Download Mergers

INDUSTRIAL ECONOMICS II
Prof. Davide Vannoni
Handout 6
Mergers
Content of lecture
1.
Broad introduction on mergers
2.
The changing antitrust emphasis
3.
Unilateral versus coordinated effects
4.
The literature post S-C-P
5.
Farrell and Shapiro
6.
The simulation methodology
Reading
Key articles
Werden G and Froeb L (1994), “The effects of mergers in differentiated products industries:
Logit demand and merger policy”, Journal of Law, Economics and Organisation”, vol 10,
no.2.
Nevo A, (2000) Mergers with differentiated products: the case of the ready-to-eat cereal
industry”, Rand Journal of Economics, vol.31 no.3, Autumn 2000, pp. 395-421
Pinkse J and Slade M, (2004) “Mergers, Brand Competition, and the Price of a Pint”, European
Economic Review, 2004, vol. 48, issue 3, pages 617-643
Farrell, J. and C. Shapiro, 1990, “Horizontal Mergers: An Equilibrium Analysis”, American
Economic Review 80, 107-126.
Salant S, Switzer S and Reynolds R, (1983) “Losses from horizontal merger: the effects of an
exogenous change in industry structure on Cournot-Nash equilibrium” Quarterly Journal
of Economics, vol. 98 pp. 185-99
Textbooks
Cabral L, (2000), Introduction to Industrial Organization, MIT, chapter 15, pp. 277-288
Church J & Ware R (2000), Industrial Organization: a Strategic Approach, McGraw-Hill,
chapter 23
Martin, Advanced Industrial Economics, Blackwell Publisher, parts of ch.9
Recent issue of International Journal of Industrial Organization: Merger Control in
International Markets (December 2005)
1
1.
Broad introduction on mergers
Various parts of the literature
 causes of mergers
 effects of mergers on concentration
 cross border M&A
 testing the success of mergers
 merger waves (& the stock market?)
 effects of mergers on price and welfare
I will concentrate on the last mentioned – i.e. the antitrust angle
2.
The changing antitrust emphasis
 Convenient starting point is Structure-Conduct-Performance
(SCP). Broadly speaking, within this traditional paradigm, the
main concern was that a merger, leading to increased
concentration, would raise the chances of collusion - what we now
refer to as the coordinated effects.
 This was enshrined in the original US merger guidelines and the
structural emphasis on market shares and concentration. As late as
1991, Scherer argued that merger policy should mainly focus on
the potential for coordinated effects since, where they occur, these
2
are likely to be much more important quantitatively than unilateral
effects.
 Over the last 15-20 years, however, the emphasis in the literature
has switched to the unilateral effects. A unilateral effect occurs
merely as a consequence of individual firms unilaterally (i.e. not
cooperatively) choosing to revise their price/output/product
positioning etc choices as a consequence of the new market
structure of the industry post-merger. It will occur even without
any change in the mode of conduct, since price/output equilibria
will typically change post-merger.
3.
Unilateral versus coordinated effects in a simple model
To formalise the distinction, consider a purely hypothetical example:
 symmetric triopoly, in which three firms each sell a differentiated
brand of a product.
 the demand system is symmetric - demand for each brand depends
in linear way (positively & identically) on prices of competing
brands and (negatively and identically) on the brand’s own price.
 Each firm has constant marginal costs of 0.67.
 We ignore entry/exit.
3
Qa = 1 – Pa + Pb/6 + Pc/6
Qb = 1 + Pa/6 – Pb + Pc/6
Qc = 1 + Pa/6 + Pb/6 – Pc
Q and P refer respectively to quantity and price, and subscripts a,b,c
refer to each of the brands of the three firms.
Pre-merger equilibrium
Prior to merger, each firm sets price independently of the others to
maximise its own profits in a one-shot game. This yields a price-setting
Nash equilibrium with the following outcomes1.
 Price of all three brands = 1
 The market is shared equally between the three brands
 At this price, each brand has an own price elasticity of –3, and all
cross-price elasticities are +0.5.
Post-merger equilibrium: unilateral effects
 Suppose A and B merge to form (A/B), which is now in Bertrand
competition with just one outsider, C.
 A/B and C still maximise their own profits, given the price of their
rival. However, the crucial difference is that the prices of A and B
are now set cooperatively, taking into account (internalising) the
cross-effects of the price of A(B) on the demand for B(A)2.
1 Algebraically, this equilibrium is calculated by maximising each firm’s profits with respect to its own price,
holding rivals’ prices constant. This yields a best response function for each firm, which shows its optimal price,
corresponding to the prices of its two rivals. Since there is a best response function for each firm, the equilibrium
reflects the solution to a system of three simultaneous equations.
2 Algebraically, this equilibrium is calculated by maximising firm A/B’s profits with respect to both P a and Pb,
holding Pc constant, and maximising C’s profits with respect to Pc, holding Pa and Pb constant.
4
 The new equilibrium prices are:
Pa = Pb = 1.0339,
Pc = 1.00565
Thus, the prices of the merged brands have risen by 3.4%, and the
outsider uses the opportunity to raise its own price by 0.56%.
Post-merger equilibrium: coordinated effects
 Alternatively, suppose that, post-merger, tacit collusion replaces
Bertrand competition3: and that firm A/B and firm C are able to set
their prices up to the full joint maximising levels. At the limit, it is
as if there is just one firm, selling 3 brands and setting prices to
internalise all cross-effects. Equilibrium now becomes4:
Pa = Pb = Pc = 1.0833
Comment
Here, both unilateral & coordinated effects are quite large (but, of
course, the values are arbitrary.) Note that, quite generally:
 can’t be both unilateral & coordinated effects – one or the other
 the price of the outsider will always rise (absent cost savings)
 all three prices will rise by more with coordinated effects.
 higher cross-price elasticities would ensure greater effects
 the difference between unilateral and coordinated effects could
become more pronounced if we allow for asymmetries in the cross
price elasticities.
3 For example, suppose that the game is a repeated game, in which the parameters are such that, while tacit collusion
is not the equilibrium when there are three players, it is the equilibrium with only two players.
4 Algebraically, this equilibrium is calculated by maximising joint profits with respect to each of the three prices.
5
4.
The literature post S-C-P
4(i) The catalyst for the renewed look at the potentially anticompetitive
effects of mergers was Salant, Switzer and Reynolds (1983).
Their model assumed






