Technical Appendix to Accompany “Measuring Market Power and the Efficiency of Alberta’s Restructured Electricity Market: An Energy-Only Market Design” by David P. Brown and Derek Olmstead In this Technical Appendix, we consider an array of extensions to our baseline model. In particular, we (i) define MW supply cushion thresholds instead of year-specific percentile supply cushion thresholds; (ii) use coal price data from Wyoming’s Powder River Basin and coal unit-specific heat rates to impute marginal cost of coal units; (iii) assume industrial demand is perfectly price-inelastic; (iv) adjust our aggregate quantity-weighted market power measures by assigning zero market power to high supply cushion hours where the measures are negative; and (v) investigate the impact of transmission capacity constraints within Alberta and from imports from neighboring provinces on our baseline analysis.
B1
MW Supply Cushion Thresholds
Throughout the analysis, we decompose our estimates by the level of supply cushion using the 5th, 25th, 75th, and 95th percentile for each year of our sample. Because these percentile cutoffs can vary by year in MW terms, we also decompose our estimates by absolute MW thresholds to ensure that we are comparing hours with similar supply and demand conditions across the years in our sample. In particular, we use the supply cushion thresholds 748 MW, 1,294 MW, 1,738 MW, 2,235 MW, and 2,870 MW which corresponds to the average 5th, 25th, 75th, and 95th percentile across all years in our sample (rounded to the nearest integer), respectively. Table B1: Average Observed and Competitive Outcomes and Market Power by MW Supply Cushion
Observed Price ($/MWh) Observed Output (MWh) Competitive Price ($/MWh) Competitive Output (MWh) M P (l) (%)
Supply Cushion Bottom 5 % Bottom 25 % 390.10 162.41 8,188 8,129 150.31 68.51 8,360 8,177 60.3 57.5
IQR 37.10 7,812 30.09 7,844 18.9
Top 25 % 19.45 7,425 20.11 7,459 -3.2
Top 5% 15.25 7,336 17.60 7,325 -14.4
Total 65.94 7,800 37.93 7,837 42.8
Notes: For each measure, the values reflect averages across all hours within a particular supply cushion. IQR reflects the interquartile range. Total is the average across all hours. M P (l) represents our market power measure.
Tables B1 - B4 present the results of our analysis when the absolute MW thresholds are used. While adjusting our definition of the supply cushion thresholds alters the quantitative numbers, the key results and conclusions of our main analysis are unchanged. Tables B1 and B2 demonstrate that firms exercise a sizable amount of market power in the low supply cushion hours. Tables B3 and B4 illustrate that the market power execution results in sizable rent transfers and production inefficiencies in the low supply cushion hours. When comparing our results across years we observe similar trends to our baseline analysis. In the post-2010 OBEGS period, market power estimates and firms’ observed rents increase relative to the competitive benchmark (see Table B2 in particular). This provides us with further confidence that the OBEGs impacted firm behavior and subsequently, firm’s profits. Lastly, the conclusions regarding firms’ long-run investment incentives by generation technology are unchanged when looking across all hours within a given year. 1
Table B2: Observed and Competitive Outcomes and Market Power by MW Supply Cushion and Year Supply Cushion
Measure
2008
2009
2010
2011
2012
2013
2014
PObs PComp M P (l) PObs PComp M P (l) PObs PComp M P (l) PObs PComp M P (l) PObs PComp M P (l) PObs PComp M P (l)
299.53 157.29 46.7 152.22 91.99 38.9 49.43 40.63 16.9 21.61 19.48 10.6 14.67 15.66 -6.0 91.19 60.92 32.6
293.41 116.66 59.1 99.37 50.26 49.1 35.82 27.17 23.5 17.33 18.51 -6.8 12.45 16.44 -31.1 48.09 31.32 35.2
308.20 128.71 56.4 114.65 58.71 48.1 36.16 31.75 11.7 21.25 22.68 -7.6 15.17 19.51 -28.3 51.45 36.18 29.6
471.42 160.25 64.1 175.53 64.27 62.2 28.37 26.40 6.7 16.30 19.06 -18.0 12.94 16.96 -31.9 76.16 38.02 50.4
433.21 134.44 67.4 192.40 55.57 70.1 29.17 21.62 25.7 14.71 16.02 -9.5 10.89 13.78 -25.4 64.02 28.20 56.1
508.39 175.18 64.6 218.06 77.06 63.6 37.50 30.58 18.9 20.16 22.57 -12.6 16.64 20.16 -22.1 80.72 40.98 48.7
425.57 142.36 76.9 214.65 59.86 72.0 45.22 33.60 26.2 22.71 20.96 8.3 17.19 17.90 -3.2 50.05 30.19 40.8
Bottom 5%
Bottom 25%
IQR
Top 25%
Top 5%
Total
Notes: For each measure, the values reflect averages across all hours within a particular supply cushion and year. Total is the average across all hours within a year. M P (l) represents our market power measure.
