Analytical Screens for Electricity Mergers (with David Newbery)

Transcription

1 University of California, Berkeley From the SelectedWorks of Richard J Gilbert 2008 Analytical Screens for Electricity Mergers (with David Newbery) Richard J Gilbert Available at:

2 Rev Ind Organ (2008) 32: DOI /s Analytical Screens for Electricity Mergers Richard Gilbert David Newbery Published online: 20 July 2008 Springer Science+Business Media, LLC Abstract Electricity mergers pose distinct challenges for competition policy. Electricity demand is highly inelastic in the short run, storage is limited, and transmission constraints limit the ability to substitute generation at other locations. As a result, a merger can affect prices in many different markets and even generators with small market shares may be able to exercise market power. The U.S. Federal Energy Regulatory Commission s approach for screening horizontal mergers, based on the concentration thresholds in the Department of Justice/Federal Trade Commission Horizontal Merger Guidelines, can fail to identify mergers that lessen competition, and mergers that fail the FERC screen may have no significant anticompetitive effect. We propose competitive residual demand (CRD) analysis, which examines the supply curves of the markets affected by a merger and considers the ability and incentive of firms to raise prices before and after a proposed merger. CRD analysis is a relatively easy way to address the incentives for generators to exercise market power and relies on data that are often available. Vertical (convergent) mergers between electricity and gas raise additional concerns, and we propose a methodology to screen vertical mergers. Keywords Merger policy Electricity Gas Convergent mergers Vertical integration JEL Classification G34 L22 L94 L95 R. Gilbert (B) Department of Economics, University of California, Berkeley, CA 94720, USA gilbert@econ.berkeley.edu D. Newbery Faculty of Economics, Cambridge University, Sidgwick Avenue, Cambridge CB3 9DE, England dmgn@econ.cam.ac.uk

3 218 R. Gilbert, D. Newbery 1 Introduction Electricity mergers pose distinct challenges for competition policy. A merger of electricity suppliers can affect prices in a very large number of relevant antitrust markets corresponding to different points in time and different geographic regions. Demand for electricity is highly inelastic, and electricity is not easily stored; hence separate product markets for electricity can exist even at close points in time. Furthermore, separate geographic markets can exist at geographic points that are not distant from each other if transmission constraints limit the ability of a consumer to substitute electricity from a different location. The Pennsylvania-New Jersey-Maryland (PJM) Interconnection and other wholesale electric power pools operate markets for electricity in which generators submit day-ahead bids for electricity supplies for each hour of the next day. The PJM also solicits real-time bids to balance supply and demand in each hour of the day. A merger could raise competition concerns in any of these 8,760 hourly day-ahead and 8,760 hourly real-time markets for each year. In addition to these 17,520 annual product markets operated by the PJM, transmission constraints can create separate geographic markets, each of which can have 17,520 separate product markets per year. If transmission capacity limits the flow of electricity into a region, or if transmission losses impose significant costs on electricity flows, then the existence of competitive electricity producers in one region may not suffice to limit the ability of a merged firm in another region to increase prices above pre-merger levels. For these reasons the analysis of mergers of electricity suppliers is often a daunting exercise. We discuss some of the approaches that may be used to evaluate the risk of price increases from electricity mergers and some shortcuts to simplify the analysis. Our focus in this discussion is on estimation of wholesale prices for electricity. Implicit in our discussion is the existence of markets for wholesale electricity in which suppliers make bid offers to a central pool. To simplify matters we ignore markets for generation capacity and transmission capacity and for other ancillary services such as voltage stability and spinning reserves, all of which add additional possible relevant markets that could be affected by a merger in the electricity industry. 2 Approaches to Electricity Merger Analysis A sophisticated approach to evaluate possible price impacts from an electricity merger is to estimate an equilibrium model of price formation for each relevant market. If generators submit hourly bids to a centralised pool, then an equilibrium model would estimate each supplier s bid price and quantity for each hour. The estimated equilibrium bids have the Nash property that each supplier s bid maximises its expected profit given the bids that are optimal for all other suppliers. Such supply function equilibrium models were developed by Green and Newbery (1992) to compare alternative market structures of the England and Wales generation market that were potential alternative choices facing the UK Government when restructuring the Central Electricity Generating Board before privatisation. Hortacsu and Puller (2006) construct