Cournot behaviour,
homogeneous product,
symmetric firms,
constant marginal costs,
not merger to monopolise,
no cost savings.
 Surprising result: under many conditions, mergers in such an
industry will be typically unprofitable for the merging parties.
Recall the Cournot Model in handout 1: Profits of firm i are
  c 
i  

 (n  1) 
2
In the case in which m firms merge (the total number of firms passes
from n to n-m+1) profits for each firm (including the merging firms
are:
  c 
i  

 (n  m  2) 
2
Profits are higher for the merging firms if :
2
2
  c 
  c 
m


 (n  m  2)  that is if m (n-m+2)2> (n+1)2
(
n

1
)




6
For all n ≤ 6 the only profitable merger is that in which m=n
(monopoly). For n > 6 there are situations in which the merger is
profitable but only if it involves most of the firms (unrealistic). For
example if n =12 the merging is profitable if involves more than 10
firm and so on.
 With hindsight, it is not this headline result that was so significant
- others writers quickly established how it could easily be
overturned by modifying the assumptions.
 But what was seminal was the understanding it provided on why
the merger would be less profitable for the interested parties, in
this type of model: in a Cournot equilibrium, the ‘insiders’ will
choose to contract their output (i.e. the merged firms supply less
post-merger than the sum of their outputs pre-merger), but,
because of this, it will be profitable for the ‘outsiders’ to increase
their output (strategic substitutes.)
4(ii) Literature taken forward significantly by Farrell & Shapiro (1990).
Generalising to asymmetric firms, they established:
 Crucially, unless the merger leads to cost savings (and/or entry), it
must result in increased price.
 Even where there are cost savings, these need to be very significant if
they are to offset the forces working towards an increased price.
7
 But this does not mean that aggregate welfare will necessarily fall.
The realignment of market shares following the merger may lead to a
higher proportion of industry supply in the hands of low cost firms.
For example, if the non-merging firms are relatively more efficient,
because they expand output post-merger, the industry will have a
lower weighted average marginal cost, and thus higher producer
surplus. Since more efficient firms have larger market shares in the
Cournot model, it is more likely that aggregate welfare will increase
the smaller are the market shares of the merging firms.
4(iii) Models of product differentiation (Bertrand, rather than Cournot,
see previous example in section 3.)
Deneckere and Davidson (1985) established two early important results:
 without cost savings or entry, the merger must lead to a price
increase: consistent with the above Cournot result, but
 in this case, non-participants also raise their prices: important
difference between Cournot & Bertrand models (best response
functions slope downwards in the former, but upwards in latter.)
8
5 Farrell & Shapiro - formal treatment
We now examine the three bullet points on the previous page.
assumptions/definitions
 n-firm Cournot oligopoly.
 P(X)
market demand for homogeneous product;
 X = Σxi
total industry output;
i
 C (xi)
total costs of firm i;
i
 cx (xi) marginal costs of firm i.
(i) without cost savings, price must rise
First, they show a key general point: if best reply curves are downwards
sloping, when one firm changes its output, aggregate industry output
moves in the same direction, but by less: if firm 1 reduces output, the
others will react by raising their output, but their aggregate increase <
firm 1's reduction.
They then use this result to consider the price effects of a merger
between a group of firms which will construct one firm with costs c M.
The neat thing about the above result is that we need only look at the
output response of the insiders to the merger to ascertain how the
industry’s output (and thus price) will change in the new post-merger
equilibrium.
Let pre-merger variables be indicated with *. Consider the first-order
condition of the merged entity at the pre-merger output levels
9
P(X*) +P(X*).x*M – cxM(x*M) where x*M = x*1 + x*2
{1}
We wish to examine whether the merged firm would want to increase
output from these pre-merger level (leading to a fall in prices). For {1}
to positive, we require:
P(X*) + P(X*).x*M > cxM(x*M)
{2}
But note x*i must have been optimal for each insider5, before merger:
P(X*) +P(X*).x*1 – cx1(x*1) = 0
{3a}
P(X*) +P(X*).x*2 – cx2(x*2) = 0
{3b}
Adding the two, we get:
2P(X*) +P(X*).x*M = cx1(x*1) + cx2(x*2)
{4}