Table B3: Hourly Production Inefficiencies by MW Supply Cushion
Production Inefficiencies Internal ($/hr) Internal Per-Unit ($/MWh) External ($/hr) External Per-Unit ($/MWh)
Supply Cushion Bottom 5 % Bottom 25 % 12,595 10,543 1.74 1.43 9,331 1,845 1.17 0.23
IQR 9,191 1.19 909 0.12
Top 25 % 8,562 1.07 707 0.10
Top 5% 6,476 0.79 637 0.08
Total 9,453 1.22 990 0.17
Table B4: Mean Hourly Payments, Costs, Rents, and Deadweight Losses by MW Supply Cushion
Observed Measure Payments ($/hr) Rents ($/hr) Production Costs ($/hr) Deadweight Loss ($/hr) Deadweight Loss ($/MWh)
Bottom 5 % 3,235,562 [ 152.0 ] 3,091,952 [ 166.7 ] 149,590 [ 17.7 ] 15,086 1.84
Supply Cushion Bottom 25 % 1,343,424 [ 135.3 ] 1,238,465 [ 161.1 ] 108,960 [ 11.0 ] 1,345 0.17
IQR 293,482 [ 23.2 ] 210,113 [ 30.3 ] 83,368 [ 8.4 ] 1,039 0.13
Top 25 % 147,372 [ -3.1 ] 78,003 [ -12.3 ] 67,368 [ 10.7 ] 738 0.10
Top 5% 116,202 [ -12.6 ] 50,109 [ -29.4 ] 64,593 [ 7.9 ] 494 0.07
Total 530,214 [ 73.9 ] 438,951 [ 95.9 ] 91,263 [ 13.0 ] 988 0.13
Notes: For each of the observed measures, the values reflect averages across all hours within a particular supply cushion. The numbers in the brackets denote the percentage change in the observed outcome relative to the competitive benchmark.
2
B2
Distribution of Monte Carlo Results
Each iteration of the Monte Carlo Coal marginal cost analysis yields a value for each of these central measures. This establishes a distribution of estimates. Because observed price and payments represent a deterministic outcome, they do not have a distribution. Alternatively, oligopoly (observed) rents vary because the underlying cost structure varies for each iteration of the Monte Carlo Simulation. Table B5 presents the distribution of results of our analysis for the key variables of interest. While the quantitative values vary by iteration, the distributional results exhibit strong patterns discussed in our main analysis. Table B5: Detailed Distribution of Central Model Estimates by Supply Cushion
Bottom 5% Bottom 25% IQR Top 25% Top 5% Total
Bottom 5% Bottom 25% IQR Top 25% Top 5% Total
Bottom 5% Bottom 25% IQR Top 25% Top 5% Total
Bottom 5% Bottom 25% IQR Top 25% Top 5% Total
Bottom 5% Bottom 25% IQR Top 25% Top 5% Total
Competitive Price ($) Std Dev Min Max 21.17 140.81 246.75 5.89 64.00 93.47 1.63 27.12 34.16 1.53 16.81 24.32 1.30 14.48 20.81 2.46 34.88 47.79 M P (l) Mean Std Dev Min Max 60.0 5.8 35.0 77.9 57.3 3.9 40.5 61.2 18.0 4.4 6.8 25.8 -2.1 6.8 -20.4 12.2 -18.9 9.1 -45.5 -1.1 42.8 3.9 27.2 47.4 Internal Ineff ($/hr) Mean Std Dev Min Max 15,831 9,987 8,952 37,215 10,564 7,812 7,454 30,171 9,221 6,536 6,861 21,464 8,509 3,836 6,714 14,430 6,355 947 5,394 9,608 9,453 2,367 7,059 21,394 Competitive Payment ($/hr) Mean Std Dev Min Max 1,322,981 182,638 1,197,447 2,154,258 567,845 50,064 531,224 793,529 235,454 12,207 214,618 269,457 150,125 11,115 127,209 183,266 128,828 9,626 110,181 158,580 304,879 19,867 281,467 389,878 Competitive Rent ($/hr) Mean Std Dev Min Max 1,173,877 175,779 1,059,028 1,975,532 456,962 47,156 430,013 675,126 155,771 9,925 146,683 191,915 87,737 9,526 74,394 121,678 74,324 8,478 63,179 105,309 224,117 17,760 208,823 303,391 Mean 155.36 68.36 29.90 19.97 17.01 37.93
3
