginning in 2002. Deliveries and mine characteristics are therefore connected at
the county-supplier level.
6
Coal mining employment is reported under the four-digit NAICS code, “2121.”
4
Figure A.1: Total Heat Content and Cost of Coal Deliveries by Rank, 1990-2009
(a) Total Heat Content of Coal Deliveries
0 5 10 15 20
Quads (10e15 btu)
1990 1995 2000 2005 2010
year
Bituminous Sub−Bituminous
Other
(b) Total Cost of Coal Deliveries
0 10 20 30 40
Billions of $
1990 1995 2000 2005 2010
year
Bituminous Sub−Bituminous
Other
Note: Vertical lines denote the year in which divestitures begin (1998) and when reporting for non-utilities
commences (2002). Source: Forms EIA-423,923 and FERC 423.
5
