Cost and Benefits of Merger Control in the Netherlands 1998-2002
An Applied Game Theoretic Analysis of Prices*
Bas Postema**
Peter van Bergeijk***
Marie Goppelsröder ****
Summary
This paper analyses the costs and benefits of the new Dutch merger control legislation that was
introduced in 1998. Using an applied game theoretic model we analyse the effects of proposed
mergers in cases where the original merger proposals were not accepted by the competition authority
(or where substantial remedies had to be taken in to account by the proposed mergers).

First we describe the new legislation and the main characteristics of the mergers during this
period. We also pay attention to the costs (red tape, personal) of this legislation and review
published estimates for the US and UK.

We then continue presenting the basic model underlying all simulation models discussed in
this paper. After that, we outline the most common demand models that are used for merger
simulation, and discuss the respective applicability of these models.

Next we discuss our data set and apply the models to the empirically relevant context of four
markets. We also perform sensitivity analyses with respect to key parameters such as cost
savings and demand patterns. This enables us to estimate average expected post merger
price increases that where prevented by rejecting merger proposals by the competition
authority.

Finally we estimate the net benefits of merger control at €80 million per year.
*
Preliminary version
Tilburg University
*** Erasmus University (OCFEB) and Netherlands Competition Authority (NMa)
**** Netherlands Competition Authority
**
1
Introduction
Merger control legislation was introduced in the Netherlands in 1998 at the top of a merger wave.
From that year on, mergers for which the involved companies’ annual turnovers exceeded a certain
threshold (€30 million per company, and €113,45 million combined) had to be approved by the Dutch
Competition Authority. Over these past six years, more than two in three of the mergers that were
subject to merger control concerned sizeable transactions where the involved companies’ combined
annual turnovers exceeded €500 million. Looking at the most recent merger figures, we even see that
between 2000 and 2002 in almost 50% of the Dutch merger cases the combined annual turnovers
exceeded €2 billion. When divided into categories based on economic activities, most of the (almost
400) mergers that were filed in this period are in construction and in retail & wholesale. In the years
2002 and 2003 the absolute and relative amount of mergers in these sectors have decreased while
the main point of the activities distribution shifted towards the concentrating energy sector.
After these first years of merger control, it might be interesting to look at ways to quantify potential
benefits. The intuitive way to do this is by comparing revenues to costs. Where revenues are hardly
concrete, the costs of merger control are somewhat easier to estimate. First of all, there are costs for
the government. These costs are budgeted, and have been €2 million per year, on average. Then
there are also costs for the companies involved in merger proposals. Based on a recent report by
Pricewaterhouse Coopers1, these costs can be estimated to lie between €10 million and €20 million
per year, on average. The benefits of merger control come from the mergers that were not realised (in
original form) as a result of merger control (either direct or indirect). There are different ways these
benefits are estimated by competition authorities. The easiest way to do this is by computing
consumer savings. These are defined as the amount of money the consumers did not have to pay in
price changes as a result of merger control preventing a merger. Among the competition authorities
quantifying the benefits of their merger control through consumer savings are the Federal Trade
Commission (US) and the Office of Fair Trading (UK). They estimate a prevented price increase of at
1
A tax on mergers? Surveying the time and costs to business of multi-jurisdictional merger reviews, PwC, June
2003
2
least 1% over 2 years for companies that were involved in a prevented merger.2 The reason that they
arbitrarily choose these numbers is that they expect the real prevented price changes to be
considerably higher, such that the 1% estimate guarantees a conservative estimate. The Antitrust
Division of the Department of Justice (US) uses somewhat more sophisticated ways to estimate the
benefits of their merger control. They use various techniques to estimate case-specific price changes
that were prevented.3 The table below gives an overview of recently published results of estimations of
the benefits of merger control.
Competition authority
Annual revenue merger control Prevented pricechange Pricechange estimated through
Federal Trade Commission, US
$1,2 billion
1%
Assumption
Department of Justice, US
$3,3 billion
1%-20%
Modeling/Simulation
Office of Fair Trading, UK
1%-7%
Assumption/Modeling
Table 1 Estimations of revenues of merger control
For the US markets, the FTC and the DoJ estimate collective annual revenues of their respective
merger control of (at least) $4,5 billion, for instance. This number is based on the consumer savings
computed by the FTC based on an assumed price increase and the DoJ’s revenues that are based on
price increases that they estimated through modelling.
In order to estimate the revenues of merger control in The Netherlands, we used a merger simulation
model to quantify the price effects of prevented mergers on their respective relevant markets.
Seminal articles on the use of those models have been published by Hausman (1997), Werden
(1997b,c) and Epstein and Rubinfeld (2001). Their models quantify the price effect as a result of a
merger, using prices, quantities or market shares and elasticities of demand.
This paper proceeds as follows: firstly, the basic model of merger simulation common to all models is
explained. Then, the most cited (demand) models are outlined and discussed briefly. We then show
and discuss the empirical results of the paper. Finally, we present our conclusions.
2
3
Davies, S. & A.Majumdar (2002)
Nelson,P. & S.Sun (2001)
3
Basic Model with Different Demand Forms
Merger simulation models share a number of common characteristics. As Werden (1997b) has
highlighted, they are all based on the same oligopoly model. They all assume Bertrand competition in
differentiated goods markets, non-cooperative one-shot games where firms set prices to maximise
profit. In addition, they also commonly assume that marginal costs stay constant throughout the
relevant output range.
Lastly, they all assume a certain functional form of demand. This form of demand is the main
difference between the models discussed. We will now present the model underlying all the merger
simulation models discussed in this paper. We will refer to this in the following as the basic model.
The Basic Model
It is assumed that each product j is produced by a single firm i. It is possible though that a firm
produces more than one product and F() assigns the products to the firms that produce them. As
mentioned before, the marginal costs cJ are assumed to stay constant. The costs are allowed to differ
from firm to firm. Firms play a Bertrand game and thus set prices to maximise profits (i). Prices are
indicated by pj, the vector of all prices is indicates by p and quantities are indicated by qj.
i 
( p
j , F ( j ) i
j
 c j )q j (p)
(1)
The first order condition for each good k produced then yields
0
 i
 q k (p) 
p k

