VanKoten Bidding Behavior JRE 20 2008.12.12

8/3/2019 VanKoten Bidding Behavior JRE 20 2008.12.12

1/48

1

1. Introduction 3

2. Literature 6

3. The model 83.1 Assumptions 8

3.2 The second-price auction 113.3 The first price auction 163.4 Alternative models 20

4. The procurement auctions in New Jersey and Illinois. 23

5. Conclusion 26

6. Appendix 29

7. Parameter overview 38

8. Literature 39

XX. Figures and tables FOR QUICK NAVIGATION ONLY DELETE BEFORE RESUBMISSION 43


2/48

2

Effects of vertical integrations on auction outcomes in the EU and US

electricity markets.*

Silvester van Koten**CERGE-EI

UNDER REVISION VERSION 12.12.2008

Abstract

With the deregulatory reforms in the electricity industry, production stages have been split up and are

performed typically by different companies that compete for inputs and/or customers in decentralized

markets. Often goods are sold by auction in such markets. As the extant EU and USA regulatory framework

allow integrated electricity holding companies to have ownership of firms active in generation, distribution

and transmission, these holding companies often own both the seller and one of the buyers in such

decentralized markets. A holding company that owns both a buyer (called the integrated buyer) and a seller

in an auction has distorted bidding incentives. Specifically, the holding company will make the integrated

buyer bid more aggressively to increase the auction revenue. As a result, the integrated buyer is more likely

to win the auction and the good is sold for a higher price. However, since the auction is now inefficient,

efficiency is decreased. Moreover, independent companies are less likely to win the auction, and, in any

case, pay a higher price.

Keywords: asymmetric auctions, bidding behavior, electricity markets, regulation, vertical integration.

JEL classification code: L43, L51, L94, L98, R39.

*I am grateful to Levent elik, Libor Duek, Dirk Engelmann, Peter Katuk, Jan Kmenta, Thomas-Olivier Lautier, Avner

Shaked, Sergey Slobodyan, the participants of the EEA-ESEM 2008 conference in Milano, the participants of the Econometric

Society Winter Meetings 2008 in Cambridge, and two anonymous referees of the Journal of Regulatory Economics for theirhelpful comments. Special thanks to Andreas Ortmann. I thank the REFGOV Integrated project funded by the 6th European

research framework programme, CIT3-513420, for financial support.**

Email:[email protected], [email protected].

CERGE-EI is a joint workplace of the Center for Economic Research and Graduate Education, Charles University in Prague,and the Economics Institute of Academy of Sciences of the Czech Republic.


3/48

3

1. Introduction

Liberalization

The electricity supply industries in the USA and the EU are being reformed. In the past the production

stages in the electricity supply industry, most notably generation, transmission, and distribution,1

used to be

performed by Vertically Integrated Utilities (VIUs) that often were national or local monopoly producers of

electricity. Now, production stages have been split up and are performed typically by different

companies that compete for inputs and/or customers in decentralized markets.

In many such markets, competition is organized by running auctions, as auctions have, in theory,

features that have been judged highly desirable for electricity markets such as non-discrimination (the

highest bidder wins regardless his identity), efficiency (the bidder with the highest value makes the highest

bid and thus wins), and selling at efficient prices (prices that reflect the scarcity of a good).2

For example, in

the European Unions, the right for generators or suppliers to use capacity on cross-border transmission lines

is often allocated by explicit auction.3

In the USA, contracts for electricity supply by generators have been

awarded by procurement auctions, such as in New Jersey (Loxley and Salant, 2004; Reitzes, 2007) and

Illinois (Illinois Commerce Commission 2006; Negrete-Pincetic and Gross, 2007).

For the liberalization of the electricity supply industries to be successful, it is essential that such

decentralized markets are competitive. However, national markets are often dominated by large holding

companies, often the incumbent VIUs that own companies that are involved in different steps of the

electricity production process. As a result, in a market sometimes the seller and one of the buyers are owned

by the same holding company; I will refer to such buyers and the integrated buyer.

For example, in Illinois and New Jersey, distribution firms awarded contracts for electricity delivery in

procurement auctions to generator companies. Some of these generators where integrated buyers; they were

owned by a holding company that also owns the seller of the contracts.4

In the EU, the capacity on cross-

border transmission lines (also called interconnectors) is mostly sold by auction to generators (ETSO, 2006).

In many instances, one of the generators buying capacity is an integrated buyer; he is owned by a holding

1I focus on the three main production steps of generation, transmission, and distribution: generation is the production of

electricity in power plants, transmission is the transport of electricity over long distances, and distribution is the transport of

electricity over short distances, mostly to the final consumer.2

See for example Consentec (2004). However Joskow and Tirole (2003) argue that auctions result in prices that undervalue the

benefits of transmission line.3

In 2007 explicit auctions were used to allocate capacity for international transmission lines at 21 border crossings (Commission

of the European Communities, 2008b, p.30)4 See, for the case of New Jersey, http://bgs-auction.com, and, for the case of Illinois, Illinois Commerce Commission (2006, p.8).


4/48

4

company that also owns the interconnector. For example in 2006 in 12 of the 27 EU member states, VIUs

that were involved in generation and/or distribution also owned the transmission and interconnector

networks.5

The typical pattern in such EU states is that a large dominant electricity generator, in which the

state has a majority stake, fully owns the transmission networks(see Commission of the European

Communities, 10.01.2007).

Legal unbundling

The holding company could have incentives to instruct the integrated seller to stifle competition by

selling only to the integrated buyer. Regulation in EU and USA aims to prevent integrated sellers from

discriminating against independent entrants by favoring integrated buyers. EU laws therefore mandate that

the integrated seller must be legally unbundled from the holding company (Directive 2003/54/EC and

Regulation 1228/2003).6

While the seller may still be fully owned by the holding company, the seller must

be a legally independent company with an autonomous management, and the holding company is not

allowed to give day-to-day instructions to the seller. Legal unbundling implies that the holding company

will not be able to make the seller discriminate against independent buyers in favor of the integrated buyer.

Auctions organized by such an integrated, but legally unbundled, seller are non-discriminatory in the sense

that the highest bidder wins, regardless of the identity or affiliation of the buyer.

USA laws mandate a comparable form of separation called functional unbundling that should

guarantee such non-discriminatory outcomes in auctions (FERC Order 888, P.21552). In the procurement

auctions in New Jersey and Illinois distributors selling electricity delivery contracts were not allowed to

own generators that could participate in the auction (Loxley and Salant, 2004; Illinois Commerce

Commission 2006; Negrete-Pincetic and Gross, 2007). However, it was allowed for distributors to be part of

a holding company that owned distributors and generators. This liberty did not go wasted; all four

distributors in New Jersey and both two distributors in Illinois were part of a holding company that also

owned a generator that participated in the auction.

5 VIUs own transmission networks, including the interconnectors, in the following countries: Austria, Belgium, Bulgaria, Cyprus,

Germany, Denmark, Estonia, France, Greece, Hungary, Ireland, and Luxembourg (Commission of the European Communities,2008b, p.38-39).6

Recently the European Commission has proposed new laws with a stricter requirements on unbundling. However, also in these

new laws VIUs would still be allowed to own generation and network activities, provided the network activities are legally

unbundled and operated by an independent System Operator (Commission of the European Communities, 19.9.2007, p.5).Access to the Network for Cross-Border Exchanges in Electricity (OJ 2003 L 176/1).


5/48

5

While legal or functional unbundling might accomplish that the seller behave non-discriminatory,7

the

holding company is not restricted in changing the bidding behavior of its integrated buyer. I will argue that

the ownership of the seller gives the holding company incentives to make its integrated buyer bid more

aggressively in auctions. The intuition is that the price the integrated buyer pays for the good is not a net

cost to the holding company as (a part of) the payment returns to the holding company through its

ownership of the seller. The holding company will thus instruct the integrated buyer to adapt his bidding

behavior to consider the lower cost of bidding and bid more aggressively.

