Slides Montero

The role of forward contracting in the competition of energy markets Juan-Pablo Montero PUC-Economics Universitat de ...

0 downloads 192 Views 263KB Size
.

The role of forward contracting in the competition of energy markets Juan-Pablo Montero PUC-Economics Universitat de Barcelona, IEB Energy Symposium January 29, 2013

Motivation

• to what extent does the possibility of contracting forward affect competition in oligopolistic (energy) markets?

• forward contracting (sell part of own’s production in advance) can be used as a credible commitment device to change competition in the spot market

• we will exclusively focus on the strategic reasons associated to forward contracting and NOT on hedging reasons coming from demand/supply uncertainty (which are also present in perfectly competitive markets)

• most models assume complete information and no uncertainty, but not all

More motivation and organization of the talk

• important in commodity markets with few players • forward markets are generally seen to be pro-competitive (Allaz and Vila, JET 1993)

• based on this argument, forward trading has been advanced as an important mechanism to mitigate eventual market power problems in electricity markets (e.g., Joskow, Enery J 2008)

• where is this pro-competitive result coming from? • what does the evidence tell us?

• how robust and general is the pro-competitive result to different assumptions (e.g., repeated interaction)?

• some open research questions — empirical test of firms’ strategic incentives to sell forwards — how intensively do firms compite for contracts? Does it matter whether is through centralized mechanisms like auctions or decentralized bilateral trading? Preliminary evidence from long-term power contracts in Chile suggests that it does.

Building block: The Allaz and Vila (JET 1993) pro-competitive result • Cournot market preceded by a forward market • consider two periods  = 0 1 and two firms  and  • with constant marginal costs  • forward sales at  = 0 are denoted by   (think of contracts for differences where positions are close at the spot price;   0 denotes a short position while   0 a long position) • production at  = 1 are denoted by   • inverse demand in the spot market is given by  =  − ( +  )

Subgame at  = 1 given forward positions (  )

• Spot profits are:   = ( +  )( − ) −    0 makes marginal revenue to jump up

 −  + 2 −    =⇒  =   3  + 2 −  −   =⇒  =    3

Subgame at  = 0 • forward sales chosen: (  ) = [(  )(  ) − (  )] + [ − (  )] | {z } Cournot profit

|

{z } arbitrage profit

where  is the forward price. But arbitrage profit=0 (financial intermmediaries are Bertrand players), so in equilibrium:  =  = ( − )5  0 • firms end up selling forward contracts, i.e., taking short positions • firms face a prisoners’ dilemma: since each wants to get a bigger share of the oligopoly rents –the Stackelberg share– that makes the market to become more competitive as a result (i.e., greater production)

Extension I: what if the are M  2 firms in the market?

• since oligopoly rents in the spot market decrease in  , one could conjecture that firms would have less incentive to sell forwards

• the opposite is true, competition is fiercer (the incentive to get the Stackleberg share of a smaller pie increases).

• it is not difficult to show that the amount of contracting is increasing in  (and total in the limit)  1 =1−  

Extension II: what if there are N  1 forward openings before production? back to  = 2; the equilibrium solution is given by µ



(1)

µ



(2)

− 3 1− () = 2 3 + 2 − 1 1− () = 2 3 + 2 () =  +

− 3 + 2

(3)

P

where  = 0=−+1  is firm ’s cummulative position at the opening of the spot market/production As  −→ ∞, the equilibrium goes competitive, i.e.,  = , and all production is sold forward.

Predictions of Allaz&Vila 1. regardless of whether contracting is endogenous or exogenous to firms, the larger the contract coverage (short positions) the more competitive the spot market is 2. firms have strategic incentives to sell forwards; and possibly a large amount 3. and the amount of contracting is increasing in the number of firms • most welcome predictions for market performance! • (shortly we will see how more recent theories may depart from some of these predictions)

What does the evidence tell us?

• mixed results • Wolfram (AER 1998) for the the UK power pool in the early 90s: — found little evidence of prediction 1 despite firms were highly (but exogenously) contracted; she looked at whether spot prices increased as older contracts were expiring. — as older contracts expire, firms didn’t replace them with new ones, which goes against prediction 2 (went down from 75% coverage to 50%) — are these contracts supply (i.e., generators) driven or demand driven (e.g., retail companies)? it matters for A&V’s second prediction

• Also for the UK power pool: Green (JIE 1999) disagrees with Wolfram in that the large contract coverage in the early 90s was responsabile for the lower than Cournot prices (no empirical test though)

• Sweeting (EJ 2007), also for the UK pool, finds evidence of collusion for the late 90s in that prices were higher than what predicted by static models with a large contract coverage (above 90%).