Observed Price ($) Std Dev Min Max Deadweight Loss ($/hr) Mean Std Dev Min Max 14,247 39 7,387 18,421 1,333 32 1,187 1,375 1,042 13 1,009 1,080 720 25 647 775 502 38 376 576 988 11 960 1,015 External Ineff ($/hr) Mean Std Dev Min Max 9,981 388 8,287 11,414 1,829 125 1,733 2,799 903 53 792 1,080 701 66 597 995 632 82 497 1,019 990 54 881 1,166 Observed Payment ($/hr) Mean Std Dev Min Max 3,315,142 1,333,392 289,068 144,585 109,002 530,214 Oligopoly (Obs) Rent ($/hr) Mean Std Dev Min Max 3,143,766 11,587 3,028,896 3,151,701 1,211,494 12,221 1,093,061 1,229,112 204,037 12,366 97,436 212,526 74,774 11,501 48,052 82,468 44,158 9,295 16,343 51,250 438,951 22,694 337,380 452,233 Mean 399.69 161.21 36.59 19.10 14.18 65.94
B3
Coal Marginal Cost Approximation
In Alberta, coal generation units sign unobservable bilateral contracts with coal mines near the generation facility. To check the robustness of our Monte Carlo simulation approach used to impute the marginal cost of coal units in our baseline analysis, we use weekly coal price data (pC t ) from Wyoming’s Powder River Basin (PRB) from the Energy Information Administration. Using Bank of Canada exchange rates, we adjust the PRB prices from USD to CAD. PRB coal is classified as “sub-bituminuous”, which is the type of coal primarily mined and used for electricity generation in Alberta. In particular, coal in Alberta and PRB contain similar heat rate contents (Alberta Energy, 2014). For all coal units in Alberta, we obtained unit-specific heat rates from CASA (2004). The Specified Gas Emitter Regulation (SGER) imposed a requirement that fossil fuel generators pay a $15/tC02e or buy offsets (SGER, 2007). For coal units, this results in compliance costs of approximately $1.92/MWh (Pfeifenberger and Spees, 2011). Data on technology-specific variable O&M costs were obtained in EIA (2013) for the two types of coal technologies in Alberta: IGCC and Advanced PC. We assume that the O&M and the environmental compliance costs grow at the average inflation rate in Alberta over our sample. We model marginal cost of coal units as the summation of its fuel costs (HRi × pC t ), the environmental compliance costs, and estimated variable O&M costs by coal unit technology. We reestimated the analysis using the computed coal unit marginal cost proxy. The short-run marginal cost from the Monte Carlo procedure and the imputed cost using the PRB weekly prices yield similar results. Specifically, the short-run marginal cost of coal units range from $10.24 to $23.29 per MWh using the PRB coal prices and $8.60 and $25.06 per MWh using our Monte Carlo simulation method. The spread of implied marginal cost in both approaches reflects the diverse efficiency rate of coal units in Alberta. Table B6: Average Observed and Competitive Outcomes and Market Power by Supply Cushion with Coal Marginal Cost Estimated Using PRB Coal Prices
Observed Price ($/MWh) Observed Output (MWh) Competitive Price ($/MWh) Competitive Output (MWh) M P (l) (%)
Supply Cushion Bottom 5 % Bottom 25 % 399.69 161.21 8,189 8,128 164.80 68.64 8371 8177 57.3 57.2
IQR 36.59 7,800 26.53 7826 27.4
Top 25 % 19.10 7,421 16.71 7472 4.3
Top 5% 14.18 7,345 15.81 7322 -9.5
Total 65.94 7,800 35.55 7838 46.3
Notes: For each measure, the values reflect averages across all hours within a particular supply cushion. IQR reflects the interquartile range. Total is the average across all hours. M P (l) represents our market power measure.
