----------------------- Page 1----------------------THE JOURNAL OF FINANCE VOL. LXV, NO.

. 1 FEBRUARY 2010 Product Market Competition, Insider Trading, and Stock Market Ef ciency JOEL PERESS ABSTRACT How does competition in rms product markets in uence their behavior in equity markets? Do product market imperfections spread to equity markets? We examine these questions in a noisy rational expectations model in which rms operat e under monopolistic competition while their shares trade in perfectly competitiv e markets. Firms use their monopoly power to pass on shocks to customers, thereby in sulating their pro ts. This encourages stock trading, expedites the capitalization o f private information into stock prices and improves the allocation of capital. Sever al implications are derived and tested. HOW DOES COMPETITION in rms product markets in uence their behavior in equity markets? Do product market imperfections spread to equity markets? These questions are increasingly of interest as product markets are becoming more competitive in many countries thanks to the relaxation of impediments to trade and barriers to entry.1 In this paper, we analyze these questions usin g a noisy rational expectations model in which rms operate under monopolistic competition while their shares trade in perfectly competitive markets. The model is guided by recent empirical work showing that stock re turns are affected by the intensity of product market competition. Gaspar and Massa (2005) and Irvine and Pontiff (2009) document that more competitive rms have more volatile idiosyncratic returns, and Hou and Robinson (2006) show that such rms earn higher risk-adjusted returns. The model is also guided by a direct examination of the data. Our starting point is the nding in Gaspar and Massa (2005) that analysts earnings forecasts about rms operating in more competitive industries are more dispersed. Since differences in opinions Joel Peress is with INSEAD, Department of Finance. I thank for help ful comments Patrick Bolton, Philip Bond, Markus Brunnermeier, Murillo Campello, James D ow, Bernard Dumas, Lily Fang, Michael Fishman, Denis Gromb, Charles Jones, Massimo Massa, Jacques Olivier, Jose Scheinkman, David Thesmar, Laura Veldkamp, Bernard Yeung, and semi nar participants at Columbia Business School, HEC Paris, NYU Stern School of Business, Princeton University,

HEC Lausanne, the 2007 WFA meeting, Goldman Sachs Asset Management, the 2006 CEP R Summer Symposium in Financial Markets, the 2007 Adam Smith Asset Pricing meeting, the 2008 European Winter Finance Conference (Klosters), and the Caesarea Center 5th Annua l Academic Conference. I am also grateful to an anonymous referee and the editor, Campbell Harvey, for many insightful comments and detailed suggestions. 1Such changes may have an impact on equity markets. For example, the rise in idiosyncratic return volatility (Morck, Yeung, and Yu (2000), Campbell et al. (2001) ) may be related to the deregulation of the economy (Gaspar and Massa (2005), Irvine and Pontiff (2009)) . 1 ----------------------- Page 2----------------------2 1 . 0 r e 1 v . o n r u T 2 . 3 . 3 Weak market power 4 Strong market power 2 The Journal of FinanceR

Figure 1. Market power and turnover. This gure shows stock turnover acro ss market power groups. Turnover is de ned as the log of the ratio of the number of shares traded during a year to the number of shares outstanding. Firms are sorted every year from 1996 to 20 05 into market power quintiles. Market power is measured as the excess pricecost margin (PCM). T he PCM or Lerner index is de ned as operating pro ts (before depreciation, interest, special i tems, and taxes) over sales (Compustat annual data item 12). Operating pro ts are obtained by subtr acting from

sales the cost of goods sold (item 41) and general and administrative expenses ( item 178). If data are missing, we use operating income (item 178). The excess pricecost margin is c onstructed as the difference between the rms PCM and the PCM of its industry. The industry PCM i s the valueweighted average PCM across rms in the industry where the weights are based on ma rket share (sales over total industry sales) and industries are de ned using two-digit SIC cl assi cations. are usually a motivation for trading, we expect to nd a greater volume of trade for these rms. To analyze this conjecture, we sort rms on their market power and measure the average trading volume in each group.2 As Figure 1 shows, we nd the opposite of our conjecture. Stocks in the bottom market power quintile are traded less frequently than those in the top quintile. A possible explanation for the mismatch between belief heterogeneity and trading volume is that the opinions examined in Gaspar and Massa (2005)analysts forecastsare not representative of the overall market but of informed investors who trade differently. We explore this possibility by studying trades initiated by insider s of cers of rms who presumably have access to privileged information. Figure 2, in the spirit of Figure 1, reveals that their trading volume is again larger in rms with more market power. Thus, it appears that investors scale 2Our data and methodology are described in Section V, where we con rm that the results we presented graphically in this Introduction are statistically signi cant and robust to the inclusion of other factors including rm size. ----------------------- Page 3----------------------Product Market Competition and Stock Market Ef ciency 3 8 0 0 . 6 0 0 . 4 0 0 . 2 s n I 0 0 .

r e v o n r u t r e d i

0 Weak mkt power Strong mkt power 3 . s e d a r t r e d i s n i f o r e b m u N Strong mkt power Figure 2. Market power and insider trading. This gure shows insider trading a ctivity across market power groups over the 1996 to 2005 period. In the top panel, insider t rading activity is measured as the log of the ratio of a rms annual total insider trading dollar volu me to the rms market capitalization, and it is denoted Insider turnover. In the bottom panel, it is measured as the log of the ratio of the rms annual number of insider trades to the rms number of act ive insiders and is denoted Number of insider trades. Insider trades are open market transact ions, excluding sells, initiated by the top ve executives of a rm (CEO, CFO, COO, President, and C hairman of board). Active insiders are de ned as executives who have reported at least one tr ansaction in any of the sample years. Firms are sorted every year from 1996 to 2005 into market p ower quintiles. Market power is measured as the excess pricecost margin or Lerner index (see Figu re 1). down their trading of more competitive stocks even when they have superior information. The enhanced trading activity, especially among informed investors, for rms enjoying more market power raises the possibility that fundamental info rmation is more quickly capitalized into their stock price. To investig ate this 0 Weak mkt power 2 3 4 1 . 2 . 2 3 4

hypothesis, we measure the stock price reaction to earnings announcements across market power groups. Earnings of closely followed rms are anticipated long before their of cial release so their prices do not react to announcements. In contrast, announcements by remotely followed rms provide useful infor mation that causes investors to revise their valuation and stock prices to adjust. Figure 3 shows that rms with more market power experience smalle r price changes at announcements after controlling for standard risk fact ors, ----------------------- Page 4----------------------4 8 . 6 . n r u t e r t n e v e 4 . . n b a . s b A 2 . 0 4 Weak market power Strong market power 2 3 The Journal of FinanceR

Figure 3. Market power and stock price informativeness. This gure shows stock price informativeness across market power groups. Informativeness is measured as the a bsolute abnormal return surrounding an earnings announcement. Abnormal returns are the residu als from the FamaFrench three-factor model. For every rm, we regress stock returns on the marke t, size, and book-to-market factors over an estimation window extending from t = 250 to t = 5 r elative to the earnings announcement day 0. We estimate the residuals over an event window

ranging from t = 2 to t = +2. Then, we sum their absolute value on each day of the event windo w to measure the stock price reaction to the announcement. Finally, we average the absolute a bnormal returns estimates obtained from each announcement during a year to get an annual measure . Firms are sorted every year from 1996 to 2005 into market power quintiles. Market power is measured as the excess pricecost margin or Lerner index (see Figure 1). suggesting that their prices are more informative. This is consistent with our previous nding that insiders in these rms trade more aggressively, which

3In a similar vein, Hoberg and Phillips (2009) document that, in less compe titive industries, analyst forecasts are less positively biased and stock returns comove less with the market.

Product Market Competition and Stock Market Ef ciency 5 to most models of trading under asymmetric information in competitive stock markets (e.g., Grossman and Stiglitz (1980)) but for one difference: Our rms sell goods in an imperfectly competitive product market. Speci cally, they oper ate under monopolistic competition. Each rm owns a unique patent to produce a good, the demand for which is imperfectly elastic, so the rm enjoys some market power. Firms are subject to random productivity shocks. Investors do not observe these shocks but are endowed with private information. Trading causes private information to be re ected in stock prices but only part ially because noise precludes their full revelation. Product market power plays an important role in an uncertain environment. It allows rms to insulate their pro ts from shocks by passing the shocks on to their customers. Firms that face a captive demand for their good can hedge their pro ts effectively . But more competitive rms yield more risky pro ts, even though they face the same amount of technological uncertainty (the variance of productivity shocks is independent of the degree of competition). As Hicks (1935 p. 8) puts it, the best of all monopoly pro ts is a quiet life. This insight drives our results. Because the pro ts of rms with more market power are less risky, investors

Page 5

dence, monopoly power in product markets reduces the dispersion of earnings forecasts (Gaspar and Massa (2005)) but stimulates trading, including that by insiders, and enhances the informativeness of stock prices (Figures 13). In ad dition, it lowers risk adjusted expected returns (Hou and Robinson (2006)) and idiosyncratic return volatility (Gaspar and Massa (2005), Irvine and Po ntiff (2009), Chun et al. (2008)). The contribution of this paper is to present a rational expectatio ns model that explains these observations and provides further insights into how product market competition interacts with information asymmetries. Ours is similar

speeds up the incorporation of information into prices.3 To summarize the evi

trade their stock in larger quantities (even though their private signals about productivity are just as accurate). These larger trades, in turn, expedite the i n corporation of private information into prices. The improved accuracy of public informationstock prices are more informativefurther encourages investors to trade. It also makes their pro t and productivity forecasts less dis persed as they rely more on public information and less on their own privat e sig nals. Thus, investors disagree less but trade more, and stock prices are more informative, in line with the evidence presented above.4 Furthermore, rms with more market power have less volatile, and on aver age lower returns. These effects obtain in imperfect competition models regard less of information asymmetries simply because their pro ts are less risky. The novel aspect emphasized here is that monopoly power also exerts an indirect in uence through the informativeness of prices, which further reduces volatility and expected returns, even after adjusting for risk. Indeed, stock prices of mor e monopolistic rms track future pro ts more closely, allowing returns to absorb a smaller fraction of shocks.5 The ratio of expected excess returns to their st an

4This nding may explain why stock picking appears to be declining in the Unite d States since the 1960s (Bhattacharya and Galpin (2005)) as competition in product markets int ensi es (Gaspar and Massa (2005), Irvine and Pontiff (2009), Chun et al. (2008)). 5When information is perfect, for example, prices re ect technology sh ocks perfectly while returns equal the risk free rate. 6These ndings are consistent with the dramatic increases in idiosyncr atic return volatility (Morck et al. (2000), Campbell et al. (2001)) that occurred in the U nited States following the deregulation of product markets (Gaspar and Massa (2005), Irvine and Pontiff (20 05), Chun et al.

The effects considered so far are essentially nancialthey involve trades, prices, and returns. An important contribution of the paper is to show that they extend to real variables when rms raise new capital. In that case, investors not only value rms, but also determine how much capital rms are to receive. We nd that fresh capitalthe proceeds from share issuancesis more ef ciently distributed when rms have more market power. The reason is that their stock prices are more informative so investors can easily identify the

Page 6

The Journal of FinanceR

dard deviationa measure of expected returns adjusted for risk, known as the Sharpe ratiois reduced too, indicating that the informational effect of mar ket power is stronger on the risk premium than on risk. These results suggest that product market deregulation ampli es return volatility not only because it deprives rms from a hedge but also because public information, conveyed by stock prices, is less accurate.6

better technologies to channel more funds to. In other words, informational ef ciency feeds back to real ef ciency. Thus, competition, rather than the lack thereof, generates an inef ciency when it interacts with information asymmetries. That is, product market imperfections, rather than spreading to equity markets, tend to limit stock market imperfections.7 Our paper relates to several important strands of literature. It is part of the research agenda that links industrial organization to nancial market s. Starting with the work of Titman (1984) and Brander and Lewis (1986), schol ars have established, both theoretically and empirically, that rms capita l structure and the intensity of competition in rms product market are jointly determined.8 In particular, debt can be used strategically to relax informa tional constraints. In Poitevin (1989), for example, debt signals to i nvestors that a rm entering a market dominated by a monopoly has high value, while in Chemla and Faure Grimaud (2001) it induces buyers with a high valuation to reveal their type to a durable good monopolist. Less is known about how other nancial variables such as trading volume and the informativeness of stock prices are related to market power. P erotti and von Thadden (2003) argue that a rms dominant investors can limit the informativeness of its stock price by being opaque, which in turn mit igates product market competition. In Stoughton, Wong and Zechner (2001), con

sumers infer product quality from the stock price, so a high quality entrant has an incentive to go public to expose itself to speculators attention. Tookes (2007) is most closely related to our work. She examines trading and informa tion spillovers across competing stocks and shows that informed agents prefer to trade shares in a more competitive rm, even if their information i s not speci cally about this rm but about a competitor. In her setting, agents are risk neutral and capital constrained so they seek the stock with the greatest sensitivity to shocks. In contrast, we assume that agents are risk averse and characterize how the risk return trade off varies with a rms market power. Our paper also belongs to the large body of research on trading under asymmetric information. This literature studies the impact of information on (2006)). They are also in line with Hou and Robinson (2006), who document that rm s in more competitive industries earn higher returns after adjusting for risk. 7In our setting, product market power does not generate a net social gain. Ra ther, it reduces the social loss. This is because our solution technique assumes that shocks are small. Hence, stock prices and investments differ only slightly from those obtained in a riskless ec onomy. 8For example, rms may choose low leverage ratios to guarantee that they will b e able to service their products (Titman (1984)) or high leverage ratios to commit to aggressive o

perating strategies through limited liability provisions (Brander and Lewis (1986)).