The holding company will order the integrated buyer to bid more aggressively only when the integrated

seller can keep a part of the profit of the auction and send it on to the holding company. This is the case

when the holding company is residual claimant of the income of the integrated seller; for example, the new

EU regulations allow the building of for-profit interconnectors (cross-border transmission lines), where the

owner can keep the full profits generated by auctions.8

Even when the income of the seller is regulated, the

seller is often allowed to keep at least a part of the increased profit due to cost reductions. Often regulation

allows transmission and distribution owners, in order to provide incentives to innovate and implement

efficiency gains, to keep a part of the profits. For example, if the regulated price for which a distributor

delivers electricity is fixed, then the distributor keeps the full gain of discounts he manages to obtain in the

buying process from electricity generators. In the paper I will assume that the seller can keep a certain

proportion of the profits. I refer to this portion as the (effective) ownership share and denote this by the

symbol K .9

The main driving question in this paper is if legal unbundling is a sufficient measure to assure a

competitive market where allocations and prices are non-discriminatory and efficient. This is an important

question; if legal unbundling puts independent buyers in a disadvantaged positionthen this makes it less

attractive for new, independent entrants to enter the energy market. This is highly relevant for EU electricity

generation markets: as national electricity generation markets in the EU are very concentrated,10

they need

7 There is evidence that legal separation is not a sufficient measure to guarantee non-discriminatory behavior of VIUs. For

example, in the EU the European Commission Competition DG (6.02.2006, p.144-148) found several concrete examples of

legally unbundled VIUs that curb competition through their combined ownership of generation and transmission or distribution

networks.8 While no merchant line has been built yet, it seems likely they will be built in the future; beginning 2007 the European

Commission had received two announcements of plans to built a merchant line (Commission of the European Communities,2008b, part 2, p.117)9

It is understood that the effective ownership share might be smaller than the stakes a buyer has in the seller due to regulation

or profit sharing with other parties.10 Efficient and non-discriminating outcomes in interconnector capacity auctions are indeed main objectives of the EuropeanCommission, especially as interconnection is seen as a means to increase competition in the highly concentrated markets for


6/48

6

to attract new entrants to make the liberalization reforms successful.So far legal unbundling has been

regarded as a sufficient measure in the EU and the US. However, up till now the effect of integrated

ownership on bidding behavior and auction outcomes in electricity markets has not been studied. I formulate

my research question as follows:

What is the effect of a buyer having an ownership share in the seller on auction outcomes under legal

unbundling? Specifically, are the auction outcomes still efficient and non-discriminating?

To answer this question I choose to model a very simple set-up with two bidders that have private values

that are independently and uniformly distributed.11

I also assume that the good on sale is sold in one piece,

and not in many divisible units. This simplification allows me to derive explicit solutions that enable

efficiency comparisons. The results from this simplified model give a suggestive answer to the effects of

unified ownership of buyer and seller in auctions.

The remainder of this paper is organized as follows. In the next section I will describe the setup of my

model, and discuss its connection to the literature on legal separation and toehold auctions. Then I analyze

the first- and second prize formats of the main model and the efficiency implications. To show the limits

and robustness of the effect of unified ownership, I also present models that employ the same setting, but

under different assumptions. I then present empirical data on the procurement auctions in Illinois and New

Jersey where the one of the effects predicted in the model, discrimination of independent buyers, seems to

have occurred.

2. Literature

On legal unbundling

The effect of legal separation has been studied in three earlier papers : Hffler and Kranz (2007a), Hffler

and Kranz (2007b), and Cremer, Crmer and De Donder (2006). Hffler and Kranz (2007a) claim that legal

unbundling can have superior qualities over ownership unbundling. Hffler and Kranz (2007a) model

electricity generation by enabling foreign generators to access to such markets (Consentec, 2004, p.I). National markets for

generation in the EU are indeed highly concentrated; in 2006, out of 20 EU member states, 7 were highly concentrated (HHIbetween 1800 and 5000), and 8, among which Belgium and France, were very highly concentrated (HHI above 5000)

(Commission of the European Communities, 2008b, p.11).11

The bidders might have in addition to their private value a publicly known value component that is identical (common) for both

bidders. As long this common value component is identical and publicly known, such a value component does not affect theanalysis.


7/48

7

competition between generators that buy transmission capacity for a fixed, regulated rate from a

transmission company to transport their electricity to consumers. The capacity on the transmission network

is unlimited in the relevant range. The transmission company is owned by one of the generators. Hffler and

Kranz (2007a) compare market outcomes for legal unbundling, full integration (where the generator can

give the transmission company instructions), and ownership unbundling. They show that the output of

generators is weakly higher under legal unbundling than under full integration or ownership unbundling,

and include an example of duopolistic price competition where the output is strictly higher. Hffler and

Kranz (2007b) extend the results of their earlier paper (2007a) by considering partial ownership and

imperfect legal unbundling.

My model resembles that of Hffler and Kranz (2007a); the regulated transmission company in their

model is an integrated seller, and the generator that owns the transmission company is the integrated buyer.

The main difference is that Hffler and Kranz (2007a) assume that the transmission company has an

unlimited capacity and has a vested interest to sell as much as possible of the capacity. The auctions I study

have a limited capacity (such as a limited supply of interconnection capacity or a limited supply of

electricity contracts in procurement auctions). In this setting, my model leads to conclusions opposite to

those of Hffler and Kranz (2007a): auction outcomes under legal unbundling are worse in terms of

competition and efficiency than under ownership unbundling.

Cremer et al. (2006) study the effects of legal unbundling of the buyer: in their model a downstream

firm (a buyer) is restricted to maximize his own (buyer) profit. This is different from my model where the

seller is the one who is legally unbundled and thus restricted to maximize his own profit (the auction

revenue), while the buyer is the one who can be instructed by the holding company to behave in ways that

do not maximize the buyer profit (but the total profit of the holding company).12

On toehold auctions

In the model setup, it will become clear that auctions with an integrated seller and an integrated buyer

are mathematically identical with so-called toehold auctions. Toehold auctions have been analyzed mostly

in the context of financial takeovers, where two buyers compete to buy a company and one or both buyers

already own, by holding shares, a fraction of the company (Bulow, Huang and Klemperer, 1999; Burkart,

1995; Ettinger, 2002). The fraction of the company owned by the potential buyer(s) is called a toehold.

12Hffler and Kranz (2007a) call the form of separation where the buyer is unbundled: reverse unbundling. I analyze a regime

that combines the regimes of legal unbundling andreverse unbundling in VanKoten (2007), and show that auction outcomesunder such a regime are still inferior to auction outcomes under ownership unbundling.


8/48

8

Burkart (1995) analyzed second-price private value toehold auction and finds that the bidding function

of the bidder with a toehold becomes more aggressive for larger toeholds, while the bidding function of the

bidder without a toehold is unaffected. Ettinger (2002) compares first-price and second-price auctions with

symmetrical toeholds and notes that, for strictly positive toeholds, the revenue equivalence theorem doesnt

hold. Bulow et al. (1999) analyze common value toehold auctions, where both bidders can have a toehold

(and at least one bidder a strictly positive toehold) and show that the bidder with a larger toehold has a much

larger probability of winning the auction. Bulow et al. (1999) also show that the winning price is

dramatically affected by the toeholds.

As Burkart (1995) uses very general assumptions, he cannot make an analysis of the effect of a toehold

on efficiency, I use his results to determine, under more restrictive assumptions, the effects on competition

and efficiency for second-price auctions. I then generalize this model for an arbitrary number of bidders. I

further use models of Ettinger (2002) and Bulow, Huang and Klemperer, 1999, to asses the effects under

slightly different assumptions. As first-price toehold auctions have not been analyzed before, I present a

general result for first-price auctions with two bidders, one of which an integrated buyer that fully owns the

integrated seller. Under more restrictive assumptions, I numerically solve such first-price auctions with

partial ownership, and I show that the revenue equivalence theorem doesnt hold in such auctions. As a

default case I also present a model without uncertainty.

3. The model

3.1 Assumptions

In the main applications for my model, a generator competes to obtain a good, service or contract, such

as capacity on an interconnector or a contract for electricity supply, which it needs to be able to perform a

profitable transaction. The profitability of the transaction depends especially on the costs of generating

electricity. I will assume that the cost of generating electricity differs among the buyers.13

This implies that

13The value of the good to a generator is dependent on the costs of generating electricity. As a generator does not know the costof his competitors, he treats it as a random variable, drawn from a distribution that for sake of simplicity I will assume to be

uniform. The random costs drive the dynamics of the bidding behavior. In electricity generation, there is also a common cost

component, mainly gas or oil prices. I assume that the size of these common cost components are common knowledge and that

they are identical for both generators. As a result, these common cost components are inconsequential for the bidding behavior;this is determined by the unknown private value factors. (MORE EXPLANATION OR IS OK SO? Like:


9/48

9

the buyers value the good on auction differently.