Tables B6 - B10 present the results of our analysis using the coal marginal cost computed via PRB prices and the methodology described above, holding all other aspects of our analysis constant. These results illustrate that the key qualitative conclusions of our analysis remain in the setting where PRB coal prices are used. This demonstrates the robustness of the Monte Carlo methodology. In particular, the coal marginal cost proxy approach yields: (i) similar market power measures across all supply cushion hours (Tables B6 and B7), (ii) close estimates on counterfactual prices (Tables B6 and B7), (iii) and slightly lower estimates on production inefficiencies (Table B8). In addition, the market 4
power increasing impact in the post-2010 OBEGs period persists as demonstrated by Table B7 and B10. While we observe a slight increase in aggregate observed oligopoly rents (see Total Average Hourly Rents in Table B9), the capacity investment incentives observed in the baseline model are unchanged. The competitive benchmark provides insufficient revenues to cover estimates on levelized capacity costs. Further, observed rents are often exceed the levelized cost of capacity for natural gas and cogeneration technologies, but are below levelized capacity costs for coal and wind. Table B7: Observed and Competitive Outcomes and Market Power by Supply Cushion and Year with Coal Marginal Cost Estimated Using PRB Coal Prices Supply Cushion Bottom 5%
Bottom 25%
IQR
Top 25%
Top 5%
Total
Measure
2008
2009
2010
2011
2012
2013
2014
PObs PComp M P (l) PObs PComp M P (l) PObs PComp M P (l) PObs PComp M P (l) PObs PComp M P (l) PObs PComp M P (l)
460.17 241.64 46.4 197.93 111.81 42.9 65.07 49.55 22.9 30.08 22.89 14.9 19.11 16.72 6.7 91.19 59.31 34.5
242.89 106.12 55.2 96.88 50.36 47.6 37.05 25.78 29.5 19.17 15.85 12.0 13.04 14.12 -8.2 48.09 29.69 38.2
276.43 121.36 54.2 109.74 56.86 47.5 35.95 28.51 20.1 21.56 17.99 6.1 15.89 16.64 -4.5 51.45 33.24 34.1
519.32 198.11 60.0 216.50 73.72 64.7 34.24 26.43 22.7 17.76 17.53 0.8 13.97 16.39 -17.7 76.16 36.17 52.4
448.23 158.04 68.4 182.94 55.43 62.8 27.47 18.94 30.8 13.89 14.86 -6.8 9.28 14.04 -47.6 64.02 27.30 57.2
584.12 239.82 62.8 219.50 79.14 57.0 35.72 22.99 35.9 18.98 15.81 5.9 14.71 15.21 -4.8 80.72 36.26 54.4
289.51 77.43 72.8 111.26 44.53 60.1 33.05 23.10 30.2 18.10 17.03 6.4 12.87 16.75 -27.1 50.05 27.28 46.4
Notes: For each measure, the values reflect averages across all hours within a particular supply cushion and year. Total is the average across all hours within a year. M P (l) represents our market power measure.
Table B8: Hourly Production Inefficiencies by Supply Cushion with Coal Marginal Cost Estimated Using PRB Coal Prices
Production Inefficiencies Internal ($/hr) Internal Per-Unit ($/MWh) External ($/hr) External Per-Unit ($/MWh)
Supply Cushion Bottom 5 % Bottom 25 % 10,831 7,359 1.53 0.89 9,857 1,780 1.23 0.22
IQR 6,948 0.85 867 0.11
Top 25 % 5,845 0.71 883 0.12
Top 5% 4,629 0.58 1,018 0.19
Total 6,843 0.83 1,010 0.13
Notes: For each measure, the values reflect averages across all hours within a particular supply cushion and year.
5
Table B9: Mean Hourly Payments, Costs, Rents, and Deadweight Losses by Supply Cushion with Coal Marginal Cost Estimated Using PRB Coal Prices
Observed Measure Payments ($/hr) Rents ($/hr) Production Costs ($/hr) Deadweight Loss ($/hr) Deadweight Loss ($/MWh)
Supply Cushion Bottom 5 % Bottom 25 % 3,315,142 1,333,392 [ 134.46 ] [ 132.57 ] 3,140,640 1,205,921 [ 147.9 ] [ 161.3 ] 174,501 127,471 [ 18.5 ] [ 13.9 ] 11,684 1,386 1.43 0.17
IQR 289,068 [ 37.84 ] 198,712 [ 58.5 ] 90,357 [ 7.2 ] 1,071 0.14
Top 25 % 144,585 [ 14.93 ] 69,800 [ 27.1 ] 74,785 [ 5.5 ] 648 0.09
Top 5% 109,002 [ -8.69 ] 38,861 [ -26.9 ] 70,142 [ 6.0 ] 379 0.05
Total 530,214 [ 86.07 ] 438,849 [ 120.3 ] 96,721 [ 9.1 ] 973 0.11
Notes: For each of the observed measures, the values reflect averages across all hours within a particular supply cushion. The numbers in the brackets denote the percentage change in the observed outcome relative to the competitive benchmark.