Product Market Competition and Stock Market Ef ciency 7 nancial variables. To the best of our knowledge, the role of product market power has not yet been examined in this context. Our paper contribute s in particular to the subset of this literature that emphasizes the real bene ts of informational ef ciency. In our model, stock prices re ect the quality of rms investment opportunities and help investors channel capital to the better ones.9 The remainder of the paper is organized as follows. Section I describes the economy. Section II solves for the equilibrium and Section III studies how it is affected by competition. Section IV considers extensions to the baseline model. Section V confronts the model with the data. Section VI concludes. Proofs are provided in the Appendix. I. The Economy Ours is a standard rational expectations model of competitive stock trading under asymmetric information (e.g., Grossman and Stiglitz (1980)) but for one important difference: Firms in our setting enjoy monopoly power in their prod uct market. The economy consists of two sectors, a nal and an intermediate goods sector. Intermediate goods are used as inputs in the production of the nal good. They are produced by rms operating under monopolistic competition and subject to technology shocks. These shocks are not observed but investors receive private signals about them. Monopolies stocks trade competitively on the equity market. Their prices re ect private signals but only partially because of the presence of noise. Prices in turn guide investors in their portfolio allo ca tions. Time consists of two periods. In the investment period (t = 1), markets open and investors observe their private signals and trade. In the production period (t = 2), intermediate and nal goods are produced and agents consume. The model is further de ned as follows. A. Technologies A.1. Intermediate Good Sector There are M monopolies operating in the intermediate good sector. Monopoly m (m = 1 toM) is the exclusive producer of good m. Its production is determined by a risky technology that displays constant returns to capital: Ym AmK0 where Am is a technology shock speci c to lue of for all m = 1, . . . , M rm m and K 0 is the book va

its capital stock. Firms are endowed with an arbitrary capital stock K 0, which cannot be adjusted. Our analysis focuses on the interplay between competition in the product market and information asymmetries in the equity market, for which the initial capital stock is irrelevant. We allow rms to change their capital stock in the last section of the paper, where they raise fresh capital.

Page 7

We 9See, for example, Dow, Goldstein, and Guembel (2006) and the references ther ein. ----------------------- Page 8----------------------8 The Journal of FinanceR

assume that the M intermediate monopolies are entirely nanced with public equity. Goods are produced and then rms are liquidated in the production period m m (t = 2). Thus, if rm m sells Y goods at a price of Q , its val ue at t = 2 is m m m m = Q Y . The technology shocks A (m = 1 to M) are assumed to be lognormally distributed and independent from one another. Market power allows rms to insulate their pro ts from productivity shocks. They increase goods prices in times of shortage (bad shock) and decrease them in times of abundance (good shock). This behavior complicates modeling under rational expectations because it leads to stock payoffs that are not linear in shocks. As a result, the extraction of information from equilibrium stock prices can no longer be solved in closed form.10 For this reason, we resort to a small-risk expansion. W e assume that the productivity shocks are small and log-linearize rms reactions to these shocks. Speci cally, we assume that ln Am amz, where am can be interpreted as the growth rate of technology m and z is a scaling factor, and amz is normally distributed with mean zero and precision h /z (variance z/h ). a a The model is solved in closed form by driving z toward zero. Peress (2004) demonstrates the convergence and the accuracy of such an approximation in a noisy rational expectations economy. Throughout the paper, we assume that the scaling factor z is small enough for the approximation to be valid. A.2. Final Good Sector

Intermediate goods are used as inputs in the production of the nal good. Many identical rms compete in the nal good sector and aggregate to one representative rm. The nal good is produced according to a riskless technology, M G (Ym)1m , m=1 m here G is nal output, Y is the employment of the m th type of intermedi ate good, and m is a parameter bet een zero and one.11 The parameter m is the key parameter of the model. It measures the degree

of market po er enjoyed by

rm m. To see this, note that

nal good producers

10Rational expectations models of competitive stock trading under asy mmetric information typically assume that preferences display constant absolute risk aversion (CARA) or risk neutrality and that random variables, including payoffs and signals, are normally distribut ed. Equilibrium stock prices are conjectured to be linear functions of these random v ariables. The preference assumptions generate stock demands linear in expected payoffs and price s hile the normality assumption leads to expected payoffs linear in signals including prices, thus va lidating the initial guess. The canonical examples are Grossman and Stiglitz (1980) ith competitive traders and Kyle (1985) ith strategic traders. Alternative assumptions are used, for example, by Ausubel (1990), Rochet and Vila (1994), Barlevy and Veronesi (2000), and Peress (2009). Bernardo and Judd (2000) and Yuan (2005) use numerical solutions to solve more general models. 11The nal good technology provides a convenient ay of aggregating the differe nt goods produced by the monopolies. It is in the spirit of Spence (1976), Dixit and Stiglit z (1977), and Romer (1990), among others, and is used in much of the industrial organization literat ure. ----------------------- Page 9----------------------Product Market Competition and Stock Market Ef ciency 9 M set their demand for inputs to maximize pro ts, G m=1 Q king in m as given (we use the price of the nal the termediate goods prices Q m m 1/m 12 numeraire). The resulting demand for input m is Y = [(1 . Its m m m elasticity, d ln Y /d ln Q , equals 1/ and declines hen Thus, the higher m, the less elastic the demand for good m and the arket po er rm m exerts. When m is identical across rms (m = for all m), m Y m , ta

more m it can

be interpreted as (the inverse of) the degree of competition in the intermediate goods sector. Indeed, the elasticity of substitution bet een any t o g oods m m m m m

good as m )/Q m gro s. ]

and m , d ln(Y /Y )/d ln(Q /Q ), equals 1/. So the inverse of measure s the extent to hich inputs are substitutes for one another, a lo er indicating more substitutability and a more competitive input market. In the limit hen = 0, inputs are perfect substitutes and the intermediate goods sector is perfectly competitive. The main characteristic of market po er is that it makes monopoly pro ts less sensitive to technology shocks. Substituting the demand for intermediate goods into the expression for these pro ts yields m = (1 m)(Ym)1m . Thus, 1 m also measures the elasticity of pro ts to shocks, (ln m)/(ln Am), for a given stock of capital K0 . B. Assets Monopolies equity trades on the stock market. We normalize the numbe r of shares outstanding to one perfectly divisible share. The price of a share of rm m is denoted Pm. To avoid the GrossmanStiglitz (1980) paradox, e assume that some agents trade stocks for exogenous random reasons, creating the noise that prevents prices from fully revealing private signals. We denote by m the aggregate demand for stock m emanating from these noise traders as a fraction of investors wealth, that is, m is the number of shares noise traders purchase multiplied by the price of stock m and divided by wealth.13 We assume that mz is normally distributed with mean zero and variance 2z, and i independent from all other random variable and acro tock . Thi formulation implie that the level of noi e trading i identical acro ector and doe not bia our re ult . There are no hort- ale con traint . A ri kle a et i available in perfectly ela tic upply, allowing inve tor to borrow and

The main deci ion maker in our economy are inve tor . There i a continuum of them, indexed by l in the unit interval [0, 1]. They derive utility from the con umption of the nal good g . Utility di play con tant relative ri k aver ion 12 The demand for input m i independent of the demand for input m becau e th e nal good production function i eparable. Thi impli e the analy i ub tantially. 13We derive inve tor demand for tock at the order 0 in z when we olve the model. Accordingly, m repre ent the order-0 component of noi e trader demand. ----------------------- Page 10----------------------10 (CRRA): 1 1 The Journal of FinanceR

C. Inve tor

lend freely. The ri kle

rate of return i

denoted Rf

= 1 + r f z.

U( ) =

where > 0 measures relative risk aversion and = 1 corresponds to lo util ity. Investors are endowed with a portfolio of stocks and bonds. We denote by w and f m (m = 1 to M) a ent ls initial wealth and the fraction of wealth initiall y 0,l invested in stock m respectively. We assume for simplicity that investors start with the same initial wealth w, thou h its composition (the f m s) may vary 0,l arbitrarily. Investors choose new portfolio wei hts flm in the investment period (t = 1) and consume in the production period (t = 2). D. Information Structure Investors do not observe technolo y shocks in the intermediate ood sector when they rebalance their portfolio (t = 1).14 But they are endowed with some private information. Speci cally, investor l receives a private si nal sm about l s technolo y shock: rm m

sm = am + m, l l wh r m is an rror t rm ind p nd nt of th rms pro t m and across ag nts. l Th t rm mz is normally distribut d with m an z ro and pr cision h /z (varil s anc z/hs ). W assum for simplicity that pr cisions ar id ntical across stock s and inv stors. E. Equilibrium Conc pt W d n th quilibrium conc pt for this conomy, starting from individual maximization (conditions (i) and (ii)) and proc ding to mark t aggr ga tion (conditions (iii) and (iv)).

m t rm diat goods to maximiz pro ts taking pric s Q (m = 1 to M) as giv n. As shown abov , this l ads to a d mand for input m qual to m m m 1/m Y = [(1 )/Q ] . (ii) In the investment stage, investor l sets her portfolio eights flm g uided

(i) In th for in-

production stag ,

nal good produc rs s t th ir d mand

by stock prices Pm(m = 1 to M) and her private signals s . Investors ar e l 14The model assumes that investors have private information about the ects hile rms prosp

managers do not. Accordingly managers, unlike investors, do not make any decisio n. This assumption allo s us to focus on the in uence of product market competition on investors incentives to trade and on the aggregation of their dispersed private signals through price s. An alternative interpretation of the model is that managers possess private information that th ey have already conveyed (possibly imperfectly) to the market. This information is encoded in th e prior distribution of technology shocks used by investors. Our focus is on the additional trading a nd informativeness generated by investors private signals. ----------------------- Page 11----------------------Product Market Competition and Stock Market Ef ciency 11 atomistic and take stock prices as given. Their problem can be expressed formally as: M w, max (1) {flm,m=1 toM} m=1 where c l and Fl mption and information set, and Rm le and lo returns on stock m. Firm m enerates a pro t m before bein liqui dated, yieldin a ross stock return of Rm = m/Pm (there is one share outstandin ). Investors hold a position in every stock, be it lon or sh ort, s. The return on their portfolio equals Rf problem + M since there are no transactions costs nor short sales constraint f m Rm Rf . Their m m sl , P for m = 1 to M E[U(c ) | F ] l l subject to c l = Rf + f m Rm Rf l

and rmz = ln(Rm) denote the simp

m=1 l is simpli ed by notin that the nal ood production function is separa ble. This implies that the demand for input m is independent from the (as stated in condition (i)), and therefore quantity employed of input m

denote a ent l s consu

15 that the return on stock m is independent from the return on stock m . m (iii) Intermediate oods prices Q (m = 1 toM) clear the market for inte rme diate oods: [(1 m m 1/m m )/Q ] = A K0 for m = 1 to M, here the left-hand side is the demand for good m and the right-hand side its supply. (iv) Stock prices Pm(m = 1 to M) clear the market for stocks:

Pm

Pm

where the integral and wm/Pm repre ent, re pectively, the number of and

ide i the number of hare out tanding. We are now ready to olve for the e uilibrium. II. E uilibrium Characterization

measure the value of P , , and K when z = 0 (in which case Am = Rf 1 for

out ut to favor another. Their o timal o erating strategy is to maximize ro ts in all rms. ----------------------- Page 12----------------------12 The Journal of FinanceR , , and k are

functions of a and that ca ture the order-z erturbation induced by the shocks and the riskless rate. Our focus throughout the a er is on the interaction of market ower with the shocks, which is re ected in the order-z term, that is,

all m, that is, there is no risk and no value to time). The terms

15Inde endence across f limiting one rms

rms im lies that shareholders are not better of

We di cu the trading and pricing of monopolie tock. From now on, we con ider a generic tock and drop the uper cript m when there i no ambiguity to implify the notation. We gue that tock price are approximately (i.e., at the order z) log-linear function of technology and noi e hock , olve for portfolio , derive the e uilibrium tock price , and check that the gu e i valid. We expre tock price , pro t , and capital a P exp(pz ), exp(z), and K ex (kz) at the order z. Note that P , , and K are deterministic constants that

hare

demanded by inve tor

and noi e trader , and the right-h

flm dl +

m = 1

for m = 1 to M,

librium when technolo y shocks are not perfectly observed.