j , F ( j ) i
(pj cj)
The price-cost margin is then described by
4
q j (p)
p k
(2)
pj cj
pj
where

1
 jj (p)
(3)
 jj (p ) is defined as the own price elasticity of demand for product j. It depends on the prices
set by all other products in the market.
The analysis is split up in two stages. In the first stage there is a market situation where a number of
firms each sell at least one product. This stage is used to analyse the pre-merger market situation. For
this stage the required data consist of product prices, quantities and (cross-) price elasticities.
Given the current prices and elasticities the only unknowns in this problem are the companies’ costs.
They are generated using the first order conditions (FOC’s) in (2). So in the initial step the companies’
marginal costs are estimated.
In the second stage a new market situation is analysed. Here (at least) two of the firms in the initial
market have merged. In this new market situation, the price setting problem for the merged firms will
change; their products will now be priced by only one instead of two (or more) firms. For the merged
firm m the profit maximisation problem will now change to:

 m
q
 q j     pi  ci  i

p j
p j
i , F ( i ) m 

0


, j, F
 j  m
(4)
Possible expected efficiency gains can be taken into account here.
To complete the model, assumptions have to be made about the demand side. Possible demand
models include linear demand, constant elasticity demand, Logit demand and AIDS. The curvature
and the (cross-) price elasticities together define the demand side equations from the initial
equilibrium. Knowing supply and demand side of the market, the post merger market equilibrium can
now be solved.
5
The model can be used to predict the consequences of a proposed merger on a market. These can be
translated into consumer costs and welfare effects resulting from the merger. If this concerns a merger
of which the proposal was not accepted by the competition authority, the effects can be seen as the
benefits of merger control.
Linear Demand
The linear demand system is one of the simplest demand systems used in merger simulation models.
Following Crooke et al. (1999) the demand function for product i is then denoted as
q(p)  q 0  M(p  p 0 )
(5)
Here, demand is simply a linear function of price. The superscript o denotes the initial situation
(premerger indices) and M denotes the matrix of price slope coefficients. 4
The attractiveness of the linear demand system stems from the fact that it is easy to implement. This
computational convenience comes at a cost however: the oversimplification may lead to coefficients
that have the wrong signs. Another problem the modeller can be confronted with when using linear
demand as a working hypothesis is negative quantities in the equilibrium. This is likely to happen if the
merging entities are of very different size. A solution to this problem is the adding of non-negativity
constraints to the model.
Constant Elasticity
From the linear demand estimation one can derive the constant elasticity demand system.
An example where this assumption has been used is in a merger in the beer market (see Baker and
Bresnahan, 1985).
n
n
j 1
j 1
log( qi ( p))  log( qi0 )    ij log( p j )    ij log( p 0j )
4
with
mij 
qi
q0
  ij i0
p j
pj
6
(6)
The constant elasticity model is very convenient. Other than in many functional forms (e.g. AIDS) the
elasticities do not have to be estimated econometrically. The regression coefficients estimated in (6)
are already the elasticities in the respective market.
There are however disadvantages due to the parsimony of the model. First of all, as with the linear
demand system, the coefficients of the elasticities might have the signs inconsistent with economic
theory. In addition, the simulation model demand system might overstate the price rise following a
merger due to the assumption of constant elasticities.
Logit Demand5
The Logit model is based on random utility theory. In this context, consumers make a discrete choice
from a choice set maximising their utility. The simplest model to determine the utility of consumer k of
choosing option i is defined by:
Vik   i  pi  vik
(7)

