Transcript Coordinated effects in mergers King’s College Postgraduate

Pan Fish/Marine Harvest:
Three theories of harm, two competition authorities,
and a remedy
Association of Competition Economists
2007 Annual Meeting – Thursday 29th November
Speaker: Diana Jackson
Other team members: Cristina Caffarra, Penelope Papandropoulos,
Hugh Wills, Valter Sorana, Rameet Sangha, Dan Donath, Paul Reeve
Overview
Several theories of harm in relation to fish farming
• Merger between the two largest salmon farming companies in the
world: both with significant operations in Scotland and Norway
• Market definition was important:
– Is the market global (as the parties believe): do frozen Chilean and wild
Alaskan imports compete with European production?
– Does Norwegian salmon compete with Scottish salmon?
• Several theories of harm raised by the UK Competition
Commission and DGCCRF in France:
– “Short term” unilateral output restriction (e.g. harvest early)
– “Long term” unilateral output restriction (e.g. buy fewer eggs) – key issue
– Post merger price discrimination
• UK cleared unconditionally: France required divestments in
Scotland (although less than had been offered in Phase I)
2
ACE Meeting – Toulouse, 2007
The deal
The merging parties: Pan Fish, Fjord Seafood and Marine Harvest
• The parties account for:
– around 30% of all Atlantic salmon farmed in Europe (global share is similar)
– around 45% of the Scottish salmon harvest
• There are many small competitors in the UK and Norway
3
Estimated 2005
harvest
Share
Marine Harvest
Pan Fish
Fjord Seafood
141,156
54,177
43,855
18%
7%
6%
Merged firm
239,188
30%
Lerøy Midnor /Aurora AS
Salmar AS
Seafarm Invest AS*
Mainstream
Scottish Seafarms
Others
63,000
27,000
26,000
22,700
23,000
390,112
8%
3%
3%
3%
3%
49%
Total
791,000
100%
ACE Meeting – Toulouse, 2007
A Norwegian Salmon Farm
Per Eide. Norwegian Seafood Export Council
4
ACE Meeting – Toulouse, 2007
Key industry features/production cycle
Fish farming looks a lot like Cournot
• From the purchase of an egg it takes 2-3 years to harvest:
– First 12 months (eggs >> fry >> smolt) in fresh water
– Smolt are then put to sea for up to 24 months before being harvested
• Scotland and Norway are the two main European producers:
– Scottish salmon traditionally seen as superior: small premium
– Scottish production costs are significantly higher than those in Norway
• Fixed/sunk equipment costs are relatively low:
– Main financing requirement is working capital (feed costs)
– Small producers are able to expand at similar cost to large
• Capacity constrained by environmental regs and site availability:
– Absolute constraints are not likely to be binding but
– Unit costs are increasing in output (as less optimal sites are used).
• Sales arrangements
– Fish are sold on a weekly cycle over the phone
– Some contract (40 percent), a lot spot
– Contract prices often linked to the published Oslo spot price (even in
Scotland)
5
ACE Meeting – Toulouse, 2007
Market definition
Findings in previous cases
• Product market:
– European Commission in Nutreco/Stolt/Marine Harvest JV (2005):
- All farmed salmon.
– UK Competition Commission Nutreco/Hydro (2000):
- Salmon farmed in Norway, Scotland, Ireland, Faroes.
– Anti-dumping case:
- Implies Norwegian farmed salmon must compete with EU farmed salmon.
– Some academic studies suggest even wider (incl. frozen, Pacific, wild)
• Geographic market (customer location):
– European Commission (2005) and UK CC (2000): EEA wide
– Some academic studies suggest global influences
• Pricing analysis (correlation, stationarity)
– Standard problems with price analysis less pressing here
– Supply is predetermined, so usual factors such as variable input costs are
unlikely to play a major role
6
ACE Meeting – Toulouse, 2007
Market definition
3
0
1
2
Euro par kg
4
5
Correlation between Norwegian and Scottish salmon: 0.89 (UK and France)
Jan-03
Jul-03
Jan-04
Jul-04
Ecossais
7
ACE Meeting – Toulouse, 2007
Jan-05
Jul-05
Norvegien
Jan-06
Jul-06
Market definition
.9
1
1.1
1.2
1.3
1.4
Relative prices of Norwegian and Scottish salmon are stationary
Jan-03
8
ACE Meeting – Toulouse, 2007
Jul-03
Jan-04
Jul-04
Jan-05
Jul-05
Jan-06
Market definition
Interpreting results from correlation and stationarity analysis
• Relative pricing stability has survived strong market shifts
– Key example: steep fall in Scottish salmon production
- Scottish production fell by 25% from 160kt in 2003 to 120kt in 2005
- Over the same two years Norwegian production increased by 11% from 508kt
tonnes to 572kt
– UK consumers appear to have switched from Scottish to Norwegian:
- UK consumption of Scottish salmon fell: 76kt to 62kt
- Norwegian imports increased: 15kt to 40kt
- Total UK consumption of salmon increased by about 12%
– No impact on relative prices of Scottish/Norwegian salmon
– Implication: Scottish and Norwegian salmon extremely close substitutes
• CC carried out demand estimation:
– Technical difficulties – colinear Scottish and Norwegian prices (high SEs)
– But also led to a conclusion of a broad geographic and product market
9
ACE Meeting – Toulouse, 2007
Unilateral effects modelling
Short and long term withholding theories
• Short term withholding theory
– Merged firm reduces output immediately (e.g. by destroying adult fish, or
harvesting fish early/at smaller size);
– Benefit: immediate price impact, and rivals can’t respond very strongly
(limited scope to increase harvest weight);
– Cost: at this stage most costs are sunk: the cost of the strategy is
practically the entire value of the withheld fish.