(1 )2 p0() (2) p () (1 ) 1 a a ha h()

(1 )K 1 0 r f ,

0,

p () (1 )p (),

h() h () (4) p s (1 ) 2 2 2 a h2 s , and h() h

fl = (5)

Proposition 1 con rms our initial guess that prices are approximately loglinear functions of technology and noise shocks. This is illustrated in Figures 4 and 5, hich depict p , pa , and p . The technology shock a a ears directly in the rice function because individual signals sl , once aggregated, colla se

(1 )

l +

(1 )K 1 0

Inve tor l

allocate a fraction fl of her wealth to each

hs + h () + h . p

tock uch that

Shares trade at a = p0() + p ()a + p () , a

price

P = (1 )K0

PROPOSITION 1: There equilibrium characterized as follows.

exists a lo linear

rational

expectations

1 exp(pz ), here p

, , and k. We begin with a brief discussion of a benchmark economy in which technology shocks, A = ex (az), are observed erfectly. The e uilibrium in this riskless economy is solved in closed form without any a roximation, unlike the general case. Given its ca ital stock K0, a mono oly generates a ro t = (1 )K 1 exp[(1 )az] at t = 2 (see Section I.A.2). Its stock trades at 0 t = 1 for P = /Rf = (1 )K 1 exp[(1 )az r f z]. Investors earn the risk 0 less rate on a riskless investment. The followin proposition describes the equi

to their mean, a. The noise shock enters the rice e uation because it re resents noise traders demand. The rice P reveals a + = (a + (1 )/h ), a a a s signal for a with error (1 )/hs . Investors cannot tell whether the valuation of an ex ensive stock is justi ed by a good technology (a large) or by large noise 2 2 2 2 trades ( large). The function var[ (1 )/hs ]/z = (1 ) /h the noi ine of tock price and it inver e, hp , it informativene . Note th at h = z/var (az | F ) mea ure the total preci ion of an inve tor informat ion. l l She u e information about pro t from three ource , namely, her prior (the ----------------------- Page 13----------------------Product Market Competition and Stock Market Ef ciency 13 3.5 3 2.5 ) P ( e c i r p k 1.5 c o t S 1 0.5 0 10 5 10 0 0 -5 -5 -10 Noi e hock () a) -10 5 2

Figure 4. The e uilibrium tock price. The tock price i plotted aga in t the realization of the technology hock a and noi e hock . The parameter are = 0.8, ha = 1, 2

Technology

hock (

mea ur

= 0.1, r f = 0.02, M = 10, w = 0.3, = 1, hs = 0.1, and z = 0.1. ha term), the stock price (the hp term), and her private si nal (t he hs term), and their precisions simply add up (equation (4)). The rst two sources of in formation are public and their total precision equals ha + hp . The equilibrium coincides with that obtained in the riskless benchmark economy when informa tion is perfect (h = hs = , p0 = r f , pa = (1 ), and p = 0). Investors hold

a osition in every stock, be it long (fl > 0) or short (fl < 0). Their ortfolio shares are ex ressed in e uation (5) as the average weight across inv estors (the order-0 com onent of the rms ro t (1 )K 1 divided by investor 0 ealth ), minus noise trades tilted by their rivate signal errors l , scal d by risk av rsion , the precision of their si nal hs , and one minus market power . III. The Impact of Market Po er In this section e examine ho po er in rms product market affects the equilibrium outcome. We start by analyzing trades. From trades follo informativeness of stock prices, dispersion of investors forecasts, distributi on of returns, liquidity, and allocative ef ciency. A. Trading Volume We study the impact of product market competition on investors tradi ng activity. Trading volume is measured as the value or the number of s hares ----------------------- Page 14----------------------14 The Journal of FinanceR

e c i r p k c o t s f o

) a 1 p ( s k c o 0.5 h s

y t i v i t i s n e S 0.2

. h c 0 e t o t 0 0.8 0.6 1

Weak

Strong

e c ) 2 i r

k c o t s f o

s k c 1 o h s

y e t s i i 0 v o i t i n 1 s n o 0 t

Private signal 0.4 precision (hs)

0.2

0.8

0.4 0.6 Mkt po er ()

e S 0.2 Weak

0.8 0.6 0.4 Private signal recision (hs) 0.4 0.2 0 0.8 1 Strong technology and noise shocks. and to noise shocks p (bo 0.6 Mkt

ower

()

w = 0.3, = 1, and z = 0.1.

traded, conditional on the distribution of stock endowments (the f m s) as in 0,l Holthausen and Verrecchia (1990). The number of shares traded coincides with the stocks turnover given that there is one share outstanding. The following ro osition characterizes the relation between trading volume and market ower.

----------------------- Page 15----------------------Product Market Com etition and Stock Market Ef ciency 15 0 10 8 e m u l o v g n i d a 0.2 0.4

0.6

0.8

1 100 80 s s e n e v i t a m r o 10 8

60

40

the

to

left

The ro osition establishes investors to trade. This is illustrated in 6. Intuitively,

that

market

ower

encourages of Figure

anel

PROPOSITION 2:

Trading volume is larger for rms with more market

Figure 5. The sensitivities of the stock price to The sensitivities to technology shocks pa (top panel) ttom anel) are lotted against the recision of rivate information hs and The parameters are ha = 1, 2 = 0.1, r f = 0.02, M = 10, K0 = 1,

the degree of market

ower.

ower .

f r T 2 0 0 0.4 0.6 Mkt po er () 0.07 n 0.06 o i s 0.05 r e p s 0.04 i d s t s 0.03 a c e 0.02 r o F 5 0.01 0 0 0.4 0.6 Mkt po er () 0.4 0.03 0.2 0.8 0.4 0.6 0.8 1 1 Mkt po er () 0.4 0.8 1 0.1 1 i t 0.025 0.08 0.8 r a 0.25 0.02 y t 0.6 h S 0.06 d a r e p 0.3 o 0.35 0 0 0.2 i L 10 t i d i u 15 q 20 y 0.2 0.8 0.4 0.6 0.8 1 1 Mkt ower () 20 0 n I 2 0 0 0.2

0 0.6

0.2

30 25

0.6

0.8

0.2

i l i t 0.015 a l o V 0.4 0.01

0.2 n a m 0.15 u i m e r p k s i R 0 0 0 0.2 0.6 Weak 0 0.6 1 Mkt po er () Strong 0.8 0.4 0.8 Strong 1 0 0 Weak 0.2 M 0.1

0.04

0.005 0.2

0.02

0.05

0.4 kt po er ()

Figure 6. The impact of market po er on the equilibrium. on the x-axis meas ures the rms po er in its product market. The top left panel displays the share turnover (solid curve, right scale) and the dollar trading volume (dashed curve, left scale). The top r ight panel displays the informativeness of stock prices hp . The middle left panel displays the disp ersion of investors forecasts about the rms pro t (solid line, right scale) and technology sh ock (dashed line, left scale). The middle right panel displays liquidity 1/p . The bottom left anel dis lays the variance of log ro ts (solid curve, right scale), the variance of log rices (dashed curve , left scale), and the variance of stock returns (dotted curve, right scale). The bottom right anel di s lays a rms risk remium (dashed curve, left scale) and its Shar e ratio (solid curve, right scal e). The arameters are ha = 1, 2 = 0.1, r f = 0.02, M = 10, K0 = 1, w = 0.3, = 1, hs = 0.1 , and z = 0.1.

mono olies are less vulnerable to roductivity shocks because they can ass these shocks on to their customers. This makes their ro ts less risky. Investors are more con dent in their ro t forecasts (though they trust the ir roductivity forecast just as much) so they trade more aggressively on their ri vate information. Thus, com etition erodes insiders informational advantage. ----------------------- Page 16----------------------16 The Journal of FinanceR

This is a key im lication of the model, from which the next ro ositions will follow. It may seem sur rising that stocks of more com etitive rms are not more intensely traded given that their highly sensitive ayoff offers a more effectiv e avenue for trading on rivate information. The reason is that investor s are risk-averse and these stocks are also more risky.16 Were investors risk-neutral

B. Informational Ef ciency The following ro osition describes how the increase in informed trading for rms with more market ower affects the informativeness of stock rices.

ket ower. The ro osition establishes that market ower increases the amount of information that is revealed through rices. Above we show that investors scale u their trades of more mono olistic stocks because of their reduced risk. Conseuently, their rivate signals are more fully ca italized into rices. Thus, a l ess ef cient roduct market (in the sense that rms enjoy more mono oly ower, that is, face a more ca tive demand for their roduct) leads to a more ef cient stock market (in the sense that stock rices are more informative). Putting it differently, stock market im erfectionsthe extent of information asymmetriesare mitigated by roduct market im erfectionsmarket ower.18 Pro osition 3 is illustrated in the to right anel of Figure 6. C. Dis ersion of Investors Forecasts The following ro osition shows how the increased accuracy of ublic information affects the dis ersion of investors forecasts. The dis ersion of investors 16In our setting, a nite number of stocks with inde endent returns trade on th e stock market. Therefore, risk amounts to the variance of these returns. The variance can be inter reted as a covariance with the market if these stocks are only a subset of available securi ties. For exam le, su ose that the return on the market (which encom asses other ublicly traded s ecurities, rivate assets, human ca ital, etc.) is rmarketz = M (1 m)amz + bz, here b i s a random variable m=1 independent of the am market m m 2 m m . Then cov(r z,r z) = (1 ) var(a z) = var(r z ). 17Formally, stock returns can be expressed as r = ln(/P ) = (1 )az pz , which depends on

PROPOSITION 3:

Stock

rices are more informative when rms enjoy more mar-

and ca ital-constrained, they would rms as in Tookes (2007).17

refer to trade stocks of more com etitive

productivity shocks throu h (1 )a. As 1 rises (less market po er), returns are mo re sensitive to these shocks. Expected returns increase by a factor (1 ) hile their variance increases by a factor (1 )2 . Thus, the ratio of expected excess returns to their variance, h ich determines investors trades, is magni ed by a factor 1/(1 ) > 1. It follo s that ding volume and turnover increase ith market po er . 18Formally, prices provide a signal for technology shocks a ith error so their recision, which measures the informativeness of rices, e uals h 2 2 2 = 1 ) ]. It increa e the dollar tra (1 )/hs 2 h /[ (

when inve tor trade more ( lower or higher) or hen noise traders are less acti ve ( 2 lower).

----------------------- Page 17----------------------Product Market Competition and Stock Market Ef ciency 17 productivity a, pro t , and return r forecasts are measured for any gi ven rm, conditional on the realization of the roductivity and noise shocks a and : var E(a | F ) | a, , var E(ln | F ) | a, , and var . l l

rms enjoy more market ower.

Pro osition 4 establishes that the dis ersion in investors roductivity, ro t, and return forecasts is reduced by market ower. Indeed, investors ass ign a smaller weight to their rivate signals and a greater weight to stock rices when their informativeness im roves. This results in less disagreement among 2 investors. For exam le, the dis ersion of roductivity forecasts e uals hs /h , the ratio of the recision of the rivate signal to the s uared recision of tot al information (h h + h + h ). As market ower strengthens, h rises, induca s ing investors to rely less on their rivate signal. Figure 6 shows the forecast

PROPOSITION less dis ersed when

4: Investors

roductivity,

In particular, price

are perfectly revealing when is close to one.

E(r | F ) | a, l

ro t, and return forecasts are

dis ersion (middle left anel) at different levels of market ower. D. Stock Returns Market ower acts as a hedge that allows rms to ass shocks on to their customers. Hence, ro ts uctuate less when rms enjoy more market ower. So do stock rices, a discounted version of ro ts, and returns, which ca ture the difference between ro ts and rices. Ex ected returns are lower too to com ensate investors for bearing less risk. These effects obtain in im erfect com etition models, regardless of information asymmetries, sim ly because ro ts are less risky. The novel as ect em hasized in this a er is that market ower also exerts an indirect in uence through the informativeness of rices. We focus on these informational effects. The following ro osition summarizes the im act of market ower on the distribution of stock returns. We consider the volatility of stock returns (unconditionally and conditional on stock rices ), the conditional volatility of ro ts, the ex ected excess stock return and the Shar e ratiothe ratio of ex ected excess returns to their standard deviation. PROPOSITION 5: Firms enjoying more market ower have less volatile returns (unconditionally and conditional on ublic information), lower ex ected returns , and higher Shar e ratios. They also have less volatile ro ts. Their return and ro t volatility, ex ected return, and Shar e ratio are reduced further by the im roved informativeness of their stock rice. We know from Pro osition 3 that market ower enhances how much investors can learn from stock rices. Im roved information in turn makes ro ts conditional on rices less variable. Thus, the informational effect of market ower works to dam en ro t volatility. The behavior of returns mirrors that of rofits. They too are less volatile for more mono olistic rms as a result of the direct effect of market ower. Market owers indirect effect through the informativeness of rices decreases return volatility further. This is because, with ----------------------- Page 18----------------------18 The Journal of FinanceR

better information, rices track future ro ts more closely, leaving returns to absorb a smaller fraction of shocks. In the case of erfect information for exam le, rices re ect technology shocks erfectly while returns e ual the risk-free rate so their variance is zero. The informational im act of market ower also reduces ex ected stock returns and the average Shar e ratio, indicating that it is stronger on the risk remium than on risk. Thus, the indirect effect of market ower through the informativeness of rices magni es the decrease in ro t and return volatility, ex ected returns, and the Shar e ratio. The bottom anels of Figure 6 illustrate these ndings. E. Li uidity

We analyze next the im act of market ower on li uidity. We use the sensitivity of stock rices to (uninformative) noise shocks, = (ln P )/(z), to ca ture li uidity as is common in asymmetric information models. PROPOSITION 6: Firms enjoying more market ower have stock rices that are less sensitive to noise shocks. Pro osition 6 extends to noise shocks the intuition we develo ed for roductivity shocks a: Firms use their market ower to shield their ro ts from shocks, whatever their source. Pro ts and therefore rices of more mono olistic rms are less vulnerable to noise shocks. Figure 6 (middle right anel) shows gra hically that decreases with . F. Allocative Ef ciency The interplay bet een imperfections in the product and equity markets has implications that are not only nancial as discussed so far but also r eal. To illustrate this point, e consider rms that raise fresh capital through an equity issuance. Investors not only value these rms, but also determine ho much capital they are to receive. An ef cient allocation of capital requires that investors channel more funds to more productive technologies (i.e., those ith a higher technology shock A ), and less funds to less productive technologies. In this section, e investigate ho competition in uences investors ability to perform this allocation. As before, rms start ith K 0 units of capital and one share outstanding. We assume that rms issue new sh res ( n rbitr ry positive number), the proceeds of which will serve to exp nd their sset b se. We denote K = P the mount of c pit l r ised, where P g in represents the stock price (P nd K re determined endogenously in equilibrium). As before, we st rt with the benchm rk perfect-inform tion economy. We denote prices, pro ts, nd c pit l in this economy with superscript P . Th nks to its exp nded c pit l stock K0 + PP , monopoly gener tes riskless pro t P = (1 )(K0 + PP )1 exp[(1 )az] at t = 2. Therefore, its stock trades at t = 1 for PP = /Rf = (1 )(K0 + PP )1 exp[((1 )a r f )z].