, the price coefficient
i ,
the perceived quality
As shown in (7) the utility for consumer k from choosing product i depends on
that is assumed to be the same across all consumers and products, and
differences among products. The latter ones are all put under the roof of the i’s. The disturbance term
vik is a component of utility specific to each consumer, uncorrelated with price p i . Given that the vik ’s
are distributed according to the extreme value distribution, one can express the choice probabilities as:
 ik 
exp( i  pi )
 exp( j  p j )
(8)
jC
where C indicates the choice set. Werden et al. (1996) adapt the Logit model to typical antitrust
situations.6 One product is defined as the choice “none of the above”; it is thus the outside good and
7
its price is assumed to be zero to make the utility of this good a constant. The choice probabilities of
the remaining goods, “the inside goods”, are defined as shares. Technically, they are the choice
probability for the inside good conditional on the choice being an inside good. The inside goods are
closely related to the concept of the relevant market, but they do not necessarily have to coincide.
The Logit model is characterised by the Independence of Irrelevant Alternatives Assumption (IIA).
Anderson et al (1992) show that this property has implications for the behaviour of the elasticities in
the Logit model. It means that the cross elasticities of demand of every inside good i with respect to
the price of a given inside good j is the same for all
i  j . This means that the substitution patterns
between brands are determined by their relative market shares. To illustrate this, assume that there
are three firms in the market, A, B and C with respective market shares of 30%, 60% and 10%. If C
were to increase its price, this would imply that firm A would catch half as many of the lost sales of C
than firm B.
If the classic Logit model with its inherent IIA assumption does not properly reflect the market
analysed, it is possible to re-estimate the model using a more complex demand specification. This can
be done by using the nested Logit model which is characterised by a hierarchical decision making
process (see Werden, 1997c).
The attractiveness of using Logit demand when analysing merger cases comes from its simplicity and
tractability. This makes it easy to implement and perform merger simulation, especially in the many
cases where there is a lack of data.
Yet this simplicity is not costless. The IIA assumption is restrictive and might not always hold in real
world situations. In fact, Hausman and Leonard (1997) argue that econometric tests of the IIA
assumptions have very high rates of rejection. Citing Hausman and Leonard (1997, p. 322):
“Since the size of the post-merger unilateral effect depends crucially on the pattern on demand
substitution of the merging firms, restricting the demand substitution patterns would seem to defeat the
5
This section draws from Werden (1997c)
8
purpose of performing the unilateral effects analysis. Indeed, the entire framework of differentiated
products arises because some products are more closely competitive than others.”
The AIDS Model
Hausman and Leonard (1997) have used an adaptation of the multistage budgeting approach by
Gorman (1995) when they applied the AIDS model in the context of mergers. The estimation of
demand is split up in two stages that are later combined to generate the overall own and cross
elasticities of demand for the product.
The top level of demand refers to the overall demand for a respective product. It can be estimated
using the theory on price indices and is generally used to estimate the overall price elasticity of a
product. The bottom level of demand however specifies the competition among the brands producing
a product. This level is characterised using AIDS (Almost Ideal Demand System 7). The demand
system for the bottom level is specified in terms of market shares. Following Hausman and Leonard
(1997) the bottom level is defined as:
Y 
sint   in   i log  nt  
 Pnt 
where

