Short Term Hedging, Long Term Vertical Integration

Short-term Hedging, long-term Vertical Integration?
An
Analysis of Corporate Risk Management.

ADOUA, Imane
1
, DE BAZELAIRE, Stanislas
2
, STOFER, Michel
3
, YANY-ANICH, Andrs
4
.

Objectives:
Conventional wisdom suggests that short-term corporate risk management is achieved through
market instruments such as forwards and futures while long-term corporate risk management relies
on corporate arrangements such as vertical integration or conglomerate acquisitions. The objective
of this work is to build a model of corporate risk management with a nite horizon and to investigate
the interaction of hedging and vertical integration as two strong tools of corporate risk management.

Issues:
The model will need to combine a simple way of specifying rms concern for risk management,
a simple modeling of derivative pricing and an analysis of corporate arrangements such as vertical
integration.
Abstract
The purpose of this article is to develop a model to study the interaction and compare two
tools of corporate risk management: vertical integration and short-term hedging. We begin by
studying some important articles in the literature of Corporate Risk Management. In particular,
we focus on two hedging models that we extend for the application of vertical integration. We
build a two-period model with two markets, an upstream and a downstream. In every market,
rms can use hedging with forward contracts to protect their initial wealth against some risky
factor. We develop two cases, with and without vertical integration. We compare the optimal
hedge ratio using each approach. We observe that wealth and correlation eects have a strong
impact on hedging. Moreover, we argue that vertical integration can reduce the need of forward
contracts under certain conditions given by (a) initial wealth, (b) correlations and volatility
with respect to a risky variable and (c) a volume eect. We then add more elements to the
analysis, such as a (i) long-term perspective and (ii) the presence of more actors. We try to
illustrate our results by particular examples taken from some commodity markets. Then, we
develop a specic framework to study vertical integration in the electricity market and see under
which conditions the optimal level of forwards can be reduced.

Keywords: Optimal hedging, vertical integration, changing investment opportunities, forward
contracts, hedging in multiple periods.

Option : Corporate Finance
Group Tutor : Gilles CHEMLA, [email protected]
ENSAE tutor : Romuald ELIE, [email protected]
Dates : November 15th to May 29th
1
Master of Science at ENSAE Paris-Tech in Corporate Finance
2
Master in Management at ESCP Europe and Master of Science at ENSAE Paris-Tech in Corporate Finance
3
Third year student at ENSAE Paris-Tech with major in Corporate Finance
4
Engineer at Ecole Polytechnique and Third year student at ENSAE Paris-Tech in Corporate Finance
1/41
CONTENTS
Contents
1 Introduction 3
2 Corporate risk management in the literature 4
3 A rst paradigm of the benets to hedge 6
3.1 The model setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.2 Stochastic investment function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
3.3 Solving the model by a recursive method . . . . . . . . . . . . . . . . . . . . . . . . . 8
4 Interaction between vertical integration and hedging 11
4.1 The non-integrated case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4.2 The integrated case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
4.3 Comparison of optimal levels of hedging in the two cases . . . . . . . . . . . . . . . . 13
5 Extensions 15
5.1 An endogenous weighted average . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
5.2 A model with several actors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
5.3 Intertemporal issues: the inuence of long-term . . . . . . . . . . . . . . . . . . . . . 17
6 Illustration of the model 20
6.1 Testing the sign of and the hedging ratio in an oil company . . . . . . . . . . . . . 20
6.2 Delta Airlines: an example of vertical integration . . . . . . . . . . . . . . . . . . . . 22
7 Comments, critics 23
7.1 The cost function of raising external funds . . . . . . . . . . . . . . . . . . . . . . . . 23
7.2 The time horizon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
7.3 Perfect price anticipation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
8 A model for the electricity market 24
8.1 Vertical integration in the electricity market . . . . . . . . . . . . . . . . . . . . . . . 24
8.2 An adapted model for the electricity market . . . . . . . . . . . . . . . . . . . . . . . 25
8.2.1 The setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
8.2.2 Problem and optimal solution . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
8.2.3 Extension : Vertical integration . . . . . . . . . . . . . . . . . . . . . . . . . . 27
9 Conclusion 29
10 Appendix 31
10.1 Equations : First model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
10.2 Equations of the model solution in the case of Cobb-Douglas functions . . . . . . . 32
10.3 Eects on the optimal hedging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
10.3.1 The eect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
10.3.2 The endogenous weighted average . . . . . . . . . . . . . . . . . . . . . . . . 34
10.4 The electricity market . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
10.5 The Envelope theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
10.6 Rubinsteins rule (1976) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
10.7 Data of regression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2/41
1 INTRODUCTION
The best Wall Street minds and their best risk-management tools failed to see the crash coming.
New York Times, January 2009.
1 Introduction

The scale of losses in the nancial crisis that began in 2007 is unprecedented. The International
Monetary Fund estimates these losses to $4.1 trillion and $1.1 trillion to help x them. These huge
numbers show the depth of the worldwide crisis and the new challenges that are coming. Since then,
corporations have become more risk averse and have given strong relevance to risk management. In
fact, nothing in the past resembles what we are currently seeing. We are in the presence of events
that we have not seen since World War II. This is a period of absolutely exceptional uncertainty
that calls for responses that match the events from both the public and private sector (Jean-Claude
Trichet, Head of the European Central Bank).
There are several ways to manage corporate risk. Vertical integration has long been considered
as a major tool for that. Causes for vertical integration are diverse: strong volatility of intermediate
goods price and demand, the lack of exibility in some markets, a strong risk aversion. It also
restores the symmetry between upstream and downstream rms exposure to price and demand
volatility. However, as Carlton (1979) points out, the study of vertical integration has never been
a completely-understood phenomenon. Researchers and managers have often concluded that the
prominent incentive for vertical integration is a stable supply of inputs for retailers and/or stable
cash-ows for producers. For example, DeltaAirlines has recently acquired an oil renery, justifying
this strategic move by a signicant reduction in fuel costs, the bane of the airline industry.
In a world of increasing uncertainty, corporations have also relied on derivative instruments
to hedge their risk in a short-term perspective. From now on, by hedging we refer to the use of
forward contracts to protect a wealth from a risk. The use of these nancial hedging instruments has
strongly risen in the past few decades (Hull (2003)). A rm can hedge its wealth in order to reduce
the need of external nancing to fund investments. External nancing can be costly for several
reasons well documented in the literature: direct or indirect costs of nancial distress, informational
asymmetries, transaction costs and agency costs between managers and outside investors. A good
hedging policy can improve a rms performance by better coordinating nancing and investing
activities. It does not mean necessarily that the rm will fully hedge to avoid risk. For instance, if a
rms investment opportunities are positively correlated with a risk factor then the rm can engage
in partial hedging. It can also engage in under-hedging to increase the exposure to the risk factor
if investment opportunities are negatively correlated with the risk.
To which extent can vertical integration reduce the need for forward contracts? The interaction
between these two risk management tools has not been much studied in the past literature. The
conventional wisdom is that short-term risk management is achieved through hedging with nancial
instruments while long-term risk management is achieved through vertical integration. The objective
of this paper is to investigate the interaction between vertical integration and hedging. This issue is
of paramount importance for industries where the price of the input is highly volatile, like commodity
and electricity markets. For that purpose, we develop simple models based on reference models that
combine hedging and vertical integration.
The paper is organized as follows. In section 2 we review the existing literature on hedging
and vertical integration. In section 3, we present the basics of Froot, Scharfstein and Steins model
(1993), a prominent model of the coordination between hedging and investment. In section 4 we
extend this model to a vertical integration framework and study the interaction between these two
tools. In section 5, we test a hypothesis and a prediction of our model on two oil companies. In
section 7 we comment the limits of our work. Section 8 ends by studying an equilibrium-model for
the electricity market. Section 9 concludes.
3/41
2 CORPORATE RISK MANAGEMENT IN THE LITERATURE
2 Corporate risk management in the literature
This section describes dierent rationales for corporate risk management in the past literature. We
analyze dierent approaches developed in the last years.

Hedging in the economic literature
Although there are stories to explain why rms might wish to hedge, it seems there is not yet
a single, accepted framework which can be used to guide hedging decisions (Froot, Scharfstein and
Stein, 1993).
Managerial motives have been pointed out by several authors. Stulz (1984) argues that hedging
reects the risk aversion of managers who hold a large portion of their wealth in the rms stocks.
Breeden and Viswanathan (1990) and DeMarzo and Due (1992) oer a dierent perspective based
on information asymmetry. They argue that managers may want to inuence how they are perceived
by the labor market by taking hedging decisions that reduce the variance of the rms value.
Smith and Stulz (1985) have put forth taxes, costs of nancial distress and debt capacity as
reasons to vertically integrate. If taxes are a convex function of earnings, a more volatile earnings
stream leads to higher expected taxes. If there are costs of nancial distress and there is an advantage
to having debt, then hedging can be used as a means to increase debt capacity because it reduces the
rms probability of default. In the same spirit, Myers (1977) and Stulz (1990) argue that hedging
can create value by reducing investment distortions which debt nance sometimes induces.
Close to these debt overhang explanations is the paper by Froot, Scharsfstein and Stein (1993)
which we use for our purpose. Their model features a single rm faced with an internal wealth
hedging decision and an investment to undertake. The rationale for hedging internal wealth is that
it reduces the variability of external funds to be raised. Reducing the variability of external funds
makes it possible to increase the rms expected prot provided the cost of external funding is
convex and the production function is concave.
In line with empirical evidence which shows that rms hedge in order to coordinate their in-
vestment opportunities with their investment capacities, Geczy, Minton and Schrand (1997), show
that rms with greater investment opportunities and tighter nancial constraints are more likely to
hedge. Empirical results of Brown (2001) and Nain (2004) support the fact that the hedging policy
of a rm depends on its competitors hedging policies. Indeed, a competitors hedging policy aects
its investment opportunities, thus its behavior in the product market, therefore the rms investment
opportunities and ultimately its hedging policy. Loss (2002) studies how the interaction between
rms, i.e. the fact that investments are either strategic substitutes or strategic complements, aect
their hedging policies in a setting of imperfect competition and nancial constraint. In equilibrium,
when investments are strategic substitutes
5
(complements), the level of hedging increases (decreases)
with the correlation coecient between shocks that aect rms internal funds.
Bessembinder and Lemmons model (2002) oers a radically dierent perspective on hedging.
The model features a nite set of producers and retailers in the electricity industry. Hedging is done
through forward contracts between producers and retailers who also have access to a spot market.
The model derives the equilibrium forward price (as electricity cannot be stored, the traditional no
arbitrage cost of carry approach is not valid) and the resulting optimal forward positions.