For example, when bidding for transmission capacity on an interconnector, the value of the good for sale

is the profit that could be made by selling electricity abroad. This profit is equal to the difference between

the price abroad and the costs of the generator.14 When generators compete for an electricity supply contract

in procurement auctions, the generators actually place as bids the price they will charge for the electricity

supply and the lowest price wins (Loxley and Salant, 2004). For ease of modeling I will transform such a

procurement auction, without loss of generality, into an equivalent discount auction where a given

electricity supply price is set and the generators make bids that represent the discount they will offer on the

price. In such a discount auction the highest bidder wins. For a generator the (private) value of the contract

is equal to the set electricity supply price minus the (private) cost of electricity generation. For example, a

bidder with low costs of electricity generation has thus a high value for the contract, and will be willing to

bid high discounts in the discount auction. The high discount bid in the discount auction translates to a low

bid (which reflects the price for which the bidder is willing to supply electricity) in the procurement auction.

I will assume that a buyer knows his own value, but not the value of the competing buyer. In my model

this implies that a buyer does not know his competitors marginal cost of producing electricity (except for a

common, identical cost factor such as gas or oil prices). In older models stemming from the time electricity

generator markets were tightly regulated, it was usual practice to assume that marginal costs are common

knowledge, however, since the electricity industry has become competitive, information on the cost

structure of electricity generation has strategic value and is therefore carefully guarded (Lautier, 2001,

p.34). Parisio and Bosco (2003, p. 8) add: generators frequently belongto multi-utilities providingsimilar

services often characterized by scope and scale economies (Fraquelli et al., 2004, amongothers). The cost

ofgeneration therefore can vary across firms because firms can exploit production diversities in ways that

are not perfectly observable by competitors. Parisio and Bosco (2003, p. 8). In this line of thought,

_ a _ a_ a _ a

[ ; ] Pr [Y wins;b] Pr [Y looses;b]

[ ; ] Pr [Y wins;b] ( Pr [Y looses;b]

[ ; ] Pr [Y wins;b] (

(1 ) [payment|Ywins;b] [payment|Y looses;b]

(1 ) [ payment|Ywins;b] [ payment|Y looses;b]

(1 ) [

b v v

b v R v

b v R v

E E

E R E R

E

T

T

T

K K

K K

K

!

!

!

_ a _ a _ a _ a

Pr[Y wins;b])

[ ; ] Pr [Y wins;b] ( Pr [Y wins;b])

[ ; ] [ ; ]

[ ; ] [ ; ]

payment|Ywins;b] (1 [ payment|Y looses;b]

(1 ) [payment|Ywins;b] (1 [payment|Y looses;b]b v R v

b v b v

d b v d b v

R E R

R E R E

R

db db

T

T T

T T

K

K K

K

!

!

!

14 In line with the empirical evidence, I assume that, as transmission capacity is fixed and small relative to the total demand,buyers cannot influence the final price in the distant location (see e.g. Consentec, 2004).


10/48

10

competitors can only make an estimate of each others marginal costs. While earlier literature assume that

the cost of generating electricity by a generator is publicly known to all its competitors (Green and

Newbery, 1992; von der Fehr and Harbord, 1993), I find the arguments of Lautier (2001) and Parisio and

Bosco (2003) the most compelling, and I therefore assume that there is uncertainty about the competitors

costs. However, for completeness I also consider a deterministic configuration, where generators know the

costs of electricity generation for competitors.15

I model the competition of the two generator as two risk-neutral bidders who have private values that are

independently and uniformly distributed on ? A0,1 . Both bidders are thus, at the outset, symmetrical; they

have identical, independent value distributions. I assume that the good on sale is sold as one indivisible

good.16

As usual in auctions, the highest bidder wins the good, which reflects that the integrated seller does

not favor the integrated buyer and thus the legal separation of the integrated seller is working as intended by

the regulators.

One of the bidders is an integrated buyer; a holding company fully owns the integrated buyer and (a part

of) the integrated seller. I denote with parameter 1k the proportion of the integrated seller that the holding

company owns. I denote with parameter 2k the proportion of the auction revenue, which the integrated

seller can retain. For example, when the integrated seller is unregulated, he can keep all of the auction

revenue and2

1k ! . When the integrated seller is regulated, he can often still retain a part of the profit due

to incentive regulation (and possibly by creative accounting), and thus20 1k e . The relevant parameter in

the model is the proportion of the auction revenue that is received by the holding company, which I denote

by 1 2k kK ! .

I assume that the values of both buyers are independently and uniformly distributed on ? A0,1 . At the

outset, the buyers are therefore symmetrical. Given his value realization, the integrated buyerY chooses his

optimal bid,Y

b . In line with the literature, I assume that there exists a differentiable, strictly increasing

bidding strategy [ ]Y

b that maps the integrated buyers realized value ? A0,1Yv into his bid [ ]Y Yb v . The

bidding strategy [ ]Yb has an inverse [ ]y such that ? A[ ]Yy b v v! . Analogously, the optimal bid of the

independent buyer X, Xb , is determined by her bidding strategy [ ]Xb that maps her realized value

15I thank this suggestion to an anonymous referee.

16 While transmission capacity and electricity supply procurement auctions are usually multi-unit auctions, I restrict my focus tosingle-unit auctions to simplify the analysis and focus on the effect of integrated ownership.


11/48

11

? A0,1Xv into her bid [ ]X Xb v . The s

trategy [ ]X

b has an inverse [ ]x , such that ? A[ ]Xx b v v! . Using these assumptions, the following lemma

holds (Krishna, 2002).

Lemma 1: The integrated buyerY wins (looses) the auction with bid b if [ ]X

v x b ( [ ]X

v x b" ), which has

probability [ ]x b ( 1 [ ]x b ) for the uniform distribution.

Proof: Ywins the auction if his bid b is larger than the bid ofX, [ ]X X

b v , thus when [ ]X X

b v b . Applying

the inverse biddingfunction [ ]x on both sides of the equation gives ? A ? A1X Xv b b x b

| . When values are

uniformly distributed, the probability of ? AXv x b is equal to [ ]x b . The argument forYloosingthe auction

is symmetrical (Krishna, 2002).

3.2 The second-price auction

For second price auctions17

with two buyers, an independent buyer and an integrated buyerY who owns a

share of the seller, where values are independently and identically distributed with a twice continuously

differentiable distribution function G(.) that satisfies the monotone hazard rate condition, Burkart (1995)

gives characterizations of the bidding functions of X and Y:

1 ( )

( )

Y

Y Y

Y

G bb v

g bK

! ,

X Xb v! .

Burkart (1995) shows that Y overbids his valuation. The amount of overbidding increases in the ownership

share K and decreases in the valueY

v . The intuition for this is as following (Burkart, 1995). When Y looses

the auction, Y is not indifferent to the price for which transmission is sold as Y receives the toehold times

his bid,Y

bK , when loosing. By bidding higher than his value but lower than the value of X,Y Y X

v b v , Y

has an increased gain in loosing the auction. However, when Y bids higher than X while the value of X is

higher than Y,Y X Y

v v b , the resulting profit ofY (winning the auction for a high price) is lower than

when he bids his true value (loosing the auction). The optimal amount of overbidding balances these two

opposite effects on profits. The gain of loosing with a higher bid is increasing in the toehold K , hence Y

17 In the second price auction the highest buyer wins and pays the bid of the second highest bid.


12/48

12

overbids more for a higher K . The probability of winning is increasing in the value ofY, and therefore the

overbidding ofY is decreasing in his valueY

v . Furthermore Burkart (1995) shows that an inefficient

allocation happens with a positive probability and that the expected profit ofY increases with the size of

ownership.

The rather general assumptions under which Burkart (1995) analyzes the problem, do not allow him to

give an estimate of the size of the effects of integrated ownership. For this reason I explicitly determine

efficiency and competition effects under the assumption that the values of X and Y are identically,

independently, and uniformly distributed on [0,1], an assumption that can be thought of as a reasonable

approximation of reality. The bidding functions of X and Y then simplify to:

(1 )

1

YY Y

vb v K

K

!

,

X Xb v! .

The overbidding ofY is now given by the explicit term,(1 )

1

YvKK

, and the same intuition as explained

above for the general case applies

Figure 1. The bidding function of buyer Y in second-price auctions

Figure 1 illustrates the bidding by the integrated buyer and the independent buyer. Because of his

ownership holding in the seller, the integrated buyerY bids more aggressively. This has several interesting

effects, summarized in proposition 4.