Substituting {4} into {2}, we find that a necessary and sufficient
condition for the merger to lead to lower prices is that:
cx1(x*1) + cx2(x*2) - P(X*) > cxM(x*M)
{5}
5 I’m assuming that the merger is between only two firms (insiders), but the result generalises to any number.
10
To see the implications most clearly, denote price minus marginal cost
pre-merger by m, then simple rearrangement of {5} gives:
mM > m1 + m2
{6}
 This reveals that a price reduction is impossible if the marginal costs
of the merged firm are the same as even the costs of the more
efficient partner pre-merger (mM = m1)
(ii) even with cost savings, a price reduction is unlikely
Indeed, it is impossible unless the costs of the newly merged firm are
lower than those of the more efficient of the two merging firms (i.e. c x1>
cxM and thus mM > m1).
 Even this condition is not sufficient: numerical example: suppose
cx1=cx2 =0.67 and P(X*) = 1, then price will rise unless cxM < 0.33,
i.e. MC would have to halve before the merger would lower price.
 More generally, since (m1 + m2)/P = (s1 + s2)/e, it is less likely that
price will fall, the larger are the merging firms.
(iii) welfare implications
Note that these results are about whether a merger increases the
equilibrium price. If we care only about consumers, then the mergers
which lead to higher prices should all be blocked.
11
However, if we care about aggregate welfare (Consumer + producer
surplus), then mergers may be welfare increasing. To show this, as a
background, we first digress to establish one of the classic ‘traditional’
results (e.g. Cowling-Waterson Economica 1980)
Cournot, Concentration and Welfare
 write profits for firm i as
Πi(xi, X-i) = P(X).xi - Ci(xi)= 0
 derive first order condition as:
P + P(X).xi - cxi(xi) = 0
{7}
 this can be rewritten, to derive firm’s optimal price-cost margin:
{P- cxi}/P = -P(X).(xi/P) = si/e
{8}
 Aggregating across firms, the weighted) average industry price
cost margin can be written as:
si.{(P- cxi)/P} = si2/e = H/e
industry profits are
concentration index.
positively
related
{9}
to
the
Herfindahl
Now consider a merger between two firms, i and j
 If their market share after the merger is unchanged (= the sum of
their pre-merger shares), then it is easy to show, arithmetically,
that the Herfindahl index increases by an amount, 2sisj.
12
 From {9}, profits (and presumably prices) must increase. Should
we therefore conclude that mergers should not be allowed?
Not necessarily. The first thing to note is that outputs, and hence market
shares, will not remain unchanged by the merger. In particular, the
newly merged firm will tend to want to reduce output, and (because
homogeneous Cournot implies strategic substitutes) the ‘outsiders’ will
increase theirs.
Consider the change in overall welfare = consumer + producer surplus.
If outputs are changed by a little, dxi, the change in overall welfare will
be the area between the price and marginal costs for that little change,
i.e.
dW = (P- cxi). dxi
{10}
Using the f.o.c., we can write this as:
dW = -P(X).xi. dxi = -P(X) xi.dxi
Now, note that: d(xi2) = 2 xi dxi
and: d{X2.(xi2 /X2)} = d{X2.(si2)} = d (X2.H) = 2.X.H.dX + X2.dH
13
Putting all this together we get
dW = -X2.H.P(X).{(dX/X) + (dH/2H)}
{11}
Since demand is downwards sloping (P(X) < 0), the sign of dW is the
same as the sign of {(dX/X) + (dH/2H)}. Thus welfare increases with a
small increase in output only if (dX/X) + (dH/2H)>0. This is more
likely to increase if concentration increases!
To help understand how this might happen, consider again the first order
conditions of any two firms (1 and 2) under Cournot. Subtracting one
from the other gives:
P(X).(x1- x2) – (cx1 - cx2) = 0
{12}
This shows a well known feature of Cournot: if cx1 > cx2, then x1< x2: the
more efficient firm has the higher output.
Now consider a change, such that the large firm's output increases by as
much as the small firm's output decreases: then dX=0 and dH> 0, so
concentration has risen, but so has welfare.
14
6.
The simulation methodology
Most academic work over the last decade has focused on mergers in
industries with product differentiation, with many of the main
contributors keen advocates of simulation methods.
To see why
simulation has assumed a prominent role, consider some of the
implications of a merger within a market which comprises firms selling
differentiated products:
 One point is fairly obvious: in these circumstances data on
concentration and market shares alone are now not nearly so
informative. With all firms producing a homogeneous product, a