• Long term withholding strategy
– Merged firm produces less fish by buying less eggs/smolt, leading to lower
harvest levels and higher prices 2-3 years later;
– Cost: The costs of this strategy are more limited, as most of the costs of
production (feed, vaccinations, etc.) are saved;
– Benefit: The benefit of the strategy is also limited, however, as rivals have a
chance to respond to the lower anticipated output by increasing their own
egg/smolt purchases (total industry purchases are transparent).
10
ACE Meeting – Toulouse, 2007
Unilateral effects modelling
CC panel members indicate the need for economic modelling
• During the site visit, the CC panel indicated that they expected to
see unilateral effects modelling
• Fundamental data availability is good:
– Pricing data – very good (frequent trading)
– Quantity data – reasonable (production data from Kontali)
• Industry has strong Cournot characteristics:
– Set production, then 3 years later quantity determines price
• Cournot modelling for merger simulation is rarely used:
– Current output patterns may have counterfactual implications for cost
structure (several cases thrown out in the US on this basis)
– Large merger synergies are needed to explain the profitability of the merger
– The assumption of product homogeneity makes market definition crucial
• The main focus was on the long term theory:
– Short term theory rejected on basis of fundamentally high costs
11
ACE Meeting – Toulouse, 2007
Unilateral effects modelling
Simple version of long term withholding model
• We used a simple residual demand model and a more complex Cournot
model to assess the likely effects of the merger
Merged firm
marginal cost
2,200
2,100
price =
marginal revenue
price =
marginal cost
Price (£/tonne)
2,000
1,900
Residual demand
1,800
1,700
Marginal revenue
1,600
1,500
0
100,000
200,000
300,000
Output by merged firm (tonnes, wfe)
12
ACE Meeting – Toulouse, 2007
400,000
Unilateral effects modelling
Simple version of long term withholding model: results
• Predicted price increases not large (even assuming no
efficiencies)
– The base case for EEA production predicts price increases of around 1% if
costs are fully recoverable, and a negligible increase if costs are only
partially recoverable;
– Even narrowing to Scottish demand alone, price increases are not much
over 1% (elasticity of -3 is below the CC estimate; -2 also tested);
– We also ran a sensitivity assuming costs are not fully recoverable – in that
case price increases are negligible.
Elasticity
Market
EEA produced (baseline)
Scottish produced
13
ACE Meeting – Toulouse, 2007
-1.5
-3.0
-4.0
-5.0
Costs fully
recoverable
1.1%
1.3%
1.1%
1.0%
Costs 95%
recoverable
0.1%
0.0%
0.0%
0.0%
Unilateral effects modelling
Version of long term withholding model allowing strategic reaction
• Incorporate strategic behaviour by some rivals (share > 1%)
• Modelled as a homogeneous Cournot industry and calibrated
against 2005 market outcomes
• Predict post-merger outcomes in two steps:
– Pan Fish/Fjord Seafood (already completed, but not reflected in market
outcomes), and then this new enlarged Pan Fish merging with Marine
Harvest (the merger in question)
• Small price impact predicted from base case (even with zero
efficiencies) – equivalent to 10 to 9 symmetric merger
2005
Parties (tonnes)
Rival (tonnes)
Industry (tonnes)
Price (£/kg)
14
ACE Meeting – Toulouse, 2007
239,188
551,812
791,000
1.932
Premerger
212,971
574,913
787,884
1.937
Postmerger
148,000
632,162
780,162
1.950
Change
-64,971
57,249
-7,722
0.013
%
Change
-30.5%
10.0%
-1.0%
0.65%
Unilateral effects modelling
Sensitivity testing
• Demand elasticity:
– Ranging from -1 to -2, based on academic literature,
– Linear and constant elasticity functional forms.
• Range of industry costs/capacity constraints:
–
–
–
–
Costs at the best/average/worst sites (shallower/steeper mc curve)
Functional forms (affine, linear, powers)
Forms of heterogeneity (same slope, same mc at 0, same mc range)
Sensitivities where smaller rivals have no ability to expand at all (to address
concerns in France)
• Product market:
– All salmon farmed in Europe
– All salmon farmed in Scotland (with a higher elasticity: -2 to -5)
– Global salmon production (including Chile, in particular)
• Synergies:
– Parties expected significant cluster synergies
– Used model to calculate “critical” synergies
15
ACE Meeting – Toulouse, 2007
Unilateral effects modelling
Modelling seen as useful in UK, but Lerner index preferred in France
• UK CC: cleared the deal without remedies
“even though substantial costs can be avoided by restricting output, we
think that rivals have the potential to increase production, particularly in
Norway, if prices were anticipated to increase because the merged entity
was planning to reduce future output, which would largely offset the effects
of any such strategy. Further, demand appears to be reasonably elastic,
suggesting that large output reductions would be required by the parties to
realize significant price increases. ”
• French authorities: carried out their own estimate of impact:
– A Lerner index calculation based on pre-merger HHI;
– No scope for rival response: unsurprisingly this generates significant price
increases (around 5%);
– Appears to have been influential in the Decision, despite the availability of
Cournot modelling incorporating rival responses;
– Concerns about smaller rivals’ (in)ability to expand were also tested in the
Cournot modelling: results relatively insensitive.
16
ACE Meeting – Toulouse, 2007