Proposition 1:As the ownership share K increases, a) the price of the good on auctiongiven by the

auction revenue increases, b) the probability to win for the integrated buyerYincreases while that for

buyerXfalls, c) the strategic profit of the integrated buyerYincreases, and d) total efficiency falls.18

Proof: See appendix.

The intuition for proposition 4 is as follows: Ad. a) The price of transmission capacity increases as the

losing bid ofY is higher, and hence X pays more for the good. When Y wins, Y either pays the same (Y

18 The effects are all described by ex-ante expectedmeasures (before bidding and before concrete values have been realized).


13/48

13

would have won with or without the ownership share) orY pays more (Y would have lost without the

ownership share). Ad. b) The probability ofY winning the auction is higher as Y bids more aggressively

than X. Consequently, the probability of X winning the auction falls. Ad. c) The strategic profit of generator

Y increases as Y changes his bid, while the old one is still available. As there is a unique optimal bid, this

argument reveals that Y must be better off with his new bid. Ad. d) Total efficiency falls as the auction is

now asymmetric and therefore inefficient. While Y has the same value distribution as X, Y now wins in

some cases when he does not have the highest value for the good, because Y now bids more aggressively

than X.

Figure 2. Effects of ownership share on outcomes in second-price auction .

Figure 2 shows the effect of ownership share on auction outcomes relative to the case with no ownership

share. There is a considerable efficiency loss19, up to 6.25% for full ownership. The gain for the VIU, given

by the strategic profit20

is also considerable; a VIU can, by bidding more aggressively, increase its profit

with up to 16.7% for full ownership. The price of the good is strongly affected; it can increase with up to

37.5% for full ownership. However, this might also be considered a positive effect in the case of

transmission line capacity auctions; Joskow and Tirole (2003) show that in general auction revenues are too

low to incite the building of an efficient amount of merchant transmission lines. This is what we see in this

model as well; the expected value of the transmission line, which is equal to the price difference between

the locations connected by the transmission, is equal to 23

, but the expected auction revenue is 13

. To have

an efficient level, the auction revenue should increase with 100%, and the 37.5% we observe under full

ownership is therefore a considerable step closer to its optimal value. Also in procurement auctions this

could be seen as a positive effect, as then the price paid reflects the discount generators are giving to the

distributor that buys electricity. An increase in the price means that the generator gives a larger discount

than without ownership integration; electricity is thus bought by the distributor at a lower price.

However, ownership integration creates strong discrimination against independent generators favoring

the VIU is a negative effect. As can be seen in Figure ownership integration increases the expected

19The efficiency loss percentage is calculated as

? A ? A

? A

0

0

W W

W

K, which is equal to

2

2

25

1

K

K.

20 The strategic profit percentage is calculated asYStrategic

YPassive

T

T.


14/48

14

probability for the integrated generator to win the auction with up to 50%.21

Also at low levels of ownership

integration discrimination is considerable; even with a an effective ownership share of only 10%, the

independent generator already has a 10% higher expected probability to win the auction. This violates one

of the key principles which the EU intends to apply to the electricity markets : creating fair competition in

electricity generation. Moreover, the fact that ownership integration creates strong discrimination against

independent generators might discourage investment into generation by independent investors and thus lead

to a lower, suboptimal, level of competition. The dynamic cost of such a suboptimally low level of

competition are not determined here, but are likely to be considerable.

Auction with n independent buyers

An interesting question for policy is to what extend the above analysis generalizes to auctions with more

than one competing independent buyer; in many countries in the EU and the USA there are several

independent buyers in transmission and procurement auctions. When the integrated buyerY faces n

independent buyers, has value v for the good on auction, and makes bid b, then his expected profit is given

by:

( )

nd

th th

2

[ ; ] Pr [ wins] ( (1 ) E[the highest bid from buyers| wins])

Pr[ has the 2 highest bid]

Pr[ has the i highest bid] E[the2nd highestbid from buyers| has the i highest bid]

n

Y

n

i

b v Y v n Y

Y b

Y n i Y

T K

K

K!

!

The expression in the first line gives the part of the profit when Y wins; in that case Y receives his value vminus the money he must pay that doesnt go to the integrated seller, this is equal to 1 K times the highest

expected bid from the n competing independent buyers. The expression in the second line gives the part of

the auction revenue Y receives when he has the 2nd

highest bid. In this case, Y sets the price to be paid by

the winner of the auction; Y thus receives the effective ownership share, K , times his bid b. The expression

in the third line gives the expression when Y has a bid lower than the 2nd

highest bid and thus does not set

the price. When Y has the ith

highest bid (with 2 i ne e ), the expected payment by the winner is the 2nd

highest bid from the (n-i) bidders that have a higher bid than Y. The total expected profit forY in this case

is thus his effective ownership share, K , times the summation of the probability ofY having the ith highest

21 In the model with two buyers, this implies that the integrated buyer is expected to win three times more often (75%) than theindependent buyer (25%).


15/48

15

bid times the expected 2nd

highest bid from the (n-i) bidders.

The effect of having more independent buyers participating in the auction on the bidding function of the

integrated buyerY is not immediately clear. On the one hand, more buyers lowers the risk for the integrated

buyerY to win the auction with a bid higher than his value (the first line in the equation), and thus gives Y

an incentive to bid more aggressive. On the other hand, having more independent buyers lowers the

probability that Y will be setting the price by having the 2nd

highest bid (the second line in the equation),

and thus gives Y an incentive to bid less aggressive. Interestingly, for values being independent and

uniformly distributed on [0,1] the two opposite effects cancel out, and the integrated buyerY bids the same

as in an auction with 1 competing independent buyer:

Proposition 2:For any 1n u , in a second-price auction with n+1 bidders, n independent bidders and one

integrated bidder who receives a share K of the auction revenue, where values are distributed

independently and uniformly on [0,1], the independent bidders bid their value, and the integrated bidder,

bids(1 )

1

YY Y

vb v K

K

!

.

Proof:See appendix.

It is intuitive that the efficiency loss becomes smaller when the number of competing bidders goes up.

Interestingly, the discrimination effect keeps rather strong even for a high number of competing bidders.

Proposition 3:For any 1n u , in a second-price auction with n+1 bidders, n independent bidders and one

integrated bidder who receives a share K of the auction revenue, where values are distributed

independently and uniformly on [0,1], the integrated bidder has a higher expected probability of winning

the auction. The percentage increase in the expected probability of winninggoes to 100K % when n goes to

infinity.


Figure 3 gives a graphical illustration of the remarkable strength of the discrimination effect of integratedownership. The left graph shows the relative increase in expected probability of winning, which increases

with the number of competing bidders. The right graph shows the negative effect that each independent

bidder experiences. As we have seen before, a single independent bidder has a probability of winning 50%


16/48

16

less for full ownership. Each competing independent bidder has, with full integrated ownership, a

probability of winning 37.5% less with two competing independent bidders. Even with three competing

independent bidders, which is a rather generous assumptions as the markets for electricity generation are

rather concentrated in the EU,22 each of them experiences a decrease in the probability of winning of 29%

for full integrated ownership. Even for low effective ownership shares the discrimination effect is rather

strong; for example when 0.15K ! , each independent buyer experiences a decrease in the probability of

winning between 5% (with 3 competing independent buyers) and 13% (with 1 competing independent

buyer)

Figure 3: Effects on discrimination with several independent buyers in second-price auctions

3.3 The first price auction

In this section, I will analyze the effect of the integrated buyerY owning a part of the integrated seller in

first price auctions.23 When Y fully owns the integrated seller in first price auctions, a general result can be

established.

Proposition 4: When the values ofXandY,X

v andY

v , are independently distributed without any further

restrictions on the possible distribution, then when the integrated buyerYhas toehold 1K ! , Ybids in a

first-price auction his own value.

Proof: When 1K ! , Yreceives the full amount of any bid paid. Therefore Ydoes not have to take bidding

costs into account and has a lower bound on the expected profit of min[ , ]Y X

v b . Now an ar gument similar to

that for truthful biddingin second-price auctions applies. Suppose Yhas valueY

v . IfYmakes a bid lower

than his valueY Y

b v , then with a positive probabilityXwins with a bid,X

b ,which is higher than the bid of

Ybut lower than the value ofY,Y X Y

b b v . In this case Ycan guarantee himself a higher profit at no costs

by biddinghis value,Y Y

b v! . A similar argument establishes thatYwill not make a bid higher than his

22For example, in a survey of the European Commission the average share in total generation of the largest generator in 2006was, for the 18 countries that reported, 61%22 (Eurostat)

23 In a first price auction the highest buyer wins and pays his own bid.


17/48

17

value. Hence, YbidsY Y

b v! and has an expected profit of max[ , ]Y X

v b .