Product Market Competition and Stock Market Ef ciency 19 P P We search for a solution to this equation of the form P = P exp(p z), where we ne lect terms of order lar er than z. It is useful to de ne a rms dilution factor as P /(K0 + P ). This f ctor equ ls zero when no sh res re issue s in the prece ing sections, n one when the rm h s no other c pit l but th t newly 1 raised. The term P is the implicit solution to P (1 + ) = (1 )(K0 + P ) n c n be expresse in terms of K 0 n s

Pa e 19

 P = [K0 Moreover, no shares are Sec P

/(1 )]1 [(1 ) (K /(1 ))]. 0

0 P P = P = P exp(p z

tion II. The mount of c pit l r ise is K ). The sc ling f ctor 1/(1 + ) in the expression for pP that the

accounts for the fact

ne ly issued shares allo rms to expand their asset base: A 1% increase in the amount of capital raised generates a (1 )% increase in pro ts, of hich ne shares claim a fraction . Therefore, 1% incre se in current stock prices re uces investors return by less th n 1%, n mely, by (1 + )%. Formally, the stock return is rP = (1 )(a + pP ) pP (given th t kP = pP ). Since the investment is riskfree, rP = r f n pP follows. The el sticity o f investments to technology shocks, (ln KP )/(ln A), equ ls pP (, ) (1 )/(1 + ), a hich is positive, indicating that more capital o s to better rms. We turn to the analysis of the imperfect-information economy.

p (, ) , P = K0 1 1 f p0(, ) , r (1) (1 ) (1 )2 1 K0 ,

1 +

h()

and h() is de ned in Proposition 1. Firms raise K = P exp(kz) units of c pit l, where k = p .

p (, )

1 + 1

h() h

2 (1 )(1) 0,p (, ) (1 )p (, ), hs

Sh res tr e t p (, ) +

price P = P exp(pz ) such th t p = p

r tion l expect tions equilibrium ch r cterize

s follows. (, ) + 0

1 (1 )K(1) 0

PROPOSITION 7:

Assume that rms issue new sh res. There exists

log line r

issued

( = 0), PP

= (1 )K 1 exp[((1 )a r f )z]

s i

pP = [(1 )a r f ]/(1 + )19.

We

can

check

that

he

The only ifference is th t the coef cients p0 , p , n are 1/(1 + ) to account for the capital base expansion as in

the benchmark perfect-information economy. If no ne shares are issued, then = 0 n the equ tions coinci e with those of Proposition 1. The equilibrium 19 P is uniquely e ne by this equ tion.

coll pses to th t of the benchm rk economy when inform tion is perfect (h = hs = , p0 = r f /(1 + ), pa = (1 )/(1 + ), and p We analyze next how the accuracy of information h affects the ef cie ncy of investments, holding xed the degree of market ower . This ill pro ve useful for understanding the role of market po er. The elasticity of investments to technology shocks, (ln K)/(ln A) = pa(, ), me sures the economys lloc tive ef ciency: A l rger el sticity me ns th t more (less) pro uctive rms ttr ct more (less) c pit l. The following proposition est blishes link betwee n inform tion l n lloc tive ef ciency.

The el sticity of investments to technology shocks p incre ses with the lev el of inform tion, hol ing xed. Hence, better-informed economies distribute capital more ef ciently across rms. In the perfect information limit (h = ), the elasticity collapses to pP = (1 )/(1 + ) as derived above. It falls to a p = (1 )/(1 + ) 1 h /h when inform tion is imperfect. A worsening ( ) of inform tion (lower h) pushes it w y from its v lue un er perfect inform tion (1 h /h further from one). In the limiting c se of no inform tion (h = h ),

We procee to the imp ct of m rket power. Since its intensity in uen ces the inform tiveness of prices (Proposition 3), m rket power ffects the ef ciency of investments. To ssess its imp ct, we nee to neutr lize th e irect effect of m rket power, which c n be i enti e in the perfect inform ti on c se. The el sticity of investments with respect to technology shocks

= 0 so investments re in epen ent from technology shocks.20

PROPOSITION 8: e ccur te.

C pit l is more ef ciently lloc te when inform tion is mor

20

P ge 20

The Journ l of Fin nceR

The pricing se of Proposition 1. p now scaled by

equ tions presente

in Proposition 7

re simil r to tho

= 0).

equ ls pP (1 )/(1 + ) hen information is perfect. We are interested in the a indirect effect of market po er on allocative ef ciency, hich e measure relative to the perfect-information benchmark as pa/pP . We establish the follo ing a result.

Proposition 9 sho s that imperfect competition has an ef ciency impact through the distribution of capital across rms. It combines Proposition 2, hich establishes that the informativeness of stock prices improves as market po er strengthens, ith Proposition 8, hich sho s that information improves the quality of investments. Formally, pa/pP rises ith market po er . Thus, a the social loss of market po er is tempered by improvements in the c apital 20Proposition 8 also makes apparent the informational role of the stock marke t. It can best be understood by comparison to an economy in hich prices do not convey any informa tion. In such an economy, investors total precision (the combined precisions of price and private signals) is reduced to h + h < h and the elasticity of investments to technology shocks t o [1 1/(1 + h /h )](1 s s )/(1 + ) < pa . (The equilibrium in this economy can be derived from the general case by driving the volatility of noise, 2 , to in nity to make tock price uninformative .) The allocation of capital i not a ef cient, though the ame private ignal were ob erved. ----------------------- Page 21----------------------Product Market Competition and Stock Market Ef ciency 21 1 0.9 0.8 0.7 y c 0.6 n e i

PROPOSITION more market po er.

9:

Capital is more ef ciently allocated

hen

rms enjoy

c i f f E 0.5 0.4 0.3 0.2 0.1 0 1 Weak rong Figure 7. The impact of market po er on the economys allocative ef ciency. The p lot displays the ef ciency of investments pa/pP . on the x-axis measures the rms po er in its a product market. The dilution factor is = 0.1, n the other p r meters re h = 1, 2 = 0.1, r f = 0.02, M = 10, K0 = 1, w = 0.3, = 1, hs = 0.1, and z = 0.1. Mkt power () St 0.2 0.4 0.6 0.8

results in an inef ciency. This effect, illustrated in Fi ure 7, results from the interaction of monopoly power in the product market with informational fric tions in the equity market. The proposition implies that the dere ulation of product markets has an additional effect that operates throu h the stock market. Openin produ ct markets reduces the information content of stock prices, which dama es the ef ciency of the capital allocation within these markets. This ndin has im plications for policy desi n. It su ests that product market reforms should not be conducted in isolation but in combination with stock market reforms. Since product market competition can hurt stock markets, policies aimed at improvin nancial ef ciency, such as the liberalization of the nancial sector, should be implemented simultaneously.22 21The social loss stems from the fact that monopoly power induces ce fewer oods rms to produ

and sell them at prices that exceed their mar inal cost. On the other hand, a li terature initiated by Schumpeter (1912) ar ues that competition is detrimental to innovation becaus e it reduces the monopoly rents that reward it. Our ndin s reinforce the Schumpeterian view: Compe tition implies that ood ideas stru le to attract capital, which further weakens the incentives to innovate.

allocation.21 ck thereof,

Puttin

it differently, competition, rather than the la

Empirically, Chun et al. (2008) show that competition boosts the volatility of rm s productivity. 22This recommendation echoes that of interest roup models of nancial development such as Rajan and Zin ales (2003). They su est that incumbent rms oppose nancial devel opment because it breeds competition. They ar ue that dere ulation should take place in both product and nancial markets to overcome the resistance from these roups.

22 IV.

In this section we explore various extensions of the model. First, we allow investors to learn from rms past performance. Next, we discuss the role of noise tradin . Finally, we examine whether our ndin s eneralize to factors other than product market competition such as levera e.

am = am + m, (7) 0 w e e is a positive pa amete and m is an e o te m, t at is independent of am, of all ot e andom va iables, and ac oss ms, and is no mally dist ibuted

cont ol t e pe sistence of s ocks: T e co elation between am nd am equals 2 a a e un elated to past s ocks if o equals ze o.



1/

1 +

/(

), w ic inc eases wit o

. In pa ticula , cu ent s ocks



wit mean ze o and p ecision

/z (va iance z/

). T e pa amete s and

a 0

So far, investors information consists of their private si nals and stock prices. In this section, we allow investors to learn about technolo y shocks from rms past pro ts. We assume that rms operate at t = 0, durin a period that precedes the tradin round (t = 1). The pro t rm m enerates at tha t m m time is m = (1 m)(AmK )1 = (1 m)(K )1 exp[(1 m)amz], 0 0 0 0 0 Am exp(amz) denotes its technology shock at t = 0. 0 0 To connect past and future pro ts, e make t o assumptions about rm ms technology shock and pro t in period 0. First, e assume that technology shocks display some persistence. Therefore, past shocks are informative about future shocks. Speci cally, e assume that

  

 

A. Learnin

from Past Pro ts

Pa e 22

The Journal of FinanceR Discussion and Extensions

here

 

Second, we assume t at p o ts in pe iod 0 a e impe fectly epo ted. T us, t e past p o t p ovides a noisy signal fo (1 m)am, denoted m (all the other 0 0 com onents of the ast ro t are deterministic). S eci cally, rm m re orts m m m m 0 = (1 )a0 + , (8) where m is a error term that is i depe de t of am, of all other ra dom vari0 ables, a d across rms, a d is ormally distributed with mea zero a d precisio h /z (varia ce z/h ). Observi g is e uivalent to observing a signal 0 am + m/(1 m) about am. This signal is less accurate for a rm that enjoys 0 0 m 2 m more market po er (the precision of the signal (1 ) h is lower w he is larger). This is once again because rms use their market po er to insulate pro ts from shocks, thus eakening the link from productivity to pro ts. As before, e consider from no on a generic stock and drop the superscript m to simplify notations. Substituting equation (7) into equation (8) yields 0 = (1 )a + (1 ) + . Thus, observi g 0 is e uivalent to observing a signal about a, namely, a + u were u ( + /(1 ))/. T e p ecision of t is signal is denoted 0 and e uals ----------------------- Page 23----------------------Product Market Com etition and Stock Market Ef ciency 23

2 0 () = 1/h 2 . + 1/[(1 ) h ]

Note that h0 increases with and as past s ocks a e mo e co ela ted to cu ent s ocks, and wit as the reporti g error shri ks, but it decreases with as rms exert their market po er to hedge pro ts. In particular, the signal is uninformative hen = 1 because the pro t at t = 0 is unrelated to the shock at t = 0 and hence to the shock at t = 1. Finally, e note that this extension reverts to the model solved so far hen , , o equals zero. I that case, past pro ts do ot provide a y i formatio about future pro ts. The followi g propositio describes the equilibrium. PROPOSITION 10: There exists a log-li ear ratio al expectatio s equilibri um i which shares trade at a price P = (1 )K 1 exp(pz ), here 0 p = p () + p ()u + p ()a + p (), 0 0 a u a + ( + /(1 ))/,



 

 

 



(1 )2 1 p () 0 () 2 h()

(1 )K 1 0 f

ha

(10) p () (1 )2 h() p () 2 h2 s 1 + 2 hs 2 2 , p0 () (1 )h0 h() 2 2 2 , h0 () = 1/h a d 2 + 1/[(1 ) h ] , ,

(1 )

(12)

(1 )

h() h (13) a

+ h 0

() + h () + h . s

The ricing e uations resemble those in Pro osition 1. They are altered in two ways. First, the rice is now a function of the ast ro t. The corres onding signal error u enters with a weight 0 . Naturally, 0 rises with the recision of the signal h0 , so it decreases when market ower rises. In parti cular, p0 = 0 when = 1 because the past pro t is then uninformative. Second, investors total precision h is larger by the amount h0 , the recision of the new signal (com aring e uation (13) to e uation (4)). This extra term makes h nonmonotonic in market ower . Hence, t o opposing forces are at or k as market po er strengthens. On the one hand, more market po er implies a more informative stock price (hp is higher, as in Proposition 3). On t he other hand, it means a less informative past pro t (h0 is lower). This trade-off is illustrated in the to left anel of Figure 8, which shows that market ower is not unambiguously bene cial to stock market ef ciency. Informational and ----------------------- Page 24----------------------24 .2 0.4 10 8 80 s s e e m 0.6 0.8 1 10 100 8 The Journal of FinanceR 0 0

p () (1 ) a

n e v i t a m r o f n I

6 60

u l o v g n i d 40 a r T

2 20 0 .2 0 0.4 0.2 0.6 0 0.4 0.6 0.8 1 Mkt ower () 0.4 0.8 1

2 0 0 0

Mkt po er () 0 0.014 0.012 n o i s r e p s i d s t 0.006 s a c e r o F

0.2

0.01 2 y c 0.008 1.5 n e i c i f f E 0.8

0.004 0.7 0.002 0 0.5 -3 0 0.4 0.2 0.6 0.4 0.6 0.8 1 Mkt po er () 0.4 0.8 x 10 0 1 0.6 0 0

