Price Ceilings as Focal Points for Tacit Collusion:
Evidence from Credit Cards题记摘抄大全
Christopher R.Knittel and Victor Stango¤
10/30/01
Abstract
We test whether a non-binding price ceiling may rve as a focal point for tacit collusion.
Our sample contains data from the credit card market during the1980s;in the sample,most曼听公园门票
credit card issuers face a state-level interest rate ceiling,and well over half match their ceiling.
Our empirical model explicitly allows for the possibility that ceilings may have been binding.
The model yields evidence in favor of tacit collusion:a statistically signi¯cant proportion of
issuers match their ceiling even though it is not binding.Within a state,tacit collusion is less
likely as the ceiling ris,more likely as concentration or costs ri,and less likely in periods
of high demand.We also¯nd that entry into credit cards is higher where we¯nd evidence of
tacit collusion,and lower where we¯nd evidence that a ceiling is binding.It appears that tacit
collusion became less prevalent over the1980s,as entry into credit cards surged nationwide.
The results highlight a largely unconsidered adver e®ect of price cap regulation.
Keywords:Focal Points,Tacit Collusion,Price Ceilings,Double Hurdle.
¤Knittel:Department of Finance and Economics,Boston University,Boston MA02215.Email:knittel@bu.edu. Stango:Emerging Payments Department,Federal Rerve Bank of Chicago,230S.LaSalle St.,Chicago IL60604. Email:victor.stango@Thanks to Ron Johnson,Sharon Oster,Ariel Pakes,Rob Porter,Peter Reiss,Alan Sorenn,and participants at the2001NBER Summer Institute for constructive comments and suggestions.
1
.艾籺
..The Michigan Citizens Lobby asrted that the failure of virtually all VISA and Mastercard issuers in the state,including the10largest,to reduce their rates from
the maximum18%allowed by law may indicate\potentially illegal activities."\Since
smaller banks have assured us that they are making pro¯ts charging interest rates of
15%and below,it is clear that this uniformity is not justi¯ed by actual costs.We fear
the alternative may be tacit or explicit collusion,"said the Citizens Lobby director.
-from The American Banker,March26,1987
1Introduction
Price ceilings are a common form of economic regulation.While debates over their welfare and distributional e®ects are far-ranging and often contentious,one commonly held conception is that their e®ect on prices can only be negative.At the heart of this conception is the assumption that a price ceiling has no price or output e®ects when it is not binding.While a small body of work exists challenging this view,it has been to this point anecdotal or experimental;empirical evidence suggesting that non-binding price ceilings a®ect prices is largely non-existent.1
In this paper,we empirically test the hypothesis that a non-binding price ceiling may lead to higher prices{by rving as a focal point for tacitly collusive price-tting.We test the focal point hypothesis using data from credit card issuers during the1980s.During our sample period,most credit card issuers face state-level price ceilings that could plausibly rve as focal points.The price ceilings vary across and within states;there is also a group of states with no ceiling.More importantly for our purpos,many issuers match their ceiling{particularly in the early years of the sample.Finally,the states themlves vary in other market characteristics thought to a®ect the sustainability of tacit collusion.The data therefore display a great deal of heterogeneity in¯rm behavior,focal points and market characteristics.This allows us to conduct a variety of tests of the focal point hypothesis.
The novelty of our empirical approach is that it can parately identify the instance in which an issuer matches the ceiling of its home state becau it is binding,and the instance in which it 1See Isaac and Plott(1981)and Smith and Williams(1981)for experimental evidence suggesting that nonbinding price ceilings a®ect prices.
There is also a ca precedent supporting the view that horizontal agreements¯xing maximum prices can facilitate tacit collusion.In Arizona v.Maricopa County Medical Soc.,457U.S.332(1982),the maximum-fee schedule ud by a medical association was found\to have the e®ect of stabilizing an
公园野战门d enhancing the level of actual charges by physicians."
matches the ceiling even though it is not binding{a result that we interpret as tacitly collusive.The likelihood function for the data explicitly allows ceilings to be binding,by incorporating features of a standard censored model of pricing.It then extends the model to allow for tacit collusion by introducing an independent probability that an issuer matches its ceiling even though it is not binding.Our full speci¯cation us issuer-,state-and time-speci¯c covariates to allow the probability of tacit collusion to vary across issuers and time.