Corollary 1:A buyer in a first-price auction who receives the auction revenue in full bids as a buyer (who

does not receive the auction revenue) in a second-price auction. This is true for any independently

distribution of values and for any number of competingbidders as longas these bidders are not

beneficiaries of (a part of) the auction revenue.

The Corollary follows from the fact that the equilibrium bidding function in second-price auctions is

bidding the own value and from the fact that the proof above does not depend on the distribution of Y or X,

nor on the number of competing bidders. DELETE? OR IS INTERESTING FACT?

To further analyze the bidding functions of X and Y, I assume that the values of X and Y, ,X Yv v are

independently and uniformly distributed on [0,1].

Following lemma 1, Y wins the auction with bidY

b for value realizations ? AX Yv x b and looses for value

realizations ? AX Yv x b" . The expected compound profit ofY can thus be calculated by integrating over all

possible value realizations of X;

1) ? A ? A

? A

1

0

Y

Y

x bY

Compound Y Y Y Y X X X x b

b v b b dv b dvT K K!

The first integral is the expected profit when Y wins; the profit ofY is equal to the value of transmission

minus his bid plus the fraction of his bid that he actually pays to himself, K times his bid. The second

integral is the expected profit when Y looses; the profit ofY is then equal to the share of the payment by the

winner of the auction, K times the winning bid. Solving the first integral and substituting [ ]X X

v x b| in the

second integral and integrating by parts results in,

2) ? A ? A 1 [ ] [ ]Y

bY

Compound Y Y Y Y Y Y b

b x b v b b b x b x d T K K F F ! ,

where b is the maximum bid.

To determine the first order condition for profit maximization forY, differentiate equation 2) with respect to

Yb , set it equal to zero and substitute [ ]y b for

Yv :


18/48

18

3) ( [ ] ) '[ ] (1 ) [ ]y b b x b x bK ! .

The profit maximization problem for X is identical to that forY with the ownership share set to zero, 0K ! ,

therefore the first order condition for profit maximization for X is:

4) ( [ ] ) [ ] [ ]x b b y b y bd ! .

Under no ownership of the seller, when 0K ! , the problem is symmetrical for X and Y and both have

bidding function 12b v! . Under full ownership, when 1K ! , Y bids his value, and thus, using 4), X bid is

given by 12X Xb v! . The more aggressive bidding ofY has several interesting effects on the profits of X and

Y and on efficiency. Interesting indicators are the compound, the passive and the strategic24

profits ofY.

The compound profit ofY consists of the transmission auction profit and the generation profit together.

Proposition 2 summarizes the main effects.

Proposition 5: When integrated buyerYhas full ownership, 1K ! , then relative to the case of no

ownership 0K ! : a) the price of the good on auction given by the auction revenue is higher, b) the

probability to win forYis higher while that forXis lower, c) the strategic profit of buyerYis higher, and d)

total efficiency is lower.25


Quantitatively, when Y has full ownership of the integrated seller, the auction revenue increases with 62.5%

from 13

to 1324

, the probability of winning forY increases from 50% to 75%, the probability of winning for X

falls to 25%, the strategic profit increases from 0 to 124

, and efficiency falls with 4.2% from 23

to 1524

.

Corollary 2:Revenue equivalence between first and second-price auctions does not hold.

24 The compound profit ofY, the profit of the transmission auction and the generation together, is influenced by the ownership

share K has a direct and a strategic effect on the compound profit. The direct effect translates into what I will refer to as the

passive profit and is due to the fact that Y receives proportion K of the auction revenue. The passive profit is the profit that Y

would receive when he owns the proportion K , but bids as i f his ownership share was zero. The strategic effect translates into

what I will refer to as the strategic profit and is due to Y changing his bidding schedule. The strategic profit can be found by

subtracting the passive profit from the compound profit.25 The effects are all described by ex-ante expectedmeasures (before bidding and before concrete values have been realized).


19/48

19

Proof: Usingthe above biddingfunction the auction revenue can be calculated to be equal to 1124 for the

case of full ownership. In section 3.1 I found that the auction revenue in a first-price auction is equal to 1324

for the case of full ownership.

Outcomes for :0 1K K lie in between the extremes of no ownership, 0K ! , and full ownership,

1K ! . Equations 3) and 4) can be solved numerically for [ ]x b and [ ]y b for : 0 1K K .26 Figure 3

therefore shows numerical approximations of the bidding functions for 0 1K together with the highest

bid b and the bidding function for 0K ! and 1K ! .27

Figure 4: the bidding functions for independent buyer X and integrated buyer Y in first-price

auctions.

The bidding functions in Figure 4 demonstrate that an increased ownership share in the seller results in

the integrated buyerY bidding more aggressively. Y maximizes profits given by

Pr[ wins | ] ( (1 ) ) Pr [ looses | ] ( )Y Y Y Y X

Y b v b Y b bK K . A positive ownership share, 0K " , increases the

gain of winning, (1 )Y Yv bK . This gives Y the incentive to sacrifice a part of this gain, by bidding

stronger, to increase his probability of winning. This incentive is partly countered by the income Y earns

when he looses; the ownership share times the bid of X,X

bK .28 All in all, Y bids stronger.

26To my best knowledge there exists no explicit analytical solution for the bidding function in first-price auctions with

: 0 1K K . Proposition 6 in the appendix lays out the necessary restrictions that the bidding strategies must fulfill.27

Note that there is a discontinuity at 1K ! . If and only if 1K ! , then biddingY Y

b v! is a weakly dominant strategy forY.

Suppose 1K H! (for small 0H " ), then if X sticks with her strategy 12X X

b v! , then Y would never bid more than 12

I (for

small 0I " ). At 12Y

v I! there would be a mass point which in turn would create an incentive for X to try to overbid it

whenever her value is larger ( 12X

v I" ). Therefore, once 1K , biddingY Y

b v! cannot be an equilibrium strategy forY. For

an equilibrium in pure strategies to exists at all, the bidding functions of X and Y must have the same bid for 1Y X

v v! ! . This is

the case in the strategies shown in Figure 3; there are no mass points, and the density ofYs bids is continuous, excluding the

possibility for X to improve her profits by deviating from her strategy. This implies that the maximum bid b converges to 1 when

the ownership share K goes to 1.28

Compare the bidding ofY with the much stronger bidding of a comparable buyerY in Van Koten (2007), where Y is a

subsidiary of a holding company and the pay-off ofY is given by' ' '

Pr[ ' wins | ] ( (1 ) )Y Y Y

Y b v bK , while 0 1Ke e is a

parameter set by the holding company in order to strategically delegates decision powers to Y. When Y looses he has noearnings and he bids stronger than buyerY in the present paper.


20/48

20

The stronger bidding ofY lowers the profits of X, Pr[ | ] ( ) X X X

X wins b v b , by lowering the

probability of X to win the auction. This gives X the incentive to sacrifice a part of her earnings, by bidding

stronger, to increase her probability of winning. The relevant effects are, firstly, that there is a negative

efficiency effect; the ownership share makes the auction inefficient. The loss can be as large as 4.2% for full

ownership. Secondly, there is a discrimination effect; the independent generator, who has thesame value

distribution as the allied generator, has a lower probability of getting access to the transmission, which can

be as low as 25% for full ownership. Thirdly, there is a strong price distortion effect; transmission capacity

will be sold for a higher price. The increase in price can be up to 62.5% for full ownership.

3.4 Alternative models

In this section I analyze two alternative cases that might be relevant in electricity markets. The cases are

very similar to the setup I analyzed before, but make different assumptions concerning information. In the

first case I assume that there is no uncertainty; generators know the exact value of their competitors. In the

second case I assume that generators have the same value for the good on auction, but dont know this

value; they only have available an estimate of this value. This case can be modeled as a (unknown) common

value auction.

No uncertainty

While I assumed that generators have private information about their values (allowing a common value

factor that is publicly known), as they have to make an estimate that is subject to error, it could be useful to

look at an idealized situation where generators can estimate the exact value of their competitor without

error. It is shown that in such an ideal case of perfect information, most of the effects found before

disappear; there is no inefficiency and the probability of the independent buyer to win is not negatively

affected. Due to the fact that such a model has a continuum of Nash-equilibriums, the effect on the expected

price cannot be determined.