Vertical integration in the economic literature
Until the mid 1970s, economic thinking on vertical integration was informal. Malmgren (1961)
applied Coases theory of the rm (1937) to argue that vertical integration could reduce a rms
transaction costs. Activities which tended to uctuate, causing uctuations in prices and outputs in
the market, could be integrated and balanced against one another. When discussing the reasons for
the formation of the largest companies in the US, Chandler (1969) argues that the initial motives
for expansion or combination and vertical integration had not been specically to lower unit costs
5
When investments are strategic substitutes (complements), a marginal increase in a rms strategic variable
decreases (increases) a competitors marginal prot.
4/41
2 CORPORATE RISK MANAGEMENT IN THE LITERATURE
or to assure a large output per worker by ecient administration of the enlarged resources of the
enterprise. The strategy of expansion had come...from the desire...to have a more certain supply
of stocks, raw materials and other supplies.... In the same spirit, Porter (1980) argues that it
is to assure their supply of inputs that rms decide to integrate upstream - this has become the
traditional business explanation.
Green (1974) and Arrow (1975) are considered to be the rst economists to have analytically
investigated the eect of uncertainty on the incentives for vertical integration. Arrow analyzed a
model where vertically integrated rms obtain information on the input supplies conditions earlier
than non-integrated rms. This information advantage creates an incentive to fully vertically in-
tegrate. Green and Carlton (1979) consider that the inputs price is xed while demand for it is
stochastic. Therefore, rationing can occur and to avoid it, rms integrate to some extent.
Bolton and Whinstons model (1993) features a single rm producing an input used by two
downstream rms. Upstream rms have a random capacity, possibly insucient to meet down-
stream rms demand, resulting in ex post bargaining to transfer the input (incomplete contract-
ing). Downstream rms do not appropriate the entire surplus as the upstream rm gets part of it
resulting in ex ante ineciently low investments. Ineciently low investments create an incentive
to vertically integrate.
Emons model (1996) features a nite set of upstream and downstream rms whose input require-
ment is stochastic. Vertical integration is modeled through a downstream rms input production
capacity, which leaves space for partial vertical integration. The input is transferred downstream
through a spot market organized by an auctioneer who chooses between two prices depending on
whether downstream rms demand is larger than upstream rms capacity or not. The main result
is that full integration is an equilibrium market structure: downstream rms will always want to
integrate to some extent in order to cut down on aggregate demand and depress prices.

Discussion

In the following sections we develop two models which are extensions of two existing risk manage-
ment models: the Froot, Scharstein and Steins model (1993)
6
and the Bessembinder and Lemmons
model (2002). In these models rms use nancial instruments while they do not in the vertical
integration literature. What literature on vertical integration shows however is that a model dealing
with hedging and vertical integration should ideally include an industrial organization dimension as
well.

In FSSs hedging model future wealth is a weighted average of a random component (which
depends on a risk factor) and a non-random one whose proportions are xed in advance. In their
setup, a rm hedges to maximize the return on its investment. This return depends on a production
function which takes into account the correlation between investment opportunities and the risk
factor. In the authors setup, there is only one rm. We decided to extend the model by adding
another rm in order to have an upstream rm and a downstream one. However, literature on
vertical integration suggests that modeling vertical integration should involve a set of upstream
and downstream rms. The reason is that vertical integration aects supply and demand of the
intermediary good, therefore its price.This price eect of vertical integration explains why in Emons
model downstream rms always choose to integrate vertically to some extent.

Bessembinder and Lemmons model (2002) has this industrial organization dimension we think
is necessary. In their framwork, rms hedge not only to maximize their prot from production,
but also because hedging can be a source of revenue in itself. Contrary to FSSs setup where the
rm has no utility function (the increasing cost of marginal external funds is a way to include risk
aversion directly in the prot function), in Bessembinder and Lemmons setup market players have a
6
From now on, we call FSS to Froot, Scharstein and Stein
5/41
3 A FIRST PARADIGM OF THE BENEFITS TO HEDGE
mean-variance utility function. The model derives the equilibrium forward price of the intermediary
good. Therefore, when market players are not risk averse, they can speculate on the dierence
between the forward price and the expected spot price by entering into forward contracts.

In the following section we start with FSSs basic model and extend it to deal with vertical
integration. We then present and extend Bessembinder and Lemmons model.
3 A rst paradigm of the benets to hedge
In this section we present several important results of FSS (1993) that we will use later on for our
application on vertical integration. We consider a rm facing an investment and nancing decision
in a three-period model (we called them period zero, one and two), where we suppose the existence
of a forward market and a costly credit market.
3.1 The model setup

The rm enters period one with internal wealth w. In period one, the rm chooses its investment
I and raises the dierence e externally, therefore:
I = w +e (1)
For the sake of simplicity we assume that the interest rate is zero, that is, that outside investors
require an expected repayment of e in the second period. However, we assume there are deadweight
costs C(e) associated with external nancing. These costs could arise from direct or indirect costs of
nancial distress, informational asymmetries, transaction costs or agency issues between managers
and outside investors. For further details see Myers and Majluf (1984), Jensen and Meckling (1976),
and Myers (1977). We assume that not only these costs are an increasing function of the amount
raised externally, but also that these costs increase at the margin, that is, C
e
> 0 and C
ee
> 0.
By f(I) we denote the level of output resulting from investing I and assume that the production
technology is such that f
I
> 0 and f
II
< 0, that is, that there are technological decreasing returns
to scale. By F(I) we denote the net value of the investment
7
F(I) = f(I) I (2)
In period one, The rm chooses its investment level in order to maximize its prot (w):
(w) = max
I
F(I) C(e) (3)
The optimal investment level I
satises:
f
I
(I
) = 1 +C
e
(e
) (4)
In period two there are no optimization problems. It is only a period where output from investment
is realized and external investors are repaid.
External nancing induces therefore an optimal investment level below the rst-best level reached
in the absence of external nancing (f
I
(I) = 1). In this sense, external nancing results in under-
investment. Using the envelope theorem on (3) and the implicit function theorem on (4), the second
derivative of prots is given by:
ww
(w) = (
dI
dw
)
2
f
II
(I
) (
dI
dw
1)
2
C
ee
(e
) (5)
7
For the sake of simplicity, we denote F the net value investment function and f the production technology
6/41
As f
II
< 0 and C
ee
> 0,
ww
< 0. Besides, P
w
> 0 because
w
(w) = C
e
(e
). Thus, the prot

function is increasing and concave.

The issue of hedging arises when wealth w is random (uncertain) at period one (until now, it was
considered to be given). To understand why full hedging of wealth raises the net expected prot
of the rm, lets imagine that w is a random variable with possible states of the world {w
1
; ...; w
n
}
and associated probabilities {p
1
; ...; p
n
} such that

p
i
= 1. Since is concave:
(
p
i
w
i
) >
p
i
(w
i
) (6)
This last equation clearly shows that before period one occurs, the rm will prefer to hedge fully
to ensure wealth w =
p
i
w
i
at period one, rather than leave this wealth random.
Figure 1: Raise in value through hedging when the prot function is concave
Proposition n1: full hedging increases the net expected prot of a rm which net value invest-
ment function is concave and external nancing cost function convex.
3.2 Stochastic investment function

In this part, introduce a linear hedging decision in period zero (forward contracts in practice).
We add a period, the period zero, because the rm makes now a hedging decision before raising
external funds and investing in period one. Let h denote the hedge ratio chosen by the rm in
period zero, h [0; 1]. The wealth of the rm in period 1 is therefore:
w = w
0
(h + (1 h)) (7)
where is a normally distributed random variable representing the source of uncertainty with
mean 1 and variance . For an oil company, could be a shock on the price of oil for instance.
7/41
Figure 2: Timeline in the three-period model
To introduce changing investment opportunities, which is far more realistic, we rewrite the net
value investment function as follows:
F(I) = f(I) I (8)
where = ( )+1 and is a measure of the correlation between investment opportunities and
the risk to be hedged. For instance, if stands for a shock on the oil price, a positive means that
the higher the oil price the more investment opportunities the company has. Here the returns on
investment are related to the same random variable aecting the value of the assets. This hypothesis
is a strong one and deserves some explanations. The variable is a random variable that depends on
the result of the risky variable. It can symbolize a shock to a given production function or in other
words, a measure of the randomness of investment opportunities. In their article, Froot, Scharfstein
and Stein (1993) dene as a variable measuring (or related to) the investment opportunities of
the rm. They also consider as the variable intermediary price of a product minus a marginal
cost.
8
3.3 Solving the model by a recursive method
We solve the model with a recursive method. In period 0 the rm chooses h to maximize expected
prots E
[(w)] where (w) = max

I
F(I) C(e), which is the prot earned at period 1. This can
be rewritten as:
max
h
E
[max
I
F(I) C(e)] (9)
That is why we say we solve the model by a recursive method. We solve rst for t=1 and then
for t=0. Thus for t=1 the optimal level of investment is given by :
i
f
I
1 = C
e
(10)

At period 0, the rst order condition of this problem is:
E
[
w
dw
dh
] = 0 (11)
Using the expression of w, this equation simplies:
E
[
w
(1 )] = 0 (12)
This can be rewritten as:
cov(
w
, ) = 0 (13)
But on the other hand, using the Rubinsteins rule (1976), we get :
8
Yet, it is quite unrealistic to think that a the price of a product sold in period 2 can be known one period before.
8/41
cov [
, ] = E [
] E(
)cov [, ] (14)
Then we get :
E [
] = 0 (15)
Using the envelope theorem and the implicit function theorem it can be shown that :
=
0
(1 h)
+f
I
I
(16)
then taking the expected value we get :
E
_
0
(1 h)
+f
I
I
_
= 0 (17)
Solving for h we get :
h
= 1 +

0
E(f
I
I
)
E(
)
(18)
and using the envelope theorem we can show that
I
/f
II
, then
h
= 1 +
E[f
I
/f
II
]
0
E[
]
(19)
where the expressions inside the expected value are evaluated at I = I
*
().

Interpretations

First, this expression corroborates the optimal hedge ratio of one (i.e. full hedging) found
previously in the case of a non stochastic investment function (h
= 1 when = 0).
Second, if investment opportunities and the risk variable are positively correlated ( > 0), the
rm will not hedge fully. Recall that f
I
> 0, f
II
< 0 and
ww
< 0. To understand why, imagine
the rm is an oil company with representing the source of uncertainty arising from volatile oil
prices. If is low, the oil rm might be worse-o in terms of internal funds. Nevertheless, because
investment opportunities have deteriorated due to lower oil prices (for instance, investing in some
oil elds where extraction costs are relatively high would be not protable because of low oil prices),
the rm can accept this situation at least to some extent, i.e. it does not hedge fully (h
[0; 1[).
In fact, if investment opportunities are extremely sensitive to the risk variable , it could be that
h
< 0, i.e. the rm increases its exposure to the risk variable procyclically. In the oil rm example,
this would be the case if high oil prices are associated with so large investment opportunities that
the sole increase in oil prices would not be sucient to generate enough internal funds to nance
these investments.
Third, if investment opportunities and the risk variable are negatively correlated ( < 0), it could
be that h
> 1, ie: the rm increases its exposure to the risk variable countercyclically. This would
be the case if low values of the risk variable are associated with so large investment opportunities
(think about low prices pushing some competitors into bankruptcy) that the sole hedging of the risk
variable would not be sucient to secure enough internal funds to nance these investments.

Proposition n2: a rm with a stochastic and concave investment function will hedge partially
(h
[0; 1[) if investment opportunities are positively correlated with the risk variable.
Proposition n3: a rm with a stochastic and concave investment function will underhedge
(h
< 0), if investment opportunities are strongly positively correlated with the risk variable.
9/41
Proposition n4: a rm with a stochastic and concave investment function will overhedge
(h
> 1) if investment opportunities are negatively correlated with the risk variable.