jB , P
ij
log p jnt   int , n=1,…,N t=1,…T
(9)
sint is the share of brand i of total industry expenditure in city n in period t. Ynt is the overall
expenditure in the industry,
Pnt is an overall price index and p jnt t is the price of brand j in city n and
time t. Different to the Logit model, the
 ij ’s allow the cross-price elasticities to behave flexibly.
After having estimated the demand structure, the effects of the merger can be analysed along the lines
of the basic model. Since this involves solving a system of non-linear equations, Hausman and
Leonard (1997) apply a linearisation method by assuming that that the elasticities and shares of the
merger stay constant after the transaction.
6
They call this model Antitrust Logit Model (ALM)
9
If the data needed for the estimation of the AIDS model is lacking, it is possible to use an adaptation of
the AIDS model called Proportionality-Calibrated AIDS model (PCAIDS). Different to AIDS, it does not
need scanner data or data on premerger prices (For a discussion, see Epstein and Rubinfeld, 2001).
The advantage of using the AIDS demand form in comparison to the other three demand forms is its
flexibility. At the point of estimation, the demand elasticities are unconstrained and therefore the
behaviour of the elasticities is determined by the data that is used. The complexity of the model has its
shortcomings however. It is often difficult to estimate the large numbers of parameters. In addition,
scanner data that is needed to feed the AIDS model might in many cases not be available for the
market in question. As Epstein and Rubinfeld (2001) have argued, scanner data tracks retail prices
whereas mergers often take place at the wholesale level. The implementation of the AIDS model is
then only possible if the modeller makes additional assumptions on margins between the wholesale
and the retail level. In addition, the linearisation method proposed may be unfeasible if the price
changes following a merger are large. Hausman and Leonard (1997, p 332) state:
“The quality of the approximation decreases with the size of the post-merger price changes. Thus, for
small or moderate changes, the approximation is likely to be reasonably accurate. For larger changes,
the benefits of the exact calculation may outweigh the costs of increased complexity.”
Summarising, for each demand model there is a trade-off between simplicity and accurateness. The
simple models require less data and are thus often the only model implementable; they are however
not always as accurate as a more complex model would have been.
The demand structure seems to be a crucial dimension of the models. As Crooke et al (1999) have
shown, the functional forms presented here have inherent shape characteristics that trigger predictions
in price increases that differ significantly between the models. By using Monte Carlo experiments for
the different demand systems presented here, they show that the predicted price increase of a merger
that moves the market from duopoly to monopoly is highest with constant elasticity demand, closely
followed by AIDS. The predicted price increases are much smaller and roughly equal using linear and
7
For a comprehensive overview on AIDS, see Deaton and Muellbauer (1980)
10
Logit demand functions. What is more, the own price elasticities of demand of all models with the
exception of the constant elasticity model are getting increasingly elastic as prices increase. They
show that the largest increase in elasticity results when using the linear demand model. As they point
out, the rate of change in those elasticities distinctive for each demand model causes considerable
differences in predicted price increases
Data and Empirical Results
To estimate the benefits of merger control, we used the basic model and ran the merger simulations
on Mathematica using all four different demand forms described in the last section. These benefits
should come from the mergers that were not realised as a result of merger control. While there is also
a deterrent effect of merger control legislation, the only mergers that can be proved to be prevented
are those not accepted (in original form) by the competition authority. So we will focus on these cases
when estimating the benefits, but keep in mind there are more, possibly higher, benefits resulting from
merger control.
We estimated the revenues of merger control for the first five years of merger control in the
Netherlands (1998-2002). During this period there were eleven cases where the merger was rejected
by the competition authority or substantial remedies had to be taken into account. For these eleven
cases we checked whether for the relevant market(s) the assumptions of Bertrand competition and
