THE JOURNAL OF FINANCE VOL.L VIII,NO.4 AUGUST2003
Presidential Address:Liquidity and Price Discovery
MAUREEN O’HARA n
ABSTRACT
This paper examines the implications of market microstructure for ast pri-
cing.I argue that ast pricing ignores the central fact that ast prices evolve
in markets.Markets provide liquidity and price discovery,and I argue that
ast pricing models need to be recast in broader terms to incorporate the
transactions costs of liquidity and the risks of price discovery.I argue that
猪肉萝卜包子symmetric information-bad ast pricing models do not work becau they
assume that the underlying problems of liquidity and price discovery have
been solved.I develop an asymmetric information ast pricing model that
incorporates the e¡ects.
T HIS P APER EXAMINES THE IMPLICATIONS of market microstructure for ast pricing. Both rearch areas focus on the behavior and evolution of ast prices,but the microstructure implications have been largely missing from the ast pricing literature.Such an omission is unimportant if ast pricing models work well in the n of explaining the obrved behavior of ast prices,but this is not the ca.The proliferation of anomalies,momentum,and the changing cast of factors needed to explain even partially the behavior of ast prices all suggest that success is not yet within our grasp.
I will argue in this paper that ast pricing ignores the central fact that market microstructure focus on:Ast prices evolve in markets.Markets have two important functions F liquidity and price discovery F and the functions are important for ast pricing.1I will link the two concepts to our more basic constructs of risk and expected return,and I will suggest that ast pricing mod-els need to be recast in broader terms to incorporate the transactions costs of liquidity and the risks of price discovery.I will argue that information is not sym-metric nor is equilibrium revealing.The symmetric in
formation-bad ast pri-cing models do not work becau they assume that the underlying problems of liquidity and price discovery have been completely solved.I suggest a di¡erent ast pricing framework of asymmetric information that requires rethinking n Johnson Graduate School of Management,Cornell University.I would like to thank David Easley,Franklin Allen,Campbell Harvey,Gideon Saar,and John Campbell for helpful com-ments.
1The market functions,in turn,allow markets to play many di¡erent roles,among them that of allowing individuals to reallocate their ast holdings.This allocational a role results in risk sharing among investors,something that also may a¡ect ast prices.周而复始造句
徐字组词大全
1335
the role of uninformed traders F the infamous noi traders of years past.My talk will draw heavily on my work on the topics with David Easley (,Easley and O’Hara (2001)).Thus,my ideas here are really better viewed as a joint product (with the good ideas clearly evolving from my coauthor,as has so often been the ca).
Some might argue that what I propo is simply wrong becau it is inconsistent with traditional CAPM.Others will asrt that it is technically correct,but practically unimportant.I think both objection
s are incorrect,but as they say ,the proof is in the pudding.I will end with examples of how this information-bad approach has implications for the cross ction of expected returns,the equity premium puzzle,and the concept of market e⁄ciency .
I.Ast Price F ormation
育儿知识0到1岁
Consider the standard economics explanations of how ast prices emerge.The basic W alrasian story abstracts from the actual mechanics of markets.Instead,traders turn in demands to the ¢ctitious auctioneer who aggregates the traders’buy and ll desires.The auctioneer then ts a price to clear markets.Demands depend upon consumption decisions,so ast prices re£ect the consumption decisions as well.This is a symmetric information story F all traders share the same information regarding the ast’s expected risk and return.Note that in this world,buyers and llers are all prent at the same time,so the auc-tioneer need only aggregate the expresd trading desires to ¢nd the equilibrium price.
曹操语录
Now introduce information asymmetries.The Grossman^Stiglitz (1980)model and critique provides the starting point for this analysis.In the simple version of this world,the informed traders know more than the uninformed;the unin-formed know there are informed traders but not what they know;the uni
nformed make inferences about this information from the price;everyone turns in demands (to the auctioneer);and the equilibrium price emerges.If the equili-brium is fully revealing,then the uninformed learn the information from the equilibrium price,and symmetric information characterizes the market.W e are now back in the world depicted above,where all traders face the same decision problems,but where the informed have no incentive to gather information (the Grossman^Stiglitz critique).