To illustrate these propositions, we follow the work by Spano (2001) and evaluate the model using
Cobb-Douglas production and cost functions. The analytical solution is calculated as a second-order
local approximation around the expected level of investment and external nancing
9
. With this, we
can plot the optimal hedging as a function of . This is given by :
Figure 3: Optimal hedge ratio as a function of using Cobb-Douglas functions for the cost and
production.
The intuition concerning proposition 2, 3 and 4 is that the optimal hedging is decreasing with
the correlation. Using the expression of the optimal hedge ratio in (19) it is not possible to conclude
about the sign of the derivative. We must dene the form for the cost and the production function
and then solve the model. We can do this using Cobb-Douglas functions for the cost and the
production. To summarize the previous results in the case of Cobb-Douglas functions, we can say:

Proposition n5: The optimal hedge ratio of a rm with a stochastic and concave investment
function is decreasing in for a small volatility.

Proof : The intuition comes from proposition 2, 3 and 4. The idea is that
h

E[f
I
/f
II
]
0E[]
<
0 because of the concavity of the prot function. A more detailed demonstration is given in appendix.

Concerning the initial wealth, the intuition about (19) is that in a rm with a larger initial
wealth, the inuence of investment opportunities on the optimal hedge ratio will be small. Larger
rms will tend to care less about the interactions between investment opportunities and nancial
hedging. Again, we can demonstrate this for Cobb-Douglas functions following Spano (2001). The
proposition, for this type of function, is the following

Proposition n6: The optimal hedge ratio of a rm with a stochastic and concave investment
function is increasing in
0
if > 0 and decreasing if < 0.
9
See appendix for a complete list of equations
10/41
4 INTERACTION BETWEEN VERTICAL INTEGRATION AND HEDGING
4 Interaction between vertical integration and hedging

This section describes an application of the afore-presented model. We extend the use of the
FSSs model (1993) to vertical integration by considering two rms facing a three-period investment
decision. We consider an upsteam rm (producer) and a downstream rm (distributor) in an certain
industry. We use the same notations as before and again the rms have to manage their risk using
short-term hedging. We compare the dierent optimal hedgings in two cases: with and without
vertical integration. For the case where rms are separated (we call that the non-integrated case),
the optimal hedge ratio for the two rms is measured as the weighted average of the optimal hedge
ratios of each rms:
h
U+D
= h
U
+ (1 )h
D
(20)
where [0; 1] is the coecient on the weighted average
10
. This coecient can represent a sort
of relative volume eect in the forward market. In the integrated case, we call h
V
the optimal
hedge ratio when the two rms merges. Our purpose is to see under which conditions vertical
integration reduces the need of short-term hedging (i.e. when h
U+D
> h
V
).

Here, we consider only one source of uncertainty for the two rms, given by . As we said, this
risk can be interpreted in several ways. For example, as a shock on a intermediary price of the two
markets (for instance, the price of a commodity)
11
. Also, we could think the intermediary price
as a random variable which is determined by a random negotiation between the two rms. This
intermediary price uctuates between a constant marginal cost of an upstream rm and a constant
nal price of the downstream market.
Either the interpretation we consider, our model stays the same. We would like to summarize
the idea of the model built in the following gure
Figure 4: Vertical Integration in a upstream/downstream market
4.1 The non-integrated case

In this subsection we present the non-integrated case. As before, each rm (upstream and
downstream) faces a three-period investment decision (period zero, one and two). For the sake of
simplicity, we assume that the production function f is the same for both rms. The initial wealths
are noted
i
0
.
10
We have chosen an arbitrary coecient in the weighted average. Nevertheless, a simple case is to analyze =
1
2
, i.e, the case of a mean.
11
Nevertheless, the reader must understand that this is only one possible interpretation of the model. The idea of
investment opportunities can have several interpretations.
11/41
We follow a recursive route, which corresponds to the resolution of the model. The little
U
stands for Upstream and the little
D
stands for Downstream. We have i = D, P. Again, in period
0 the rm chooses its optimal hedging strategy given by h
i
in order to maximize the expected prots:
_
_
Max
h
i E
_
i
_
i
_
i
_
i
_
= max
I
i F
i
(I
i
) C(e
i
)
I
i
=
i
+e
i
w
i
= w
i
0
(h
i
+ (1 h
i
))
(21)
And the investment function are given by :
_
F
U
(I
U
) =
U
f(I
U
) I
U
U
=( ) + 1
(22)
_
F
D
(I
D
) =
D
f(I
D
) I
D
D
= ( ) + 1
(23)

where is the risk variable which we assume to be normally distributed with mean and a
variance of
2
. The same risky shock aects both rms. Here and are a measure of the
correlation between the investment opportunities and the risky variable.
The random shock
We assume = and > 0. In fact, following the interpretations given in section 2, the invest-
ment opportunities of the upstream rm and the downstream rm are correlated in the opposite
way with respect to the random variable. We can see
U
as the dierence between the intermediary
price and a constant marginal cost and
D
as the constant nal price minus the intermediary price.
Then, if we consider the intermediary price to be the only source of uncertainty (like the price of
oil for example), then we would have = and > 0 (an upstream rm positively correlated
with the intermediary price).
4.2 The integrated case

In the integrated case, the rm has to solve
_
_
_
Max
h
V E
_
V
_
V
_
= max
I
V F(I
V
) C(e
V
)
w
V
= (w
U
0
+w
D
0
)(h
V
+ (1 h
V
))
(24)
and the investment function is given by
F
V
(I
V
) =
V
f(I
V
) I
V
(25)
Since we have = then
V
=
1
2
D
+
1
2
U
= 1
12
the integrated rm faces a non-stochastic
investment function. We consider its initial wealth to be the sum of initial wealth of the downstream
and upstream rms.
12
We take a weighted average between
D
and
U
. For the sake of simplicity, we give the same weight to every
shock since nal results are not inuenced
12/41
4.3 Comparison of optimal levels of hedging in the two cases
In this subsection we use the results given in 3.1 and 3.3 to deduce the following propositions
concerning vertical integration and hedging. As we already said, we focus on comparing the weighted
average of hedging in a non-vertical integration situation h
U+D
= h
U
+ (1 )h
D
to the level
of hedging of an integrated company h
V
.

Proposition n7: Since an integrated rm faces a non stochastic investment function, it will
hedge fully, i.e. h
V
= 1.

Proof: Using proposition n1, the result is trivial.

Proposition n8: In a non integrated market, the weighted average of optimal hedging will
depend on the interaction between anticipated optimal investments, on initial wealths and the weight
coecient. In particular, we will have :
h
U+D
= 1 +
_
E(f
I
P
/
U
f
II
)
U
0
(1 )
E(f
I
P
/
D
f
II
)
D
0
_
(26)

Proof: Using the results of section 3.3, we add the optimal hedging ratios for every rm.

Proposition n9 In a non integrated market, since investment opportunities of the downstream
rm are negatively correlated with the risk variable (i.e. = is negative), the downstream rm
will overhedge (h
> 1).

Proof: We set = and > 0 then is negative and using proposition n4, the result is
trivial.

Proposition n10: In a non integrated market, if investment opportunities of the upstream
rm are not strongly positively correlated with the risk variable, then the upstream rm will hedge
partially (h
[0; 1[).

Proof: We set > 0 then using proposition n2, the result is trivial.

Thus we want to compare a weighted average of optimal hedge ratios (one larger than 1, one
smaller than 1) in the non-integrated case h
U+D
to h
V
= 1. As we said in proposition 8, several
elements aect the weighted average. For a theoretical proof of the inuence of these elements,
we need to rely on a specic form for the production and cost functions. For this, as before, we
follow the work by Spano(2001) and use Cobb-Douglas functions. The proofs are given on appendix.
However, we give an idea of every eect.

Proposition n11: If investment opportunities of the upstream rm are not strongly positively
correlated with the risk variable and is suciently small, the weighted average of the hedge ratios
of non integrated rms will be higher than that of the integrated one and thus vertical integration
will reduce the need for hedging :
h
U+D
> h
V
(27)

Proof: Using proposition 2, we get that h
U
[0; 1[ and we know that h
D
> 1 because of
proposition 9. Then, if the downstream rm has a bigger weight in the forward market than the
upstream rm, we get the result.
13/41

Proposition n12: Even if investment opportunities are strongly positively correlated with the
risk variable, for a given , the weighted average of the hedge ratios of non integrated rms will be
higher than that of the integrated one if the upstream rm has a suciently larger initial wealth than
the downstream rm (
U
0

D
0
).

Proof: Using proposition 3 and 6 we get the result.

Interpretations

We can see that there are three key elements in the analysis of corporate risk management
through vertical integration and hedging: initial wealths, correlations between investment opportu-
nities and the risk variable, the volatility of the risk variable and the volume eect (weight in the
forward market). In the following paragraph, we describes this eects.

Initial wealth eect

The initial wealth (which is a constant in the model) has an prominent role in the comparison
between the weighted average of hedgings and the vertical integrated hedge ratio. One condition to
have a benecial vertical integration (i.e., a smaller vertical integrated hedging) is that the upstream
rm has to have a suciently larger initial wealth than the downstream rm. In fact, a larger
initial wealth will provoke a decrease in the inuence of investment opportunities on hedging (i.e.
E[f
I
/f
II
]
0E[]
decreases with an increase in
0
). Then, downstream rms with small intial wealth
will tend to prefer vertical integration as a way to manage risk. Using Cobb-Douglas functions, we
solve the model and plot the hedge ratios as a function of
13
. We can see the eects of initial
wealth on the optimal hedge ratio (gure 5). If we increase the initial wealth of the upstream rm
with respect to the downstream rm, the weighted average goes up.
Figure 5: Initial wealth eect : Optimal hedge ratios (upstream in red, downstream in blue and
weighted average in green) as a function of . Left: Same initial wealth. Right: The initial wealth
of the upstream rm is 4 times larger than the downstream rm one (
U
0

D
0
). The weighted
average of separated rms is larger than 1 (i.e.h
U+D
> h
V
).
Correlation and volatility eect

The correlation between investment opportunities and the risk to be hedged is fundamental for
the comparison of hedging and vertical integration. A good measure of this correlation is given
13
3 curves: each rm separated and the weighted average. The optimal hedge ratio in the integrated case is constant
and equals to 1 (proposition 7)
14/41
5 EXTENSIONS
by (recall that
U
=( ) + 1). First, as the integrated rm has no correlation with the risk
variable, this will mean that
V
does not depend on the risk variable. In other words, the volatility
of the risk to be hedged has no impact on investment opportunities of the integrated rm. Thus,
as w
V
= w
V
0
(h
V
+ (1 h
V
)), the integrated rm will hedge totally to avoid completely the risk
variable.
Secondly, the larger is , the smaller will be h
U
. In other words, the optimal hedge ratio h
U
is a decreasing function of . The sensitivity of the correlation eect will depend mainly on the
volatility
2
of the risky variable. The higher is the variance of the risky variable, the higher is the
sensitivity of the hedging ratio to the parameter
14
. This can be demonstrated in the particular
case of Cobb-Douglas functions for the cost and the production. We illustrate this in gure
Figure 6: Volatility and the correlation eect: Optimal hedge ratios (upstream in red, downstream
in blue and weighted average in green) as a function of . At left :
2
= 3. At right :
2
= 6.
With a larger volatility the optimal hedge ratio h
becomes more sensitive to . (Other parameters:

richer downstream rm, upstream rm with bigger weight = 0.6)

The eect

Recall that we took an arbitrary and exogenous coecient for the weighted average of hedging.
This coecient measures the relative importance of every rm in the forward market. It could
illustrate several elements : a volume eect, a nancial wealth eect. It has a strong impact on
the comparison of the hedging levels as we can see in gure 14 in appendix. A lower will give
relatively importance to the downstream rm and h
D
> 1. Thus, in that case, it will be better to
engage in vertical integration.