4 Analytical Screens for Electricity Mergers 219 a supply function equilibrium model for the Electricity Reliability Council of Texas (ERCOT) wholesale electricity market. They estimate a bid function for each supplier that determines the supplier s optimal bid price conditional on expected bids by other suppliers, using as data the actual demand and observed bids supplied to the System Operator. The advantage of the supply function equilibrium approach is that it is firmly grounded in price theory. It is a model in which each supplier is doing all it can to maximise its profit given the behaviour of other suppliers. The approach, however, has a number of disadvantages. It is complex, because it requires an estimation of optimal bids for each relevant product market and a computational search for bids that are best responses to all other bids. More troubling, there may be a continuum of equilibrium bids, so that observing that one set of bids is an equilibrium does not tightly constrain possible future equilibria even if circumstances do not change, let alone if they were to change, as they would with a merger. 1 Furthermore, the approach is static; bid strategies do not account for the possibility that current bids may affect future bidding behaviour. A key variable for the analysis of market power is the extent of forward contracting, which has a potentially very significant impact on spot market bids. 2 If a producer is completely hedged (i.e., he has sold as much power forward as he expects to generate), then his optimal spot bid is marginal cost, regardless of other bids. If the spot price is less than his short-run marginal cost, he is better off reducing output and buying in the spot market to meet his contractual obligations, avoiding the higher marginal cost of generation, and vice versa. If he has sold forward more than his expected generation, he has an incentive to drive the spot price below marginal cost, as he will be a net buyer in the spot market to discharge his contract position. Only if he is under-contracted does he have an incentive to drive the spot price above marginal cost, and his profit is limited to the difference between his output and contract position. If this is small, then the incentive to exercise market power is also small. Hortacsu and Puller (2006) estimate current contract positions from the point of intersection of bids with marginal costs. 3 It is possible to estimate dynamic bidding strategies, but these add additional layers of complexity. Allaz and Vila (1993) model the choice of forward sales in a Nash- Cournot duopoly model and show that the extent of forward cover (that is, forward contracts as a proportion of output) depends on the number of contract rounds and on 1 There are various ways of reducing the set of possible equilibria, most directly by assuming that incumbents are restrained by the threat of entry (Newbery 1998). However, entry in electricity markets can be difficult to predict and may not be effective in the short to medium run. Of course, predicting the impact of a merger is in itself a complex exercise even without concerns over ill-defined current equilibria. 2 Most spot markets are actually cleared some time before supply and demand are realised (either day-ahead or within the day at least an hour before gate closure ). 3 Yet another complication is that most spot markets require bids and offers to take the form of a ladder or step function, rather than a smooth curve from which optimal bids are most readily derived. Hortacsu and Puller (2006) smooth the step function to derive differentiable curves, but there are concerns that this may not produce optimal bids. This problem can be handled once it is accepted that market participants bid in expectation of the realised residual demand schedule, but the econometrics are considerably more demanding, as Wolak (2003) shows.

5 220 R. Gilbert, D. Newbery the assumption of full disclosure of contract positions after each round. In the limit as the number of rounds increases, market power diminishes to zero. If contract positions are not revealed, then market power remains. It is possible to extend this model to any number of firms. In the symmetric case with constant costs and linear demand, when contract positions are not revealed, the contract cover will be 1 H, where H is the Herfindahl-Hirshman Index (HHI) expressed as a fraction. 4 A further complication is that in repeated games (and electricity markets are repeated at high frequency) contract positions can be used to support collusive behaviour (Green and Le Coq 2006). In merger analysis the problem of repeated games is even more complex, as the aim is not just to understand past price-setting behaviour, but to predict the future with and without the merger, to determine the merger s potential price-raising effect. One would expect contract cover to change and with it the incentive to bid above marginal cost. In the simple static Allaz and Vila model contract cover would fall from 1 H 0 to 1 H 1 where H 0 is the pre-merger HHI and H 1 the post-merger HHI. Thus if the pre-merger HHI were 1,667 and post-merger were 2,000, the predicted contract cover would fall from 83 percent to 80 percent. The effect on the predicted price-cost margin of ignoring the change in contact cover would be to understate the margin by 14 percent (Newbery 2006). 5 Sophisticated equilibrium bidding models have a mixed record in tracking actual bidding behaviour in real markets. Hortacsu and Puller (2006) found that the bids of smaller firms in the ERCOT wholesale market differed significantly from the theoretical benchmark of static profit-maximisation in their bidding model. For this and other more conceptual reasons (such as non-uniqueness) simpler models have often been preferred in merger analysis, of which the leading example is the Nash-Cournot model. 6 This workhorse of industrial organisation economics assumes that each supplier chooses its output level (in each hour) to maximise its profits given the residual demand it faces (total demand less net imports and the outputs of other suppliers) assuming that other suppliers are simultaneously and independently determining their output levels. The obvious objection to this assumption is that suppliers actually offer to supply amounts at various prices, and leave the market to determine how much of their offers to accept. One way of finessing this mismatch between the strategy choices assumed and observed is to suppose that suppliers offer amounts at prices (which might be variable costs or a mark-up on these to cover various fixed costs) up to a fixed amount that they determine. This maximum amount offered to the market would be the key short-run decision variable for strategic suppliers, and would be the total available capacity for the competitive fringe. Given the marginal cost function of each supplier, the strategies would 4 The HHI is normally measured as the sum of the squared market shares of the firms in the market, with a maximum value of 10,000 for a monopoly. Here the market shares are taken as a fraction, with the monopoly value =1, and for n equal-size firms H = 1/n. 5 The ratio of the price-cost margin ignoring changes in contract cover to the correctly calculated margin is 1 H 0 (1 H 0 ), so if H 0 = , the ratio is and the understatement is 14 percent. 6 A recent example of the use of such models in potential merger analysis is Moselle et al. (2006), undertaken for the Dutch Competition Authority, NMa. An earlier example applied to California is given in Borenstein et al. (1997).

