Transcript BE501

Determinants of Cartel
Duration
Margaret Levenstein
Valerie Suslow
ESRC Centre for Competition Policy
University of East Anglia
June 2009
1
AGENDA
• Sample of Contemporary International
Cartels
• Determinants of Cartel Duration
–Economic Theory
–Empirical Results
2
International Cartel Prevalence
• 1930s:
– 30 to 40% of world trade controlled by international cartels
• 1950s through 1980s: Don’t ask, don’t tell
– “we didn’t have any evidence that international cartels
continued to be a problem” (Klein, 1999)
• 1990s: Vigorous extraterritorial application of U.S. antitrust
law
– Mark Whitacre and ADM lead DOJ to revise Corporate
Leniency Policy
» Amnesty from all U.S. criminal penalties automatically
granted to first firm to confess (if no pre-existing
investigation). EU changed to automatic amnesty in 2002.
» Amnesty Plus Program: Substantial reduction in
penalties by admitting to cartel practices in other
markets that are not being investigated.
3
Cross-Section Sample
Sample Criteria
1. involve more than one producer
2. include producers from more than one
country
3. attempt to set prices or divide markets
4. convicted by US or EU in 1990s or 2000s
• Yields a sample of 81 cartels for which we have
been able to collect cartel and market
characteristics.
Sample selection bias: Are only the fragile
or careless cartels caught?
4
Who Colludes?
• Answer #1: Everyone, but it varies by time and
place
» Previous studies: industry composition reflects legal
and institutional environment, extent of market
integration, and sample selection criteria
» This sample: primarily sophisticated manufactured
goods and services
• Answer #2: Firms in highly concentrated
industries
» Average C4 in our sample is 80%
» Mean # firms = 7.1 and Median = 5
• Maximum # firms = 35, but of the 12 cartels with
more than 10 members, 10 of these involved a
trade association
5
Cartel Duration
• On average, cartels in our sample last 8 years
12
Number of Cartels
10
8
6
4
2
0
1
2
3
4
5
6
7
8
9
10 11 12 13 14 16 17 18 19 21 22 23 29
Cartel Duration (Years)
6
Basic Theory of Cartel Duration
• In a market with identical price-setting firms, infinitely
repeated interaction among these firms, and perfect
information, collusion can be sustained if


t 0
t 1
t
i
M
M
i
D
M
t
i
C
C


(
p
,
p
)

(
p
,
p
)