change in the supply from one firm has the same effect on the price
received by all others (there is only one price.)
But with
differentiated products, some firms will supply closer substitutes than
others. It follows that there may be some mergers between quite
large firms which will have little effect on prices (where their
products are not close substitutes), while other smaller merging firms
(supplying close substitutes) may have much more effect on price.
 In other words, to calculate the likely effects on price, we need to
know the pattern of cross-price elasticities - in principle between all
products within a market.
15
 If we also want to allow for the possibility that the merger affects
costs, and if we want to take into account how non-participants react
to the merger (eg by repositioning their products or by new entry), the
analysis becomes very complicated.
Traditional analytical methods (such as those used by Farrell and
Shapiro) become less powerful.
So what is simulation?
Typically, simulation analysis of mergers has three stages.
Stage 1: choice of model for the oligopoly game
 Are products homogeneous or differentiated?
 Do firms act unilaterally or in a coordinated (collusive) way?
 If products are differentiated, are they symmetrically so, or do
brands tend to be closer for some competitor brands than others
(localised competition.) Note that this question has both product
space and geographic space dimensions. This requires choice of a
particular model of differentiation, e.g. logit versus nested logit.
Once these decisions have been made, we can derive expressions for the
equilibrium (prices and market shares) before the merger
16
Stage 2: calibration of the model’s parameters.
The equilibrium derived in stage 1 will include various unknown
structural (demand and cost) parameters.
We next need to assign
specific numerical values to these parameters.
In order to do this, we use data on the firms’ actual market shares and
prices and estimates of the demand parameters. The demand parameters
might be derived from an econometrically estimated demand system (see
earlier lecture on empirical models of product differentiation).
By
substituting these into the equilibrium solution from stage 1, we can
retrieve estimates of all remaining unknown structural parameters.
In some cases, where other information is also available (e.g. on
marginal costs and, perhaps, elasticities of demand from previous
studies), there may be no need to econometrically estimate own and
cross price elasticities, in order to complete this stage.
Stage 3 is simulation of the post-merger equilibrium.
The equilibrium is then recomputed, using the calibrated parameters, but
now allowing three things to change:
 coordination in the price-setting of brands of newly merged firm
 any efficiency (i.e. synergy) effects, e.g reductions in the marginal
costs of the merged firm
 reactions by non-participants: in general, they will now want to
change their price and in response to the price changes of the
17
merged firms.
This is covered, automatically when the new
equilibrium is computed. Rather more difficult is the possibility
that they will also want to reposition their products (similarly, the
potential for new entry becomes important.)
Some examples of where simulation has been used in practice
 The best known case is the proposed (but actually blocked) merger
between Interstate Bakeries and Continental Baking (1995). This
involved brands of bread, and copious scanning data were
available. In this case, simulations suggested price increases of
between 5% and 15% in Los Angeles and Chicago
 Gillette (1993) potential substitution between pens, ball points etc,
 cereals merger (1995) between Kraft & Nabisco,
 L’Oreal acquisition of Maybelline
 Froeb and Werden (1994) explained their own simulation approach
in the context of long distance carriers
 Kimberley-Clark and Scott: Hausman and Leonard estimated the
merger would lead to price increases in the range of 1 –2%
Advocates of simulation often stress the increased accessibility of these
techniques, given the speed of present computing technology.
Nevertheless, they are still data- and time-intensive, calling for high
18
technical competence in economic and econometric modelling.
As
Nevo explains:
‘The analysis was performed without time pressure and using
(almost) ideal premerger data. Time and data constraints might
limit the ability to perform this analysis in real time’
As such, it is sometimes useful to have easy “back of the envelope”
equivalents to simulating how much prices might rise following a
merger.
19