• Wolak (IEJ 2000, JAE 2007) finds significant effect of forward contracting in the Australian market together with a large contract coverage — support for prediction 1 but 2 and 3 were not tested — again, what is driving contracting?

• Fabra and Toro (IJIO 2005) found evidence of collusion and price wars in the Spanish power market — the role of contracts is less clear there; they say that they probably make collusion easier — contracts were exogenously imposed upon firms (see also Fabra and de Frutos, EER 2011)

• Bushnell et al (AER 20008) found strong effect of forward contracting (mostly in the form of vertical arrangements between generators and distributors) for three power markets in the US: California, New England and PJM (Penn., New Jersey, Maryland, DC..). — they found that Cournot behavior, corrected by contract coverage, is a good predictor; why then firms don’t sell more forwards?

• van Eijkel and Moraga-Gonzalez (2012) use data from the Dutch natural gas wholesale market to specifically test predictions 2 and 3 (first attempt I’m aware of), for which they find support, but there are some problems with the empirical test: — forward positions are not entirely observable (an opacity parameter is estimated) — what if forward trading is demand driven (by industrial buyers or power retailers)? pred 1 will still work

• Experiments — Le Coq and Orzen (JEBO 2006) find support for pred 1 but not as strong — Ferreira et al (2010) find support for pred 1 even as players are engaged in repeated interaction

— Schubert (2012) finds that allowing firms to sell forward does indeed facilitate collusion

• In sum, predictions 2 and 3 has not been tested, and if anything, the empirical evidence available points otherwise

Some figures from Bushnell et al (AER 2008)

Figure 1:

Figure 2:

Figure 3:

Figure 4:

Figure 5:

Relaxing some of A&V assumptions

• demand is uncertain: Allaz (IJIO 1992) • forward positions are not observable: Ferreira (TTE 2006), van Eijkel and Moraga-Gonzalez (2012)

• firms compete in prices as opposed to quantities: Mahenc and Salanie (JET 2004)

• there is repeated interaction: Liski and Montero (JET 2006) • firms are selling a exhaustible commodity (e.g., energy in hydro-power systems, storable pollution permits): Liski and Montero (2013)

• heterogeneity in costs, stock endowments and/or capacity constraints: de Frutos and Fabra (EER 2012), Liski and Montero (2013), Bushnell (JRE 2007).

• but what if cost asymmetries are small but uncertain? (Montero, in progress)

Demand uncertainty

• Allaz (IJIO 1992) considers  =  +  − , where [] = 0 and [2] = 2

• risk-neutral producers contract as before • risk-averse producers contract more (hedging rationale)

No observability of forward positions

• how difficult is to observe firms position? • if the they cannot be observed, according to Hughes and Kao (IJIO 1997), we are back to Cournot (forwards don’t work as a commitment device)

• see also van Eijkel and Moraga-Gonzalez (2012) • firm ’s best spot response ∗( ) =

 +  −  2

where i anticipates that  follows the same reaction  + 0 −   = 2 so ’s equilibrium quantity is given by  =

 + 23  + 13 0 − 0

2 where 0 is ’s conjecture of ’s position. • in the first stage firms decide  and 

 − 0  = =0 2 • Ferreira (TTE 2006) challenges this result because on and off the equilibrium price  = 

• he allows  6=  and argues that in equilibrium  and  large enough that  =  = , back to Allaz and Vila with many forward openings!