heterogeneous goods could be made, and whether enough data was available to make a reliable
estimate of the benefits using the simulation program. For four markets these conditions were met,
and the simulation model was applied using the different demand models. A brief summary of the
results is presented in table 2.
11
Price changes
Market
weighted unweighted median
standard maximum
average
average
deviation
1
24,5%
21,7%
25,3%
12,2%
40,7%
2
5,6%
4,0%
1,4%
4,1%
11,5%
3
14,5%
13,6%
13,2%
1,4%
15,8%
4
21,8%
15,3%
9,3%
12,2%
36,4%
Total
14,0%
13,6%
11,8%
11,1%
40,7%
Table 2 Simulation outcome merger cases: prevented price changes
minimum
modal
4,0%
0,9%
11,5%
4,7%
0,9%
10-15%
Table 2 summarises the price effects the merger simulation model estimated for the 4 markets we
analysed. For example, in market 1 prices increase with 21,7% on average, and with 24,5% on
average weighing over market shares. All prices on this market increased as a result of the merger
(with at least 4%), and for one product price increased even more than 40%.
Together with the estimated changes of demand on the market and the firm profits (that were
maximised through the model) we can now estimate the effects of merger control on the 4 analysed
markets. Using the results we estimated consumer savings, and, where possible, the change in
consumer surplus. We then compared these to the estimated rise in firm profits after the merger and
found that the net result of rejecting the merger proposals amounts to 1-5% of the markets.
To be able to quantify the total effect of rejecting mergers in The Netherlands we still had to estimate
the effect for the markets we were not able to analyse with the merger simulation model. We therefore
generalised the effects we found on the four markets that were analysed with the model, and used the
estimated price increases and welfare changes to estimate the effects on the other markets. This way,
a (very) rough estimate of the annual benefits of merger control in The Netherlands equals €80 million.
The above results were obtained under the assumption of no efficiency gains (supported by Gugler et
al.8), and a market specific demand form. To see how these assumptions influence the outcome, we
performed a sensitivity analysis with respect to these parameters. As we expected, the form of
demand chosen turned out to be very important for the outcome. For each market we ran the
simulation using the various demand models. The choice of demand form that was used for the
outcomes presented in table 2 was based partly on earlier research on demand on these markets and
partly on applicability of the different demand models. For example, in some cases, when using AIDS
to describe the demand side of the market, we found predicted price increases of more than thousand
12
percent combined with a highly unlikely size of demand. It seems that in these cases the shift of the
equilibrium is such that the linearisation assumption is violated. Also, no (plausible) results were
obtained using Logit demand; the substitution data that were available for the markets in question
showed that the IIA-assumption was too restrictive for the model to be applicable. Therefore, the
results that are presented in table 2 are partly based on linear demand and partly on constant
elasticity of demand.
The assumption of no efficiency gains, on the other hand, turns out to be not very important for the
outcome of the benefits estimation. Even a cost reduction of 10% through efficiency gains from the
merger does not make the extra profits outweigh the extra costs for consumers.
Conclusion
Looking at the applicability of the model we used to analyse the merger effects in The Netherlands, we
observe the following. We compared the marginal costs that were estimated in the first stage of the
model to estimates of the mean marginal costs for the relevant products by the Central Bureau of
Statistics (CBS). The estimated costs turned out to be very close to the ones reported by the CBS,
which indicates that the model might be a good way to describe the markets we analysed. Concerning
the functional form of demand, our empirical results point into the direction that the assumed demand
form matters in predicting price changes following a merger. This is consistent with the theoretical
findings of Crooke et al. (1999). The linearisation approach used for AIDS seems to be difficult to
implement in cases where the expected price increase might be considerable. On top of that, the IIA
assumption of the Logit model was too restrictive in the mergers we analysed to be applicable.
Using our results we can also make some statements about costs and benefits of merger control in
