Investor Sentiment

8/3/2019 Investor Sentiment

1/37

THE JOURNAL OF FINANCE VOL. LXI, NO. -1 AUGUST 2006

Investor Sentiment and the Cross-Sectionof Stock Retums

MALCOLM BAKER and JEFFREY WURGLER-ABSTRACT

We study how investor sentiment affects the cross-section of stock retums. We pre-dict that a wave of investor sentiment has larger effects on securities whose valua-tions are highly subjective and difficult to arbitrage. Consistent with this prediction,we find that when beginning-of-period proxies for sentiment are low, subsequent re-turns arerelatively high for small stocks, young stocks, high volatility stocks, un-profitable stocks, non-dividend-paying stocks, extreme growth stocks, and distressedstocks. When sentiment is high, on the other hand, these categories of stock earnrelatively low subsequent returns.

CLASSICAL FINANCE THEORY LEAVES NO ROLE FOR INVESTOR SENTIMENT. Rather, thistheory argues that competition among rational investors, who diversify to opti-mize the statistical properties of their portfolios, will lead to an equilihrium inwhich prices equal the rationally discounted value of expected cash flows, andin which the cross-section of expected retums depends only on the cross-sectionof systematic risks.' Even if some investors are irrational, classical theory ar-gues, their demands are offset by arbitrageurs and thus have no significantimpact on prices.In this paper, we present evidence that investor sentiment may have signifi-cant effects on the cross-section of stock prices. We start with simple theoreticalpredictions. Because amispricing is the result of an uninformed demand shockin thepresence of a binding arbitrage constraint, wepredict that a broad-based wave of sentiment has cross-sectional effects (that is, does not simplyraise or lower all prices equally) when sentiment-based demands or arbitrage

*Baker is at the Harvard Business School and National Bureau ol Economic Research; Wurgieris at the NYU Stern School of Business and the National Bureau of Economic Research. We thankan anonymous referee, Rob Stambaugb (the editor), Ned Elton, Wayne Ferson, Xavier Gabaix,Marty Gruber, Lisa Kramer, Owen Lamont, Martin Lettau, Anthony Lynch, Jay Shanken, MeirStatman, Sheridan Titman, and Jeremy Stein forhelpful comments, as well as participants ofconferences or seminars at Baruch College, Boston College, the Chicago Quantitative Alliance,Emory University, the Federal Reserve Bank of New York, Harvard University, Indiana University,Michigan State University, NBER, the Norwegian School of Economics and Business, NorwegianSchool of Management, New York University, Stockholm School of Economics, Tulane University,the University of Amsterdam, the University of British Columbia, the University oflllinois, theUniversity of Kentucky, the University of Michigan, the University of Notre Dame, the Universityof Texas, and the University of Wisconsin. We gratefully acknowledge financial support from theQ Group and the Division of Research of the Harvard Busine.ss School,


2/37

1646 The Journal of Financeconstraints vary across stocks. In practice, these two distinct channels lead toquite similar predictions because stocks that are likely to be most sensitive tospeculative demand, those with highly subjective valuations, also tend to bethe riskiest and costliest to arbitrage. Concretely, then, theory suggests twodistinct chann els throug h which the sha res of certain firms newer, smaller,more volatile, unprofitable, non-dividend paying, distressed or with extremegrowth potential, and firms with analogous characteristicsare likely to bemore affected by shifts in investor sentiment.

To inves tigate th is prediction empirically, and to get a m ore tangible s ense ofthe intrinsically elusive concept of investor sen tim ent, we st ar t with a s um m aryof the rises and falls in U.S. m ark et se ntim ent from 1961 thro ugh the In ter ne tbubble. This summary is based on anecdotal accounts and thus by its naturecan only be a suggestive, ex post characte rization of fluctua tions in s ent im ent.Nonetheless, its basic message appears broadly consistent with our theoreticalpredictions and suggests that more rigorous tests are warranted.