• the problem with such equilibrium is that not selling is weakly dominant

Prices vs quantities vs both

• In Mahenc and Salanié (JET 2004) firms compete in prices (with some product differentiation) instead of quantities (still in a static context); less relevant for many commodity markets like power pools

• they obtain a "puppy dog" result (Allaz and Vila is a "top dog" result) that reverts prediction 2 — firms go long in the forward market = underinvestment — =⇒ softer competition in the spot market — anti-competitive

• there are indeed some markets (e.g., Pancafe cartel) in which have been found colluding using long positions (but in a dynamic context)

• In Green (JIE 1999), firms bid supply functions () into the spot market as opposed to  or . — in a linear world the supply-function equilibrium strategy is independent of rival’s contract coverage=⇒contracts play no role

Repeated interaction and collusive behavior

• In Liski and Montero (JET 2006) firms interact repeatedly in both the forward and spot markets

• they show that forward contracting can help firms to sustain collusion that otherwise would have not been possible; more precisely, it lowers the critical discount factor for maximal collusion

• results do not depend on the form of competition (whether is in prices or quantities)

• why? the interaction of two effects: 1. pro-competitive effect: forward contracting makes future collusion less attractive by reducing the amount of non-contracted sales

2. pro-collusive effect: deviation from the collusive aggrement is less attractive because (i) the available demand at the period of deviation is (non-contracted consumers) and (ii) the existence of harsher punishment (reversion to the Allaz & Vila one-shot equilibrium)

• the amount of contracting is endogenously determined, so firms can always make the second effect dominate the first one

• however, if firms are exogenously required (by some regulatory authority, for example) to maintain a substantial amount of forward sales and for many years, the pro-competitive effect can dominate the pro-collusion effect making collusion harder to sustain (but not in the UK according to Sweeting or Spain according to Fabra and Toro)

• a technical annotation: the amount of (collusive) contracting follows the exact opposite logic of the static models of Fudenberg and Tirole (AER 1984): — firms sell forwards (go short) when competing in prices — firms more likely to buy forwards (go long) when competing in quantities

Contracting for an exhaustible-resource

• In Liski and Montero (2013) we allow oligopolists selling an exhaustible resource to sell forward contracts

• relevant for — power markets with (large) hydro producers (Kauppi and Liski, 2008) — pollution permit markets in which permits are storable (Liski and Montero, Econ J 2011)

• one may conjecture that the Allaz and Vila mechanism does not work here because the total supply is fixed

• but it does: firms have incentives to use forward contracts to move production originally allocated to the future to the present, which is most profitable (correction in Hotelling rule: in oligopoly prices grow less than the interest rate)

Heterogeneity in costs, endowments and/or capacity no time!

Equilibrium price path for an exhaustible resource (spot) market

p ----- perfect competition ----- oligopoly

p dp / dt dMR/dt r; r p MR

dp / dt r p

0

T

*

Figure 6:

T

t

Small cost asymmetries but uncertain

• In Montero (in progress) we mantain all assumptions except that there are (small) cost asymmetries:  ∈ { }, where ∆ =  −  • (you may instead want to assume a little bit of private info  ∈ [ ] according to  ())

• firms learn their types only right before production (not at the time when forwards are signed a year ago; Pr( = ) = 1 − Pr( = )) • A&V predictions 2 and 3 revert: — firms don’t end up fully contracted as we increase the number of forward openings (there is a limit to contracting that depends on ∆)

— there is a number of firms, given one round of contracting, above which contracting drops to zero for a given ∆ — collorary of all this  =0 lim →∞  ∆→0

• Why? Cost differences nullify the Stackelberg incentives to sign forwards (for both high and low types)

Competition for long-term contracts: auctions vs bilateral trading

If we run  =  0 +  1 +  2 × _   + 3   +  4 +  we obtain that  1 = 103∗∗∗ and  2 = 045∗∗∗

Prices of new long‐term contracts (large buyers) 400,0 350,0

•price of first delivery vs MgC •US$/MWh •6‐month windows

300,0 250,0 200,0 150,0 100,0 50,0 0,0 0

5 10 15 20 O02                           A05                         O07                          A10

Figure 7:

Clearing prices in distribution contrats 250

200 2006‐1 A 2006‐1 B

h W150 M / $ S U  E M100 P

2006‐2 A 2006‐2 B 2008‐1 A 2008‐1 B 2010‐1 CMg

50

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 O00  O01    O02   O03   O04    O05   O06   O07   O08   O09   O10 

Figure 8:

350,0

300,0

250,0

200,0

150,0

100,0

50,0

0,0 12

14

16 PME_1obs Contrato

18

20

22

PME_Adj Lic

Figure 9: Contract prices in retail auctions (red) vs prices from bilateral trading (blue) with large buyers

.