.2

Mkt po er () 0

0.2

0.6

0.8

1 3 2.5 0.9 1

0.6

0.8

0.8

0.1 0.08

5 4.5

0.6 4 y t i l i t 0.4 a l o V 0.2 0.02 2.5 0 .2 0 0.4 Weak Mkt po er () 0.2 0.6 0.4 0.6 0.8 1 Mkt po er () Strong 0.8 Strong 1 0 2 0 Weak 0 0.06 y t i d i u q i L 3.5

0.04

Figure 8. The impact of market po er on the equilibrium hen a rms past pro t i s informative about its technology shock. on the x-axis measures the rms po er in its product market. The top left panel displays the total precision of public signal s (solid line) and its breakup bet een the information revealed by stock prices (hp , dashed line) and that revealed by rms past performance (h0 , dash-dotted line). The to right anel dis lays the shar e turnover (solid curve, right scale) and the dollar trading volume (dashed curve, left sca le). The middle left anel dis lays the dis ersion of investors forecasts about the rms ro t (solid line, right scale) and technology shock (dashed line, left scale). The middle right anel dis lays allocative ef ciency P a/ a . The bottom left anel dis lays the variance of log ro ts (sol id curve, right scale), the variance of log rices (dashed curve, left scale), and the variance of stock ret urns (dotted curve, right scale). The bottom right anel dis lays li uidity 1/ . The arameters are ha = 1, 2 = 0.1, r f = 0.02, M = 10, K0 = 1, w = 0.3, = 1, hs 8, and z = 0.1. = 0.1, h = 0.1, h = 0.01, = 0.

allocative ef ciency a e u t by ma ket powe for lo values of because much information from the past pro t is lost. They improve for high values of because much information from the stock price is gained. Similarly, volatility and liquidity are nonmonotonous functions of market po er. The exception is

 

----------------------- Page 25----------------------Product Market Competition and Stock Market Ef ciency 25 trading volume, hich continues to gro ith . This is because the past pro t is a public signal so it does not generate differences in opinions a nd hence trading.23 B. Noise Trading Our analysis of market po er assumes that it cannot in uence the intensity of noise trading, that is, that 2 i not a function of . One may ask ho our ndings are affected by this assum tion. It is not clear a riori how noise trading would change with market ower. On the one hand, if noise stems from the trades of rational agents subject to li uidity needs, then it may strengthen in rms with more market ower as these agents build u a recautionar y osition in less volatile stocks. On the other hand, if noise is generated by ri skneutral ca ital-constrained investors whose rivate signals contain systematic errors, then it may be larger in stocks with less market ower as these agents trade more aggressively stocks that are more sensitive to their signal. If we su ose that noise trading is more intense among rms with mo re market ower (e.g., noise originates in li uidity shocks), then s eculative trad es in these stocks are more easily concealed, which encourages informed trading. In this case, our ndings on trading activity are strengthened: The vol ume of informed trading and total volume increase even more than when 2 i independent of . But the impact of market po er on informativeness is no ambiguous since both informed and noise trading are more intense. It follo s that the impacts on dispersion of forecasts and allocative ef ciency are also ambiguous. If e suppose instead that noise trading is less intense among rms ith more market po er (e.g., noise originates in correlated signal errors), the n the results are reversed: Informed and total trading gro ith less than hen 2 i independent of and may even be reduced in more monopolistic rms. The net effects on informational and allocative ef ciency are again ambiguous. C. Leverage Product market com etition in uences investors trading behavior because it affects the robability distribution of the cash ows rms generate. M ore com etitive rms offer ayoffs that are more sensitive to shocks and therefore riskier (their variance is larger), so their stock is less actively t raded. We ex ect the same argument to carry over to rms with higher o erational or

nancial leverage. It is well known that rms for which o erating costs are redominantly xed (e.g., rms with large R&D ex enditures) or those that nance themselves mostly with debt offer ayoffs, are more sensitive to shocks and therefore more volatile. Investors will therefore be less rone to trade the ir stock. 23Indeed, investors ortfolio shares do not de end on u, the error in the ast ro t signal. They

are still given by e uation (5). ----------------------- Page 26----------------------26 The Journal of FinanceR

The argument a lied to nancial leverage has interesting im lications for security design. Asymmetric information is generally an im ortant concern for issuers. As Boot and Thakor (1993) show, rms nd it o timal to s lit claims to their cash ows into informationally insensitive a d i formatio ally se sitive claims, such as debt a d equity. This partitio stimulates i formed tradi g (i the i formatio ally se sitive security) a d the collectio of costly private

i vestors i Boot a d Thakor (1993) favor the more i formatio ally se sitive securities, they shy away from them i our framework. This is because traders i Boot a d Thakor (1993) are risk- eutral a d capital-co strai ed, whereas here they are risk-averse a d free to borrow. I Boot a d Thakor (1993), splitti g claims avoids the eed for traders to tie their limited fu ds to securities with k ow payoffs, from which they have little to gai , a d allows them to co ce trate i stead o assets with the greatest i formatio asymmetries. Putti g it differe tly, they trade the most i formatio ally se sitive securities because they are the riskiest, while they avoid tradi g them i our model precisely because they are the riskiest. These co trasti g results illustrate the importa ce

V. Empirical Evide ce I this sectio , we test whether some of the models predictio s are supported empirically. We describe i tur the sample formatio , the methodology a d variable co structio , a d the results. A. Sample Our sample co sists of over 5,000 U.S. rms followed over a decade. It starts from all NYSE-, Amex-, a d NASDAQ-listed securities that are co tai ed i the CRSP-Compustat Merged database for the period 1996 to 2005. We retai stocks with share codes 10 or 11, remove a cial compa ies a d regulated i dustries, a d wi sorize variables at the 1% level. The resulti g sample co tai s 28,172 rm-year observatio s a d 5,497 differe t rms with a average of 5 years of data for each rm. We obtai corporate data a d ear i gs a ou ce-

of i vestors attitude toward risk a d a ci g co strai ts to the desig of securities. I the ext sectio , we co fro t the model with data.

 

 



 

  

  

 

 

 

i formatio .24 Our

di gs differ from those of Boot a d Thakor (1993). While

 

 

 

 

me t dates from Compustat, daily stock retur s from CRSP, a d i sider trades from Thomso Fi a cial. B. Methodology a d Variables We co duct a test of Propositio s 2 a d 3, which predict that rms operati g i more competitive i dustries have a lower volume of trade a d less 24Our paper treats the precisio of traders i formatio as exoge ous, u like B oot a d Thakor (1993). Nevertheless, i vestors prope sity to trade is suggestive of how valuable private i formatio is. ----------------------- Page 27----------------------Product Market Competitio a d Stock Market Ef cie cy 27 i formative stock prices. For that purpose, we eed proxies for market power, tradi g volume a d stock price i formative ess. We describe them i tu r . Table I prese ts descriptive statistics. Market power: We proxy for a rms market power usi g its pricecost margi or Ler er i dex, de ed as the rms operati g pro t margi (sales mi us costs divided by sales).25 Followi g Gaspar a d Massa (2005 ), we subtract the i dustry average pricecost margi to co trol for structural differe ces across i dustries u related to the degree of competitio .26 T he resulti g excess pricecost margi (the differe ce betwee a rms operati g pro t margi a d the average of its i dustry) captures a rms ability to price goods above margi al cost, adjusti g for i dustry-speci c factors u related to market power. A larger pricecost margi i dicates stro ger market power (weaker competitio ). Tradi g volume: To measure tradi g volume, we use a stocks tur over , de ed as the log of the ratio of the umber of shares traded d uri g a year to the umber of shares outsta di g. We also exami e the tr ades i itiated by i siders to capture i formed tradi g. The Thomso Fi a cial I sider Fili g database compiles all i sider activity reported to the SEC. Corporate i siders i clude those that have access to o -public, material, i sider i formatio a d are required to le SEC form 3, 4, a d 5 whe they trade i their compa ies stock. We follow most studies (e.g., Seyhu 1986, Lako ishok a d Lee (2001), Be eish a d Vargus (2002)) by limiti g i sider trades to ope market tra sactio s i itiated by the top ve executi ves (CEO, CFO, COO, Preside t, a d Chairma of Board), as they are more likely to possess private i formatio , a d by excludi g sells, because they

are more likely to be drive by hedgi g rather tha i formatio

motives

 

 

 

 

  

 

 

 

 

 

  

 

 

 

   

 

    

 

 

adi g

activity i two ways: rst, as the log of the ratio of a rms a ual total i sider tradi g dollar volume to the rms market capitalizatio , de oted I sider tur over; a d seco d, as the ratio of the log of the rms a ual umber of i sider trades to the umber of its active i siders, d e oted Number of i sider trades. Active i siders are de ed as executives who

a d most of the empirical i dustrial orga izatio literature. Alter ative proxie s based o asset or sales co ce tratio such as the Her dahlHirschma i dex are i dustry- rather th a rmspeci c. Moreover, because data are limited to U.S. public rms, they do ot accou t for private or foreig rms. This is especially problematic give that our sample (1996 to 2005) covers a period of i te se global competitio . 26Speci cally, the pricecost margi (PCM) is de ed as operati g pro ts (before depr eciatio , i terest, special items, a d taxes) over sales (Compustat a ual data item 12). Operati g pro ts are obtai ed by subtracti g from sales the cost of goods sold (item 41) a d ge eral a d admi istrative expe ses (item 178). If data are missi g, we use operati g i come (item 178). Th e excess pricecost margi is co structed as the differe ce betwee the rms PCM a d the PCM of its i d ustry. The i dustry PCM is the value-weighted average PCM across rms i the i dustry where t he weights are based o market share (sales over total i dustry sales) a d i dustries are d e ed usi g twodigit SIC classi catio s. ----------------------- Page 28-----------------------

Table I Descriptive Statistics This table prese ts summary statistics for the variables used i the empirical s tudy. The sample starts from all NYSE-, Amex-, a d NASDAQ-listed securities that are co tai ed i the CRSPCompustat Merged database for the period 1996 to 2005. We retai stocks with share codes 10 or 11, remove a cial compa ies a d regulated i dustries, a d wi sorize variables at the 1% level. Market power is measured as the excess pricecost margi (PCM) or Ler er i dex. The PCM is de ed as operati g pro ts (before depreciatio , i terest, special items, a d tax es) over sales (Compustat a ual data item 12). Operati g pro ts are obtai ed by subtracti g from sales the cost of goods sold (item 41) a d ge eral a d admi istrative expe ses (item 178). If d ata are missi g,

 

28

The Jour al of Fi a ceR



 

 

25The pricecost margi a d Massa (2005),

is used i

Li de berg a d Ross (1981), Gaspar





 

 

 

  

(e.g., whe

optio s vesti g periods expire). We measure i sider tr

 

 



  

 

  

 

     

we use operati g i come (item 178). The excess pricecost margi is co structed as the differe ce betwee the rms PCM a d the PCM of its i dustry. The i dustry PCM is the value-weighted average PCM across rms i the i dustry, where the weights are based o market sha re (sales over total i dustry sales) a d i dustries are de ed usi g two-digit SIC classi catio s. Size is measured as the log of rms assets. Illiquidity is measured usi g the Amihud (2002 ) illiquidity ratio a d equals the ratio of a stocks absolute retur to its dollar tradi g volu me i a day, averaged over all days i a year, a d scaled by 106 . Retur o assets is de ed as i come before extraordi ary items (item 18) over total assets. Leverage is computed as total lo g-term debt (item 9) divided by total assets (item 6). Market-to-book is the ratio of the market value of equ ity (year-e d stock price times the umber of shares outsta di g) to its book value. Book equity is co structed as stockholders equity (item 216, or 60 + 130, or 6-181, i that order) plus bala c e sheet deferred taxes a d i vestme t tax credit (item 35) mi us the book value of preferred stoc k (item 56, or 10, or 130, i that order). Tur over is de ed as the log of the ratio of the umber o f shares traded duri g a year to the umber of shares outsta di g. I sider trades are ope marke t tra sactio s, excludi g sells, i itiated by the top ve executives of a rm (CEO, CFO, COO, Preside t, a d Chairma of Board). I sider tradi g activity is measured i two ways: rst, as the log of the ratio of a rms a ual total i sider tradi g dollar volume to the rms market capitalizatio , de oted I sider tur over; seco d, as the ratio of the log of the rms a ual umber of i si der trades to the umber of its active i siders, de oted Number of i sider trades. Active i si ders are de ed as executives who have reported at least o e tra sactio i a y of the sample ye ars. Stock price i formative ess is measured as (the i verse of) the absolute ab ormal retur surrou di g a ear i gs a ou ceme t. Ab ormal retur s are measured as the residuals from the F amaFre ch three-factor model, obtai ed by regressi g for every rm stock retur s o the mark et, size, a d book-to-market factors over a estimatio wi dow exte di g from t = 250 to t = 5 elative to t e ea nings announcement day 0. We estimate t e esiduals ove an event wind ow anging f om t = 2 to t = +2. T en, we sum t e absolute value of abno mal etu ns on eac day of t e event window. Finally, we ave age t e absolute abno mal etu ns estimates obtain ed f om eac announcement du ing a yea to get an annual measu e. We also epo t t e ave age absolute aw etu n ove t e event window. Mean Median Std. Dev. Min.