Our main empirical approach is as follows. Because cross-sectional patternsof sentiment-driven niispricing would be difficult to identify directly, we ex-amine wh ether cross-sectional predictability pat tern s in stock retu rns dependupon proxies for beginning-of-period sen tim ent. For example, low future re tu rn son young firms relative to old firms, conditional on high values for proxies forbeginning-of-period sentiment, would be consistent with the ex ante relativeovervaluation of young firms. As usu al, we are m indful of the joint hypo thesisproblem that any predictability patterns we find actually reflect compensationfor systematic risks.The first step is to gather proxies for investor sentiment that we can use astime-series conditioning variables. Since there are no perfect and/or uncontro-versial proxies for investor sentiment, our approach is necessarily practical.Specifically, we consider a number of proxies suggested in recent work andform a composite sentim ent index based on their first p rincipal com ponent. Toreduce the likelihood that these proxies are connected to systematic risk, wealso form an index based on sen tim ent proxies tha t have been orthogonalized toseveral macroeconomic conditions. The sentiment indexes visihly line up withhistorical accounts of bubbles and crashes.We then test how the cross-section of subsequent stock returns varies withbeginning-of-period sen tim ent. Using m onthly stock re tu rn s between 1963 and2001 , we sta rt by forming equal-w eighted decile portfolios b ased on seve ral firmcharacteristics. (Our theory predicts, and the empirical results confirm, thatlarge firms will be less affected by sentiment, and hence value weighting willtend to obscure the relev ant p atte rns .) We the n look for pa tte rn s in the averag ereturns across deciles conditional upon the beginning-of-period level of senti-m ent. We find th at when sen tim ent is low (helow sam ple average), small stocksearn particularly high subsequent returns, but when sentiment is high (aboveaverage), there is no size effect at all. Conditional patterns are even sharperwhen we sort on other firm characteristics. When sentiment is low, subsequent


3/37

Investor Sentiment and the Cross-Section of Stock Returns 1647these p atte m s completely reverse. In other words, several characteristics thatdo not have any unconditional predictive power actually display sign-flippingpredictive ability, in the hj^jothesized directions, once one conditions on sen ti-ment. These are our most striking findings. Although earlier data are not asrich, some of these patterns are also apparent in a sample that covers 1935through 1961.The sorts also suggest that sentiment affects extreme growth and distressedfirms in similar ways. Note that when stocks are sorted into deciles by salesgrowth, book-to-market, or extemal financing activity, growth and distressfirms tend to lie at opposing extremes, with more "stable" firms in the middledeciles. Wefind ha t when sentiment is low, the subsequent retu rns on stocks atboth extremes are especially high relative to their unconditional average, whilestocks in the middle deciles are less affected by sentiment. (The result is notstatistically significant for book-to-market, however.) This U-shaped pattemin the conditional difference is also broadly consistent with theoretical pre-dictions: both extreme growth and distressed firms have relatively subjectivevaluations and are relatively hard to arbitrage, and so they should be expectedto be most affected by sentiment. Again, note that this intriguing conditionalpattem would be averaged away in an unconditional study.We then consider a regression approach, which allows us to control for co-movement in size and book-to-market-sorted stocks using the Fama-French(1993) factors. We use the sentiment indexes to forecast the retu rns of varioushigh-minus-low portfolios (in terms of sensitivity to sentiment). Not surpris-ingly, given that our decile portfolios are equal-weighted and several of thecharacteristics we examine are correlated with size, the inclusion of SMS asa control tends to reduce the magnitude of the predictability, although somepredictive power generally remains.We then turn to the classical altem ative explanation, namely, that they sim-ply reflect a complex pattern of compensation for systematic risk. This expla-nation would account for the predictability evidence by either time variationin rational, market-wide risk premia or time variation in the cross-sectionalpattem of risk, that is, beta loadings. Further tests cast doubt on these hy-potheses. We tes t the second possibility directly and find no link between thepattems in predictability and pattems in betas with market retums or con-sumption growth. If risk is not changing over time, then the first possibilityrequires not just time variation in risk premia, but also changes in sign. Putsimply, it would require that in half of our sample period (when sentiment isrelatively low), older, less volatile, profitable, and/or dividend-paying firms ac-tually require a risk premium over very young, highly volatile, unprofitable,and/or nonpayers. This is counterintuitive. Other aspects of the results alsosuggest that systematic risk is not a complete explanation.The results challenge the classical view of the cross-section of stock pricesand, in doing so, build on several recent them es. First, the resu lts complementearher work that shows sentiment helps to explain the time series of retums(Kothari and Shanken (1997), Neal and Wheatley (1998), Shiller (1981, 2000),


4/37

1648 The Journal of Financethe effects of conditional systematic r isks; here we condition on investor sen-timent. Daniel and Titman (1997) test a characteristics-based model for thecross-section of expected returns; we extend their specification into a condi-tional characteristics-based model. Shleifer (2000) surveys early work on sen-timent and limited arbitrage, two key ingredients he re. Barberis and Shleifer(2003), Barberis, Shleifer, and Wurgler (2005), and Peng and Xiong (2004) dis-cuss category-level trading, and Fama and French (1993) document comove-ment of stocks of similar sizes and book-to-market ratios; uninformed demandshocks for categories of stocks with similar characteristics are central to ourresults. Finally, we extend and unify known relationships among sentiment,IPOs, and small stock returns (Lee, Shleifer, and Thaler (1991), Swaminathan(1996), Neal and Wheatley (1998)).