The interaction between the correlation, the wealth eect and the eect will determine if ver-
tical integration is benecial from a risk management perspective, compared to short-term hedging.
5 Extensions
5.1 An endogenous weighted average
An interesting case to analyze is when rms are symmetric and we set =

U
0
U
0
+
D
0
i.e. we calculate
the weighted average taking as coecient the proportion of initial wealth. This measures denitely
a wealth eect. Again, we plot the results in appendix in gure 15. We see that the optimal hedging
ratio of an integrated rm is, for every , larger than the endogenous weighted average of non
14
See appendix for a proof.
15/41
5 EXTENSIONS
integrated hedgings (i.e. h
U+D
< h
V
). This result has a prominent interpretation: hedging is in
this case a better risk management approach in a short-term perspective. This follows the line of
conventional wisdom, as we saw in the introduction.
5.2 A model with several actors

There is not much research around the extension of FSS to a framework with an outside market.
Loss (2012) studies how the interactions between rms aect their hedging strategies. In his model,
all the rms hedge their risk and so optimal hedgings are inuenced by the interaction with the
competitors level of hedging. Here we take a dierent route and suppose the existence of other
rms in order to justify that our integrated rm continues to sell and buy from other rms in the
downstream and upstream market.
In the previous section we built a model with one upstream rm and one downstream rm that
vertically integrate. The optimal hedge ratio of a integrated rm is full hedging since a situation
with full hedging is equivalent to a situation where the rm knows exactly his future wealth. This
is true in an integrated rm with no competitors since there is no risk : the intermediary price
has been completely covered. Nevertheless, this is not the case with more actors.
In the following, we consider the same model as before and we add one upstream rm and one
downstream rm, which are not vertically integrated. Hence, the model is as following :
Figure 7: Vertical Integration in a upstream/downstream market with more competitors
We consider x (resp. 1x) to be the proportion of demand bought by the integrated downstream
rm to the integrated upstream rm (resp. to the other downstream rm). In the same way, we
consider y (resp. 1 y) to be the proportion of supply sold by the integrated upstream rm to the
integrated downstream rm (resp. to the other upstream rm).

Here, the random shock
V
will be given by two consecutive linear combinations between invest-
ment opportunities (depending on the sign of and on the proportions sold):
V
=
1
2
_
x
V
Nonstochastic
+ (1 x)
V
stochastic,
+
1
2
_
y
V
Nonstochastic
+ (1 y)
V
stochastic,
(28)
We get:
V
=
(x y)
2
( ) + 1 (29)
Thus, the optimal hedge ratio of the integrated rm will be given by :
h
V
= 1 + (x y)
E[f
I
/f
II
]
2
0
E[
]
(30)

16/41
5 EXTENSIONS
Proposition n13 : An integrated rm will reduce its level of hedging with respect to the weighted
average of non-integrated hedgings by selling to the competitor downstream rm and avoiding buying
from the competitor upstream rm (vertical foreclosure).

Proof

Recall that
E[f
I
/f
II
]
0E[]
< 0. The optimal hedge ratio of the integrated rm depends on the
interaction between the proportions sold and bought by the downstream rm and upstream rm.
Recall that our objective is to prove a decrease in h
V
with respect to the weighted average of non
integrated hedgings under certain conditions. Thus, an integrated rm will reduce its optimal hedge
ratio by selling to the other downstream rm (i.e. increasing 1y thus increasing y) and avoiding
buying from the other upstream rm (i.e. decreasing 1 x, thus increasing x).

Interpretations

An integrated rm will reduce its level of optimal hedging by engaging a vertical foreclosure.
This strategy consists in selling the majority of the production in the vertical integrated system and
thus avoiding the entrance of competitive upstream rms. The rm engages in an abusive use of
its power market, considered an anti-competitive strategy. Also, if the rm buys and sells the same
proportion outside the vertical system (i.e. x = y), then it will engage in an operational hedging
that will cause a full hedging strategy.
5.3 Intertemporal issues: the inuence of long-term

In this subsection, we assume = , > 0 and the same hypothesis as before. We do not add
an outside market. The issue we analyze now is whether vertical integration reduces the need for
hedging in a long-term perspective. For that purpose, we extend the three-period model by adding
more periods: Each rm will perform two hedging decisions in two dierent periods, to protect its
wealth from two independent risk variables. We will give special importance to the initial wealth
eect, which is the link between the two hedging periods. As a result, we get a ve-period model
(period zero to four), which is summarized in the following gure:
Figure 8: Five-period model timeline: the problem of each rm
To summarize, this model is equivalent to two of the previous models, one after another. The
link between the two periods of hedgings is initial wealth (since initial wealth at a period t is the
prot made by the rm at t 1). We wish to compare the optimal hedge ratios and we give special
importance to the initial wealth eect. We illustrate the new comparison in the following gure:
17/41
5 EXTENSIONS
Figure 9: Comparison of optimal hedge ratios in the two cases: Integrated and non-integrated
At a given point in time, we know that the optimal hedge ration depends on the rms wealth
and the optimal investment level. Therefore, h
2
depends on I
3
and
1
:
h
2
= 1 +
E[f(I
3
)P
ww
(I
3
)/f
II
(I
3
)])
1
E[P
ww
(I
3
)]
(31)
with I
3
such that:
f
I
(I
3
) = 1 +C
e
(e
3
) (32)
and
1
such that:
1
= F(I
1
) C(e
1
) (33)
Adjusting those results to the upward and downstream rms yields:
h
U
2
= 1 +
E[f
I
(I
U
3
)P
ww
(I
U
3
)/
U
f
II
(I
U
3
)]
[F(I
U
1
) C(e
U
1
)]E[P
ww
(I
U
3
)]
(34)
h
D
2
= 1
E[f
I
(I
D
3
)P
ww
(I
D
3
)/
D
f
II
(I
D
3
)]
[F(I
D
1
) C(e
D
1
)]E[P
ww
(I
D
3
)]
(35)
where > 0 is a measure of the (positive) correlation between investment opportunities of
the upstream rm and the factor risk . We see that the second optimal hedging depends on
the prots made previously period (the expression F(I
i
1
) C(e
i
1
) in the denominator). Recall
that
E[f
I
(I
i
3
)Pww(I
i
3
)/
i
f
II
(I
i
3
)]
E[Pww(I
i
3
)]
< 0 therefore the sign of h
i
2
1 will only depend on the sign of
i
1
= F(I
i
1
) C(e
i
1
), i = U, D.
We may think of the sign of
i
1
as the ability of a rm to make protable investments (investments
which output covers the cost of the investment and the cost associated with external funding).
Note that in this setup a not protable rm will have debt. Again, the hedge ratio h
U+D
2
of the
non-integrated rms is the average of their hedge ratios weighted by an arbitrary coecient ( i.e.
h
U
2
+ (1 )h
D
2
). Besides, the investment function of the integrated rm being non stochastic
we get h
V
2
= 1.

Then, we analyze under which conditions h
U+D
2
< 1 and h
U+D
2
> 1. It is important to
understand how the optimal hedging ratio varies as a function of the sign of the initial wealth and
the correlation to the risky variable. We summarize this is the following table:

18/41
5 EXTENSIONS
initial wealth before hedging>0 initial wealth before hedging<0
corr>0 (upstream rm) h
< 1 h
> 1
corr<0 (downstream rm) h
> 1 h
< 1
Table 1: Optimal hedging ratio variation and the sign of initial wealth and correlation

Thus we get 4 cases depending on the sign of the initial wealth before hedging i.e.
i
1
=
F(I
i
1
)C(e
i
1
), i = U, D. To nd the conditions under which h
U+D
2
> 1, we rely on the propositions
of the previous sections.

Downstream rm\Upstream rm
U
1
> 0
U
1
< 0
D
1
> 0 h
U+D
2
> 1 if
U
1

D
1
h
U+D
2
> 1
D
1
< 0 h
U+D
2
< 1. h
U+D
2
> 1 if

D
1
U
1
Table 2: Optimal hedging ratio depending on the sign of previous prots

Case n1:
U
1
> 0 and
D
1
> 0. Then h
U
2
< 1, h
D
2
> 1, and so h
U+D
2
> 1 if
U
1

D
1
Case n2:
U
1
> 0 and
D
1
< 0.Then h
U
2
< 1 and h
D
2
< 1, and so h
U+D
2
< 1.
Case n3:
U
1
< 0 and
D
1
> 0. Then h
U
2
> 1 and h
D
2
> 1, and so h
U+D
2
> 1.
Case n4:
U
1
< 0 and
D
1
< 0. Then h
U
2
> 1 and h
D
2
< 1, and so h
U+D
2
> 1 if

D
1
U
1
Hence we summarize the case 1, 3 and 4 in the following proposition.

Proposition n14: vertical integration reduces the need for hedging in the long-term in the
following three cases:
the upstream rm is not protable in the rst period while the downstream rm is;
both rms are not protable in the rst period and the downtream rms debt is high enough;
both rms are protable and the upstream rm is wealthier enough than the downstream rm.

Symetrically:

Proposition n15: vertical integration increases the need for hedging in the long-term in the
following three cases:
the upstream rm is protable in the rst period, the downstream rm is not;
both rms are not protable and the upstream rms debt is high enough;
both rms are protable and the downstream rm is wealthier enough than the upstream rm.
19/41
6 ILLUSTRATION OF THE MODEL
Intuitions
15
The upstream rm is not protable while the downstream rm is. The not protable
upstream rm (whose investment opportunities are positively correlated with the risk factor) over-
hedges its debt (negative wealth) because it wants to make its level sensitive to investment oppor-
tunities in such a way that debt decreases (resp. increase) when investment opportunities increase
(resp. decrease). The protable downstream rm (whose investment opportunities are negatively
correlated with the risk factor) overhedges its wealth in order to increase it when investment oppor-
tunities increase. Recall that the downstream rm overhedges because its investment opportunities
are negatively correlated with the risk factor and that overhedging makes the downstream rms
level of wealth countercyclical with the risk factor. Therefore, two such rms will have a weighted
average optimal hedge ratio greater than one, that is, greater than the optimal hedge ratio of
vertically integrated rms.
Both rms are not protable and the downstream rms debt is high enough. Now,
if the downstream rm is not protable as well (i.e: it has debt), it will either hedge partially or
underhedge depending on how strongly (negatively) correlated are its investment opportunities with
the risk factor. If the downstream rms debt is high enough, the overhedging of the upstream rm
will outweigh downstream rms partial hedging or underhedging, making therefore the optimal
hedge ratio of the vertically integrated rm lower than the weighted average of optimal hedge ratios
of separate rms.
both rms are protable and the upstream rm is wealthier enough than the down-
stream rm. Finally, if both rms are protable (i.e.: they have some wealth), the upstream rm
will hedge partially or underhedge because its investment opportunities are positively correlated
with the risk factor, while the downstream rm will overhedge because it investment opportunities
are negatively correlated with the risk factor. If upstream rms wealth is high enough (high enough
given the dierence in the optimal hedge ratios and the dierence in wealth), the optimal hedge
ratio of the vertically integrated rm is lower than the weighted average of optimal hedge ratios of
separate rms.
6 Illustration of the model
6.1 Testing the sign of and the hedging ratio in an oil company
In this section we try to test some of our results with data taken from an oil company. We examine
the sign of of a downstream oil rm and the hedge ratio of an integrated oil rm. Recall that
we have assumed that was negative for a downstream rm (and positive for an upstream rm);
we have shown that the optimal hedge ratio for a vertically integrated rm was 100%. We test this
model assumption and this proposition.