For generality, I assume that both buyers may have an ownership share; buyerY has ownership share

: 0 1Y Y

K Ke e and X has ownership share : 0 1X XK Ke e . To guarantee the existence of Nash-equilibriums,

I make the assumption that if both buyers make the same bid, then the auction then is won by the buyer withthe highest value (and in case of equal values the winner is chosen at random).


21/48

21

Proposition 7:When a buyerYwith ownership share : 0 1Y Y

K Ke e and valueY

v , and a buyerXwith

ownership share : 0 1X X

K Ke e and valueX

v , are competingin an auction and the values are common

knowledge, then there exists a continuum of Nash-equilibriums. WhenX Yv v , then Ybids Yb p! andX

bidsX

b p! with [ , ]X Y

p v v . Ywins and earns both in first-price and second-price auctions

(1 )Y Y Yv pT K! . Xlooses and earns X XpT K! . In case X Yv v" , the Nash-equilibriums can be

expressed symmetrically.

Proof: When the price for transmission is equal to p, then a buyer with ownership share K and value v

receives when winning (1 )v p v p pK K ! and when loosing pK . From the relationship

p v v p p pK K " it follows that when the price is lower (higher) than his value, the buyer prefers to

win (loose) the auction and receive v p pK ( pK ).

There is a continuum of Nash equilibriums, in all of which the buyer with the highest value wins the

auction; all Nash equilibriums are thus efficient. However, because of the multiplicity of equilibriums it is

not clear which equilibrium will be chosen, and neither is it clear whether the buyers will be able to

coordinate on a Nash equilibrium. To determine the effects on efficiency, competition and price, further

assumptions must be made on how buyers set their bid or possibly bargain. I assume that the buyers are able

to coordinate on a Nash equilibrium. This implies that the auction is efficient, as all Nash equilibriums in

this auction are efficient. Furthermore, as the buyer with the highest value wins the auction, both buyers

have equal probability to win the auction; 50% each, which indicates that there is no discrimination againstthe independent buyer concerning winning the auction. However, it is possible that an independent buyer,

without an ownership share, earns less profit than an integrated buyer. For example, imagine that when the

value of the independent buyer is the higher (lower), the Nash equilibrium with the higher (lower) price is

selected. In that case the independent buyer earns zero profits, while the independent buyer earns positive

profits. Such a result would discourage investment in new independent generation. Stronger assumptions

would be needed to determine the precise effects. The price of the good on auction, as reflected by the

auction revenue, will be between the lowest and the highest value, which means that the price is weakly

higher compared to an auction where values are not common knowledge and no buyer has an ownership


22/48

22

share.29

As explained before, a higher price can be interpreted as a positive effect as the resulting price in an

auction is lower than the value of the good.

The case of no uncertainty thus shows that with complete information, the negative effects are likely less

pronounced for integrated ownership of generation and transmission. To determine the precise effects more

structure needs to be added to the model. It is however clear that when information is complete, integrated

ownership seller does not result in inefficient allocations, and neither is the independent buyer disfavored in

winning the auction, while it is unclear whether the independent buyer possibly earns less profits.

Unknown common values

While I allowed in my model for an identical common value component in the valuations of the bidders,

I assumed that this component is common knowledge to both two buyers, thus preventing this component to

affect bidding strategies; these are determined by the unknown private value factor. A setup without a

private value factor and where the size of the common value component is unknown to both buyers, can bemodeled as a common value auction. Bulow et al. (1999) model such common value auctions where buyers

own a share of the seller. Both buyers have the same value for the good on auction, but the exact value of

the good is only known with certainty after a buyer has won the auction. Both buyers have private

information (called a signal) that allows them to make an estimate of the value of the good.30

From the

results of Bulow et al. (1999) for the case where only one buyer, the integrated buyer, has an effective

ownership share, and under additional assumptions similar to the ones I use in my model, signals are

uniformly distributed on [0,1], and the common value component is equal to the average of the signals,

similar conclusions to the ones in my model can be drawn.

While efficiency is not an issue in such a common value auction by definition (the good has the

same value for each buyer), ownership integration has, like in my model, a strong discrimination effect

(against the independent buyer) and an upwards effect on prices. Under the mentioned additional

29It is a standard result that in an auction with two buyers, where values are not common knowledge and no buyer has an

ownership share, the expected price is the lower of the two values (see for example Krishna, 2002).30

Such an analysis might be relevant for the electricity markets. For example, generators that have the same costs in producing

electricity might both need transmission capacity to sell electricity in a distant location. The exact price the generators will receive

in the distant location is not certain, and each generator makes an estimate of this price given his private information. The value oftransmission capacity to the distant location is then the same for both generators, but each has a different estimate of this value.

This situation can be translated into the model of Bulow et al. (1999). A similar situation could be in the case of procurement

auctions. Negrete-Pincetic and Gross (2007) argue that there was to a high extend uncertainty over the value of the contracts on

sale in Illinois in 2006. If the value of the contracts was indeed uncertain, and if generators had more or less the same cost ofproducing electricity, then the auction could be modeled by a common value auction, as done in Bulow et al. (1999).


23/48

23

assumptions the probability of winning of the independent buyer is1

[ ]2

IP X winsKK

!

in first-price, and

[ ] 0IIP Y wins ! in second-price auctions. The discrimination effect is thus stronger in such common value

auctions; the probability of winning for the independent buyer in second-price auctions iszero, even if the

integrated buyer has only a small effective ownership share, and in first-price auctions goes to zero when the

effective ownership share ofY goes to one. The expected price of the good on auction when the integrated

buyer has a strictly positive, but possible very small, effective ownership share cannot be compared with the

price when the integrated buyer has no effective ownership share. In the latter case such a common value

auction has a multiplicity of equilibriums (Bulow et al., 1999). However, it can be determined that the

expected price is increasing in the ownership share of the integrated buyer; the expected auction revenue in

second-price auctions lies in between 0, when the ownership share K approaches zero, and 34

for full

ownership (Bulow et al., 1999), and in first-price auctions in between 13

, when the ownership share K

approaches zero, and 58

for full ownership.31

The model of Bulow et al. (1999) show that integrated ownership has similar effects on competition

as my model, while efficiency is by definition not an issue and the effect on expected price cannot be

determined due to indetermination of the model when the integrated buyer has no ownership share.

4. The procurement auctions in New Jersey and Illinois.

The procurement auctions held in New Jersey from 2002 till 2008 and in Illinois in 2006 are

examples of cases where distributors and generators figured as integrated sellers and buyers. In 2002, New

Jersey organized its first procurement auction where distribution companies sold one-year forward contracts

to ensure the electricity needs of their default service customers for a one-year period (Loxley & Salant

2004).32

The contracts were sold in procurement auctions as fixed percentages of load, called tranches. All

31The expected auction revenue in second-price auctions is equal to

2 1( )

4 4

IIm

KK K

K

!

, which lies in between 0, when the

ownership share K approaches zero, and3

4 for full ownership (Bulow et al., 1999). Using the functions in Bulow et al. (1999)with the additional assumptions mentioned above the expected auction revenue in first-price auctions can be shown to be equal to

32

1 1

2 2

1

(12 8 ) Gamma[ ] (3 )(2 ) Gamma[ ]1 1 1( )

2 6 2 6 7 2 (1 )(3 2 ) Gamma[3 ]

Ij j

mj j

K

K K

K

KK

K K K

!

.

32 See also http://bgs-auction.com.


24/48

24

four distribution companies selling contracts, Public Service Gas & Electric Company (PSE&G), Jersey

Central Power & Light Company (JCP&L), Atlantic City Electricity Company (ACE) and Rockland

Electric Company (RECO), were integrated sellers; they were owned by holding companies that also owned

generation companies. In the procurement auction in Illinois in 2006 electricity supply contracts were, like

in New Jersey, sold in tranches (Negrete-Pincetic and Gross, 2007). Both the two distributors involved,

Ameren and ComEd, were integrated sellers as they were owned by holding companies that owned

generators that were bidding in the auction. Table 1 gives an overview of the distributors and their

integrated generators in New Jersey and Illinois.