Section I discusses theoretical predictions. Section II provides a qualitativehistory of recent speculative episodes. Section III describes our empirical hy-potheses and data, and Section IV presen ts the main empirical tests. Section Vconcludes.

I. Theoretical Effects of Sentiment on the Cross-SectionA mispricing is the resu lt of both an uninformed demand shock and a limiton arbitrage. One can therefore th ink of two distinct channels through whichinvestor sentiment, as defined more precisely below, might affect the cross-section of stock prices. In the first channel, sentimental demand shocks vary

in the cross-section, while arbitrage limits are constant. In the second, thedifficulty of arbitrage varies across stocks but sentim ent is generic.We discussthese in turn .A. Cross-Sectional Variation in Sentimen t

One possible definition of investor sentim ent is the propensity to speculate.^Under this definition, sentiment drives the relative demand for speculativeinvestm ents, and therefore causes cross-sectional effects even if arb itrage forcesare the same across stocks.What makes some stocks more vulnerable to broad shifts in the propensityto speculate? We suggest that the main factor is the subjectivity of their valu-ations. For instance, consider a canonical young, unprofitable, extreme growthstock. The lack of an earnings history combined with the presence of appar-ently unlimited growth opportunities allows unsophisticated investors to de-fend, with equal plausibility, a wide spectrum of valuations, from much too low-to much too high, as suits their sentiment. During a bubble period, when thepropensity to speculate is high, this profile of character istics also allows invest-ment bankers (or swindlers) to further argue for the high end of valua tions. Bycontrast, the value of a firm with a long earnings history, tangible asse ts, and^Aghion and Stein (2004) develop a model with both rational expectations and bounded ratio-


5/37

Investor Sentiment and the Cross-Section of Stock Returns 1649stable dividends is much less subjective, and thus its stock is likely to be lessaffected by fluctuations in the propensity to speculate.^While the above channel suggests how variation in the propensity to spec-ulate may generally affect the cross-section, it does not take a stand on howsentimental investors actually choose stocks. We suggest that they simply de-mand stocks tha t have the bundle of salient characteristics tha t is compatiblewith their sentiment.* That is, investors with a low propensity to speculate maydemand profitable, dividend-paying stocks not because profitability and divi-dends are correlated w ith some unobservable firm property th at defines safetyto the investor, but precisely because the sa lient characteristics "profitability"and "dividends" are essentially taken to define safety.^ Likewise, the salientcharacteristics "no earnings," "young age," and "no dividends" mark the stockas speculative. Casual observation suggests that such an investment processmay be a more accurate description of how typical investors pick stocks thanthe process outlined by Markowitz (1959), in which investors view individualsecurities purely in terms of their statistical properties.B. Cross-Sectional Variation in Arbitrage '

One might also define investor sentiment as optimism or pessimism aboutstocks in general. Indiscriminate waves of sentiment still affect the cross-section, however, if arbitrage forces are relatively weaker in a subset of stocks.This channel is better understood than the cross-sectional variation in senti-ment channel. A body of theoretical and empirical research shows tha t arbitragetends to be particularly risky and costly for young, small, unprofitable, extremegrowth, or distressed stocks. First, their high idiosyncratic risk makes relative-value a rbitrage especially risky (Wurgier and Zhuravskaya (2002)). Moreover,such stocks tend to be more costly to trade (Amihud and Mendelsohn (1986))and particularly expensive, sometimes impossible, to sell short (D'Avolio (2002),Geczy, Musto, and Reed (2002), Jones and Lamont (2002), Duffie, Garleanu, and The favorite-longshot bias in racetrack betting is a static illustra tion of the notion tha t in vestorswith a high propensity to speculate (racetrack bettors) have a relatively high demand for the mostspeculative bets (longshots have the most negative expected returns; see Hausch and Ziemba(1995)).* The idea that investors view securities as a vector of salient characteristics borrows fromLancaster (1966, 1971), who views consumer demand theory from the perspective that the utilityof a consum er good (e.g, oranges) de rives from more prim itive cha racteristics (fiber and v itamin C),^ The implications of categorization for finance are explored by Baker and Wurgier (2003),Barb eris an d Shleifer (2003), Barb eris, Shleifer, and Wu rgier (2005), Greenwood and Sosner (2003),and Peng and Xiong (2004). Note that if investors infer category membership from salient char-acteristics (some psychologists propose that category membership is determined by the presenceof defining or characteristic features, see, for example. Smith, Shoben, and Rips (1974)), thensentiment-driven dema nd w ill be directly connected to characteristics even if sentim ental investorsundertake an intervening process of categorization and trade entirely at the category level. It is

also empirically con venient to boil key investm ent categories down into vectors of stable and mea-surable characteristics: One can use the same empirical framework to study episodes such as thelate 1960s growth stocks bubble and the Internet bubble. In other words, the term "Internet bub-