Ast pricing models typically start from here F symmetric information.The CAPM,the APT,and the consumption-bad CAPM all assume symmetric infor-mation.2A rationalization for this is simply to argue that we characterize ast 2
An exception is the Merton (1987)model that assumes incomplete information.The dis-tinctions between the incomplete and asymmetric information worlds will be addresd shortly,but it is uful to note here that the incomplete information models do not allow tra-ders to learn from the market price.Thus,prices play only an allocational role,and not the informational role that I argue will be important for ast pricing.The Journal of Finance
1336
斜风细雨Liquidity and Price Discovery1337 prices as if markets are in symmetric information equilibrium.But this is clearly a caricature F we know it is not right for individual asts,and it ems equally implausible for the market as a whole.Over time,as if became as is,and ast pricing models were bad on the notion that ast prices could be viewed as aris-ing from a symmetric information world.One justi¢cation given for this is that information only matters for the market as a whole;individual stock risk,the idiosyncratic risk,can be diversi¢ed away.So even if the symmetric framework isn’t true,this problem doesn’t matter:Hold enough stocks and the world ems symmetric.
But there are some obvious problems here.An immediate one is that the ex-pected risk^return trade-o¡envisioned here requires computing the market’s expectation.Even in a symmetric information world,this can be challenging if the underlying process generating ast returns is complex.Thus,rearch by Brennan(1998),Brennan and Xia(2000,2002),Xia(2001),and Lewellen and Shanken(2002)considers how one determines the risk-return trade-o¡in the pre-nce of model uncertainty,parameter uncertainty,and learning risk(but all of it in a symmetric information world).But what if we don’t all know the same thing? If there is di¡erential information,who expectation are we calculating?Lint-ner raid this concern in his1969paper,and Ned Elton(1999)discusd the im-porta
免加盟费的项目nt empirical implications of this in his presidential address. Alternatively,it may be that a nonrevealing equilibrium emerges,and prices do not level the playing¢eld between traders.Informed traders can now pro¢t from their information,and so their paradox of earning a return from their informa-tion-gathering e¡orts disappears.But now the problem is with the uninformed traders,who are losing what the informed are gaining.The solution here is noi traders.The traders,and the concept itlf,rescue the story.As Fischer Black (1986)pointed out in his presidential address,it is noi that allows markets to function.
But the noi traders always lo,raising the obvious question:Why are they so stupid?One explanation can ari from the rearch in behavioral ¢nance.Overcon¢dence,mistakes in updating,prospect theory,and framing issues all can explain why it is that traders remain so docile(or deluded).3 Alternatively,the same factors may in£uence the informed traders’behavior, providing the opening that the uninformed need to remain‘‘in the game.’’The behaviors will lead to ast prices that do not behave as predicted by symmetric information models,and so may accord better with obrved ast prices.I am sympathetic to the ideas developed by my behavioral colleagues,and I suspect that they will prove uful in expanding our understanding of ast price beha-vior.
3Shleifer and Summers(1990)provide an excellent overview of the noi trader approach to ast pricing.They note that‘‘Our approach rests on two assumptions.First,some investors are not fully rational and their demand for risky asts is a¡ected by their beliefs or nti-ments that are not fully justi¢ed by fundamental news.Second,arbitrage F de¢ned as trading by fully rational investors not subject to such ntiment F is risky and therefore limited’’(p.19).
More compelling to me,however,is that it is not noi that makes markets work,but rather that uninformed traders are smarter than we have allowed:They recognize risk and they want compensation for bearing it.4The unin-formed traders know they will lo to better informed agents,but they have port-folio choices to make,and the choices allow them to choo asts in which their risk of losing to better informed traders is lower.Information risk matters,and so,too,does the process by which information enters ast prices.And this ts the stage for the role of markets in ast pricing.
II.Asymmetric Information,Ast Prices,and the Role of Markets Markets provide liquidity and price discovery.The two concepts are related,but they are not the same.As each function can in£uence ast prices,I ¢rst dis-cuss how liquidity enters into ast price formation,and then turn to the impact of the price discovery process on ast price behavior.
Liquidity refers to the matching of buyers and llers.It is intertemporal in nature and it is not necessarily linked to price discovery.As a simple example of this distinction,suppo that all buyers of an ast arrive on Monday and all ll-ers on Tuesday .The buyers and llers may all agree on the‘‘fundamental value’’of the ast,but in this illiquid world,the concept of a market price is not well de-¢ned.No trade will take place on Monday in the abnce of llers,and unless the buyers stick around until the next day ,no trade will occur on Tuesday either.In this world,a role emerges for a market intermediary who will ll to the buyers on Monday and buy from the llers on Tuesday .For providing this liquidity ,a spread emerges between the buying and lling prices to compensate the middle-man.This notion of liquidity production was applied by Demtz (1968)to explain the behavior of stock exchange specialists,and it has been expanded by a legion of authors (Garman (1976),Stoll (1978),Ho and Stoll (1981),Amihud and Mendelson (1986,1988),O’Hara and Old¢eld (1986),Grossman and Miller (1988),Biais (1993),and Madhavan and Smidt (1993),to name but a few)to a wide range of issues in market microstructure.