The sign of of a downstream rm: the PKN Orlen case

The parameter is a measure of the correlation between the risk factor and investment oppor-
tunities. The risk factor we consider here is oil price because this is the risk factor which accounts
for vertical integration.
We test the sign of in the case of a specic downstream rm, PKN Orlen, a major
16
European
oil rener and petrol retailer. This company perfectly ts such an analysis as it is not vertically
integrated and data is easily available.
15
We interpret only proposition n14 since the n15 is the symetric case.
16
PKN Orlen is Central Europes largest publicly traded rm
20/41
As a proxy for a companys investment opportunities, we use Tobins Q, i.e., market value
17
of assets over their book value. Indeed, the higher Tobins Q, the higher the market values assets
in place, that is, the more it expects assets in place to generate cash-ows, that is, the more it
expects good investment opportunities. The risk factor we consider is oil price. We use Brent oil as
a benchmark for oil price.
In order to check how PKN Orlens Tobins Q varies when the oil price varies, we run a standard
linear regression analysis where ln(Q) is the explained variable and ln(Oil) is the explanatory
variable using data ranging from Q4 2004 to Q4 2011 (i.e: 29 periods):

ln(Q
i
) = +ln(Oil
i
) +
i
(36)

We obtain the following estimates of (, ) and the following t-statistics (t( ), t(
)):

= 0, 42
= 1, 29
t( ) = 3, 05
t(
) = 2, 18
(37)
We perform a test to check the sign of . The one sided pvalue of t( ) is 0, 25%. Thus, we can
reject the hypothesis that is positive with a probability of 99, 75%. The slope of this regression is
therefore very statistically signicant.
This regression suggests that when the price of oil increase by 1%, investment opportunities are
expected to decrease by 0, 4%. Our intuitive assumption about the sign of is therefore justied.
The data we took for our regression is given in the appendix. We show the regression in the following
gure :
Figure 10: Tobins Q as a function of oil price for PKN Orlen

17
market value of assets = market capitalization + market value of debt market capitalization + book value of
net debt
21/41
The hedge ratio of an integrated oil rm: the OMV case

Contrary to PKN Orlen, OMV
18
is an integrated oil (and gas) rm. Our model predicts there-
fore full hedging of oil production. To test this prediction, we simply checked what proportion of
produced oil had been hedged by OMV in the past few years (results are presented in the table
below).
The good news is that OMV has been fully hedging its oil production for the past three years.
The rms rationale for hedging, as the annual reports states it, is to secure cash-ows to nance
investments and to maintain an investment grade credit rating. Obviously, as in our model, the
investment aspect of hedging is taken into consideration by the company, as well as the cost of
external funds aspect (a credit rating usually determines the cost of external funds). A worse news
is that OMV did not hedge oil production at all in 2007, arguing that cash-ows were strong
enough. This is in contradiction with our model which stipulates that full hedging should always
occur (for an integrated rm), whatever the wealth of the company.

Oil Production (Q)
Oil Hedging
instruments secured price hedge ratio
2011 50000m bbl swaps $97.07/bbl 100%
2010 63000m bbl puts $54.20/bbl 100%
2009 65000m bbl puts spreads $80/bbl or $95/bbl 100%
2008 60900m bbl puts NC 31%
2007 59800m bbl no hedging concluded no hedging concluded 0%
Table 3: Oil hedging at OMV. Source : OMV Annual Reports.
As assumed in the model, the sign of was found to be negative in the case of an oil distributor,
PKN Orlen. As predicted by our model, the hedging ratio was found to be equal to one in the case
of an integrated oil company, OMV.
6.2 Delta Airlines: an example of vertical integration

Delta Airlines has recently bought for $180M, an oil renery from Phillips 66 located in Penn-
sylvania, USA. Delta bought the renery in order to reduce the airlines fuel costs. Indeed, 80%
of Deltas US fuel needs should be provided by such purchase. If we do the analogy with our
rst model, the oil renery represents the upstream rm whereas Delta Airlines is the downstream
rm. Delta Airlines is a large airline company (Market capitalization of 8.5 billions dollars, earnings
greater than 500 million dollars. . . ) and the renery a small one (the renerys purchasing cost is
nearly equal to a jumbo jets purchasing cost). Thus, we could say this example corresponds to the
case where the downstream rms volume is much greater than the upstream rms volume. If we
take into consideration the volume eect (given by ), our model predicts that vertical integration
is better in this case.
The aim of the vertical integration is to be protected against oils price volatility. By doing this,
the vertical integration allows to guarantee a xed level of income. Furthermore, Delta Airlines wants
to spend another $ 100M to convert the renerys operations to maximize jet fuel production. This
new spending can be seen as an investment related to good investment opportunities. Therefore,
if we follow our model, the vertical integration is totally logical and obvious (here, both wealth
and correlation eects lead to concentration) as we bear in mind that the rationale for hedging
in FSSs paper is to protect the rm against external nancing assumed to be more costly than
18
OMV is Austrias largest listed industrial company in terms of turnover and one of the largest integrated oil and
gas groups in Central Europe.
22/41
7 COMMENTS, CRITICS
internal generated funds. However, some analysts have explained they were really skeptical about
this purchase. Why? That is one of the weaknesses of the model. Indeed, the model only focuses
on the hedging of the stochastic input and not on the global industrial risk.
Therefore, on the one hand, Delta decreases his sensibility to the oil variation which is certainly
a way to be protected. But, on the other hand, such strategy increases the break-even point because
the integration of the renery will create new xed costs for the downstream rm (all the xed costs
supported previously by the upstream rm). The strategys impact on the break-even point is also
a fundamental question in Corporate Finance. The higher the break-even point is, the more risky
will be the rm. What would happen if the oil price suddenly fell? The integrated rm would stay
with his high xed costs due to the upstream rm. A way to improve the model would be to add
this xed cost.
7 Comments, critics
7.1 The cost function of raising external funds
Literature oers two dierent perspectives on the consequences of hedging on external nancing.
On the one hand, when capital market imperfections make external nance more costly than
internal funds, then hedging reduces nancing costs because it reduces the variability of external
funding. On the other hand, DeMarzo and Due (1995) show that hedging can reduce information
asymmetry problems by cleaning a rms prots from the noise related to exogenous factors.
Hedging therefore improves the quality of the signal about the managers ability.
Recent research
19
has provided empirical evidence showing that hedging rms raise more exter-
nal funds than non-hedging rms. This result contradicts somewhat FSSs model.
The cost of external funds function of FSS (1993) is a pillar of their and therefore of our results,
and yet there is no consensus about its form and its existence in the literature. Our criticism of this
cost function is rst and foremost its exogenity but also, to a lesser extent, its convexity.
Exogenity. In FSSs model (1993), the cost of external funds function is exogenous. This does
not seem very accurate as the cost function should depend on the hedging strategy and thus be
endogenous. In fact, some articles
20
consider that the marginal cost of external funds depends on
the eectiveness of hedging.
Convexity. FSS (1993) use this arbitrary characteristic because it is a condition under which
hedging increases value. However, it is dicult to imagine why it should always be the case that the
cost function is convex, although there is a large literature which supports theoretically this view
of external nancing.
7.2 The time horizon

The rationale for our extension of FSSs paper is to protect an integrated rm against a stochastic
investment opportunity and thus against external nancing. Yet, the purpose of vertical integration
in a hedging perspective is to reduce the risk arising from the inputs price volatility. This rationale
can be used for commodity prices but it is dicult to think this model as a useful modelisation for
other particular markets, such as the electricity one, where the intermediary spot price has a strong
volatility in the very short-term.
19
S.Magee, (2009), The Eect of Foreign Currency Hedging on External Financing, Macquarie University, Applied
Finance Center
20
A. Mello and J. Parsons, 2000, Hedging and Liquidity, The Review of Financial Studies, Vol 13, No 1, pp
127-153.
23/41
8 A MODEL FOR THE ELECTRICITY MARKET
The time horizon in this model is therefore that of an investment. This is an important limit of
this model. With such a time horizon, hedging is not considered as a way to reduce volatilities at
the very short-term, as was seen in the electricity market. Moreover, the simplicity of this model is
its strength but also its limitation, since markets need to rely on more specic models. In section
8 we extend a framework made exclusively for the electricity market. We will see that this model
made-to-measure is a more sophisticated way of treating our problem.
7.3 Perfect price anticipation
Another limitation is the interpretation of the random shock . Recall that the model has three
periods: the rm knows the value of the risk variable in period 1 and then can deduce the value
of . In period 2 it earns the prot given by investment, with the investment opportunity known at
t = 1. If the variable is said to be interpreted as a price (like in the case of vertical integration),
then we are perfectly anticipating prices with one period in advance. This is very unrealistic, in
particular in markets with high volatility.
8 A model for the electricity market
In this section we do not try to nd other interpretations concerning the interaction between hedging
and vertical integration. Instead, we show an extension made for a specic market : the electricity
market. In particular, we can see that results are easier to interpret. Thus, this framework proposes
another rationale to study the corporate risk management. We begin with a presentation of the
electricity market.
8.1 Vertical integration in the electricity market