(
p
,
p


i ,t
 i ,t
i ,0
 i ,0
i ,t
 i ,t )
collusive (monopoly) price charged by firm i in period t
price charged by firm i if it chooses to defect
price charged by firm i in continuation equilibrium after defection
• Permanent collusion can be an equilibrium if firms are
sufficiently patient and if the difference between the profits
earned while colluding and the profits earned after a firm
cheats is sufficiently high
7
Basic Theory of Cartel Duration
• So, what can this model tell us about how long an
existing cartel will last?
• Some theories of cartel behavior have tried to answer this
question by considering equilibria in which collusion does
last, but in which firm behavior fluctuates. These behavioral
fluctuations are often observationally equivalent to cartel
breakup.
• We also consider the possibility that, while firms expect this
inequality to hold when a cartel is formed, the constraint is
violated by future unanticipated shocks.
• Must distinguish between those that broke up because
– They fell apart
– They got caught
8
Determinants of Cartel Duration
• Hypothesis: Cartel organization matters to cartel
duration.
– Successful cartels develop mechanisms to
prevent and punish defection without expensive
price wars.
» Rarely included in cross-sections. Several
case studies provide rich detail, e.g., Genesove
and Mullin (2001), Spar (1994), Baker and
Faulkner (1993)
9
Cartel Organization
Eleven characteristics
 Monitoring of output and
pricing
Vitamin
A: “If at the end of the year a
Vitamin B2: Hoffman-LaRoche
producer
was
substantially ahead of its
 Market or customer allocationmonitored
Japanese government
quota, it had to purchase vitamins from
export data knowing
there was
 Agreeing to other terms
conditions
of salethatthem
the&
others
in order to compensate
only one cartel member, Takeda,
for
the
corresponding
in their
 Standardize product characteristics
orshortfall
quality
producing
in
that geographic
Organic Peroxides:
Producers
“agreed that
allocation.”
purchase...a (new)
 Control of distribution each of them would location.
competitor. Akzo agreed to acquire the
 Trade association participation
organic peroxide business of Nobel and
Enichem. Laporte would purchase Aztec.
 Bid rigging
Atochem would take over [...]. Only the
 Compensation scheme for
exceeding cartel quotas
latter did not occur.”
 Disciplinary actions in response to cartel violations
 Exclusionary actions to prevent entry or expansion
 Hierarchical internal structure
10
Other Determinants of Cartel Duration
• Discount rate
• External shocks
– Demand variability
– Exchange rate variability
• Industry structure
– Number of firms
– Industry concentration
11
Variable Definitions & Descriptive Statistics
N = 81
Variable
MEMNUM
MINC4
BIDRIGORG
MKTALLOCORG
STDIZEORG
OTHERTERMSORG
DISTNORG
TAORG
Min
Max
Mean
StdDev
2
24
0
0
0
0
0
0
35
100
1
1
1
1
1
1
7.1
74.8
0.24
0.80
0.08
0.45
0.15
0.31
5.9
18.4
0.43
0.40
0.27
0.50
0.36
0.47
continued...
12
Variable Definitions & Descriptive Statistics
Variable
MONITORORG
DISCIPLINORG
COMPENSORG
EXCLUSORG
HIERARCHORG
Min
Max
Mean
StdDev
0
0
0
0
0
1
1
1
1
1
0.79
0.19
0.33
0.36
0.43
0.41
0.39
0.47
0.48
0.50
13
Kaplan-Meier Curves
Nonparametric: probability of failure after time t
Shows basic shape of raw survival data
0.00
0.25
0.50
0.75
1.00
Kaplan-Meier Survival Probability, Entire Sample
0
10
20
30
analysis time
14
Kaplan-Meier Curves
1.00
Kaplan-Meier, by Number of Members
0.00
0.25
0.50
0.75
Those cartels with few members fail
at about the same rate as those with
more members (MEM_OVER5 = 1)
0
10
20
30
analysis time
mem_over5 = 0
mem_over5 = 1
15
Kaplan-Meier Curves
1.00
Kaplan-Meier, by Organizational Index
0.00
0.25
0.50
0.75
Those cartels with few organizational
mechanisms fail more quickly than
those with more sophisticated
organization (ORGINDX > 3)
0
10
20
30
analysis time
org_over3 = 0
org_over3 = 1
16
Kaplan-Meier Curves
0.00
0.25
0.50
0.75
1.00
Kaplan-Meier, by Trade Association Involvement
0
10
20
30
analysis time
taorg = 0
taorg = 1
17
Kaplan-Meier Curves
0.00
0.25
0.50
0.75
1.00
Kaplan-Meier, by Cheaters Disciplined
0
10
20
30
analysis time
discinplinorg = 0
discinplinorg = 1
18
Kaplan-Meier Curves
0.00
0.25
0.50
0.75
1.00
Kaplan-Meier, by Compensation System
0
10
20
30
analysis time
compensorg = 0
compensorg = 1
19
Kaplan-Meier, by Exclusionary Action Against Non-members
0.00
0.25
0.50
0.75
1.00
Kaplan-Meier Curves