6 Analytical Screens for Electricity Mergers 221 then be levels of maximum output to offer in each hour, and this would determine the market-clearing price. Contracting affects the equilibrium by shifting the residual demand schedule facing each strategic supplier, and can be either specified (based on observations) or determined as another strategic variable (the Allaz and Vila approach discussed above). The attraction of Nash-Cournot modelling is that it readily lends itself to merger analysis, and can also be used to define the relevant markets in the presence of transmission constraints, although a drawback is that it requires an equilibrium analysis of the bidding strategies for each firm in the market. All of these modelling approaches suffer from the additional disadvantage that they are not particularly transparent when properly calibrated to replicate the complexities of plants and markets, which limits their value in a regulatory proceeding or a court of law. The models generate point estimates of market prices and do not explain why a firm chooses a particular price or why a merger may lead the merged firm to choose a higher price. The answers to these questions lie deep within the model and do not emerge as part of its output. 3 An Alternative to FERC s Merger Screen: Competitive Residual Demand Analysis The U.S. Federal Energy Regulatory Commission (FERC) has adopted the Department of Justice (DOJ)/Federal Trade Commission (FTC) Horizontal Merger Guidelines (1992) (Guidelines) as the basic framework to evaluate the competitive effects of proposed mergers of electricity generating companies. 7 The Guidelines presume that a merger is unlikely to adversely affect competition in a relevant market if the post-merger HHI is less than 1,000, or if the post-merger index is between 1,000 and 1,800 and the increase in concentration from the merger is less than 100, or if the post-merger index is greater than 1,800 and the increase in concentration from the merger is less than The FERC requires applicants for merger authority to calculate market shares based on different measures of generation capacity and uses a delivered price test to determine the generators that compete in a relevant market. 9 If the market share analysis shows that a proposed merger does not exceed the Guidelines thresholds, and if there are no interventions that raise substantial competition concerns that cannot be resolved by the written record, the Commission does not look further at the effect of the merger on competition. 10 By employing the Guidelines market concentration screen, the Commission s position is that it will be able to identify proposed mergers that clearly will not harm competition. 11 However, given the very inelastic short-run demand for electricity aggravated by limited storage possibilities 7 US Federal Energy Regulatory Commission (1996, p. 3 and Appendix A, p. 59). 8 More specifically, according to the DOJ/FTC Merger Guidelines, mergers that fall under these thresholds are unlikely to have adverse competitive consequences and ordinarily require no further analysis. 9 US Federal Energy Regulatory Commission (2000, p. 35). Relevant product markets include, as a minimum, non-firm energy, firm energy, and long-term capacity. Ibid., p US Federal Energy Regulatory Commission (2000, p. 60). 11 See also US Federal Energy Regulatory Commission (2001, pp. 6 7).

7 222 R. Gilbert, D. Newbery and transmission constraints, this conclusion can be incorrect, and the FERC screen can permit mergers in electricity markets that risk higher prices. A merger also can fail the FERC screen even though it does not pose a risk of higher prices. To see why a merger may pass the FERC screen yet still pose a risk of higher prices, consider the following hypothetical. Firms A and B propose to merge. Firm A has 1,511 megawatts (MW) of uncommitted capacity. Firm B has 1,383 MW of uncommitted capacity. The units owned by A and B have relatively low incremental costs and would operate under most demand conditions. The electricity supply industry has a total uncommitted capacity of about 18,000 MW and is not concentrated. Figure 1 shows the industry incremental cost. The list of firms, their capacities and marginal costs are in Appendix A. (The firm capacities and marginal costs in Appendix A are based on industry data but are not intended to represent an actual electricity supply industry.) Nearly all units in Fig. 1 and Appendix A are assumed to have separate owners. The pre-merger HHI for the supply industry is only about 833. Post-merger, the HHI would be 964, which is low enough that the merger would not raise concerns about possible anticompetitive effects under the Guidelines or the FERC merger policy screen. Despite the fact that the proposed merger passes the FERC market concentration screen, there is a risk that the merger would result in higher prices. Figure 1 illustrates why. The figure shows the industry marginal cost curves for uncommitted capacity before and after the merger, along with market demand, which is assumed to be inelastic. If the industry acts competitively before the merger, the market price would be $62.5/megawatt-hour (MHr). This is the point at which the industry marginal cost curve intersects demand, shown as the vertical line at 12,000 MW. Suppose that after the merger the merged firm reduces its output by 1,500 MW. The reduction in output shifts the industry marginal cost curve to the left and if other firms continue to act Marginal Cost ($/MWh) Output (MW) Marginal Cost Demand 1,000 MW reduction Fig. 1 Marginal cost pre and post-merger and industry demand