In the electricity industry, the question about corporate risk management seems essential for
several reasons : uncertainty in the demand, spot price volatility
21
, lack of market exibility and
risk aversion. For a long time, the electricity industry has been characterized by an important level
of vertical integration. However, in the last two decades, many market regulatory reforms (Joskow
2005, Gilbert and Newberry 2006) have been made to facilitate competition in the upstream and/or
downstream market. According to Aid, Chemla, Porchet and Touzi (2011), throughout these reforms
a large number of rms went bankrupt (Enron, Pacic Gas & Electricity in 2001, British Energy in
2002). In some countries, like New Zealand where the market has not been subject to regulation,
most of retailers that were not integrated had to exit the market, since integrated producers had an
important strategic advantage in terms of risk management.
Table 4 in appendix shows a study made by the IEA
22
about the organization and the capital
structure of the electricitys industry in the 15 countries with largest consumption. We can see there
is a strong level of vertical integration in these markets. In China for instance, several regional rms
are responsible for the electricitys production, transport, local distribution, and sales in restricted
area. According to Aid, Chemla, Porchet and Touzi (2011), non-integrated retailers can nd dicult
to stay in the market, unless there is large forward market in which they can rely to manage their
risk.
In the following, we would like to know to which extent vertical integration has helped manage
the risk and how this can aect the forward market. In particular, we focus on the relation between
forward contracts and vertical integration in the electricity market.
21
According to Lemmon and Bessembinder (2002) power prices for day time delivery are typically more than twice
as high as for nighttime delivery
22
International Energy Agency, www.iea.org
24/41
8.2 An adapted model for the electricity market
We rely on the model developed by Bessembinder and Lemmon (2002)
23
to analyze how vertical
integration aects optimal forward positions. We summarize the most important results in the rst
part. Then, we extend the model to a vertical integration study, where a retailer merges with a
producer and we study how the vertical integration inuences the forward market. The purpose of
studying this model is that it is a much more accurate model, more comprehensive and has a better
description of reality. Besides, it relies on a general equilibrium between retailers and producers.
We will see later on that the simulation of the prices has the shape of the spot price of electricity
in reality.
8.2.1 The setup
Bessembinder and Lemmon analyze power production during a single future time period. The model
features N
P
identical producers who sell power to N
R
identical retailers in the wholesale market.
Retailers sell power to nal consumers at a xed unit price P
R
. The realized demand for retailer
i is an exogeneous random variable denoted Q
Ri
. It is assumed that power companies are able to
forecast demand in the immediate future with precision (Q
Ri
is known at t = 0) and hence know
the spot price in the immediate future.
Each producer i has a cost function which depends on the total produced quantity:
C
i
(Q
Pi
) = F +
a
c
(Q
Pi
)
c
(38)
where F are xed costs and Q
Pi
is the output of producer i, and c 2 (marginal production
costs increase with output). In the wholesale spot market, producers sell to retailers who in turn
distribute power to customers. Let P
W
denote the wholesale spot market, Q
Pi
W
the quantity sold
by producer i in the wholesale spot market, Q
Pi
F
and Q
F
Rj
the quantity that producer i and retailer
j has agreed to deliver (or purchase if negative) in the forward market at the xed forward price
P
F
. We also note Q
D
=
Q
Rj
and we notice that

Q
F
Pi
= 0. We summarize the model in the
following gure :
Figure 11: The electricity market by Bessembinder and Lemmon (2002)
23
Bessembinder, Lemmon, 2002, Equilibrium Pricing and Optimal Hedging in Electricity Forward Markets, The
Journal of Finance.
25/41
8.2.2 Problem and optimal solution
The model is solved in two steps. First, optimal behavior in the spot market is derived assum-
ing forward positions are given. Second, optimal forward positions are derived assuming optimal
behavior in the spot market is given

Optimal behavior in the spot market

The ex post prot of producer i is given by earning minus costs. Retailers buy on the wholesale
spot market the dierence between realized retail demand and their forward positions. The ex post
prot are then given by :
_

Pi
= P
W
Q
W
Pi
+P
F
Q
F
Pi
C
i
(Q
W
Pi
+Q
F
Pi
)
Rj
= P
R
Q
Rj
+P
F
Q
F
Rj
P
W
(Q
Rj
+Q
F
Rj
)
(39)
Thus, for a given level of forward demand, the rst-order conditions and equilibrium conditions
lead, to
24
:
_
P
W
= a(
Q
D
N
P
)
c1
Q
W
Pi
=
Q
D
N
P
Q
F
Pi
(40)
To illustrate the quality of the results, we perform a simulation using Microsoft Excel 2007
over 10 periods, with 10 producers and 10 retailers. The spot demand is supposed be uniformly
distributed over [0; 1]. In gure 12, we see that the shape of our simulations is similar to the graphic
showing Leipzig Power Exchange electricity spot-price over the period 16 June 200015 October
2001:
Figure 12: Leipzig Power exchange electricity spot-price & model simulation with 10 retailers and
10 producers
Optimal behavior in the forward market (mean-variance optimization problem)

Lets assume that the objective function of producers ans retailers are, with A 0:
_
max
Q
F
P
i
E(
Pi
)
A
2
V ar(
Pi,
)
max
Q
F
R
i
E(
Ri
)
A
2
V ar(
Ri,
)
(41)
The rst order condition yields
25
:
24
See appendix for further details
25
See appendix for further details
26/41
_
Q
F
Pi
=
P
F
E(Pw)
AV ar(P
W
)
+
Cov(
Pi
,P
W
)
V ar(P
W
)
Q
F
Ri
=
P
F
E(Pw)
AV ar(P
W
)
+
Cov(
Ri
,P
W
)
V ar(P
W
)
(42)
where
Pi
and
Rj
denote respectively the prots of producer i and retailer j with no forwards.
We can show that the equilibrium forward price is:
P
F
= E(P
W
)
AN
P
(N
P
+N
R
)ca
x
_
cP
R
Cov(P
x
W
, P
W
) Cov(P
x+1
W
, P
W
)
(43)
8.2.3 Extension : Vertical integration

We assume know that retailer R
i
merges with producer P
j
and that retailer R
i
no longer goes on
the wholesale market to buy the input when producer P
j
and retailer R
i
merge. We also assume that
producer P
j
faces no capacity constraint. Retailers demand on the wholesale market is Q
D
Q
Ri
.
As a result, producer P
j
produces the quantity of input retailer R
i
needs, that is, Q
Ri
, and like
before the merger, an even share of the supply on the wholesale market, that is, (Q
D
Q
Ri
)/N
P
.
Let denote P
W
the new wholesale equilibrium price. As before, forward contracts are in zero net
supply. Retailers show up on the wholesale market with a demand equal to Q
D
Q
Ri
(instead of
Q
D
). The following gure summarizes the idea :
Figure 13: Vertical Integration : Only the retailer leaves the downstream market. The integrated
producer continues to sell electricity to other retailers.
Also, with the previous assumptions, the optimal wholesale spot price is given by :
P
W
= a(
Q
D
Q
Ri
N
P
)
c1
(44)
Besides, following the merger, the new forward price is:
P
F
= E(P
W
)
N
P
N
ca
x
_
cP
R
Cov(P
x
W
, P
W
) Cov(P
x+1
W
, P
W
)
_
(45)
where N
= (N
R
+N
P
1)/A. The ex post prot with no forwards of the merged entity is:
Ri+Pj
= P
R
Q
Ri
+P
W
Q
W
Pj
F
a
c
(Q
Ri
+
Q
D
Q
Ri
N
P
)
c
(46)
27/41
Using equation (42), the optimal demand for forward contracts of the merged entity is:
Q
F
Ri+Pj
=
P
F
E(P
w
)
AV ar(P
W
)
+
Cov(
Ri+Pj
, P
W
)
V ar(P
W
)
(47)

In an atomistic market...

In an atomistic market, we consider that Q
D
Q
Ri
, i.e. the total demand for electricity is much
larger than the demand for electricity of the retailer i. Using (44), this implies that the wholesale
spot price is unchanged after the merger, i.e.:
P
W
P
W
(48)
Thus, the wholesale spot price depends only on agregate demand Q
D
. We also assume a large
number of economic agents so as to make the following simplication:
N N
(49)
This approximation and this assumption imply that:
P
F
P
F
(50)
As we did in the previous model, we want to see how the optimal number of forward contracts
varies with or without vertical integration. For this, we can adopt two point of views.
A macroeconomic point of view where we compare the sum of optimal forward contracts to
the integrated case, i.e. Q
F
Ri+Pj
(Q
F
Ri
+Q
F
Pj
).
A corporate point of view where the retailer compares its demand on forward contract in the
integrated case and in the non integrated case, i.e. Q
F
Ri+Pj
Q
F
Ri
.

Looking from a macroeconomic perspective at how vertical integration aects (optimal) forward
positions of two rms which integrate vertically means comparing the optimal forward position of
the integrated rm with the sum of the optimal forward positions of the separate upstream and
downstream rms and adjusting for intra-companies pre-merger forward positions. As upstream
rms are symetric and downstream rms as well, an upstream rm enters into forward contracts
with all the downstream rms equally and a downstream rm enters into forward contracts with all
the upstream rms equally. Therefore, (optimal) forward contracts of a specic upstream rm P
j
where the counterparty is a specic downstream rm R
i
amounts to
1
N
R
Q
F
Pj
and forward contracts
of a specic downstream rm R
i
where the counterparty is a specic upstream rm P
j
amounts to
1
N
P
Q
F
Ri
. The variation in forward positions from a macroeconomic perspective is Q
F
Ri+Pj
(Q
F
Ri
+
Q
F
Pj
+
1
N
R
Q
F
Ri
+
1
N
P
Q
F
Pj
).

Using the approximations P
W
P
W
, P
F
P
F
and a large number of economic agents, we can
further approximate the above formula:
Q
F
Ri+Pj
(Q
F
Ri
+Q
F
Pj
+
1
N
R
Q
F
Ri
+
1
N
P
Q
F
Pj
)
cov(
Ri+Pj

Ri

Pj
, P
W
)
V (P
W
)

P
F
E(P
W
)
AV (P
W
)
(51)

Besides, using (47), from a corporate point of view we will have :
Q
F
Ri+Pj
Q
F
Ri
=
Cov(
Ri+Pj

Ri
, P
W
)
V ar(P
W
)
(52)
28/41
9 CONCLUSION
Ri+Pj

Ri
P
W
Q
Ri
(1
1
N
P
)
a
c
(Q
Ri
+
Q
D
Q
Ri
N
P
)
c
(53)

Q
F
Ri+Pj
Q
F
Ri
will be negative if
Ri+Pj

Ri
and P
W
vary on average in opposite directions.
Recall that P
W
= a(
Q
D
N
P
)
c1
. Assuming the number of producers xed, P
W
increases (decreases) if
aggregate demand increases (decreases). Therefore, Q
F
Ri+Pj
Q
F
Ri
will be negative if the increase
in aggregate demand induces on average an increase in (variable) production costs which outweighs
the benet from higher wholesale revenues.

Proposition n 16: In an atomistic market, upstream vertical integration will reduce a retailers
optimal number of forward positions if an increase in aggregate demand induces on average an
increase in production costs which outweighs the benet from higher wholesale revenues.