Table B10: Mean Hourly Payments, Costs and Rents by Year with Coal Marginal Cost Estimated Using PRB Coal Prices Observed Measure Payments ($/hr) Rents ($/hr) Production Costs ($/hr)
2008 698,183 [ 52.80 ] 587,655 [ 64.75 ] 110,528 [ 10.27 ]
2009 371,276 [ 61.90 ] 284,072 [ 92.75 ] 87,203 [ 6.43 ]
2010 399,566 [ 54.12 ] 302,113 [ 76.02 ] 97,453 [ 11.21 ]
2011 627,563 [ 110.19 ] 526,824 [ 155.40 ] 100,739 [ 9.15 ]
2012 535,401 [ 133.82 ] 451,010 [ 194.29 ] 84,391 [ 11.45 ]
2013 674,459 [ 119.64 ] 578,820 [ 159.28 ] 95,638 [ 14.09 ]
2014 443,540 [ 86.68 ] 340,803 [ 128.19 ] 102,737 [ 16.43 ]
Notes: For each of the observed measures, the values reflect averages across all hours within a particular supply cushion. The numbers in the brackets denote the percentage change in the observed outcome relative to the competitive benchmark.
6
B4
Price-Inelastic Industrial Demand
In Section 3.4, we estimated the demand elasticity of several large industrial consumers in Alberta. Using lagged prices as our instrument in the hourly demand function estimation may result in us estimating a medium-term price elasticity of demand, instead of a short-term elasticity. Reducing the price-elasticity of industrial demand decreases the price-elasticity of our residual demand function. In our analysis, decreasing the price-elasticity of industrial demand results in a steeper residual demand function anchored around the observed market outcome (i.e., rotating residual demand inward below the observed market outcome, impacting the competitive counterfactual). Therefore, this change only impacts the competitive counterfactual equilibrium outcomes. This has the potential to elevate our market power, observed oligopoly rents, and production inefficiency measures because of the now steeper residual demand function which impacts our estimated competitive counterfactual. To provide a conservative upper bound estimate on these measures, we consider a sensitivity analysis where we assume industrial demand is perfectly priceinelastic. Table B11: Average Observed and Competitive Outcomes, Market Power, and Internal Production Inefficiency by Supply Cushion with Perfectly Inelastic Industrial Demand
Observed Price ($/MWh) Observed Output (MWh) Competitive Price ($/MWh) Competitive Output (MWh) M P (l) (%)
Supply Cushion Bottom 5 % Bottom 25 % 399.69 161.21 8,189 8,128 145.27 66.07 8,340 8,160 62.6 58.8
IQR 36.59 7,800 30.36 7,827 17.0
Top 25 % 19.10 7,421 20.34 7,455 -3.5
Top 5% 14.18 7,345 17.24 7,328 -19.3
Total 65.94 7,800 37.62 7,831 43.5
Notes: For each measure, the values reflect averages across all hours within a particular supply cushion.
Tables B11 and B12 highlight the key changes in our analysis when industrial demand is assumed to be perfectly price-inelastic. Reducing the price-elasticity of the residual demand function impacts the results of our analysis in an intuitive fashion. From Table B11 we see a small increase in the market power estimate in the lowest supply cushion (highest demand) hours due to the more price-inelastic residual demand function. Table B12 demonstrates that observed payments and rents increase compared to the competitive counterfactual in the low supply cushion hours when industrial demand is assumed to be perfectly priceinelastic. In addition, the difference between observed and competitive production costs increase across all hours. While reducing the price-elasticity of industrial consumers increases the market power, production inefficiencies, and oligopoly rents compared to the competitive counterfactual, these increases are relatively limited due to the limited size of price-responsive consumers in our sample. Thus, the capacity investment results found in Section 5 persist in this environment. Because we consider the bounding case where industrial consumers are perfectly price-inelastic, intermediate values on the price-elasticity of consumers yields similar results.
7
Table B12: Mean Hourly Payments, Costs, Rents, and Deadweight Losses by Supply Cushion with Perfectly Inelastic Industrial Demand
Observed Measure Payments ($/hr) Rents ($/hr) Production Costs ($/hr)
Bottom 5 % 3,315,142 [ 159.4 ] 3,143,766 [ 177.1 ] 171,376 [ 19.2 ]
Supply Cushion Bottom 25 % 1,333,392 [ 143.0 ] 1,211,494 [ 175.7 ] 121,898 [ 11.5 ]
IQR 289,068 [ 25.7 ] 204,037 [ 34.0 ] 85,032 [ 9.5 ]
Top 25 % 144,585 [ -4.7 ] 74,774 [ -16.9 ] 69,810 [ 13.1 ]
Top 5% 109,002 [ -16.9 ] 44,158 [ -39.8 ] 64,844 [ 12.2 ]
Total 530,214 [ 75.0 ] 438,951 [ 96.9 ] 91,263 [ 14.0 ]
Notes: For each of the observed measures, the values reflect averages across all hours within a particular supply cushion. The numbers in the brackets denote the percentage change in the observed outcome relative to the competitive benchmark.