The results support the focal point hypothesis.Our model estimates a statistically and econom-ically signi¯cant probability of tacit collusion.In the early years of the sample,we estimate that tacit collusion is quite common;a large fraction of issuers match their ceiling even though it is not binding.We con¯rm that the facilitative power of the ceiling dissipates as the ceiling ris.We also ¯nd that tacit collusion is more likely as concentration and issuer-level costs ri,and that tacit collusion is less likely in periods of high demand.Near the end of the sample period,we identify a regime change after which tacit collusion is much less likely.We attribute this to a surge in entry into credit cards during1985-86,and aggressive competition at the national level by a t of large issuers.
In the¯nal ction of the paper,we show that our estimates of state-level tacit collusion are directly related to state-level entry rates in credit cards.Entry rates are signi¯cantly higher than average when we estimate that issuers within a state are tacitly colluding,and signi¯cantly lower than average when we estimate that issuers face a binding ceiling.The link between state-level tacit collusion and entry is quite strong from1979-84,then disappears.This corroborates our¯nding of a regime change in credit card competition in1985-86.
2Price Ceilings,Focal Points and Tacit Collusion:Theory and Empirical Implications
In this ction we discuss the empirical implications of the hypothesis that¯rms are tacitly colluding at a focal point.We also discuss some general empirical implications of models of tacit collusion. Becau the natural alternative hypothesis explaining pricing at price ceilings is simply that they are binding,we discuss the empirical implications of the focal point hypothesis in the context of this alternative.We then relate our empirical approach to previous empirical work testing for collusion and tacit collusion.
花雕酒的功效与作用2.1Tacit Collusion at a Focal Point
Under quite general conditions,¯rms may sustain supercompetitive prices by interacting repeatedly a
nd constructing strategies under which they u the threat of future punishment to sustain current cooperation.2In this context,the\Folk Theorem"asrts that for su±ciently low discount rates nearly any t of payo®s may be sustained as the outcome of a repeated game.3The Folk Theorem is powerful,in the n that it provides quite general conditions under which tacit(or explicit) collusion may be sustainable.However,this generality leads to di±culty in conducting empirical tests for collusion or tacit collusion.
In practical terms,the problem of tacit collusion often reduces to one of successful coordination. Firms can resolve the coordination problem in many ways;one such way is through the u of a focal point.The theory of focal points dates at least to Schelling(1960),who noted that in simple games with many equilibria,agents can quite often recognize a focal point and u it to coordinate.In one of his more well-known examples,Schelling discuss the problem of two people simultaneously choosing a common location(in which to meet)in New York City.Given that the game posss an in¯nite number of equilibrium location-pairs,we might expect the odds of successful coordination to be quite low.However,in practice most people who play the game choo a well-known spot {such as Times Square or the Statue of Liberty{and can successfully coordinate.In situations where¯rms t prices,it is often suggested that the\clustering"of prices occurs at certain natural focal ,$9.99).
Becau¯rms may sustain tacit collusion under a variety of obrvationally equivalent mech-anisms,we do not attempt to explicitly model the process by which a focal point facilitates tacit collusion.Rather,we develop empirical implications of the focal point hypothesis by making ob-rvations regarding the patterns of pricing that we would obrve if a focal point were facilitating tacit collusion.
我不是小三
2.1.1Empirical Implications of Tacit Collusion at a Focal Point
企业邮箱怎么注册
The¯rst empirical implication of the focal point hypothesis is that if the focal point facilitates tacit collusion,we should obrve greater clustering at the focal point than would otherwi be 2Simple forms of the models are described in Tirole(1992),Chapter6.Some well-known supergame-theoretic models of tacit collusion can be found in Green and Porter(1984),Rotemberg and Saloner(1986),Haltiwanger and Harrington(1991),and Abreu,Pearce and Stacchetti(1986).
,Fudenberg and Tirole(1991)and others for discussion of the Folk Theorem.
expected.4Becau the focal point is a price ceiling,we might expect a certain degree of clustering even abnt tacit collusion.The relevant empirical test,then,is an estimate of the extent to which ¯rms match the ceiling even when it is not binding.We outline the econometrics of this test below.