The intuition is that if an increase (decrease) in aggregate demand induces on average an increase
(decrease) in production costs which outweighs the benet (loss) from higher (lower) wholesale
revenues, then the vertically integrated rms revenues will be less sensitive to wholesale spot prices
than the distributors revenues.
9 Conclusion
The purpose of this paper is to model the interaction bewteen two tools of corporate risk manage-
ment: hedging with nancial instruments and vertical integration.
We have made several extensions to Froot, Scharfstein and Steins framework. First, we have
added a rm in order to model a producer-distributor relationship in a short-term model. The
important assummption we make is that investment opportunities of the producer are positively
correlated with the risk factor, which stands for a shock on the price of the intermediary good, while
investment opportunities of the distributor are negatively correlated with this single risk factor.
Then, we have considered these two rms in a long-term setup to analyze the impact of both rms
wealth after the rst investment.Finally, we have introduced an outside market in the short-term
setup to take into account horizontal interactions between rms.
We have shown that in the short-term setup, when both rms have positive initial wealth, the
optimal hedging strategy for an integrated rm is to hedge fully, for a non integrated upstream
rm, to hedge partially (i.e: limit its exposure to the risk varible) or to underhedge (i.e: increase
its exposure to the risk variable), for a non integrated downstream rm, to overhedge (i.e: have a
counter-cyclical exposure to to the risk variable). In the long-term setup, we have shown that the
optimal hedge ratio depends not only on the correlation between the risk variable and investment
opportunities (correlation eect), but also on the protability of the rms investments (wealth
eect) and the volatility of the risk factor. A protable (resp. not protable) upstream rm will
have to hedge partially or under-hedge (resp. overhedge) while a protable (resp. not protable)
downstream rm will have to overhedge (resp. hedge partially or under-hedge). Therefore, from
a macroeconomic point of view (or that of a holding company owning both rms), how vertical
integration will aect the need for hedging will depend both on the sign of their respective wealths
and their relative size. Finally, we have shown that when the market consists of an integrated rm
facing an outside market, its optimal level of hedging will be reduced by selling to other downstream
rms and avoiding buying from other upstream rms (vertical foreclosure).
We came up with some meaningful results with this model. Some of its hypothesis and predictions
have been tested in the case of two oil companies. Nevertheless, several of its limits must be pointed
out. Some of them are a legacy of FSSs model. First, the cost function is assumed to be increasing
at the margin. Although this assumption seems intuitive in some external nancing cases, like
bank nancing, we do not really see why this should always be the case. Second, this cost function
29/41
9 CONCLUSION
is exogenous. It should instead rather be determined endogenously. An important limit of this
model is also the fact that at the time the rm invests, it perfectly knows the return it will get
on its investment, as uncertainty about wealth and investment opportunities exists only up to this
moment. Another limit is that the time horizon is that of an investment. With such a time horizon,
hedging is not considered as a tool to cope with extreme short-term volatilities such as in the
electricity market. Finally, the simplicity of the model does not make it suitable for the use in
a specic industry. To adress this issue, we have extended a model designed with the electricity
market in mind.
We have applied a prominent model by Lemmon and Bessembinder (2002) to the situation
where a retailer merges with a producer. Contrary to the previous model, this one has denitely an
industrial organization avour. An interesting and useful result is that the optimal forward position
of a market player depends on (a) the sensitivity of its prots to the wholesale spot price of the
intermediary good (b) the dierence between the forward price and the expetected spot price (the
rm can speculate on this dierence if it is not risk-averse).
We have shown that in an atomistic market, upstream vertical integration will reduce a retailers
optimal number of forward positions if an increase in aggregate demand induces on average an
increase in production costs which outweighs the benet from higher wholesale revenues. The
intuition is that if this condition is satised, the vertically integrated rms revenues will be less
sensitive to wholesale spot prices than the retailers ones.
Our work could be carried on. In every step we made we found new challenges, new ideas.
Some of them were not developed in this article. We have treated every extension (short-term
perspective, long-term perspective, outside market) separately. It would be interesting to analyze
how the extensions we made interact. We have also supposed that the producer and the retailer
had their investment opportunities aected with the same magnitude (but in the opposite way)
by the risk factor. Research could be extended to the case where rms have their investment
opportunities aected by the risk factor with dierent magnitudes. This would make the model
more realistic. Empirical research could also be developped as we have only tested one hypothesis
and one prediction in the case of two oil rms. In the long-term perspective, we could link both
shocks by some correlation, creating a new link between the periods.
30/41
10 APPENDIX
10 Appendix
10.1 Equations : First model

I = w +e (54)
F(I) = f(I) I (55)
(w) = max
I
F(I) C(e) (56)
f
I
(I
) = 1 +C
e
(e
) (57)
ww
(w) = (
dI
dw
)
2
f
II
(I
) (
dI
dw
1)
2
C
ee
(e
) (58)
w = w
0
(h + (1 h)) (59)
F(I) = f(I) I (60)
= ( ) + 1 (61)
max
h
E
[max
I
F(I) C(e)] (62)
i
f
I
1 = C
e
(63)
cov(
w
, ) = 0 (64)
h
= 1 +
E[f
I
/f
II
]
0
E[
]
(65)
_
_
Max
h
i E
_
i
_
i
_
i
_
i
_
= max
I
i F
i
(I
i
) C(e
i
)
I
i
=
i
+e
i
w
i
= w
i
0
(h
i
+ (1 h
i
))
(66)
_
F
U
(I
U
) =
U
f(I
U
) I
U
I
=( ) + 1
(67)
_
F
D
(I
D
) =
D
f(I
D
) I
D
D
= ( ) + 1
(68)
_
_
_
Max
h
V E
_
V
_
V
_
= max
I
V F(I
V
) C(e
V
)
w
V
= (w
U
0
+w
D
0
)(h
V
+ (1 h
V
))
(69)
F
V
(I
V
) =
V
f(I
V
) I
V
(70)
31/41
10 APPENDIX
V
=
1
2
D
+
1
2
U
(71)
h
U+D
= h
U
+ (1 )h
D
= 1 +
_
E(f
I
P
/
U
f
II
)
U
0
(1 )
E(f
I
P
/
D
f
II
)
D
0
_
(72)
F
V
(I
V
) =
V
f(I
V
) I
V
(73)
V
=
(x y)
2
( ) + 1 (74)
h
V
= 1 + (x y)
E[f
I
/f
II
]
2
0
E[
]
(75)
f
I
(I
3
) = 1 +C
e
(e
3
) (76)
1
= F(I
1
) C(e
1
) (77)
h
U
2
= 1 +
E[f
I
(I
U
3
)P
ww
(I
U
3
)/
U
f
II
(I
U
3
)]
[F(I
U
1
) C(e
U
1
)]E[P
ww
(I
U
3
)]
(78)
h
D
2
= 1
E[f
I
(I
D
3
)P
ww
(I
D
3
)/
D
f
II
(I
D
3
)]
[F(I
D
1
) C(e
D
1
)]E[P
ww
(I
D
3
)]
(79)
10.2 Equations of the model solution in the case of Cobb-Douglas func-
tions

_
f(I) =
I
1
1
< 1
(80)
_
C(e) =
e
1+
1+
0 < < 1
(81)
Second order approximations around expected values of investment and external nancing:
_
f(I) = aI
2
+bI +k
a < 0, b > 0,
(82)
_
C(e) =
c
2
e
2
+re +z
c > 0, r > 0,
(83)
where the values of a, b and c are calibrated using and .

Optimal hedge ratio :
h
= 1 +
(1 +r c
0

bc
a
)((a c)
2
+ 3a
2
2
)
0
(a c)((a c)
2
+ 3ac
2
2
)
(84)

32/41
10 APPENDIX

Eects on the optimal hedging ratio

Correlation eect The optimal hedging ratio, h
, is a decreasing function of the parameter

for any
2
lower than a critical value
2
=
(ac)
2
3
2
ac
Proof : Given the conditions on the constants, the expression (1 + r c
0

bc
a
) is positive.
Besides, ((ac)
2
+3a
2
2
) is also positive and (ac) < 0, . Then the sign of the fraction depends
on ((ac)
2
+3ac
2
2
) which is positive for any
2
lower than a critical value
2
=
(ac)
2
3
2
ac
. Thus,
for
2
<
2
, the optimal hedging ratio, h
, is a decreasing function of the parameter .

Eect of initial wealth Since c > 0 the higher is the initial wealth of the rm, the lower is the
sensitivity of the hedging ratio h
to the parameter . Therefore, the optimal hedging ratio is an

increasing function of the initial wealth if > 0 and a decreasing function if < 0.
Proof : The factors containing
0
can be insulated:
(1+rc0
bc
a
)
0
=
(1+r
bc
a
)
0
c. Given the
conditions on the constants, this expression is positive (see Spano (2001) for further details). Hence
it is lower as
0
is higher. Hence, the value of h
is closer to 1 (full hedging) as

0
is higher, for any
value of . In other words, h
is less sensitive to the relation parameter .

Eect of the volatility h
is more sensitive to the correlation parameter when the volatility

increases.
Proof : When the variance increases, the factor containing the variance in the numerator in-
creases by the amount 3a
2
2
, while the factor in the denominator decreases by 3a
2
2
. Hence,
for a given value of , the value of h
is farther from 1 as the variance is higher (i.e. h
is more
sensitive to the correlation parameter ).
10.3 Eects on the optimal hedging
10.3.1 The eect
Figure 14: Volume eect : Optimal hedging ratios (upstream in red, downstream in blue and
weighted average in green) as a function of . Left: Same weight i.e. = 0.5. Right: the weight of
the upstream rm is 0.7 ( = 0.7). We see that the weighted average of hedgings moves down (i.e.
h
U+D
< h
V
).
33/41
10 APPENDIX
10.3.2 The endogenous weighted average
Figure 15: Endogenous weighted average by initial wealth : Optimal hedge ratios (upstream in
red, downstream in blue and weighted average in green) as a function of . The coecient of the
weighted average of the upstream rm is

U
0
U
0
+
D
0
. We see that the weighted average of hedgings is
below 1 (i.e.h
U+D
< h
V
) for all values of . Other parameters: same initial wealth for both rms.
10.4 The electricity market

Optimal behavior in the spot market

In the wholesale spot market, producers sell to retailers who in turn distribute power to cus-
tomers. Let P
W
denote the wholesale spot market, Q
Pi
W
the quantity sold by producer i in the
wholesale spot market, Q
Pi
F
and Q
F
Rj
the quantity that producer i and retailer j has agreed to
deliver (or purchase if negative) in the forward market at the xed forward price P
F
. The ex post
prot of producer i is:

Pi
= P
W
Q
W
Pi
+P
F
Q
F
Pi
F
a
c
(Q
Pi
)
c
(85)
where
Q
Pi=
Q
W
Pi
+Q
F
Pi
Retailers buy on the wholesale spot market the dierence between realised retail demand and
their forward positions. The ex post prot of retailer j is:
Rj
= P
R
Q
Rj
+P
F
Q
F
Rj
P
W
(Q
Rj
+Q
F
Rj
) (86)
The rst-order condition of (16) gives:
Q
W
Pi
= (
P
W
a
)
x
Q
F
Pi
(87)
where
x = 1/(1 c)
At equilibrium, total retail demand Q
D
=
Q
Rj
equals total production

Q
W
Pi
and forward
contracts are in zero net supply, ie.:

Q
F
Pi
= 0. This implies:
34/41
10 APPENDIX
P
W
= a(
Q
D
N
P
)
c1
(88)
and
Q
W
Pi
=
Q
D
N
P
Q
F
Pi
(89)
Optimal behavior in the forward market (mean-variance optimization)

Lets assume that the objective function of market players is, with A 0:
max
Q
F
{P,R}i
E(
{P,R}
i
)
A
2
V ar(
{P,R}
i,
) (90)
where
{P,R}
i
=
{P,R}
i
+P
F
Q
F
P
W
Q
F
(91)
where
Pi
and
Rj
denote respectively the prots of producer i and retailer j with no forwards:
Pi
= P
W
(
Q
D
N
P
) F
a
c
(
Q
D
N
P
)
c
(92)
Rj
= P
R
Q
Rj
P
W
Q
Rj
(93)
The rst order condition of (73) yields:
Q
F
{P,R}i
=
P
F
E(P
w
)
AV ar(P
W
)
+
Cov(
{P,R}i
, P
W
)
V ar(P
W
)
(94)