B5
Zero-Adjusted Aggregate M P (l)
To ensure that our negative market power estimates do not unduly bias the aggregated quantity-weighted market power measures downward, we truncate the market power measures in the high supply cushion (low demand) hours when negative market power estimates occur to price equals marginal cost (i.e., zero market power).1 During these hours, there are limited concerns of market power execution. This approach follows the sizable empirical and theoretical electricity auction literature that illustrates that firms exercise market power largely in high demand periods (e.g., see Crawford et al., 2007; Bushnell et al., 2008; Mansur, 2008). Table B13 illustrates that the results of our quantity-weighted market power measure (M P (l)) across all hours is elevated slightly when we undertake this zero-adjusted approach. In addition, the higher post-2010 aggregated market power measure is consistent with the post-OBEG findings in our main analysis. These results, coupled with the market power measures decomposed by the level of the supply cushion, illustrate that our aggregated findings (across all hours) are not being driven by measurement error associated with our static estimate of marginal cost on coal units during high supply cushion (low demand) hours. Table B13: Mean Hourly Zero-Adjusted Aggregate Market Power Measure by Year
Aggregate M P (l)
2008 34.40
2009 36.70
2010 32.50
Zero-Adjusted M P (l) 2011 2012 2013 52.50 58.90 52.40
2014 43.20
2008-2014 45.30
Notes: The average aggregate market power measure M P (l) represents the zero-adjusted quantity weighted market power measure across all hours within a specified year. 1
Recall, in low demand (high supply cushion) hours coal generators may operate at prices that are below our static estimates of marginal cost because of ramping constraints and startup costs. This results in negative market power estimates in these hours. Due to data limitations, we are unable to fully account for these non-convex cost characteristics. Although, the focus of our analysis is on the high demand (low supply cushion) hours when market power concerns are highest.
8
B6
Transmission Constraints
B6.1
Neighboring Import Constraints
The import supply function estimation approach can yield estimated imports from neighboring provinces that exceed the intertie capacity limits. The Saskatchewan and British Columbia interties have a maximum capacity of 153 MWh and 700 MWh, respectively. To investigate the impact of transmission capacity limits from neighboring provinces, we estimate the import supply functions from each of the neighboring provinces and cap estimated imports at the transmission capacity intertie limits.2 Because imports are price-responsive, this creates a kink in the hourly residual demand curve (i.e., the residual demand curve becomes more price-inelastic) as we hit these transmission import capacity constraints. Tables B14 presents the observed and competitive equilibrium outcomes and the market power measures by supply cushion when import capacity limits are considered. Accounting for import capacity limits primarily impacts the lowest supply cushion hours when import constraints bind. As anticipated, the market power measures are elevated by a few percentage points in the bottom 5% and 25% of the supply cushion because of the steeper residual demand curve when neighboring import constraints are binding. Table B14: Average Observed and Competitive Outcomes and Market Power with Censored Imports from Neighboring Provinces by Supply Cushion
Observed Price ($/MWh) Observed Output (MWh) Competitive Price ($/MWh) Competitive Output (MWh) M P (l) (%)
Supply Cushion Bottom 5 % Bottom 25 % 399.69 161.21 8,189 8,129 153.32 66.36 8,374 8,183 62.2 59.1
IQR 36.59 7,800 29.40 7,838 18.5
Top 25 % 19.10 7,421 19.98 7,453 -2.05
Top 5% 14.18 7,345 17.01 7,325 -18.9
Total 65.94 7,800 36.51 7,841 43.6
Notes: For each measure, the values reflect averages across all hours within a particular supply cushion. IQR reflects the interquartile range. Total is the average across all hours. M P (l) represents our market power measure.
Table B15 decomposes these effects by year and supply cushion. The findings in the yearly analysis parallel the findings throughout our baseline analysis, market power increased in the post-2010 period. Although, the market power measures are elevated slightly in the lowest supply cushion hours. The increase in the estimated market power measures is largest in 2008 and 2013. This occurs because import constraints from neighboring provinces are binding most in these two years of our sample. While there is a marginal increase in our market power measures at the low supply cushion (high demand) hours, the key findings in our analysis persist because of the relatively small size of imports in our non-censored baseline model. 2
A more robust analysis would use a nonlinear censored regression model to account for the upper bound on imports. However, due to the computation complexity of our benchmark counterfactual analysis, we are unable to use such nonlinear models in our algorithm.