8 Analytical Screens for Electricity Mergers 223 competitively, the new price that equates supply and demand is $68.8/MHr, an increase of 10.1 percent compared to the pre-merger price. Figure 1 shows that the merged firm could increase the wholesale price of electricity by a not insignificant amount, but the figure does not tell us if the price increase would be likely because it is profitable for the merged firm. To answer this question we propose a simulation method that we call Competitive Residual Demand (CRD) analysis. The CRD analysis examines the residual demand facing the merging firms pre- and post-merger and computes whether the merged firm has an incentive to reduce output and raise prices relative to pre-merger levels. The residual demand is the wholesale market demand less the aggregate uncommitted supply from all other firms, computed at each price chosen by the merging firms. As a further simplification, the market demand is assumed to be perfectly inelastic. The market demand is then independent of prices, but will differ at different points in time. 12 The aggregate supply of all firms depends on their conduct specifically whether they act competitively or reduce their outputs below competitive levels in order to benefit from higher prices. The CRD approach assumes that all other firms act competitively, in which case the supply of each non-merging firm is the output that equates the firm s marginal cost to the market price. The assumption of competitive conduct for all other firms is not unreasonable when other firms are small and are unlikely to have significant market power. An alternative approach would assume that other firms add a bid margin to their marginal costs. Because our focus is on the likely price increase from a merger, the assumption of a positive bid margin need not significantly change the predicted price increase because it would elevate both pre-merger and postmerger prices. Non-competitive bidding by all other firms would change the merger predictions relative to the assumption of perfectly competitive behaviour only if the departure from marginal cost pricing differed significantly before and after the merger. To the extent that bid margins are higher after the merger, the competitive residual demand analysis would underestimate the likely price increase from the merger. 3.1 CRD Results Under Moderate Demand Scenario The CRD analysis computes the market clearing price and the merged firm s profits under different demand assumptions for different levels of output by the merged firm. Given the industry marginal cost schedule in Fig. 1 and a merger of firms A and B, Table 1 shows the market-clearing price, the percentage increase in the price and the percentage increase in the merged firm s profit for different levels of output by the merged firm relative to no supply restriction. Table 1 assumes a market demand of 12,000 MW, which corresponds to a moderate level of demand for this market with no requirement to call on output from peaking units. 12 Baker and Bresnahan (1988) estimate the slope of the residual demand faced by firms selling differentiated products before and after a merger. Their focus is on estimation of supply and firm-specific demand functions with limited cost information. The CRD approach assumes that demand is fixed at any point in time and uses known industry marginal costs to estimate residual demands for the merging firms and to compute profit-maximizing prices before and after a merger.

9 224 R. Gilbert, D. Newbery Table 1 Market-clearing price and merged firm s profit for different output levels when market demand = 12,000 MW. (Capacity of the merged firm = 2,894 MW) Output reduction Market price Profit 82,990 76,888 66,630 63,922 57,236 54,157 42,777 37,442 25,218 12,311 % Profit increase 7.4% 19.7% 23.0% 31.0% 34.7% 48.5% 54.9% 69.6% 85.2% 100.0% % Price increase 0.0% 0.3% 1.2% 1.2% 3.5% 3.6% 3.9% 5.1% 6.2% 8.3% Table 2 Market-clearing price and merged firm s profit for different output levels when market demand = 17,000 MW. (Capacity of the merged firm = 2,894 MW) Output reduction Market price Profit %Profit increase 8.9% 20.1% 18.6% 34.1% 35.2% 29.0% 41.2% 46.7% 20.4% 5.7% 100.0% %Price increase 0.0% 7.6% 20.9% 23.1% 37.2% 40.0% 41.9% 59.8% 67.4% 78.6% 100.9%

10 Analytical Screens for Electricity Mergers 225 Table 1 shows that reductions in supply by the merged firm from its full capacity level of 2,894 MW would increase the market-clearing price relative to the perfectly competitive level but are not profitable. In this particular example the merged firm would have no incentive to increase prices by reducing its output. The FERC s merger policy screen correctly shows that the merger is not likely to give the firm an incentive to raise prices at this moderate level of demand. 3.2 CRD Results Under High Demand Scenario Other demand states lead to different conclusions for this hypothetical market. Consider the incentive of the merged firm to reduce output and increase price at a level of peak demand equal to 17,000 MW. Table 2 shows the profit of the merged firm for different output levels and the corresponding market prices for this level of market demand. At this higher level of demand the merged firm maximizes its profit when it reduces its output by 878 MW below its maximum capacity of 2,894 MW, corresponding to a capacity utilization of about 70 percent. The lower output results in a market-clearing price of $143.2/MWh when other firms act competitively, which is more than 50 percent higher than the market-clearing price when all firms act as perfect competitors. Table 2 shows that the merged firm has an incentive to raise price significantly above the perfectly competitive level for high levels of demand, but the analysis does not prove that the merger would raise price above the pre-merger level. To gain a better understanding of the risk of a price increase, the CRD analysis repeats the type of calculations shown in Table 2 separately for Firms A and B before the merger. These calculations are shown in Tables 3 and 4 below. Table 3 shows that before the merger Firm A maximizes its profit by withholding 484 MW of its total capacity of 1,511 MW when all other firms, including Firm B, act competitively. Table 4 shows that before the merger Firm B maximizes its profit by withholding 91 MW of its total capacity of 1,383 MW when all other firms, including Firm A act competitively. Before the merger, the two firms have unilateral incentives to withhold a total of 575 MW of their capacity when all other firms act competitively. After the merger, the merged firm has an incentive to withhold 878 MW of its total capacity of 2,894 MW. Comparing profit-maximizing levels of output before and after the merger, the merged firm has an incentive to reduce its output by 303 MW more than the combined reductions of the two firms before the merger. Before the merger, the market price under the assumption that firms other than A and B act competitively would be $125.4/MWh. This is the price at which total supply equals demand of 17,000 MW when Firms A and B reduce output by a total of 575 MW and can be verified from the capacity and marginal cost data in Appendix A. After the merger, the market price under the assumption that all other firms act competitively would be $143.2/MWh. This is the price at which total supply equals demand of 17,000 MW when the merged firm reduces output by 878 MW and again follows from the data in Appendix A. The CRD analysis suggests that the merger could increase prices by roughly 14 percent.