  

 



 

  

 

 

 

 

 

     

 

 

  



 

 

  

   

  

Max.

N 4 0.115 5.516 0.053 0.045 1.923 0.143 0.076 0.001 0.147 3.321 0.119 1.891 44.737 0.374 555 0.206 1.006 0.025 0.337 43 90,022 0.00 6.529 0.00 0.00 1 0.107 0.50 0.00

Numbe of yea s 5.125 10 5,497 pe m Ma ket powe 0.142 3.614 26,264 Size 5.607 1 11.743 26,946 Illiquidity 4.792 0 3 194 28,103 Retu n on assets 0.029 2.188 26,972 Ma ket to book 0.380 6,365 26,495 Leve age 0.191 0 3.862 26,816 Tu nove 0.127 10.313 26,405 Inside tu nove 0.007 0 1.051 26,795 Numbe of inside 0.246 0 4.868 24,676 t ades

P oduct Ma ket Competition and Stock Ma ket Ef ciency 29 Table IContinued Mean Max. N 0.151 0.135 0.078 0.000 Abs. 5 day aw 0.845 24,827 etu n a ound ea nings announcements Abs. 5 day abn. 0.855 24,827 etu n elative to t e t ee facto model a ound ea nings announcements Median Std. Dev. Min.

0.149

ave epo ted at least one t ansaction in any of t e sample yea s (Ke, Hudda t, and Pet oni (2003)). Ea nings announcements: We use t e stock p ice eaction to ea nings an nouncements to assess t e info mativeness of stock p ices. La ge (small) p ice c anges a e indicative of emotely (closely) followed ms, as s own by nume ous studies sta ting wit Beave (1968). We measu e t e abso lute abno mal etu n ove a 5 day window cente ed on days ea nings a e

    

0.133

 

  

  

     

       

   

(continued) Page 29

0.078

0.009

ment, and take t e ave age ove all announcements in a yea to obtain an annual measu e (consistent wit ou p oxy fo ma ket powe and ot e cont ol va iables). We examine t e impact of ma ket powe in sepa ate panel eg essions fo tu nove , inside t ading, and stock p ice info mativeness. We co ect standa d e o s fo se ial and c oss sectional co elation using yea and m cluste s.28

27Speci cally, absolute abno mal etu ns su ounding an ea nings announcement a e de ned as +2 m t=2 ut , w e e t = 2, 1, 0, +1, and +2 count t ading days elative to t e announ cement day 0 fo m m, um = Rm (m + m MKT + m SMB + m HML ), Rm is the return on m ms t t 0 mkt t SMB t HML t t stock on y t, n MKT , SMB , n HML re respectively the returns on the m rket, size n t t t book to m rket f ctors on y t. The coef cients m, m , m , n m re estim t e for every 0 mkt SMB HML rm over win ow r nging from t = 250 to t = 5. 28We correct st n r errors using the proce ure outline in Thompson (2006) n C meron,

s tot l long term ebt (item 9) ivi e by ssets. The m rket to book r tio is e ne s the r tio of the m rket v lue of equity (ye r en stock price times the number of sh res o utst n ing) to its book v lue. Book equity is constructe s stockhol ers equity (item 216, or 60 + 130, or 6 181, in th t or er) plus b l nce sheet eferre t xes n investment t x c re it (item 35) minus the

T ble II M rket Power n Turnover This t ble presents results of nnu l p nel regressions of turnover on m rket power n other rm ch r cteristics over the 1996 to 2005 perio . Turnover is e ne s the log of the r tio of

30

P ge 30

The Journ l of Fin nceR

Gelb ch, n Miller (2006)). 29Firm size is me sure s the log of r ge is compute

rms tot l ssets (Compust t item 6). Leve

Ou eg essions include cont ols fo seve al facto s, suc as m size, equity ma ket to book atio, liquidity, p o tability, and leve age, t at may be associ ated wit t ading activity o wit eactions to announcements.29

 

 

 



 



  

 

    

ounce

t e event window to measu e t e stock p ice

eaction to t e ann

announced. Abno mal etu ns a e de ned elative to t e FamaF enc (1993) t ee facto model.27 We sum t ei absolute value on eac day of

 

  

Turnover M rket power 117 1.340 0.896 1.458 4.78 0.537 4.17 0.420 3.54 0.364 2.93 0.582 4.67 2. 4.55 0.270 3. 0.125 0.123 0.151 4.40 Illiqui ity 0.009 0.009 Return on 0.154 0.047 ssets 1.70 0.290 M rket to book 3.51 0.010 6.74 0.166 1.69 3.57 0.009 6.77 0.502 3.32 5.396 2.768 2.99 1.610 Lever ge 0.941 0.946 12.10 12.10 Const nt 0.840 Observ tions 791 25,732 R2 062 0.092 C. Results 0.254 4.70 25,798 25,389 0.011 0.121 0.899 5.36 25,791 0.059 0.724 4.40 25,732 0.091 0.737 4.48 25,462 0.093 0.725 4.35 25,389 0.120 1.105 5.50 25, 0. 2.11 4.54 0.009 6.81 0.006 0.030 3.511 4.84 0.101 0.103 0.130 0.

M rket power 0.156 0.148 Size 99 2.55 2.30 Size 162

As heir

prelimin ry, we sort

rms every ye r into

ve groups b se

respectively, for the two t ile See T ble I for the v ri ble e nitions.

hypothesis test th t the coef cient equ ls zero.

3.24

3.70

6.64

0.847

4.40

on t

the number of sh res tr e uring ye r to the number of sh res outst n ing. M rket power is me sure s the excess pricecost m rgin or Lerner in ex. The bsolute v lues o f t st tistics re ispl ye below the coef cient estim tes. They re b se on st n r errors cl ustere both by rm n ye r. The symbols , , n enote signi c nce t the 1%, 5%, 10% levels,

m rket power n me sure within e ch group, the ver ge turnover, insi er ctivity, n price re ction to nnouncements. The results re presente in Figures 1 to 3. The gures reve l th t tr ing ctivity, inclu ing th t initi te by insi ers, is higher for rms with more m rket power n th t the price of such rms re cts less to nnouncements. The p nel regressions, ispl ye in T bles II to IV, con rm the visu l impression of the gures fter controlling for other f ctors. The coef cient on m rket power is st tistic lly signi c nt cross ll speci c tions with the s me sign (positive for the turnover n ins i er

T ble III M rket Power n Insi er Tr ing This t ble presents results of nnu l p nel regressions of insi er tr ing on m rket power n other rm ch r cteristics over the 1996 to 2005 perio . In P nel A, in si er tr ing ctivity is me sure s the log of the r tio of rms nnu l tot l insi er tr ing oll r volu me to the rms m rket c pit liz tion, n it is enote Insi er turnover. In P nel B, it is me sure s the log of the r tio of the rms nnu l number of insi er tr es to the rms number of ctive ins i ers n is enote Number of insi er tr es. M rket power is me sure s the excess pric ecost m rgin or Lerner in ex. The bsolute v lues of t st tistics re ispl ye below the coef cie nt estim tes. They re b se on st n r errors clustere both by rm n ye r. The symbols , , n enote signi c nce t the 1%, 5%, n 10% levels respectively, for the two t ile hypothe sis test th t the coef cient equ ls zero. See T ble I for the v ri ble e nitions. P nel A: Insi er Turnover M rket power .020 0.018 0.004 0.014 2.37 0.006 3.69 0.005 3.58

0.003 1.89

0.003 1.75

0 3.92

Pro uct M rket Competition n 31

P ge 31

Stock M rket Ef ciency

price imp ct of tr es, n microstructure t (see Amihu (2002)).

is correl te with illiqui ity proxies obt ine

from

book v lue of preferre stock (item 56, or 10, or 130, in th t or er). Pro t bilit y is me sure s the return on ssets n is e ne s income before extr or in ry items over tot l ssets. Fin lly, we proxy for the l ck of liqui ity using Amihu s (2002) illiqui ity r tio, e ne s the r tio of stocks bsolute ily return to its ily tr ing volume, ver ge over ll ys in ye r n sc le by 106 . It c ptures the bsolute percent ge price ch nge per oll r of tr ing volume, th t is, the

4.

M rket power 0.002 0.002 size .39 3.55 3.18 Size 0.000 Illiqui ity 0.130 Return on ssets 0.004 1.450 6.04 0.130 0.121 8.12 7.68 Lever ge 0.000 0.000 0.050 Const nt .008 0.008 0.120 0.006 0.008 12.75 26,040 25,597 0.000 0.005 0.010 14.77 26,031 0.002 0.010 14.13 25,970 0.003 0.010 13.79 25,676 0.004 0.009 13.87 25,597 0.004 8.12 0.130 0.000 7.21 6.94 0.000 0.000 0.000 0.250 0.050 0.004 0.011 1.450 M rket to book 6.42 6.29 0.012 0.012 0.060 0.060 6.74 0.000 5.50 0.000

0.002 3 0.001 0.001 0.001 0.001 0.000 2.27

Observ tions 6,031 25,970 R2 .003 0.003

M rket power 0.484 0.391

0.230 0.327 6.41

0.228 6.52

0.204 5.67

0.160 5.35

M rket power 0.031 0.024 size .25 2.53 1.92 Size 07 2.21 Illiqui ity 0.001 51 Return on ssets 0.078 0.080 0.232 1.430 4.24 3.34 0.299 0.239

0.001 0.003 1.150 0.006 0.250 2.06

0.001 0.460 0.001

0.000 0.050 0.001 5.66

0.001 5.47 5.76

2.

0 8.29 2 0

8.

P nel B: Number of Insi er Tr es 0.184 5.93 4.85 0.043 3 0.003 1.100 0.001 0.0

4.

5.

1.430

3.27 3.987 3.651 3.86 3.43 3.52 0.107 0.108 4.83 4.83 3.761

M rket to book

Lever ge

Const nt .177

Observ tions 3,080 23,037 R2 .007 0.010

ness This t ble presents results of nnu l p nel regressions of stock price inform ti veness on m rket power n other rm ch r cteristics over the 1996 to 2005 perio . Inform tiveness is me sure s (the inverse of) the ver ge bso lute bnorm l return surroun ing n e rnings nnouncement (from t = 2 to t = +2). Abnorm l returns re the resi u ls from the F m French three f c tor mo el. M rket power is me sure s the excess price cost m rgin or Lerner in ex. The bsolute v lues of t st tistics re ispl ye below the coef cient estim tes. They re b se on st n r errors clustere both by rm n ye r. The symbols n enote signi c nce t the 1% n 10% levels re spectively, for the two t ile hypothesis test th t the coef cient equ ls zero. See T ble I for the v ri ble e nitions. veness T M rket power 0.190 0.096 0.044 0.030 0.120 0.083 0.112 8.50 4.48 7.51 o M rket power size 0.025 4.92 5.93 0.013 e 1.81 h 0.013 J 1.65 4.26 0.032

Stock Price Inform ti

0.015

0.012

0.013

0.198

0.215 0.197 13.65 23,091 22,674 0.007 0.018 P ge 32

0.212 12.57 23,080 0.007

0.223 13.60 23,037 0.010

0.214 12.92 22,743 0.015

0.216 13.17 22,674 0.018

0 9.01 2 0

10.