9
Table B15: Observed and Competitive Outcomes and Market Power with Censored Imports from Neighboring Provinces by Supply Cushion and Year Supply Cushion Bottom 5%
Bottom 25%
IQR
Top 25%
Top 5%
Total
Measure
2008
2009
2010
2011
2012
2013
2014
PObs PComp M P (l) PObs PComp M P (l) PObs PComp M P (l) PObs PComp M P (l) PObs PComp M P (l) PObs PComp M P (l)
460.17 240.18 47.8 197.93 111.46 43.3 65.07 50.96 20.2 30.08 25.74 13.5 19.11 18.13 5.7 91.19 59.08 34.8
242.89 99.18 58.7 96.88 48.93 49.3 37.05 26.96 24.0 19.17 19.30 -0.8 13.04 16.78 -28.1 48.09 30.56 36.2
276.43 116.33 56.4 109.74 56.41 47.9 35.95 30.86 11.8 21.56 22.90 -7.3 15.89 20.00 -26.0 51.45 35.14 30.6
519.32 188.49 63.8 216.50 73.39 64.9 34.24 27.64 16.9 17.76 19.94 -13.4 13.97 17.57 -27.2 76.16 36.98 51.7
448.23 140.36 70.1 182.94 51.32 70.1 27.47 21.14 22.9 13.89 15.57 -12.7 9.28 12.99 -37.6 64.02 27.24 57.5
584.12 209.09 65.3 219.50 76.48 64.3 35.72 30.04 16.2 18.98 21.87 -16.4 14.71 18.97 -30.1 80.72 38.88 51.9
289.51 69.23 76.4 111.26 44.93 59.2 33.05 27.53 16.7 18.10 18.39 -1.1 12.87 15.36 -17.9 50.05 29.68 41.4
Notes: For each measure, the values reflect averages across all hours within a particular supply cushion and year. Total is the average across all hours within a year. M P (l) represents our market power measure.
B6.2
Within Alberta Transmission Congestion
Observed production behavior could deviate from the competitive benchmark due to transmission constraints within Alberta. In our baseline model, such deviations from the perfectly competitive counterfactual would be treated as an internal productive inefficiency. Therefore, it is essential to ensure that internal transmission congestion is not driving our findings. In 2014, the MSA released a report that illustrated that congestion within Alberta had limited impacts on market outcomes. In 2012 and 2013, the MSA finds that only 0.2% of the MWh dispatched in Alberta were uneconomic due to within Alberta congestion, i.e., 0.2% of the MWh were dispatched by more expensive (less efficient) resources due to within Alberta congestion (MSA, 2014). These findings provide strong evidence that congestion within Alberta is unlikely to be a major driver of our findings. To ensure that our inefficiency results are not being largely driven by congestion due to transmission constraints within Alberta, we obtained proprietary data on transmission congestion within Alberta from the MSA. Due to data limitations, for each hour in our sample we are only able to observe a flag that indicates if some degree of Constrained Down Generation (CDG) was used to alleviate within Alberta transmission congestion (MSA, 2014). Using this internal congestion flag, we separate our sample into hours with and without congestion to investigate if there are large differences in our results and market outcomes across these hours. Figure B1 and B2 presents the cumulative distribution of the supply cushion and observed market prices across our entire sample for hours with and without congestion. For each year in our sample,
10
Figure B1: Supply Cushion Distribution
Figure B2: Price Distribution
Table B16 presents the observed market price and supply cushion for hours with and without congestion. Looking across all years in our sample, hours with congestion tend to have a lower supply cushion and a higher market clearing price. Although, there are years in which congestion arises in hours with lower average wholesale market prices (e.g., 2009, 2011, and 2014) and higher average supply cushions (e.g., 2009 and 2014). The highest amount of congestion occurs in 2010. Table B16: Average Observed Market Outcomes by Internal Transmission Congestion Status and Year Measure
Congestion Status Congested Observed Price Not Congested Congested Supply Cushion Not Congested Hours Congested (%)
2008 114.23 84.40 1,247 1,478 17.0
2009 38.65 49.75 1,742 1,712 13.6
2010 60.73 44.21 1,619 1,862 36.7
2011 66.79 78.66 1,589 1,606 18.1
2012 68.90 62.71 1,616 1,890 13.8
2013 138.13 62.77 1,532 1,949 18.9
2014 40.33 50.42 2,258 2,237 9.7
All Years 76.45 61.27 1,631 1,846 18.1
Notes: Hours are defined as congested if there was any Constrained Down Generation (CDG) used to alleviate a transmission congestion constraint within Alberta. For each measure, the values reflect averages across all hours within a particular CDG status and year.