11 226 R. Gilbert, D. Newbery Table 3 Market-clearing price and Firm A s pre-merger profit for different output levels when market demand = 17,000 MW. (Capacity of Firm A = 1,511 MW) Output reduction Market price Profit %Profit increase 5.7% 7.8% 1.7% 10.5% 8.4% 6.2% 9.9% 10.8% 79.6% 100.0% %Price increase 7.6% 20.9% 23.1% 37.2% 40.0% 41.9% 59.8% 67.4% 78.6% 100.9% Table 4 Market-clearing price and Firm B s pre-merger profit for different output levels when market demand = 17,000 MW. (Capacity of Firm B = 1,383 MW) Output reduction Market price Profit %Profit increase 5.1% 4.6% 3.0% 4.6% 1.8% 15.0% 21.2% 23.1% 99.2% 100.0% %Price increase 7.6% 20.9% 23.1% 37.2% 40.0% 41.9% 59.8% 67.4% 78.6% 100.9%

12 Analytical Screens for Electricity Mergers CRD Results Under Low Demand Scenario The potential to increase price by reducing output is high in peak demand states because the industry marginal cost curve is steep when demand is high (see Fig. 1). One might expect that the risk of price increases from a merger is less in low demand states, but that is not generally correct. Although the industry marginal cost curve is flatter in low demand states, the incentive to withhold output can be high if a firm earns a low margin on its sales and has significant amounts of inframarginal capacity. Table 5 shows market-clearing prices and profits for different output levels of the merged firm when the level of demand is 7,500 MW, less than half of the peak level of demand. Table 5 shows that the merged firm has an incentive to reduce output by 1,354 MW and raise price by about 32 percent above the market-clearing price when all firms act as perfect competitors and market demand is 7,500 MW. As in the example corresponding to the high demand state, it does not immediately follow that the merger would lessen competition because the relevant question is whether output would be lower and price higher than before the merger. Tables 6 and 7 show the pre-merger profits of Firm A and B for different levels of output reduction relative to their total capacities when market demand is 7,500 MW. Table 6 shows that Firm A has no incentive to reduce its output pre-merger. Table 7 shows that Firm B has an incentive to reduce its output pre-merger by 859 MW relative to the perfectly competitive level. Thus, the CRD analysis suggests that the merged firm has an incentive to reduce its output by 495 MW more than the combined output reductions of Firms A and B pre-merger. In the low demand state, before the merger the market price under the assumption that firms other than A and B act competitively would be $41.3/MWh. This is the price at which total supply equals demand of 7,500 MW when Firms A and B reduce output by a total of 859 MW. After the merger, the market price under the assumption that all other firms act competitively would be $47.5/MWh in the low demand state. This is the price at which total supply equals demand of 7,500 MW when the merged firm reduces output by 1,354 MW. The CRD analysis suggests that the merger could increase prices by roughly 15 percent. Thus the CRD analysis of this hypothetical merger suggests potential concerns about price increases from the merger in low as well as high demand states, but not for an intermediate level of market demand. 3.4 Discussion The CRD approach is an analysis of a firm s unilateral incentive to reduce output and increase price. It ignores possible coordinated conduct by electricity generators. Antitrust agencies have embraced unilateral effects analysis for mergers because it is more firmly anchored to economic theory and empirical techniques are available to predict price increases from a merger, although the possibility of coordinated conduct should not be discounted (Baker 2003). A firm s unilateral incentive to reduce output and raise price depends on the slope of the industry marginal cost curve in

13 228 R. Gilbert, D. Newbery Table 5 Market-clearing price and merged firm s profit for different output levels when market demand = 7,500 MW. (Economic capacity of the merged firm = 2,292 MW) Output reduction Market price Profit %Profit increase 2.1% 61.5% 169.1% 213.0% 266.7% 255.3% 246.9% 139.3% 74.8% % 100.0% %Price increase 0.0% 3.9% 14.4% 23.8% 31.6% 32.7% 34.1% 34.1% 34.1% % 60.4% The merged firm s total economic capacity is only 2,292 MW when market demand is 7,500 MW rather than its total installed capacity of 2,894 MW, because its most expensive generating unit has a marginal cost that exceeds the market price when firms act competitively. As the price rises with market withholding, its profitable supply initially increases before finally falling below the initial competitive supply of 2,292 MW. Output reduction is measured relative to this initial competitive supply