0
10
20
30
analysis time
exclusorg = 0
exclusorg = 1
20
Kaplan-Meier Curves
0.00
0.25
0.50
0.75
1.00
Kaplan-Meier, by Markets Allocated
0
10
20
30
analysis time
mktallocorg = 0
mktallocorg = 1
21
Kaplan-Meier Curves
1.00
Kaplan-Meier, by Cause of Death
0.00
0.25
0.50
0.75
Those cartels that break up on their
own fail more quickly than either
those ended by antitrust investigation
or those ended by confession.
0
10
20
30
analysis time
Death by Antitrust
Other Death
Death by Confession
22
Competing Hazards Model
Death by Anti-trust
Variable
Whitacre
Number Members
Market Allocation
Trade Association
Punishments
Implemented
Producer C4
High Customer
Concentration
Hazard Ratio
7.56**
1.08
0.12**
2.05**
0.58**
1.01
1.17
Coefficient > 1 indicates that the variable increased the probability of cartel breakup.
Coefficient < 1 indicates that the variable decreased the probability of breakup.
23
Competing Hazards Model
“Natural Death” Organizational Effects
Variable
Whitacre
Number Members
Trade Association
Punishments
Implemented
Compensation plan
Exclusive tactics
Hazard Ratio
4.43**
0.95
0.21**
3.94**
0.16**
2.14*
24
Competing Hazards Model
“Natural Death” Industry Structure Non-effects
Variable
Hazard Ratio
Hazard Ratio
Whitacre
Number Members
Trade Association
Punishments Implemented
Compensation plan
2.06**
4.07**
0.98
0.95
0.20**
0.23**
2.47*
3.21*
0.15**
0.17**
Exclusive tactics
1.81*
2.08*
Producer C4
1.00
High consumer
concentration
0.41
25
Competing Hazards Model
“Natural Death” Discount rate effects
Variable
Whitacre
Number Members
Trade Association
Punishments
Implemented
Compensation plan
Exclusive tactics
Interest coverage
T-bill
Hazard Ratio
2.82 **
0.93
0.23 **
3.43*
0.17 **
2.35*
3.30 **
0.96
26
Competing Hazards Model
“Natural Death” Macroeconomic shocks
Variable
Hazard Ratio
Hazard Ratio
Whitacre
2.68 **
0.93
0.24 **
3.42 *
0.17 **
2.41 *
2.24 **
0.91
0.24 **
3.25 *
0.15 **
2.98**
Number Members
Trade Association
Punishments Implemented
Compensation plan
Exclusive tactics
GDP change
HP Gap
1.14
1.003**
27
Conclusion
• Cartels can last
– Unexpected shocks disrupt cartels
– Increases in firm-specific discount rates disrupt
cartels
– Price wars disrupt cartels
• Organizational innovations can attenuate the
incentive to cheat
– Trade association, compensation schemes, exclusive
tactics all contribute to internal cartel stability
• Trade associations make cartels more visible to
authorities, but market division agreements
and price wars hide them
28
Comparison of Cartel Organization
Mean # firms
Median # firms
< 10 firms
Trade Association
Bid Rigging
Market Allocation
Other Terms Set
Disciplinary or
Exclusionary Actions
Monitoring
Contemp.
Intl.
Cartels
7.1
Hay &
Kelley
Fraas &
Greer
Posner
Gallo et al
7.3
16.7
29.1
3.9
5
81%
31%
23%
80%
44%
43%
7
79%
29%
29%
35%
14%
5%
8
60%
36%
19%
26%
5%
12%
6 – 10
64%
44%
14%
26%
14%
23%
30%
27%
80%
29
Cartel Duration
Comparison of Studies
Contemp. Eckbo Griffin Suslow Posner Gallo
(N=54;
(N=51;
(N=71;
(N=1551; (N=1348;
Intl.
intl,
intl,
intl,
US, 1890- US, 1955Cartels
1888-
Mean
(yrs)
Std Dev
Range
%  5 yrs
%  10
yrs
(N=72; intl,
1971-2002,
convicted
after 1990)
18191964)
8.1
5.8
1-29
30%
23%
1969)
1997)
1984)
19201939)
5.3
7.3
8.3
7.5
5.4
1-18
6.3
1-29
6.2
1-13
40%
37%
30
Examples of Cases in our Sample
INDUSTRY
DURATION
(years)
Aluminum Phosphide
1
Bromine
3
Cable-stayed Bridges
1
Carbon Cathode Block
2
Cartonboard
5
Cement
11
Citric Acid
4
Explosives
5
Ferrosilicon
2
Fine Arts
6
Graphite Electrodes
5
Graphite, Isostatic
9
Laminated Plastic Tubes
9
Lysine
3
Memory Chips
4
Methionine
11
Methylglucamine
10
INDUSTRY
DURATION
(years)
Paper, Carbonless
3
Plastic Dinnerware
1
Polyester Staple
3
Rubber Chemicals
7
Shipping, Cent. W. African 21
Sodium Erythorbate
3
Sodium Gluconate
9
Sorbates
17
Stamp Auctions
22
Stainless Steel
1
Steel Beam
6
Steel Heating Pipe
4
Steel Tube, Seamless
5
Tampico Fiber
5
Thermal Fax Paper
1
Vitamins
9
Zinc Phosphate
4
31
Sectoral Composition
Sector
Percent
Chemicals
39.51
Metal and Minerals
16.05
Non-manufacturing
19.75
Other Manufacturing
24.69
32
Cause of Breakup
Number
Average
Duration
(years)
Government-Initiated Breakup Triggered By:
Amnesty application
17
10.29
Follow-on investigation
13
8.77
Customer complaint
7
4.00
Other sources (including whistleblowers)
29
8.17
Cheating
6
7.67
Growing fringe
7
6.43
Unknown Cause of Breakup
2
4.50
Total
81
8.07
Other Breakup
33