Table 1: Distributors and their integrated generators in New Jersey and Illinois

The auctions ran in New Jersey and Illinois were, as they were multi-unit auctions and had more

than two bidders, more complicated than the auctions I modeled in this paper. A more detailed treatment ofthe auctions can be found in Negrete-Pincetic and Gross (2007) for the Illinois auctions, and in Loxley &

Salant (2004) and on http://bgs-auction.com for the New Jersey auctions. However, it is likely that the logic

in the theoretical cases treated in this paper carries over to more complicated settings. The models then

predict that non-discrimination is violated and that integrated buyers have a higher chance to win an auction

than independent buyers. A buyer will thus be more likely to win auctions when the seller and the buyer are

owned by the same holding company (they have the same affiliation), then when the seller and buyer are

owned by different holding companies (they have different affiliation). In the case of the auctions in New

Jersey and Illinois, an integrated generator is expected to acquire more tranches of his own integrated

distributor.

The raw data suggest that this might be the case; Table 2 shows the percentages of tranches won by

the generator integrated with ACE (Connective) in the auctions over 2002-2008, for the different products.

As my model suggests, the average percentage of tranches Connective won of ACE is higher than those won

of other distributors. In addition, Connective, from 2004 on, only aquired tranches from his integrated

distributor ACE, which suggests that Connective learned over time about the strategic advantage it has in

auctions for traches of ACE.

Table 2: Tranches won by the generator integrated to ACE (Connective)


25/48

25

To test if bidders with affiliation did indeed have an advantage in the BGS auctions, I compare the

(unweighted) average proportion of tranches won in auctions over 2002 till 2006 among the four integrated

generators. If affiliation has no effect, then an integrated generator should win, on average, equal

proportions for the different distributors. If affiliation brings an advantage, then the average proportion of

tranches won should be higher for a generator when the distributor has the same affiliation, then when the

distributor has a different affiliation. For example, a generator integrated to ACE should have a higher

proportion of contracts won for supply to ACE than for supply to JCP&L, PSE&G, or RECO.

Table 3: Results of the New Jersey BGS auctions over 2002-2006

In Table 3 in column no.1 I have shown the average percentages of the load (called tranches) won in

the New Jersey BGS auctions from 2002 till 2008 by the generators with the same affiliation as one of the

distributors. Numbers in bold are the percentages won when the generator and the seller had the same

affiliation. In column no.2, I have depicted the averages of the percentages won in auctions by the three

generators that have a different affiliation than the generator in the row. For example, the generator

affiliated with ACE, won over the auctions from 2002 till 2008 an average of 8,1% of the tranches of ACE,

which is higher than the average percentage of tranches he won of any of the other distributors (JCP&L,

PSE&G and RECO). Table 3 shows that percentages won by integrated generators (the bold numbers in the

first column) are higher than the average percentage won by generators with another affiliation (the second

column). The percentage of tranches won by the generator affiliated with RECO is significant.33

Table 4: Results of the Illinois auctions in 2006

Table 4 shows the average percentages of the tranches won in the Illinois auctions by the generators

with the same affiliation as one of the distributors. Numbers in bold are the percentages won when the

generator and the seller had the same affiliation. Also here the average percentage of tranches won from a

distributor is higher when the generator has the same affiliation.

For a more rigorous test, I ran separately for Illinois and in New Jersey regressions with Won, the

proportion of tranches won in auctions for the integrated generators, as dependent variable. As independent

33I compared the average of tranches of RECO won by the generator affiliated with RECO (Consolidated Edison Energy, Inc.)

with the average of tranches this generator won with the other distributors using a t-test with pooled variance. I did the same testfor the other generators, but most of them had a low significance (round 0.2 ~ 0.3).


26/48

26

variable I used indicator variableIntegrated, which takes value 1 (0) if the proportion won by the generator

was with an integrated (non-integrated) distributor. As the auctions in New Jersey took place over 2002 till

2008, I also included variable Year, indicating the year the auction took place. The theory presented in this

paper suggests that an integrated buyer will bid more aggressively in an auction and thus have a higher

probability of winning the auction; the variableIntegratedshould thus have a positive effect on the

proportion won. Table shows that in regressions with the proportion won as a dependent variable, the

coefficient onIntegratedis indeed positive and significant, both in Illinois and in New Jersey.34

Table 5: Percentage of tranches won in auctions regressed onIntegrated.

My analysis in the procurement auctions in Illinois and New Jersey shows that generators obtained

higher shares of contracts for supply, called tranches, for integrated than for non-integrated distributors.

This conforms to the intuitions developed in the theoretical models in this paper. However, an alternativeexplanation would be that there are other advantages for a generator to supply to an integrated distributor.

For example, a generator might receive information of his integrated generator which enables him to better

forecast the needed supply and thus save costs. In addition, for the theoretical models in this paper to apply,

it must be the case that the distributor at least partly benefits from the auction revenues, and that a part of the

benefit is passed on to the owner, the holding company. A more extensive study could control for such

alternative explanations.

5. Conclusion

My analyses suggest that the integrated ownership of a buyer and seller has negative effects on auction

outcomes under imperfect information. A holding company that owns both a buyer (the integrated buyer)

and (a share of) the seller, has incentives to make the integrated buyer bid more aggressively. Consequently,

the probability of winning for the integrated buyer increases at the expense of an independent buyer, thus

curbing competition. This increases the profits of the integrated buyer, while causing efficiency losses. The

aggressive bidding also drives up the price of the good on auction. This price effect can be interpreted as

positive: in transmission auctions the transmission capacity, which is generally underpriced, is then priced

slightly closer to its value, and in procurement auctions the distributor can buy electricity for a lower price.

34As a robustness test I included several sets of dummy variables in the regression. I included dummies for different products

(contracts for different duration and pricing), for years, and for generators, but the significance of the variableAffiliatedwasvirtually not influenced. See Table A in the Appendix for the regression models including the dummies.


27/48

27

Additional analyses shows that the negative effects only occur under imperfect information; when the

buyers valuations for the good are common knowledge, the allocations that result from the auction are no

longer discriminatory and inefficient.

These results should be of interest to regulators of the EU and USA electricity industries. In the EU

regulators addressed the issue of underinvestment in capacity by allowing unregulated for-profit building of

transmission lines. In such a setting, a VIU might be tempted and in fact is likely, as I have shown above

not to allocate transmission capacity in a non-discriminatory and efficient manner. Most notably auctions

lose their favorable features (non-discriminatory, market-based and efficient). As a result, the competitive

effect of new connection lines in the merchant model is smaller under legal unbundling than under

ownership unbundling. This questions the claim of Brunekreeft et al (2006) that ownership restrictions are

not much of a concern, as they assume that the owner will want to keep competitive pressure between the

generators. My model shows that this is not the case as long as only legal unbundling is applied.

This is highly relevant for EU electricity generation markets : as national electricity generation markets

in the EU are very concentrated,35

they need to attract new entrants to make the liberalization reforms

successful. Furthermore, the holding company owning the integrated seller is advantaged, and because the

holding company is often the (former monopoly) incumbent who typically still holds a dominant position in

the electricity supply industry, this does not facilitate competition.

In the USA contracts for electricity supply have been sold in procurement auctions. The distributors who

were selling the auctions were owned by companies that also owned generators that participated in the

auction. Such auctions are likely not fair integrated generators have a higher probability to win auctions.

Indeed my empirical analysis showed that integrated generators indeed obtained significantly more contracts

from integrated distributors than from other ones. This might affect efficiency negatively and discourage

new entrants. However, a positive static effect is that the aggressive bidding of integrated generators makes

the electricity cheaper for distributors, and consumers are likely to benefit from this.

There are a few possible solutions to remedy the negative results found in this analysis. Firstly,

regulators could aim their efforts at preventing that auction revenues benefit the VIU that owns distribution

or transmission networks. This would, if successful, reduce the effective ownership share to zero and thus

take away the basis for the advantaged position of the integrated generator. Enforcing ownership unbundling

would effectively achieve this goal. Alternatively, given the strong resistance against ownership unbundling

35In 2006, 20 regulators submitted data on the concentration of electricity generation in EU member states. Out of these 20, 7

countries were highly concentrated (HHI between 1800 and 5000), and 8 countries, among which Belgium and France, were veryhighly concentrated (HHI above 5000) (Commission of the European Communities, 2008b, p.11).


28/48

28

both in the EU and the USA, regulators could try to achieve this goal by means of strict regulation without

ownership unbundling, for example by using rate of return regulation for transmission and distribution

networks. However, rate of return regulation has long been known to lead to welfare losses (Averch and

Johnson, 1962). To give a network owner incentives to run the network efficiently and to add new capacity

usually a form of incentive regulation is used that allows a network owner to keep a part of the generated

profit. Moreover, preventing transmission owners from benefiting from the generated profits goes against

the EU policy of allowing the merchant (for-profit) building of new transmission lines. In addition, there is

evidence that it might be difficult in practice to regulate the auction revenues of the VIU; regulators in the

EU have not been successful in enforcing the prescribed use of auction revenues for transmission lines.