14 Analytical Screens for Electricity Mergers 229 Table 6 Market-clearing price and Firm A s pre-merger profit for different output levels when market demand = 7,500 MW. (Economic capacity of the Firm A = 909 MW) Output reduction Market price Profit %Profit increase 8.4% 5.9% 2.4% 33.0% 48.2% 82.8% 100.0% %Price increase 0.0% 3.9% 14.4% 23.8% 31.6% 32.7% 34.1% Firm A s economic capacity is only 909 MW when market demand is 7,500 MW rather than its total installed capacity of 1,511 MW, because its most expensive generating unit has a marginal cost that exceeds the market price when firms act competitively. Output reduction is measured relative to the 909 MW of economic capacity when firms act as perfect competitors Table 7 Market-clearing price and Firm B s pre-merger profit for different output levels when market demand = 7,500 MW. (Economic capacity of Firm B = 1,383 MW) Output reduction Market Price Profit %Profit increase 7.3% 85.0% 117.0% 30.5% 76.2% 100.0% %Price increase 0.0% 3.9% 14.4% 23.8% 31.6% 32.7% the neighbourhood of market demand and on the firm s capacity and marginal cost. Given other firms pricing decisions, the incentive for an individual firm to reduce its output is unrelated to the distribution of capacity ownership by other firms, and thus is independent of the HHI except to the extent that the firm s own market share affects the concentration index. In a concentrated industry other firms may have incentives to reduce their outputs, and this can contribute to a higher equilibrium price. The CRD analysis focuses only on the output incentives of the merging firms and does not consider supply incentives for other firms or equilibrium supply decisions for the industry. The CRD analysis also can demonstrate examples of potential mergers that fail the FERC merger screen, but are not likely to raise prices. This could occur in a concentrated industry with a relatively flat supply curve, so that no firm has a unilateral profit incentive to raise price, although it should be noted that CRD analysis is a screen, not a predictor of the price consequences of a merger. 13 Computationally, the CRD analysis is considerably easier than a full equilibrium analysis. Moreover, it can be repeated to estimate price increases from a proposed merger in many different time periods corresponding to different demand states and in different geographic markets. The CRD prediction of the total expected price increase from a merger in a geographic market is the price increase estimated by the CRD 13 The FERC has acknowledged that the withholding of supply is unlikely to affect price if demand falls within a flat portion of the supply curve. See, e.g., US Federal Energy Regulatory Commission (2000, pp. 27, 63).

15 230 R. Gilbert, D. Newbery analysis for each state of demand weighted by the probability for each state and summed over all demand states. The CRD analysis has its deficiencies. As noted above, it is not an equilibrium analysis and may imprecisely estimate the likely price increase from a merger even if all firms other than the merging parties act as perfect competitors. For example, suppose that the merging firms A and B act as Nash-Cournot competitors in the electricity market and all other firms are perfectly competitive. CRD analysis would understate the profit-maximizing outputs of Firms A and B relative to their pre-merger Nash-Cournot equilibrium levels. Pre-merger, if Firm B acts as a perfect competitor as assumed in the CRD analysis, then the CRD calculation underestimates the pre-merger output of Firm A compared to its Nash-Cournot equilibrium level, and would do the same for the pre-merger output of Firm B. As a consequence the CRD analysis would overstate the pre-merger price and underestimate the likely price increase from the merger. The opposite bias could occur if the supplies of the firms in the market were strategic complements, corresponding to Nash-Bertrand competition with differentiated products. 14 In this case the CRD analysis would understate the pre-merger price and overstate the likely price increase from the merger. 15 The assumption that firms other than the merger parties behave as perfect competitors is not realistic in concentrated electricity markets. Nonetheless, this is not a major flaw for a CRD analysis of possible price increases from the merger if departures from perfectly competitive pricing do not differ significantly pre-merger and post-merger. For example, CRD analysis could underestimate the equilibrium price by $20/MWh, but if the underestimate were the same before and after the merger this would not affect a conclusion about the likely price increase from the merger. Another difficulty with CRD analysis is that it does not precisely identify the premerger price even under the assumption that other firms act as perfect competitors. In the examples corresponding to low and high states of demand, the pre-merger market clearing price when Firm A acts unilaterally to maximise its profit differs from the pre-merger market clearing price when Firm B acts unilaterally to maximise its profit. Table 6 shows a pre-merger price of $36.1/MWh for Firm A while Table 7 shows a pre-merger price of $41.3/MWh for Firm B. Table 5 shows a post-merger price of $47.5/MWh. The correct estimate of the price impact from the merger is not obvious. One could assume a price increase of 32 percent corresponding to the profit-maximizing pre-merger price of $36.1/MWh for Firm A or a price increase of 15 percent corresponding to the profit-maximizing pre-merger price of $41.3/MWh for Firm B. The proposed analysis does neither and instead calculates predicted price increases by estimating market clearing prices given the total profit-maximizing supply restrictions for Firms A and B pre-merger and post-merger under the assumption that all other firms act as prefect competitors. 14 Firms pricing decisions are strategic complements if an increase in the price chosen by one firm increases the optimal prices chosen by other firms. 15 Changes in forward contracting could counteract this bias. The optimal degree of contract cover is inversely related to market concentration and therefore likely would decrease after the merger. This would increase the estimated price increase from the merger. The examples of CRD analysis that we describe do not account for contract cover, although it is relatively straightforward to include forward contracting in the analysis.

