Transcript 20061114121198

A study of cartel stability: the
Joint Executive Committee,
1880-1886
Paper by:
Robert H. Porter
Objective
• Use weekly time series data to test
empirically that observed prices reflected
switches from collusive to non-cooperative
behavior in the JEC railroad cartel from
1880-1886
The Problem
• Market does not face a consistent demand
curve, so cartel firms don’t know if a
change in price is the result of a firm
cheating or the result of outside demand
“shocks”.
Model Assumptions
• Firms set their own production level
• Firms do not know the quantity produced
by any other firms - they only observe the
market price
• Firms’ output is homogenous (they face a
common market price)
The Game
• Repeating, multi-period game
• Firms restrict output to increase overall profits
• “Cheaters” are punished by an industry-wide switch to
Cournot quantities for a fixed period of time, resulting in
lower revenues for all firms.
• Since firms do not observe one another’s output, this
switch occurs once the market price falls below a
previously decided “trigger price”.
The Cheater’s (Dis)Incentive
• In equilibrium, firms should produce at collusive output
levels so that any price wars should occur after
unexpected drops in demand - not after actual cheating.
• Many such equilibria can be supported by an
appropriately chosen “punishment pair” (trigger price,
punishment period length).
• To be effective, the “punishment pair” must offset the
marginal gains of cheating with the implicit marginal
losses of cheating.
Setting the “Punishment Pair”
• “Trigger Price” must be high enough to offset the incentive for firms
to cheat by producing more output.
NOTE: Desired effect can also be achieved by lengthening the
punishment period.
• “Trigger Price” must be low enough to where any and every outside
shock will not lead to a reversion to Cournot pricing.
NOTE: Punishment period could also be shortened to decrease the
severity of these reversions.
• Must be flexible. As industry output decreases towards perfectly
collusive levels, the price (and therefore the incentive to cheat) will
rise. So the “punishment pair” must be able to adjust in order to
combat this trend.
The Joint Executive Committee
(JEC)
• Cartel which controlled eastbound freight shipments from
Chicago to the Atlantic seaboard in the 1880s
• Because of the products shipped, shipping time was
really not a factor. Therefore the assumption of
homogeneity of goods seems to hold.
• Cartel used market share allotments rather than absolute
quantities shipped.
• Firms set rates individually, and the JEC office took
weekly accounts to see the total amount transported.
The JEC
(Continued. . .)
• Demand was variable, so any change in market share
could be from a change in a firm’s pricing (cheating) or
outside forces.
• Cartel passively accepted any new entrants and
allocated them market shares, allowing the collusive
agreement to continue
• Principal competition came from lake steamer and
sailships which were outside of the cartel agreement
• Demand faced predictable fluctuations as a result of the
annual opening and closing of the Great Lakes to
shipping.
The Econometric Model
• Time series estimation where i denotes a
particular firm and t denotes the time period (in
weeks)
• Date from 1880 -1886, so each year is
segmented into 13 four-week segments to
account for seasonal changes.
• Model assumes output shares of JEC members
are relatively stable across episodes of
reversionary conduct
The Variables
The Equations
• Demand relationship of the industry is given to be:
log Qt = a0 + a1 log pt + a2Lt + U1t
Where:
pt = market price in period t
Lt = dummy variable equal to 1 if Great Lakes are
open and 0 otherwise
• Supply relationship of the industry is given to be:
log pt = Bo + B1 log Qt + B2St +B3It + U2t
Where:
Qt = total quantity demanded
St = vector of dummies which reflect entry and
acquisitions in the industry
It = regime indicator which equals 1 for cooperative
and 0 otherwise
The Results
Interpretations
• Demand curves are very similar because they are not
affected by the differences in PO or PN.
• Supply curves show significant differences, with the
coefficient attributed to the PN estimate being larger than
PO with half the standard error.
• Each of the variables has coefficients of the expected
sign, which would indicate that the model is well defined.
However, the “fits” are not particularly good.
Interpretations
(Continued . . .)
• Setting all variables equal to their sample mean (using the PN
estimate), we get the numbers in Table 4.
• Price was 66% higher in cooperative periods and quantity
33% lower.
• When lakes were open, price fell 4.5% and quantity fell 33%.
• As a whole, the cartel could expect to earn 11% higher
revenues during cooperative periods (about $11,000 per week
in 1880 dollars).
• The opening of the lakes caused revenues to fall about 35%.
Plot of GR, PO, PN as a Function
of Time
• PO often reflects a price war before PN, but they
normally switch back to unity together. This is consistent
with GR not picking up secret price cuts, so there is a lag
in the PN estimate.
• On average, non-cooperative periods lasted about 10
weeks.
• In this sample, price wars (using either PO or PN) were
not preceded by adverse demand shocks. Normally
incidents began after entry of another firm, though they
were not immediate (average 40 week lag time).
• This is consistent with theory, as the increase in number
of participating firms leads to increased enforcement
problems for the cartel.
• Reversions also became more frequent as the number of
firms increased.
Comparison of Studies
Comparison of Studies
• PO series collected differs markedly from an
index of cartel non-adherence created by
MacAvoy.
• We can see from Table 5 that during “breakdown”
times, PN tends to resemble PO more than
MacAvoy’s results.
• Since PN was in no way constrained to
resemble PO, it is evident that PN supports the
documentation of the Railway Review and
Chicago Tribune.
Significance Test
• Porter uses likelihood ratio tests to determine whether
structural change has occurred in the industry, or if
changes in price can be attributed to outside demand
shocks.
• Tests the null hypothesis that the coefficient on It is equal
to zero (no regime change). Uses a chi-squared
distribution with 1 degree of freedom.
• Test-statistic = 554.1 - the null is overwhelmingly
rejected!
• Conclusion: Price and quantity changes cannot be
attributed solely to exogenous changes in demand and
structural conditions.
Further Evidence of Collusion
• In the “breakdown” years of 1881, 1884, and 1885 the
JEC firms captured a much higher market share as a
whole (see Table 6).
• The PO and PN series are not systematically related to
the opening or closing of the lake steamer shipping
season.
• Frequency of reversionary period increased as the
number of market participants increased.
• These facts refute the theory that price wars occurred
only as a result of outside demand “shocks” and are
consistent with a story of dynamic cartel enforcement
mechanisms.
Final Thoughts
• Do you agree with the way in which the
model was specified?
• Are there any relevant variables that may
have been omitted? Any irrelevant ones
that you think were included?