3 2 T ble IV M rket Power n Stock Price Inform tive

r n 6.15 3.55 3.20 4.75

l Size 0.021 0.018 0.018 17.06 5.50 Illiqui ity 0.001 34.61 f 0.001 0.001 0.001 0.001 0.001 0.001 F i n 6.14 6.27 6.05 5.05 6.00 5.94 5.07 0.017 0.016 0.022 16.34 0.016 o 18.27 18.06 35.35 0.016 0.020

Return on ssets 99

0.033

0.104

0.034 0.095 c

0.108

n 0.107

e 1.320 10.30 M rket to book 0.275 1.310 95.6 9.69 0.365 0.344 2.91 0.368 R 3.17 0.003 0.003 0.480 0.031 0.410 5.97 Turnover 0.026 12.69 Const nt 0.260 46.44 0.269 0.160 0.250 32.07 0.250 0.250 47.36 0.238 56.45 5.99 0.026 12.73 0.241 0.270 58.01 0.241 56.08 54.57 2.77 11.59 10.09

0.427 3.46 3.36

Lever ge 0.031

0.0

Observ tions 22,718 R2 0.375

23,417

23,432 23,417 23,415 23,113 23,415 23,040 22,718 0.023 0.198 0.237 0.255 0.202 0.239 0.256 0.376

23,040 0.255

tr ing regressions n neg tive for the e rnings nnouncements regressions). These n ings re consistent with the pre ictions of Propositions 2 n 3.30

ket power by one st n r evi tion incre ses turnover, insi er turnover, n the number of insi er tr es respectively by 5% to 10% of st n r evi tion, 1% to 3% of st n r evi tion, n 6% to 8% of st n r evi tion. Stock price inform tiveness eclines by 2% to 14% of st n r evi tion. These m gnitu es re not so surprising given th t the effects re me sure over the entire s mple of rms. Their signi c nce is likely to v ry cross rms epen ing on the extent of inform tion symmetries n noise tr ing. In p rticul r, one woul suspect the imp ct of m rket power to be stronger mong rms with more severe inform tion symmetries, such s sm ller rms; rms liste on NASDAQ; n rms with fewer n lysts, more in ivi u l sh rehol ers, l rge blockhol ers, n more R&D expen itures.31

neg tive in the positive in the

re uces tr ing, inclu ing th t by insi ers, n the inform tiveness of stock prices. These n ings re consistent with our iscussion of lever ge in Section IV.C, where we rgue th t lever ge m gni es risk in the s me w y th t compe tition oes, iscour ging tr ing n limiting the incorpor tion of inform tion into prices. VI. Conclusion

We present mo el th t links investors tr ing beh vior to the egree of pro uct m rket competition. Ours is st n r r tion l expect tions mo el of tr ing un er symmetric inform tion in competitive stock m rkets, but for one ifference: Firms enjoy monopoly power in their pro uct m rket. Pro uc tion is subject to r n om pro uctivity shocks bout which investors re ceive

symmetries over ll tr power mong The coef neg tive in

re less perv sive mong l rger rms, then informe tr ing, ing, n stock price inform tiveness re less sensitive to m rket these rms. cient estim tes on lever ge eserve some comment. Their sign is T bles II n III n positive in T ble IV, in ic ting th t lever ge

inform tiveness regression (T ble IV), suggesting th t the imp ct of m rket power shrinks with size. This is consistent with the mo el to the extent th t size is inversely rel te to the ccur cy of priv te inform tion. If inform tion

For ex mple, n inter cte term, Mkt Power Size, is inclu the regressions to ex mine how the coef cient on m rket th rm size. The sign of the correspon ing estim te coef cient is turnover n insi er tr ing regressions (T bles II n III) n

in some of power v ries wi

The economic m gnitu e of these effects is rel tively mo est. Incre sing m r

Pro uct M rket Competition n 33

P ge 33

Stock M rket Ef ciency

n lyses v il ble in the Internet Appen ix in the Supplements n D t s ets sectio at http://www.afajof.org/suppleme ts.asp. 31I deed, the model implies that market power i teracts with the precisio of i vestors private sig als i the tradi g volume a d i formative ess equatio s. Tradi g vo lume is a fu ctio of h /(1 )2 (see the Appendix), and informativeness h is a function of h /(1 ) (e quation (4)). s p s Thus, the in uence of market po er is stronger hen the precision of private signa ls hs is larger. ----------------------- Page 34----------------------34 The Journal of FinanceR

private signals. The driving force of the model is that monopolies are able to pass shocks on to customers and insulate their pro ts. We establish the follo ing results about rms that enjoy more market po er. (i) Their stock trading volume is larger. As a result, (ii) the incorporation of private information into prices is expedited. Several implications follo : (iii) in vestors productivity and earnings forecasts are less dispersed, (iv) stock liquid ity is enhanced, (v) volatility of pro ts and stock returns is dampened, and (vi) expected returns are lo er, even after adjusting for risk. Moreover, (vii) hen rms issue ne shares, capital is more ef ciently deployed across more mo nopolistic rms. Thus, product market imperfections (monopoly po er), rather than spreading to equity markets, tend to mitigate stock market imperfections (informational and allocative inef ciencies). These ndings are consistent ith existing documented facts and e present further supportive evidence. In particular, e report that trading volume, including trades initiated by insiders, and the information content of stock prices are higher for rms ith m ore market po er. Our results are of importance to policy makers and nancial economist s. They indicate that product market deregulation has implications that extend to equity markets. Therefore, these reforms should not be conducted in isolation but in combination ith reforms designed to improve the ef ciency of the nancial sector. They also shed light on some trends that have been observed in the United States. Idiosyncratic return volatility increased in the post- ar period (Morck et al. (2000), Campbell et al. (2001) and, Comin and Philippon (2006)) as competition intensi ed thanks to deregulation and globalization . Our model suggests that competition orsened the informativeness of sto ck prices, hich also contributed to the volatility increase. In our attempt to link industrial organization to the informational propertie s of stocks in a rational expectations frame ork, e omitted several points for simplicity. First, the structure of product markets is taken as given, hen in fact it is endogenous. If more productive rms raise more capital in a

30We con rm supplement ry

th t our results

re robust to

number

of ch nges

in

 

more ef cient equity market, they ill be disproportionately large and enjoy more market po er. This calls for a model in hich the degree of competition and the properties of stock prices are jointly determined in equilibrium. Second, th e precision of investors information is exogenous to the model. In practice, they may adjust their research effort to the stocks riskiness. The effect of competition on signal precision is unclear. On the one hand, information about more competitive stocks is less useful if they are traded less. On the ot her hand, increased competition exposes stockholders to more risk, making information more useful. Finally, the number of rms in the market is xed. Endogenizing the listing decision ould shed light on the joint impact of the informational and competitive environments on rms incentives to go public. Recent empirical ork suggests that rms operating in industries characterized by mo re competition and more information asymmetry are less likely to do an I PO (Chemmanur, He, and Nandy (2006)). These questions are left for future research. ----------------------- Page 35----------------------Product Market Competition and Stock Market Ef ciency 35 Appendix: Proofs

Proof of Proposition 1 (Stock prices): The proof of Proposition 1 b uilds on Peress (2004). We guess that equilibrium prices are given by equations (2) to (4) and solve for an investors optimal portfolio by driving z to ard zero. The rst step is to relate stock returns to technology shocks. Stock returns For a given stock of capital K 0, intermediate goods pric es are determined by the market clearing condition, AmK0 = ((1 m)/Qm)1/m . T he resulting monopoly pro ts equal m = YmQm = (1 m)(AmK )1m . Since there 0 is one share outstanding, the gross stock return is Rm = m/Pm . Writing m P = P exp(p z) implies that Rm m 1m m = (1 )K0 /P exp[((1 m When z = 0 (no risk), R riskless m m 1m m so P = (1 m m m m )a p )z]. m m

1m m f )K0 /P and R = 1. Stocks are m

= (1 )K0 . Thus, the log return on stock m is r z = ln(R ) = (1 m)amz pmz. T e second step is to estimate t e mean and va iance of stock

m etu ns using t e equilib ium p ices and p ivate signals sl Signal ext action We guess t at p ices a e app oximately no mally dist ibuted and given in equation (2), t at is, pmz = p z + p mz + o(z), where m am + mm, m is a 0 a constant to be determined and o(z) captures terms of order larger than z. The conditional mean and variance of amz for agent l are .

m var(a z l where 1 hm h . 0 s a m2 2 + z | F ) =

z and hm

m E(a z | F ) = a l

, hm hm + h , a hm 0 s m2 2

E(amz | F ) i a weighted average l the weight on the privat 1/(m2 2)). The conditional mean

and va ( mz | F ) = va ((1 m)amz | F ). l ----------------------- Page 36----------------------36 The Journal of FinanceR We next turn to the investors portfolio choice Individual portfolio choice c1 1 A ent l, endowed with wealth w, forms her portfolio to maximize E[ l | F ]

 

variance of tock exce

return follow mz | F = E((1 m)amz | F ) pmz E l l

The variance var(amz | F ) fall a the preci ion gl nal , peci cally h and 1/(m2 2), increa e. of prior , public and private ignal , where e ignal (the public ignal) i increa ing in h (in and

of the private and public

 

 

m m + a

m m s s l

1 = hm, p and a hm h s

1 l subject to c = w exp(r z), where r z = ln[Rf + M f m(Rm Rf )] is investor ls l l l m=1 l lo portfolio return. Note that rlz is approximately normal when z is small (e. ., Campbell and Viceira (2002)). Therefore, 1 = E w1 E c l l = w1 exp (1 )E(r z | F ) l + (1 )2 var(r z | F )/2) (1 ) + o(z), l where M E(r z | F ) = f m(E(rmz | F ) r f z) + f m(1 f m) var(rmz | F )/2 + o(z) l l m=1 and M var(r z | F ) = f m2var rmz | F + o(z). l l m=1 Maximizin E[(c1 n 1)/(1 ) | F ] with respect to f m leads to the fractio l l l l l l l l l l l 1 / 1 /(1 ) | F l l exp((1 )r z) 1 /(1 ) | F l

l l of wealth allocated to stock m (at the order 0 in z): m f m /2 E(r z | F ) r z + var r z | F l l m ). fl (A1) = var rmz | Fl

+ o(1

l f m 1 hssl m 1 + m2 2 m hm (1 m) p + r f + (1 m) 2

l = (1 m) (A2)

Substitutin

the above expressions for E(rmz | F ) and var(rmz | F ) yields l

+ o(1

Market clearing We multi ly e uation (5) by investors income w and sum over all investors to obtain investors aggregate demand for stock m (at the order 0 in z): 1 f mwdl = w mh + 1 h mdl + m 2 2 0 (A3)

(1

----------------------- Page 37----------------------Product Market Competition and Stock Market Ef ciency 37 since 1 hm dl = hm and 1 h am dl = amh . Applying the la of large num bers 0 0 s s to the sequence {hsm} of ind p nd nt random variabl s with th sam m an l z ro l ads to 1 h mdl = 0 (s H and Wang (1995) for mor d tails). Finally, 0 s l th mark t cl aring condition for stock m is ( 1 flmdl + m)w/Pm = 1. The lefthand side is the total demand for stock m, rs and noise traders demands. The right-hand side is the in the ex ression for investors demand and and above yields m = (1 m)/h . The equilibrium s follo . They are linear in am and m as guessed. (A2) leads to e uation (5). 0 which consists of investo su ly of shares. Plugging dro ing terms of order z prices given in Proposition 1 Finally, rearranging e uation

Q.E.D

Proof of Pro osition 2 (Trading volume): Since an agents informational trades are worth w|fl fl,0 |/2, t e ave age value of t ades, motivated by in fo mation, equals VI 1 w |fl fl,0 |dl. T e facto 1/2 avoids double counting 0 2 t ades. T e diffe ence fl fl,0 is app oximately no mally dist ibuted s o VI = w 2 va ( fl fl,0) (e.g. He and Wang (1995)). Replacing fl w it its exp ession 2

The

nal ste

involves aggregating stock demands and clearing the market.

s l 0 m f ) + (1 m) + o(1) 2

(1 ) m m (pm +

in e uation (5) yields VI l on fl,0 . Noi e

= w 2

2 hs

2 + 2 + o(1), conditiona

2 (1) traders generate a trading volume on average e ual to E( 1 | | w 2 2

2 (1) ing by the stocks market ca italization, (1 )K 1 + o(1) (the one share outstanding), and equals VT ss the

rm has only 0 V /((1 )K 1 ) + o(1). To asse

0 impact of market po er on trading volume, it suf ces to differentiate V and VT ith respect to . Doing so implies that V / > 0 and VT / > 0, as Proposition 2 establishes. Q.E.D. Proof of Proposition 3 (Stock price informativeness): T he informativeness 2 2 2 2 of prices is de ned as h = h /( (1 ) ). Clearly, h / > 0 so ket p s ower enhances the informativeness of rices. Q.E.D.

Proof of Pro osition 4 (Dis ersion of investors forecasts): We showed in the roof of Pro osition 1 that investors roductivity forecasts e ual E(a | Fl ) = a + a + a s . The dispersion of these forecasts across investors, for a given 0l s l rm (i.e., for a given realization of the shocks a and ) is measured by D var[E(a | F 2 ) | a, ] = var(a s l s l = (h 2 /h) /h /z = h /z/h /z. s s s 2 s l | a, ) = var(a ) = a var( ) s l

To ass ss th impact of mark t pow r on D, it suf c s to not thath is incr asing in and therefore that D is decreasing in . Similarly, investors pro t a nd return forecasts equal E( | F ) = E((1 )a | F ) and E(r | F ) = E((1 )a | l l l

volume V = w 2 ( divid-

2 h

2 + 2 + 2) + o(1). Turnover i obtained by

2 w ) = 2 . Adding information- and noi e-motivated trade lead

to a (dollar) total trading

2 Fl ) p so t ei dispe sions, conditional on a and , e ual (1 ) D. Since D is decreasing in , (1 )2D is too. Thus, investors make less dispersed forecasts about the productivity, pro t, and return of more monopolistic Q.E.D. ----------------------- Page 38----------------------38 The Journal of FinanceR Proof of Proposition 5 (Distribution of stock returns): Expected stock returns The expected excess (simple) return on a stock equals

f 2 = (p0 + (1 ) /h/2)z = (1 )3K 1z/(h ) + o(z) 0 because var(rz | F 2 l ) = var((1 )az | F ) = (1 ) z/h l

from the proof of Proposition 1. We note that E(R) Rf is identical ac oss investo s and ms wit t e same ma ket powe . T e nume ato , (1 )3K 1 , 0 re ects the direct effect of on the expected excess return. The indirect effect of operates through h in the denominator. Like the direct effect, it tends to reduce expected returns: As increases, information improves (h increases) so expected returns fall. Sharpe ratios

1 K0

z/(h )/

(1 ) /h = (1 )2K 1 z/ /h.