16 Analytical Screens for Electricity Mergers Vertical Mergers Involving Gas and Electricity Gas is an important input into electricity generation and a final energy source that competes with electricity in the retail market. When global power generation orders peaked at over 180 GW in 2001, 150 GW were for gas turbines and more than half of all orders since 1990 have been for gas turbines (OECD 2003). Liberalisation (particularly when associated with privatisation and unbundling) opens the prospect of gas and electricity companies entering each other s markets for the final product (gas or electricity supply) and acquiring firms in the other fuel market. Such mergers can be viewed as vertical integration (upstream gas and downstream electricity) or as convergence mergers (extending the company s reach from one energy source into two and reducing the difference between gas and electricity suppliers). The opportunity has clearly been attractive. Hunger (2003) cites 22 gas/electricity mergers with an asset value greater than $500 million counted by the US Energy Information Agency between 1997 and Recent mega-mergers in the EU such as E.On-Ruhrgas in 2002 and the attempted takeover of Endesa by Gas Natural in 2005 have awakened concerns both about the ability of the European Commission to rule on such mergers and to prevent mergers that are likely to be anticompetitive. In the US, the merger of Enova and Pacific Enterprises in 1998 raised significant vertical concerns. Enova was a supplier of electricity and a net purchaser of natural gas in southern California. Pacific Enterprises was the largest supplier of intrastate natural gas transport and storage in the region. 16 While there is growing agreement about the appropriate techniques to employ for analysing horizontal electricity sector mergers, that is not true for cross-fuel mergers involving an input fuel gas into the electricity sector. The approach adopted by the FERC to analyse the competitive effect of vertical mergers that involve more than de minimis sales of inputs to electricity generation requires merger applicants to define the relevant upstream and downstream markets and measure concentration in those markets. The FERC recognizes that vertical mergers can have adverse competitive consequences by foreclosing access to inputs or raising the cost of inputs for rivals, facilitating coordination, or evading regulation, 17 but does not propose a specific methodology to assess these issues. The standard argument for believing that vertical mergers are benign is that by reducing the inefficiency of transactions between upstream and downstream operations they lower costs and hence lower prices in the final market. Further, if the upstream (gas) company has market power, and if gas is used in fixed proportions with other inputs into electricity supply, the upstream gas supplier can only extract monopoly profits once and has no additional market power through ownership of the downstream (electricity) company. Indeed, if the upstream market power involved 16 The U.S. Federal Energy Commission approved the merger with nondiscrimination conditions on gas sales. The U.S. Department of Justice required divestiture of about two-thirds of Enova s electricity generating capacity and imposed limits of the reacquisition of electricity generation (see U.S. Department of Justice Competitive Impact Statement, available at accessed September 11, 2006.) 17 US Federal Energy Regulatory Commission (2000, pp ).

17 232 R. Gilbert, D. Newbery mark-ups on marginal cost in selling to the downstream firm, to the extent that these were eliminated or reduced by the merger, electricity prices would fall, not rise. In contrast to a horizontal merger, vertical mergers do not remove competitors. However, the conditions under which these arguments can be rigorously established are stringent and, as with horizontal mergers, vertical mergers require careful scrutiny to assess their welfare effects. 18 Vertical integration may make it profitable for the upstream firm to raise downstream rivals costs (Salinger 1988; Ordover et al. 1990), although Perry (1989) argues that the damage is likely to be limited. This is not the place to rehearse the debates about the anti-competitive effects of vertical integration in general, but to see whether one should be particularly concerned with gas-electricity mergers. The first point to note is that electricity and gas markets are often highly concentrated and electricity and gas are actual and potential competitors in a broader market for energy services. In the medium run, the best-placed potential entrant into the electricity market may be a gas supplier, who has access to the fuel needed for combined cycle gas turbines (the natural choice for entrants), and who can (partially) hedge gas price risk by selling gas directly or as electricity. If that potential entrant merges with an electricity supplier, a major entry threat may be reduced. If the gas company has market power and takes over an electricity company, and if it raises the price of gas, then whenever gas is the marginal fuel in electricity and the marginal electricity plant is not owned by the gas supplier, the price of electricity likely will rise. This will increase the inframarginal profits of the plants now under the control of the vertically merged company. The electricity part of the merged firm will have increased profits, additional to the normal monopoly profits of the gas part of the merged firm (and which were presumably already being maximised). Thus although the merger offers no additional opportunities to raise gas prices, it may provide an additional incentive to do so. 19 The natural way to examine this incentive is similar to the horizontal electricity merger analysis set out in section 4. The competitive supply schedule for the electricity industry can be computed for the pre-merger price of gas, and then for successive increases in the price of gas sold by the upstream (gas) part of the merged firm, taking account of the ability of electricity companies to buy gas from other gas suppliers. If the gas market is imperfectly competitive, then one can assume that the pre-merger gas price-cost margin reflects the pre-merger market power of the gas company the (residual) elasticity of demand facing the gas company will be the pre-merger inverse Lerner Index. 20 That should allow the impact of raising the gas price on gas profits to be computed (they should fall slightly), while the impact of raising gas prices on the merged firm s electricity profits can also be computed (and these should be positive whenever gas is at the margin and the marginal electricity plant is not owned by the gas supplier). 18 See e.g. Church (2004). 19 As noted below, a higher gas price lowers the firm s profits from its gas operations, but this is a secondorder effect for a small deviation in the price of gas from its stand-alone profit maximizing level. 20 The Lerner Index is (price-mc)/price, and is equal to the inverse of the residual demand elasticity facing the firm in a static non-collusive equilibrium.