The Netherlands. Based on four market simulations we expect that the price changes that were
prevented by not accepting merger proposals are 10-15% on average. Furthermore, the profits made
by the firms had the merger taken place, cannot compensate for these costs; the (positive) effect of
rejecting the mergers on total welfare (consumer plus producer surplus) amounts to 1-5% of the
market.
8
Gugler, K., D.Müller, B.Yurtoglu & C.Zulehner (2003)
13
We then used the results we found for the four analysed markets to estimate the total effect from not
accepting the eleven merger proposals (at least not in original form). Comparing these revenues to the
already presented costs, our rough estimate for the net benefits of merger control during the first five
years of merger control in the Netherlands amounts to €80 million per year.
14
References
[1] Anderson, S. P., 1992, “Multiproduct Firms: A nested Logit Approach”, Journal of Industrial
Economics, 40, 261-276.
[2] Baker, J.B., 1997, “Contemporary Empirical Merger Analysis”, George Mason University Law
Review, 5, 347-361.
[3] Baker, J. B. & T.F.B. Bresnahan, 1985, “The Gains from Merger or Collusion in Product
Differentiated Industries”, Journal of Industrial Economics, 33, 4, 427-444.
[4] Bergeijk, P.A.G. van & M.A. van Dijk, 1997, “Resource Misallocation and Mark-up Ratios: an
Alternative Estimation Technique for Harberger Triangles”, Economic Letters, 54, 165-167.
[5] Bergeijk, P.A.G. van & H.M. Godfried, 2002, “De Beëindiging van het Kartelparadijs”, ESB, 15-32002, 4351.
[6] Crooke, P., L. Froeb & S. Tschantz, 1999, “Effects of Assumed Demand Form on Simulated
Postmerger Equilibria”, Review of Industrial Organization, 15, 3, 205-217.
[7] Davies, S. & A. Majumdar, 2002, “The Development of Targets for Consumer Savings arising from
Competition Policy,” Economic discussion paper 4, OFT: London.
[8] Deaton, A. & J. Muellbauer, 1981, “An Almost Ideal Demand System”, American Economic Review,
70, 312-326.
[9] Dobson, P., 2001, “Competition and Mergers in Consumer Goods Industries: the Case of Pet
Foods”, European Competition Law Review, 10, 443-457.
[10] Epstein, R. & D. Rubinfeld, 2001, “Merger Simulation: a Simplified Approach with new
Applications”, Antitrust Law Journal, 69, 883-919.
15
[11] Froeb, L. & S.Tschantz, 2001, “How Much Information is Required to Accurately Predict Merger
Effects?” Vanderbilt University paper.
[12] Froeb, L., S.Tschantz & G.J.Werden, 2001, “Pass-through Rates and the Price Effects of
Mergers”, Vanderbilt University Paper.
[13] Gorman, W. M., 1995, “Two Stage Budgeting”, in: Separability and Aggregation: Collected Works
W. M. Gorman, vol. 1, eds: Blackorby, C., Shorrocks, A. F., chapter 2, Oxford, Clarendon Press.
[14] Gugler, K., D. Müller, B.Yurtoglu & C. Zulehner, 2003, “The Effects of Mergers: an International
Comparison”, International Journal of Industrial Organization, 21, 5, 625-654.
[15] Hausman, J.A., G. Leonard & J.D. Zona, 1994, “Competitive Analysis with Differentiated
Products”, Annales d’Economie et de Statistique, 1994, 34, 159-180.
[16] Hausman, J. A. & Leonard, G., 1997, “Economic Analysis from Differentiated Products Mergers
Using Real World Data”, George Mason Law Review, 5, 3, 321-346.
[17] Jong, H. de, 1990, “Nederland: het kartelparadijs van Europa?” ESB, 14-03-1990, 244-248.
[18] Judd, K., 1998, Numerical Methods in Economics, MIT Press: Cambridge.
[19] Khemani, R.S. & D.M. Shapiro, 1993, “An Empirical Analysis of Canadian Merger Policy”, Journal
of Industrial Economics, 41, 2, 161-177.
[20] Massey, P., 2000, “Market Definition and Market Power in Competition Analysis: Some Practical
Issues”, The Economic and Social Review, 31, 4, 309-328.
16
[21] Nelson, P. & S. Sun, 2001, “Consumer Savings from Merger Enforcement: a Review of the
Antitrust Agencies’ Estimates”, Antitrust Law Journal, 69, 921-960.
[22] Nevo, A., 2000, “Mergers with Differentiated Products: The Case of Ready-to-Eat Cereal”, RAND
Journal of Economics, 31, 3. 395-421.
[23] Werden et al., 1996, The Use of the Logit Model in Applied Industrial Organization, International
Journal of Economics of Business, 3, 1, 83-105.
[24] Werden, G.J., 1997a, “Simulating the Effects of Differentiated Products Mergers: a Practitioner’s
Guide”, University of Connecticut paper.
[25] Werden, G.J., 1997b, “Simulating the Effects of Differentiated Products Mergers: A Practical
Alternative to Structural Merger Policy”, George Mason Law Review, 5, 3, 363-386.
[26] Werden, G. J., 1997c, “The Effects of Differentiated Products Mergers: A Practitioners´s Guide”,
in: Caswell, J. A., Cotterill, R. W. eds. (1997), Strategy and Policy in the Food System: Emerging
Issues, 95-110.
17