Table B17 presents the distribution of observed and competitive benchmark outcomes and market power by supply cushion and within Alberta congestion status. We define the supply cushion thresholds using the MWh’s determined by the annual percentile thresholds used in the baseline analysis. The results are robust to using MWh thresholds discussed in Section B1. While there are differences in the equilibrium outcomes and market power measures by congestion status within each supply cushion region, the qualitative conclusions of our analysis are robust to decomposing our results by congestion status. Table B18 presents the internal and external production inefficiency measures for hours with and without congestion. These results demonstrate that there is not a sizable difference in our inefficiency measures for hours with and without some degree of within Alberta transmission congestion. These results, coupled with the MSA’s (2014) report that illustrates that congestion does not result in a sizable amount of inefficient dispatch of generation resources, provides strong evidence to support the claim that our production inefficiency measures are not being driven by congestion within Alberta.
11
Table B17: Average Observed and Competitive Outcomes and Market Power by Supply Cushion and Internal Transmission Congestion Status Measure Obs Price Obs Output Comp Price Comp Output M P (l)
Congest Status Congested Not Congested Congested Not Congested Congested Not Congested Congested Not Congested Congested Not Congested
Bottom 5% 386.08 402.99 8,004 8,268 151.70 164.74 8,193 8,432 67.5 59.6
Bottom 25% 174.20 160.07 7,953 8,192 71.79 65.38 8,016 8,234 55.9 59.3
IQR 36.02 37.34 7,642 7,850 29.12 30.77 7,670 7,884 16.8 19.1
Top 25% 18.51 19.60 7,256 7,452 19.46 20.08 7,273 7,489 -3.3 -1.3
Top 5% 14.03 15.32 7,170 7,363 16.35 17.64 7,148 7,357 -19.9 -17.8
Total 78.57 62.89 7,684 7,829 41.62 35.87 7,722 7,865 44.3 41.4
Table B18: Production Inefficiencies by Supply Cushion and Internal Transmission Congestion Status Prod Inefficiencies Internal Internal Per - Unit External External Per - Unit
Congest Status Congested Not Congested Congested Not Congested Congested Not Congested Congested Not Congested
Bottom 5% 15,617 15,975 1.95 1.99 10,016 9,121 1.26 1.19
Bottom 25% 10,995 10,446 1.50 1.41 1,997 1,825 0.24 0.22
IQR 9,459 8,830 1.24 1.13 894 931 0.12 0.13
Top 25% 8,217 8,627 1.05 1.06 757 691 0.11 0.08
Top 5% 5,542 6,864 0.69 0.82 696 607 0.09 0.08
Total 9,492 9,434 1.23 1.21 1,079 988 0.19 0.15
References Alberta Energy (2014). Coal Statistics. Available at: http://www.energy.alberta.ca/coal/643.asp Bushnell, J., Mansur, E., and Saravia, C. (2008) “ Vertical Arrangements, Market Structure, and Competition: An Analysis of Restructured Electricity Markets.” American Economic Review, 98(1): 237 266. CASA (2004). A Study on the Efficiency of Alberta’s Electrical Supply System. Project # CASA-EEEC02-04. Prepared by JEM Energy Report for the Clean Air Strategic Alliance. Crawford, G., Crespo, J., and Tauchen, H. (2007). “Bidding Asymmetries in Multi-Unit Auctions: Implications of Bid Function Equilibria in the British Spot Market for Electricity.” Internal Journal of Industrial Organization, 25:1233 - 1268. EIA (2013). Updated Capital Cost Estimates for Utility Scale Electricity Generation Plants. U.S. Energy Information Administration. Mansur, E. (2008). “Measuring Welfare in Restructured Electricity Markets.” The Review of Economics and Statistics, 90(2): 369 - 386. MSA (2014). Q4/13 Quarterly Report. Oct. - Dec. 2013. Alberta Market Surveillance Administrator.
12
Available at: http://albertamsa.ca/uploads/pdf/Archive/00-2014/Q4 2013%20140131%20Final.pdf Pfeifenberger and Spees (2011). Evaluation of Market Fundamentals and Challenges to Long-Term System Adequacy in Alberta’s Electricity Market: Update. The Brattle Group.
13