18 Analytical Screens for Electricity Mergers Extra profit Merit order Demand 25 1 SRMC Euros/MWh coal gas post-merger AVC pre-merger AVC 8 12 coal gas 5 5 Plant owned by merged company MW cumulative Fig. 2 Effect on company profits of raising the market price of gas by 5 percent Figure 2 shows the application of this technique to a market in which company A (the vertically merged gas-electricity company) owns plants 5, 4, 7 and 10 (the plant numbers are indicated above the steps in the figure). The base-load plants 11, 5, 13, 4, 2 and 3 are coal-fired; the rest, except for coal-fired plant 12, are gas-fired. Before the merger the average variable costs (i.e., the short-run marginal costs) are shown by the heavy stepped line, with price set by the marginal (gas-fired) plant 1, not owned by firm A. After the merger company A now internalises the price of gas as the true marginal cost, and thus SRMC falls (by an assumed 10 percent) for gas-fired plants 7 and 10; but if the price of gas sold to other power companies rises by 5 percent, then so does the SRMC from other gas-fired plant (but not from coalfired plant 12). The marginal (gas-fired) plant 1 continues to set the price, and A s extra electricity profit from increasing the price of gas is shown by the (arrowed) rectangles between the old price line and the new price line. The extra profit while plant 1 sets the price is company A s total capacity times the change in the electricity price shown. If other firms gas-fired plant continued to set the price as demand varied, it would be straightforward to quantify the incentives to raise the gas price, as follows: Suppose that demand for A s gas by the other electricity generators before the merger is G(g) where g is the sales price of gas to the downstream electricity supply industry. 21 In order to raise the price of gas, A reduces sales by G, and if no other gas company increases output this raises the gas price by g = g G/(εG), where ε is the firm-specific residual elasticity of demand for gas (as a positive number). The change in profits 21 This assumes that gas suppliers price discriminate for sales to electricity generators and other sectors; otherwise the analysis would need to deal with the loss in profits elsewhere from any price rise.

19 234 R. Gilbert, D. Newbery from gas sales caused by the reduction in A s gas supply will be G g + (g c) G, where c is the cost of the gas. This can be written as (1 εl)g g, where L is the Lerner Index for A s sales of gas to these firms. Pre-merger, this is zero for a small change in the gas price from the conditions of profit maximisation (and L = (g c)/g = 1/ε). If the marginal plant is owned by another company and is gas-fired with heat rate h m, then the change in the electricity price is p = h m g. If company A has a fraction α of total electricity sales Q and the gas consumption of the remaining generators is G = (1 α)βh a Q (where h a is the average heat rate of the gas plants and β is the share of gas in their electricity capacity), then the change in gas profits is (1 α)βh a Q (1 εl) g. The extra profit to company A from electricity sales is αq p = αqh m g. The change in profits of merged company A is then =[αh m + (1 α)βh a (1 εl)]q g (1) on the assumption that the demand for electricity is insensitive to its price (which has increased by p). In the pre-merger equilibrium, L = 1/ε, but after increasing the gas price L > 1/ε, so the second term is initially zero and decreasing but effectively second order, while the first term is positive and first order. 22 The incentive to raise the price of gas (assuming that this is possible) therefore increases with the merger, and the post-merger mark-up for A s gas sales will be maximised for a value of L that equates (1) to zero (i.e. when d /dg = 0): εl = 1 + αh m (1 α)βh a (2) If the elasticity of residual gas demand, ε, is constant, then the pre-merger gas mark-up would be L = 1/ε, so the ratio of the post- to pre-merger gas mark-up is then the right-hand side of (2). In general, as demand varies the marginal plant is unlikely to always be another firm s gas-fired plant (and the merit order may change here if A bids SRMC, plant 10 would move down to just above plant 7, but could stay where it is if A prices up to the bids of other firms). If another firm s gas plant is only at the margin a fraction γ of the time, then A s electricity expected profit will only rise by γαqh m g, and the ratio of the expected post- to pre-merger gas mark-up will be αγ h m (1 α)βh a (3) 22 Local concavity of the profit function, which is required for profit-maximization, implies that L(p)ε(p) is increasing in p. Hence, L > 1/ε after a small increase in the gas price above the pre-merger profit-maximizing level. 23 It is intuitive that the profit-maximizing gas mark-up for the merged company could be very high if the share of electricity that is not owned by the merged company is very small, corresponding to α close to one. But in that case the fraction of the time γ that another firm s gas-fired electricity plant would be at the margin also would be very small.