This liquidity-bad spread is a transactions cost for traders.Can this cost af-fect ast prices more generally?The ast pricing literature and the microstruc-ture literature diverge on this point.There is a long literature in ast pricing looking at the role of transactions costs (,Constantinides (19
86),Aiyagari and Gertler (1991),Heaton and Lucas (1996),V ayanos (1998),and V ayanos and Vila (1999)).In general,the authors argue that liquidity costs can only have a cond-order e¡ect on the level of ast prices becau transactions cost are just 4
One argument for limiting the trading motivations of uninformed traders has been the Milgrom^Stokey (1982)critique that becau the uninformed always lo to the informed,rationality requires that the uninformed only trade for nonspeculative purpos,or simply not trade at all.If markets are not complete,however,then the arrival of information to some traders changes the risk^return trade-o¡for all traders.Becau new risk-sharing opportu-nities ari,the uninformed are not trading for purely speculative purpos F they can and must trade,if they are rational.The Journal of Finance
1338
Liquidity and Price Discovery1339 too small relative to the equilibrium risk premium to matter.5The counter argu-ment was originally put forth by Amihud and Mendelson(1986,1988)and sub-quently expanded by numerous authors(e Brennan and Subrahmanyam(1996), Brennan,Chordia,and Subrahmanyam(1998),Chalmers and Kadlec(1998),Chor-dia,Roll,and Subrahmanyam(2000),Pastor
and Stambaugh(2001),and Amihud (2002)).The authors argue that empirically,ast prices do re£ect liquidity costs,with studies linking ast price behavior to a variety of liquidity measures such as spreads,depths,and volumes.
In this context,liquidity is akin to a tax or a cost borne by investors.It ems to me that if the costs are large enough,they should negatively a¡ect ast prices becau of their e¡ect on net ast returns.6In the same vein,reducing the costs through,for example,the introduction of a more e⁄cient trading mechan-ism should have an immediate positive e¡ect on an ast’s value.7The microstruc-ture of the market in£uences the liquidity costs,and so if the e¡ects are large enough,microstructure and liquidity a¡ect ast returns.
Can liquidity also a¡ect the risk of holding an ast?Here the issue is more complex,as liquidity would then have to be time varying,or at least be systematic in some n.There is a growing literature addressing this issue,with Chordia et al.(2000),Huberman and Halka(2001),and Amihud(2002)arguing that there are systematic factors here,while Hasbrouck and Seppi(2001)¢nd the opposite. Whether liquidity is a risk remains contentious,in part becau it is unclear what would generate commonality in liquidity.And even if such commonality exists,it may be diversi¢able across ast class.8This would suggest only a c-ondary role for liquidity in a¡ecting an ast’s risk.
But this is not the ca for the other function of markets,price discovery.Price discovery involves the incorporation of new information into ast prices,and it requires that we consider again the role of the informed and uninformed traders. For reasons given earlier,I will focus on partially revealing rational expectations
5Huang(2001)agrees that in general this is true,but that it need not be the ca if traders are constrained from borrowing against their future income stream.Holmstrom and Tirole (2001)develop a related¢rm-bad argument in which¢rms demand liquidity to meet future cash needs.In this model,asts from expected returns can be a¡ected by their covariance with market liquidity.
6Try lling a pony F this is a real cost!Indeed,for many class of asts,the transactions cost of locating buyers and llers has a signi¢cant in£uence on ast value.
7Thus,a number of rearchers have found price e¡ects for curities moving from one trading mechanism to another(,Christie and Huang(1994)or Besmbinder(2003) on curities moving from the Nasdaq to the NYSE)or when exchanges move curities from one trading platform to another(e Amihud,Mendelson,and Lauterbach(2003)).An inter-esting recent example of this liquidity e¡ect is the ri of electronic markets such as Ebay.It is now commonplace for traditionally illi
quid asts such as antiques to trade on Ebay,creat-ing a meeting place for buyers and llers.For at least some asts(such as antique Ithaca Calendar Clocks)the result has been a more active market and a more valuable ast.
8In particular,if some event caus a liquidity problem in one market,it may induce a cor-responding liquidity in£ow in another market.Examples of this could be the‘‘£ight to qual-ity’’obrved periodically in the bond markets or in the markets for emerging market debt. For an interesting analysis of liquidity linkages between stock and bond markets,e Chor-dia,Sarkar,and Subrahmanyam(2002).
天高