The average Sharpe ratio is identical across investors and the same market po er. It equals SR E[E(R | F ) Rf ]/va ( z | F ) = [E(R) Rf ]/va ( z | F ) l l l = (1 )3 2

 

 

= E( z)

f z + va ( z | F )/2 l

 

E(R) Rf

= E[E( z | F )

f z + va ( z | F )/2] l

 

 

rms.

rms

ith

0 Again, the direct effect of (the (1 )2K 1 term in the numerator) decreases 0 the Sharpe ratio and the indirect effect through the informativeness of prices (h in the denominator) decreases it further. Stock return volatility As noted above, the conditional variance of stock returns equals var(rz | F 2 l The unconditional E[var(rz | F )], l here E[var(rz | F 2 l )] = E[var((1 )az | F )] = (1 ) z/h. l l ) = var((1 )az | F ) = (1 ) z/h. l is var(rz) = var[E(rz | F )] +

variance

----------------------- Page 39----------------------Product Market Competition and Stock Market Ef ciency 39 To compute var[E(rz | Fl )], e note that E(rz | F ) = E((1 )az pz | F ) = E((1 )az | F ) pz l l l 2 2 = / + h z/h (p + p )z + o(z) s l 0 a (recall (1 )/h ). s l 2 2 2 2 2 (1 ) (hs + (1 ) )z/h + o(z). Adding the two term yield var(rz) = 2 2 2 2 2 (1 ) (h + hs + (1 ) )z/h + o(z). Again the indirect effect of through the informativeness of prices appears in the h terms. Differentiating ln var(rz) ith respect to h yields 2 + 2 s + 2 2 2 2 from Thus, the proof of Proposition

E(rz | F ) = (1 )[h ( l s l

Thu , the increa e in h generated by

trengthening market power reduce

2 2 (ln var(rz))/h =

(1 )

z/ h + h

(1 )

1 that

a + and 0

and

)]z/h p z

var[E(rz | F )

/h

the volatility of tock return . Pro t volatility

term), conditional on tock price . Here, var(ln | P ) = var((1 )az | P ) = (1 )2z/(h a p Higher informativeness hp , caused by an increase in market po er, reduces the volatility of pro ts beyond the direct effect of market po er. Q.E.D. Proof of Proposition 6 (Liquidity): In the model, liquidity re presents the sensitivity of stock prices to (uninformative) noise shocks and is measured by

To assess the im act of com etition on li uidity, we differentiate with

< 0. a a

Hence, ln p / < 0 and stock prices of more monopolistic rms are less sensitive to noise shocks, that is, are more liquid. In particular, p rices are independent from noise shocks hen is close to one. Q.E.D. Proof of Proposition 7 (Stock prices hen shares are issued): The proof follo s that of Proposition 1 except that the stock of capital is no endogenous. The amount of ne capital raised equals the value of the new sh res, th t is,

K = P . The exp n e

c pit l stock, K0 + K, llows

40

P ge 40

The Journ l of Fin nceR monopoly to gener te

2 (

) a

Since the denominator is an be written as

ositive, we focus on the numerator. It c 2

) s

= 2 (

a s

to : ln p / = 2/(1 ) + h (h/)/(1 / )/ . a a 2 2 3 2 h/ = 2hs /( (1 ) ) yield 2 2 2 2 2 ln p / = 2 (1 / ) /( (1 ) h ) /(1 a a s

2 res ect

= (1 )p

2 /h = (1 ) (1 a s

We compute the volatility of log pro t

(we take log to factor out the order-0

+ h ).

/ ())/h . a s

Plugging

/ )/(1 ).

pro t = Y Q = (1 )(A(K0 + K))1 . Since there are 1 + sh res outst n ing, the resulting gross stock return is R = /(P (1 + )) = (1 )[A(K0 + P )]1 /(P (1 + )).

R = (1 )(K

1 0 + P ) /(P (1 + )) exp[(1 )az (1 + )pz] + o(z)

here P /(K0 + P ) is the ilution f ctor. When z = 0 (no risk), R = 1 so 1 P is the solution to P (1 + ) = (1 )(K0 + P ) . Therefore, the l og stock return is rz = ln(R) = [(1 )a (1 + )p]z + o(z). The subsequent steps are identical to those that compose the proof of Proposition 1. We s olve the signal extraction and portfolio problems of an investor ho observes p and sl . We aggregate stock demands using the la of large numbers, add noise trades, and equate the total demand to the total supply of shares, 1 + . The resulting stock price p is line r in n as guessed. Its ex ression is rovided in Pro osition 7. Q.E.D. Proof of Pro ositions 8 and 9 (Allocative ef ciency): ifferentiating e uation (6) de ning a(, ) with respect to h hol ing xed yields p /h = (1 )/(1 + )/h /h . a a Thus, pa increases ith h: Investments are more ef cient hen information is more accurate. Similarly, e can measure the ef ciency of investments using p /pP = 1 / to facto out t e di ect effect of . Since h increases ith a a a (Proposition 3) and pa/pP increases ith h,pa/pP increases i th . Hence, a a capital is more ef ciently allocated across more monopolistic rms. Q.E.D.

Proof of Proposition 10 (Learning from past pro ts): The proof is identical to that of Proposition 1, except that investors observe an additional public signal 0 . We guess that the stock rice is a roximately given in e uation (9), that is, z = z + z + p z + o(z), where a + ( is a constant 0 0 0 p p

to be determined) and 0 valent to

a + u. Thus, observing

and 0

is e ui

We express stock prices We obt in

s P = P exp(pz ) n exp n returns

roun z = 0.

observing stocks},

and

. Based on her information set F l

{s , , for all

z var(az | F ) = l where 1 h h . s a + h 0 + h + h , a s p h h p = 2 2 ----------------------- Page 41----------------------Product Market Competition and Stock Market Ef ciency 41 The conditional mean and variance of exce tock return follow from rz = (1 )az pz . Investo s stock demand is given by equation (A1), w ic yields afte substitution = 1 s + + h h p + r f + (1 ) fl hs l p p 0 0 + o(1). (1 ) (1 ) 2 Aggregating stock demands and clearing the market leads to = (1 )/hs as in Proposition 1, to the expressions given in Proposition 10, and to equilibrium prices linear in p and 0 as guessed. REFERENCES Amihud, Yakov, 2002, Illi uidity and stock returns: Cross-section and time-serie s effects, Journal of Financial Markets 5, 3156. Ausubel, Lawrence, 1990, Partially-revealing rational ex ectations e uilibrium i n a com etitive economy, Journal of Economic Theory 50, 93126. Barlevy, Gadi, and Pietro Veronesi, 2000, Information ac uisition in nancial mark ets, Review of Economic Studies 67, 7990. Beaver, William H., 1968, The information content of annual earnings announcemen ts, Journal of Accounting Research 6, 6792. Beneish, Messod, and Mark Vargus, 2002, Insider trading, earnings uality, and a ccrual mis ricing, Accounting Review 77, 755791. Bernardo, Antonio, and Ken Judd, 2000, Asset market e uilibrium with general tas tes, returns, and informational asymmetries, Journal of Financial Markets 3, 1743. Bhattacharya, Ut al, and Neal Gal in, 2005, Is stock icking declining around th e world?, Working a er, Indiana University Bloomington. Boot, Anjan, and Arnoud Thakor, 1993, Security design, Journal of Finance 48, 13 Q.E.D. , a 0 h h 0 , and a h h s and h E(az | F ) = (a l p + a s )z, s l

p 0 the conditional mean and variance of az for agent l are:

491378. Brander, James, and Tracy Lewis, 1986, Oligo oly and nancial structure: The limit ed liability effect, American Economic Review 76, 956970. Cameron, A. Colin, Jonah B. Gelbach, and Douglas L. Miller, 2006, Robust inferen ce with multiway clustering, Working a er, University of CaliforniaDavis. Cam bell, John, Martin Lettau, Burton Malkiel, and Yexiao Xu, 2001, Ha ve individual stocks become more volatile? An em irical ex loration of idiosyncratic risk, Journ al of Finance 56, 143. Cam bell, John, and Luis Viceira, 2002, Strategic asset allocation: Portfolio ch oice for long-term investors (Oxford University Press, Oxford). Chemla, Gilles, and Antoine Faure-Grimaud, 2001, Dynamic adverse selection and d ebt, Euro ean Economic Review 45, 17731792. Chemmanur, Thomas J., Shan He, and Debarshi Nandy, 2006, The going ublic decisi on and the roduct market, Working a er, Boston College. Chun, Hyunbae, Jung-Wook Kim, Randall Morck, and Bernard Ye ung, 2008, Creative destruction and rm-s eci c erformance heterogeneity, Journal of Financial Economics 89, 109135. Comin, Diego, and Thomas Phili on, 2006, The rise in rm-level volatility: Caus es and conseuences, NBER Macroeconomics Annual 20, 167227. Dixit Avinash, and Jose h Stiglitz, 1977, Mono olistic com etition and o timal roduct diversity, American Economic Review 67, 297308. Dow, James, Itay Goldstein, and Alexander Guembel, 2006, Incenti ves for information ro duction in markets when rices affect real investments, Working a er, Sad Business School. Fama, Eugene, and Ken French, 1993, Common risk factors in the returns on stocks and bonds, Journal of Financial Economics 33, 356. ----------------------- Page 42----------------------42 The Journal of FinanceR

Gas ar, Jose-Miguel, and Massimo Massa, 2005, Idiosyncratic volatility and rodu ct market cometition, Journal of Business 79, 31253152. Grossman, Sanford J., and Jose h E. Stiglitz, 1980, On the im ossibility of info rmationally ef cient markets, American Economic Review 70, 393408. He, Hua, and Jiang Wang, 1995, Differential information and dynamic behavior of stock trading volume, Review of Financial Studies 8, 919972. Hicks, John R., 1935, Annual survey of economic theory: The theory of mono oly, Econometrica 3, 120. Hoberg, Gerard, and Gordon Philli s, 2009, Real and nancial industry booms and bu

sts, Working a er, University of Maryland. Holthausen Robert, and Robert Verrecchia, 1990, The effect of informedn ess and consensus on rice and volume behavior, Accounting Review 65, 191208. Hou, Kewei, and David Robinson, 2006, Industry concentration and average stock r eturns, Journal of Finance 61, 19271956. Irvine, Paul J., and Jeffrey E. Pontiff, 2009, Idiosyncratic return volatility, cash ows, and roduct market com etition, Review of Financial Studies 11491177. Ke, Bin, Steven Huddart, and Kathy Petroni, 2003, What insiders know about future earnings and how they use it: Evidence from insider trades, Journal of Accounting an d Economics 35, 315346. Kyle, Albert, 1985, Continuous auctions and insider trading, Econometrica 53, 13 151335. Lakonishok, Josef, and Inmoo Lee, 2001, Are insider trades informative? Review of Financial Studies 14, 79111. Lindenberg, Eric B., and Steve A. Ross, 1981, Tobins ratio and industrial organ ization, Journal of Business 54, 132. Morck, Randall, Bernard Yeung, and Wayne Yu, 2000, The information content of st ock markets: Why do emerging markets have synchronous stock rice movements? Jou rnal of Financial Economics 58, 215260. Peress, Joel, 2004, Wealth, information ac uisition and ortfolio choice,Review of Financial Studies 17, 879914. Peress, Joel, 2009, The tradeoff between risk sharing and information roduction in nancial markets, Journal of Economic Theory (in ress). Perotti, Enrico C., and Ernst-Ludwig von Thadden, 2003, Strategic trans arency and informed trading: Will ca ital market integration force convergence of cor orate gov ernance? Journal of Financial and Quantitative Analysis 38, 6185. Poitevin, Michel, 1989, Financial signalling and the dee - ocket argume t, RAND Jour al of Eco omics 20, 2640. Raja , Raghuram G., a d Luigi Zi gales, 2003, The great reversal s: The politics of a cial developme t i the twe tieth ce tury, Jour al of Fi a ci al Eco omics 69, 5 50. Rochet, Jea -Charles, a d Jea -Luc Vila, 1994, I sider tradi g without ormality, Review of Eco omic Studies 61, 131152. Romer, Paul, 1990, E doge ous tech ological cha ge, Jour al of Political Eco omy 98, S71 102. Schumpeter, Joseph A., 1912, Theorie der Wirtschaftliche E twichlu g, L eipzig, Du ker u d Humbolt. Tra slated by R. Opie, The Theory of Eco omic Developme t: A I qu iry i to Profits, Capital, Credit, I terest, a d the Busi ess Cycle (Harvard U iversity Press, Cambridge,

 

 

  

 

xed costs a d mo opolistic com er, 2001, IPOs a d product quali dard errors that cluster by both

----------------------- Page 43-----------------------

Titma , Sherida , 1984, The effect of capital structure o a rms liquidatio decis io , Jour al of Fi a cial Eco omics 13, 137151. Tookes, Heather, 2007, I formatio , tradi g, a d product market i teract io s: Cross-sectio al implicatio s of i formed tradi g, Worki g paper, Yale School of Ma ageme t.

----------------------- Page 44-----------------------

 

44

The Jour al of Fi a ceR

   

Yua , Kathy, A ratio al tatio s al of Fi a 411.

2005, Asymmetric price moveme ts a d borrowi g co strai ts: expecequilibrium model of crises, co tagio , a d co fusio , Jour ce 60, 379

 

Product Market Competitio 43

a d Stock Market Ef cie cy

 

Massachusetts). Seyhu , H. Nejat, 1986, I siders pro ts, costs al of Fi a cial Eco omics 16, 189212. Spe ce, Michael, 1976, Product selectio , petitio , Review of Eco omic Studies 43, 217235. Stoughto , Neal, Kit Po g Wo g, a d Joseph Zech ty, Jour al of Busi ess 74, 375408. Thompso , Samuel, 2006, Simple formulas for sta rm a d time, Worki g paper, Harvard U iversity.

of tradi g, a d market ef cie cy, Jour