Substituting the covariance term with equations (75) and (76) and using the fact that
Q
D
N
P
=
1
a
x
P
x
W
and a(
Q
D
N
P
)
c
=
1
a
x
P
x+1
W
, the optimal quantities of forward sold by the producer and the
retailer can be rewritten as:

Q
F
Pi
=
P
F
E(P
W
)
AV ar(P
W
)
+
1
a
x
_
1
1
c
_
Cov(P
x+1
W
, P
w
)
V ar(P
W
)
(95)
Q
F
Rj
=
P
F
E(P
W
)
AV ar(P
W
)
+P
R
Cov(Q
Rj
, P
W
)
V ar(P
W
)

Cov(P
W
Q
Rj
, P
W
)
V ar(P
W
)
(96)

Equating the sum of the forward positions across retailers and producers to zero, using the fact
that

Q
Rj
= Q
D
, P
W
= a(
Q
D
N
P
)
c1
yields:

Q
F
Rj
+
Q
F
Pi
= N
R
P
F
E(P
W
)
AV ar(P
W
)
+P
R
Cov(Q
D
, P
W
)
V ar(P
W
)
(97)
Cov(P
w
Q
D
, P
W
)
V ar(P
W
)
+N
P
P
F
E(P
W
)
AV ar(P
W
)
+
N
P
a
x
_
1
1
c
_
Cov(P
x+1
W
, P
w
)
AV ar(P
W
)
= 0
whereN = (N
R
+N
P
)/A

35/41
10 APPENDIX
which implies:

P
F
= E(P
W
)
N
P
Nca
x
_
cP
R
Cov(P
x
W
, P
W
) Cov(P
x+1
W
, P
W
)
(98)

Vertical integration in the electricity market in the 15 countries with largest consumption

Electricity Consumption (TWh) Vertical Integration Structure of Capital
USA 3475 Variable Private
China 1483 High Public
Japan 934 High Private
Russia 632 High Public
Germany 509 Variable Variable
Canada 504 High Public
India 418 High Public
France 408 High Public
UK 337 Medium Private
Brasil 329 Medium Private
Korea 318 High Public
Italy 291 High Public
Spain 218 Medium Private
Australia 190 Medium Private
Taipei 182 High Public
Table 4: Vertical integration of the electricity market in the 15 countries with largest consumption.
Source IEA
10.5 The Envelope theorem

The envelope theorem
26
: Consider an arbitrary maximization (or minimization) problem where
the objective function f(x, r) depends on some parameters r :
Max
x
f(x, r)
Say the function f
(r) is the problems optimal-value function. Let be the x
(r) the value, ex-

pressed in terms of the parameters, that solves the optimization problem, so that f
(r) = f(x
(r), r).
The envelope theorem tells us how changes as a parameter changes, namely:
df
(r)
dr
=
f(x, r)
r
|
x=x
(r)

10.6 Rubinsteins rule (1976)

If x and y are normally distributed and a(.) and b(.) are dierentiable functions, then:
cov(a(x), b(y)) = E
x
(a
x
)E
y
(b
y
)cov(x, y) (99)
See Rubinstein (1976) for a proof.
26
taken from wikipedia.org
36/41
10 APPENDIX
10.7 Data of regression
Figure 16: Data we collected and computed to test how PKN Orlens investment opportunities are
correlated with oil price (1/2)
37/41
10 APPENDIX
Figure 17: Data we collected and computed to test how PKN Orlens investment opportunities are
correlated with oil price (2/2)
38/41
REFERENCES

References
[1] References Aid, R., G. Chemla, A. Porchet, and N. Touzi, 2011, "Hedging and Vertical Inte-
gration in Electricity Markets", Management Science.
[2] Allaz B., 1992, "Oligopoly, Uncertainty and Strategic Forward Transactions". International
Journal of Industrial Organization, 10, 297-308.
[3] Arrow, K.J. (1975): Vertical Integration and Communication, Bell Journal of Economics,
Vol. 6, No. 1, pp. 173-183.
[4] Bessembinder, H., and Lemmon M.L., 2002, "Equilibrium Pricing and Optimal Hedging in
Electricity Forward Markets", the Journal of Finance, LVII, 3, 1347-1382.
[5] Bolton, P., and Whinston M.D., 1993, "Incomplete Contracts, Vertical Integration and Supply
Assurance". The Review of Economic Studies, 60, 1, 121-148.
[6] Breeden, Douglas, and S. Viswanathan, 1990, Why do rms hedge? An Asymmetric Informa-
tion Model. Working paper. Duke University
[7] Broll. B and Eckwert B, 2000,Market structure and multiperiod hedging, International Review
Of Economics& Finance, vol9, 291-298.
[8] Brown, Hobson. D, Rogers L. C. G. Robust Hedging of Barrier Options, Mathematical Finance
Volume 11, Issue 3, pages 285314, July 2001.
[9] Carlton D.W., 1979, "Vertical Integration in Competitive Markets Under Uncertainty". The
Journal of Industrial Economics, 27, 3, 189-209.
[10] Chandler, A. D. (1969). The structure of American industry in the Twentieth Century: A
historical overview. Business History Review 43 (Autumn): 255-298.
[11] Coase, Ronald H. 1937. The Nature of the Firm,, The Theory of the Firm: Critical Perspec-
tives in Business and Management, Vol II. London: Routledge.
[12] DeMarzo, P. M., & Due, D, 1995,Corporate Incentives For Hedging And Hedge Accounting,
Review of Financial Studies, 8, 743771.
[13] Doherty, N., and H. Schlesinger, 2002, "Insurance Contracts and Securitization", Journal of
Risk and Insurance, 69, 45-62.
[14] Donald. L and D.Shaer, 2001, Hedging Multiperiod Forward Commitments: The Case of
Period-by-Period Quantity Uncertainty, Review of Quantitative Finance And Accounting,16:
171-181
[15] Donald. L and D.Shaer, 2002, Multiperiod Strip Hedging of Forward Commitments, Review
of Quantitative Finance And Accounting,18: 345-358
[16] Elton, E., Gruber, M., Brown, S., and W. Goetzman, 2006, Modern Portfolio Theory and
Investment Analysis, Wiley, 7th Edition.
[17] Emmons W., 1996, "Good Times, Bad Times, and Vertical Upstream Integration". Interna-
tional Journal of Industrial Organization, 14, 465-484.
[18] Froot, K., A., D. S. Scharfstein, and J. Stein, 1993, "Risk Management: Coordinating Corporate
Investment and Financing Policies." Journal of Finance, 48, no. 5 (December), 1629-1658.
39/41
REFERENCES
[19] Geczy, Ch, B. A. Minton, and C. Schrand, 1997, Why Firms Use Currency Derivatives,
Journal of Finance 52, 323-354.
[20] Gibson, R., M. Habib, and A. Ziegler, 2010, "Why Have Exchange-Traded Catastrophe Instru-
ments Failed to Replace Reinsurance?", mimeo University of Zurich.
[21] Gollier C., 2004, The Economics of Risk and Time. MIT Press: Cambridge, MA.
[22] Green J.K., 1986, "Vertical Integration and Assurance of Markets". In: Stiglitz JE and Mathew-
son GF (Eds), New Developments in the Analysis of Market Structure. MIT Press, Cambridge.
[23] Green, J.R. (1974): Vertical Integration and Assurance of Markets, Harvard University DP
383, Cambridge, Mass.
[24] Gilbert. R, Newbery. D (2006) "Electricity Merger Policy in the Shadow of Regulation", MIT
Center For Energy and Environmental Policy Research.
[25] Hull.J, Nelken. I, and White.A, 2003, Mertons Model, Credit Risk, and Volatility Skews
Journal of Credit Risk, Vol. 1, No. 1, (Winter 2004/05), pp. 3-28.
[26] Jensen, M.C. and W. Meckling, 1976, Theory of the rm: Managerial behavior, agency costs
and capital structure, Journal of Financial Economics 3, 305-360.
[27] Joskow. P,2005, Regulation of Natural Monopolies, MIT Center For Energy and Environ-
mental Policy Research.
[28] Lafontaine. F, and M. Slade (2007), Vertical Integration and Firm Boundaries: The Evidence,
the Journal of Economic Literature.
[29] Loss F., 2012, Optimal Hedging Strategies and Interactions between Firms Journal of Eco-
nomics & Management Strategy, Vol. 21, Issue 1, pp. 79-129,
[30] Loss F., 2002, Optimal Hedging Strategies and Interactions between Firms, Discussion Paper
N399, Financial Markets Group, London School of Economics.
[31] Magee. S, 2009, the eect of Foreign Currency Hedging on external Financing ,Macquaire
University, Applied Finance Center
[32] Malmgren, Harold B. 1961. Information, Expectations, and the Theory of the Firm, Quarterly
Journal of Economics 75: 399-421.
[33] Meade.R and OConnor.S, 15-16 January 2009, Comparison of Long-Term Contracts and
Vertical Integration in Decentralized Electricity Markets, European University Institute.
[34] Florence. Mello. A and J. Parsons, 2000, Hedging and Liquidity, The Review of Financial
Studies, Vol 13, No 1, pp127-153.
[35] Myers, S.C., 1977, Determinants of corporate borrowing, Journal of Financial Economics 5,
147-175.
[36] Myers S and N., S. Majluf, June 1984,"Corporate Financing and Investment Decisions when
Firms have Information that Investors Do Not Have", Journal of Financial Economics, Vol.
13, No. 2 pp. 187-221.
[37] Nain, A., 2004 ,The strategic motives for corporate risk management, Working Paper, Uni-
versity of Michigan.
[38] Porter, M. 1980. Competitive strategy: Techniques for analyzing industries and competitors.
New York: The Free Press.
40/41
REFERENCES
[39] Rubinstein, M., 1976, The valuation of uncertain income streams and the pricing of options,
Bell Journal of Economics, 7, 407-425.
[40] Smith Jr, C, and R Stulz, 1985, "The determinants of rms hedging policies", Journal of
Financial and Quantitative Analysis, 20, 391-405.
[41] Spano, M, 2001, Neutral and Non Neutral Shock Eects on Hedging, Investment, and Debt,
Department of Economics and Related Studies. University of York (UK).
[42] Stulz R.M., 1990, Managerial Discretion and Optimal Financing Policies, Journal of Financial
Economics, vol 26, p 3-27.
[43] Stulz R.M., 1984, Optimal Hedging Policies, Journal of Financial and Quantitative Analysis,
Vol. 19, n 2, pp. 127-140.

41/41

Short Term Hedging, Long Term Vertical Integration

Uploaded by

Copyright:

Available Formats

Short Term Hedging, Long Term Vertical Integration

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Short Term Hedging, Long Term Vertical Integration

Uploaded by

Copyright:

Available Formats

Short-term Hedging, long-term Vertical Integration?

). Thus, the prot

[(w)] where (w) = max

becomes more sensitive to . (Other parameters:

Table 2: Optimal hedging ratio depending on the sign of previous prots

Hence we summarize the case 1, 3 and 4 in the following proposition.

, is a decreasing function of the parameter

, is a decreasing function of the parameter .

to the parameter . Therefore, the optimal hedging ratio is an

is closer to 1 (full hedging) as

is less sensitive to the relation parameter .

is more sensitive to the correlation parameter when the volatility

is farther from 1 as the variance is higher (i.e. h

(r) is the problems optimal-value function. Let be the x

(r) the value, ex-

You might also like