While EU regulations state that auction revenues of international transmission lines (interconnections)

should be spend on infrastructure projects in full, an energy inquiry by the European Commission that took

a sample of 10 transmission owners reported that, over the years 2001-2005, a mere 20% of the auction

revenues were spend on such projects (Commission of the European Communities, 2007, p.179).

Secondly, an independent generator could be awarded or sold an ownership share such that both

generators end up with equal shares.36

Ettinger (2002) has analyzed unique symmetric equilibriums for first-

price and second-price auctions with two buyers who have symmetrical shares of ownership in the seller

(called toeholds) under the assumption of private, independently distributed values. 37 In a symmetrical

equilibrium exists by definition no discrimination effect, hence the buyer with the highest value wins, which

implies there is no efficiency loss as I found in my model. Moreover, the positive effect of integrated

ownership, a price closer to the optimum is strengthened; using the solutions for the bidding functions of

Ettinger (2002), and assuming that values are uniformly distributed as in my model, it is easy to show that

when the price increases in the ownership shares. When both buyers own the maximum possible ownership

share, the increase in expected prices is 33% in first-price auctions, and 67% in second-price auction.

Giving equal shares thus provides a solution. It requires the regulator to have the authority to mandate the

VIU to sell shares in the transmission line to new independent generators. Moreover, implementation of

such a measure brings up many practical questions, such as on what legal basis should regulators be allowed

to take away ownership shares from the incumbent and for what compensation? And should ownership

shares be only given to participating buyers or also topotentiallyparticipating buyers? Giving buyers

symmetrical shares therefore does not seem a practical solution in most cases.

36A form of such co-ownership structure is used in Finland for the transmission network Fingrid (REF!)

37 Ettinger (2002) finds that also with symmetrical ownership shares revenue equivalence between the first-price and second-priceauction does not hold.


29/48

29

Thirdly, a possible remedy is to mandate the VIU to legally separate not only the integrated seller, but

also the integrated buyer. This is the form of legal unbundling that Cremer et al. (2006) consider, and for

which Hffler and Kranz (2007a) coined the term reverse unbundling. By implementing the same sort of

legal unbundling for the integrated buyer, the holding company is no longer able to give the integrated buyer

day-to-day instructions. Also, the integrated buyer is not allowed to receive take revenues of the integrated

seller or the holding company in account; it is usual that in a legally unbundled firm managers are not

allowed to receive bonuses contingent on results of the holding company.38

I take up this question in Van

Koten (2007), and show that auction outcomes are still negatively affected on the same dimensions as in this

paper, even though slightly less pronounced. Legal unbundling of the integrated buyer is therefore not a

sufficient measure.

The solution most in line with economic logic is to mandate ownership unbundling for distribution and

transmission networks. When buyers have no ownership shares in sellers, auctions are efficient and non-

discriminatory.

6. Appendix

Proposition 1:As the ownership share K increases, a) the price of the good on auctiongiven by the

auction revenue increases, b) the probability to win for the integrated buyerYincreases while that for

buyerXfalls, c) the strategic profit of the integrated buyerYincreases, and d) total efficiency falls.39

Proof: The probability forYto win the auction is given by 12[ ]2(1 )

Y winspK

KK

!

, which is increasingin

K, and thus the probability ofXwinningthe auction is decreasingin K . The above expression can be found

by usingthe Lemma 1 (page 11) and the second-price auction biddingfunctions forXandY(page 11). Y

with a realized value ofY

v wins with probability ? A[ ; ] [ ; ]1

Y wins YY Y Y

vp v x b v

KK K

K

! !

. The expected

proportion of auctions that is won by Ycan be found by integrating [ ; ]Y winsY

p v K over the realizations ofY

v :

1

0[ ] [ ; ]Y wins Y wins

Y Yp p v dvK K!

38For example, managers in legal unbundled transmission companies are not allowed to receive bonuses contingent on results of

the holding company (Directive 2003/54/EC, article 10, section 2b, and Commission of the European Communities, 16.01.2004,

p.8).39 The effects are all described by ex-ante expectedmeasures (before bidding and before concrete values have been realized).


30/48

30

1

0 1

Y

Y

vdv

K

K

!

12 1

21 2(1 )

K KK K

! !

.

a) The price of the good on auctiongiven by the auction revenue increases

Proof: The auction revenue is given by ? A 2

1 (2 )

3 6 1m

K KK

K

!

, which is increasingin the ownership

share K . The auction revenue ( ? Am K ) can be found by calculatingthe expected payment of each bidder

(( ? AYm K and ? AXm K ) and add these up.

? A ? A ? AY Xm m mK K K!

? A1 1

01

| |YWINS XWINSx x y y Y X Y X

P E b b b dv P E b b b dvKK

! " -

1 1

01

| 1 |1 1 1

y y Yx x y X Y X X

v v vE v v dv v E b v dvK

K

K K KK K

K K K

! " - -

2

1 1

01

11 | 1

2 1 1

y Yy X Y X X

v vdv v E v v dvK

K

K KK K K K

K K

! -

2

12

1

11

1+3 3 2116 1

X

X X

v

v dvKK

K K KK K K K KK

!

12 2 22 1

3

2

1

(1 )1+3 3

2( 1)6 1

X Xv v

K

K

K KK KKK

! -

2

2 2

1+3 3 1 3

6 1 6 1

K K K

K K

!

21 (2 )3 6 1

K KK! .

c) The strategic profit of the integrated buyer Y increases


31/48

31

Proof:The strategic profit ofYis given by ? A

2

6 1

Y

Strategic

KT K

K!

, which is increasingin the ownership

share K . The strategic profit is defined as the extra profitYcan earn by consideringhis ownership

share K when choosinghis biddingfunction. This is equal to

? A ? A ? A ? A ? A 0 0Y Y YStrategic Generator Generator m mT K T K T K K! . The strategic profit is the sum of the effects ofstrategic biddingon the generator profit (negative) and the auction revenue (positive) times the

ownership share K .

The generation profit ofYis equal to ? A

2

16 2

6 1

Y

Generation

KT K

K!

. It can be derived by multiplyingthe

probability of winningtimes the generation profit and integratingthe resultingexpression over the

possible value realizations:

? A ? A 1

0|Y YWINSGenerator Y X X Y Y P v E b b b dvT K !

1

0|

1 1

Y YY X X Y

v vv E v v dv

K KK K

! -

1120 1 1

Y Y

Y Y

v vv dv

K K

K K

!

12 3 2 3 21 1 1

2 3 3 2

01 2( 1)

Y Y Y Y Y

v v v v vK K K

K K

! -

21+2

6 1

K

K!

2

16 2

6 1

K

K!

.

Usingthe expressions for the generator profit and the auction revenue, the strategic profit can be

determined:


32/48

32

? A ? A ? A ? A ? A

2

1 1

6 2 6 2

2

2 2

2

0 0

1 (2 ) 1

3 36 1 6 1

(2 )

6 1 6 1

6 1

Y Y Y

Strategic Generator Generatorm mT K T K T K K

K K KK

K K

K K K

KK K

K

K

!

!

!

!

d) Total efficiency falls.

Proof:Efficiency is given by ? A

2

2

2

3 6 1W

KK

K!

, which is strictly decreasingin the ownership share

K . Efficiency can be found by addingthe profits ofXandYwith the auction revenue;

? A ? A ? A ? AY XW mK T K T K K! .

The profit ofX, ? A 21

6( 1)

XT K

K!

, can be found by multiplyingthe probability of winningtimes the

profit and to integrate the resultingexpression over the possible value realizations.

? A ? A 1

1

|X X WINS X Y X Y X

P v E b b b dvKK

T K

!

? A1

1

[ ] |1

Y X X X Y X X

vy b v v E b v dvK

K

K

K

! " -

1

1

1 | 11

X X Y X X

vv v E v v dvK

K

KK K K K

K

! -

1

1

11

211

X

X X X

vv v dvK

K

K K KK K

K

!

21

21

2

1

(1 )2( 1)

X X X v v dvK

K

KK KK

! -


33/48

33

12

3 21 16 2

1

(1 )2( 1)

X

X X

vv v

KK

KK K

K

Documents

VanKoten Bidding Behavior JRE 20 2008.12.12