A cond implication of the focal point hypothesis is that all el equal,it becomes more di±cult to sustain tacit collusion as the focal point ris.5To e the intuition behind this claim,consider ¯rst the limiting ca in which the focal point is equal to a¯rm's one-shot non-cooperative price.In this instance,it is trivially easy for a¯rm to maintain cooperation at the focal point.As the focal point ris,pro¯ts from cheating must ri faster than pro¯ts from cooperation;this must be true becau pro¯ts from cheating re°ect unconstrained re-optimization,while pro¯ts from cooperation re°ect constrained behavior.6Becau cheating becomes relatively more attractive as the ceiling ris,we should be less likely to obrve tacit collusion in markets with higher focal points for pricing.More precily,the probability that a given¯rm matches a nonbinding price ceiling should be a decreasing function of the ceiling.
A related issue is that as costs ri,cooperation at the focal point becomes easier to maintain. Again,note that cooperation is trivially easy when costs are such that a¯rm's non-cooperative price equals the ceiling.As costs fall below this level,pro¯ts from cheating ri more quickly than pro¯ts from cooperation,becau the former re°ect re-optimization.Thus,high-cost¯rms will¯nd cooperation more attractive.
In addition to the above implications of pricing at focal points,there are other general empirical implic
ations of tacitly collusive pricing that should be testable in our tting.A¯rst implication is that tacit collusion is generally viewed as easier to maintain among a low number of¯rms.Thus, we should be more likely to obrve successful tacit collusion when market concentration is high.
A cond is that we might expect larger¯rms to be more likely to cooperate than smaller¯rms. Given that cheating is attractive becau it steals business from other¯rms,a small¯rm will¯nd the gains from cheating proportionately larger than a larger¯rm.This implies that the probability that a¯rm matches the ceiling,although it is not binding,should be an increasing function of¯rm size.A¯nal implication is that tacit collusion becomes more di±cult to sustain in periods of high demand,becau high current demand increas the current gains from cheating.7Thus,we should 4We need not obrve unanimous clustering at the focal point in order to infer tacit collusion.The Folk Theorem readily admits instances of\partial"tacit collusion,in which some¯rms tacitly collude at the focal point and others play their short-run best respons given other¯rms'prices.
5In Appendix A,we show a general t of conditions under which this is true.
6The constraint under cooperation is that the¯rm's price must match the focal point.
7See Rotemberg and Saloner(1986)for a model outlining this intuition.
expect tacit collusion to be less likely when demand is high.
2.2Testing for Collusion and Tacit Collusion
Most empirical tests for collusion or tacit collusion u one of two approaches.The¯rst approach involves testing whether the distribution of prices thought to re°ect collusion or tacit collusion is di®erent from a control distribution thought to re°ect non-collusive behavior.In such tests,a central question is whether a candidate distribution of collusive prices can be identi¯ed a priori {for example,becau it comes from a group of¯rms accud or convicted of collusion.When the candidate and control distributions can be identi¯ed in this way,it is relatively straightforward to test for equality of the distributions.Rejection of equality is taken as evidence of tacit or explicit collusion;more¯nely characterized tests can often rule out plausible alternative hypothes. Examples in this line of work include Porter and Zona(1993,1999).8
When the candidate t of collusive prices can not be identi¯ed a priori,it may be possible to endogenously identify the collusive and non-collusive distributions,often using some form of mixture modeling.Porter(1983),for example,us a switching regression to endogenously classify periods of pricing into collusive and non-collusive regimes.Ellison(1994)us a similar approach that de¯nes th
e transition probabilities between collusive and non-collusive periods using a Markov process.In each of the cas,the data clearly identify periods of collusive and non-cooperative behavior.Baldwin,Marshall and Richard(1997)employ a similar model within an auction-bidding framework.They estimate a model in which winning bids may be drawn from a collusive bid distribution with probability p,or from a noncooperative bid distribution with probability1¡p. This model outperforms a model that maintains the hypothesis of noncooperative behavior.
A cond approach to testing for tacit collusion us theoretical predictions regarding the sus-tainability of tacit collusion as a function of exogenous factors{such as demand and costs.For example,Borenstein and Shepard(1999)¯nd that current retail gasoline margins fall when expected future pro¯ts fall.This result is consistent with the Rotemberg and Saloner(1986)and Haltiwanger and Harrington(1991)models of tacit collusion.Porter(1985)and Ellison(1994)both test the prediction that unanticipated demand shocks should trigger price wars,as in the model of Green and Porter(1984).
8Porter and Zona corroborate their tests in various ways.For example,in Porter and Zona(1999)they show that bids for non-cartel¯rms are correlated with¯rm-speci¯c costs in an intuitive manner,while bids from cartel¯rms are not.
甜汤圆