6/37

1650 ' 77ie Journal of FinancePedersen (2002), Lamont and Thaler (2003), Mitchell, Pulvino, and Stafford(2002)). Further, their lower liquidity also exposes would-be arbitrageurs topredatory attacks (Brunnermeier and Pedersen (2005)).The key point of this discussion is that, in practice, the same stocks that arethe hardest to arbitrage also tend to be the most difficult to value. While forexpositional purposes we have outlined the two channels separately, they arelikely to have ove rlapping effects. This m ay ma ke them difficult to d isting uishempirically; however, it only strengthens our predictions about what region ofthe cross-section is most affected by sen tim en t. Indeed , the two ch ann els can re-inforce each oth er For exam ple, th e fact th at in vesto rs can convince them selv esof a wide range of valuations in some regions of the cross-section generates anoise-trader risk th at further deters short-horizon a rbitra ge urs (De Long et al.(1990), Shleifer an d Vishn y (1997)).^ , i

II. An Anecd otal History of Investor Se ntim ent, 1961-2002Here we briefly summarize the most prominent U.S. stock market bubblesbetween 1961 and 2002 (matching the period of our m ain da ta). The reade r ea-ger to see results may skip this section, but it is useful for three reasons. First,des pite grea t inter est in the effects of investor s en tim en t, th e academ ic litera-tu re does not contain ev en the m ost basic ex post chara cteriz ation of most of th erecent speculative episodes. Second, a knowledge of the rough timing of theseepisodes allows us to make a preliminary judgment about the accuracy of the

qu antita tive proxies for sentim ent th at we develop later. Third, the discussionsheds some initial, albeit anecdotal, light on the plausibility of our theoreticalpredictions.We distill our brief history of sentiment from several sources. Kindleberger(2001) draws general lessons from bubbles and crashes over the past few hun-dred ye ars , while Brown (1991), Dre m an (1979), G rah am (1973), Malkiel (1990,1999), Shiller (2000), and Siegel (1998) focus more specifically on recent U.S.stock m ark et episodes. We tak e each of the se acco unts with a grain of salt, andemphasize only those themes that appear repeatedly.We st ar t in 1961, a yea r th at G rah am (1973), Malkiel (1990) an d B rown(1991) note as charac terized by a high d em and for sm all, young, growth stock s;Dreman (1979, p. 70) confirms their accounts. For instance, Malkiel writes ofa "new-issue mania" that was concentrated on new "tronics" firms. ", . . Thetronics boom came back to eart h in 1962. The tailspin sta rted early in the yearand exploded in a horrendous selling wave. . . Growth stocks took the brunt ofthe decline, falling m uch further th an the gen eral ma rket" (p. 54-5 7).The next major bubble developed in 1967 and 1968. Brown writes that"scores of franchisers, computer firms, and mobile home manufactures seemed'^ We do not incorporate the equilibrium prediction of DeLong et al. (1990), nam ely th at securities

with more exposure to sentiment have higher unconditional expected returns. Elton, Gruber, andBusse (19981 argu e tha t expected ret urn s are not higher on stocks that have highe r sen sitivities


7/37

Investor Sentiment and the Cross-Section of Stock Returns 1651to promise overnight w ealth [while] quality was pre tty much forgotten"(p . 90). Malkiel and Dreman also note this pattem of a focus on firms withstrong earn ings grow th or potential and a n avoidance of "the major ind ustria lgian ts, 'buggywhip companies, ' as they were som etimes contemp tuously called"(Dreman 1979, p. 74-75). Another characteristic apparently out of favor wasdividends. According to the New York Times, "during the speculat ive marketof the late 1960s many brokers told customers that i t didn't matter whether acompany paid a dividend just so long as its stock ke pt going up " (9/13/1976).But "after 1968, as it became clear that capital losses were possible, investorscame to value divid end s" (10/7/1999). In sum m arizi ng the p erforman ce of stocksfrom t he end of 1968 throu gh A ugu st 1971, G rah am (1973) w rites: "[our] com-parative results undoubtedly reflect the tendency of smaller issues of inferiorqua lity to be relatively overvalued in bull mar ke ts, and not only to suffer moreserious declines tha n the stronger issu es in the ensu ing price collapse, but alsoto delay the ir full recovery in m any cases ind efinitely" (p. 212).

Anecdotal accounts invariably describe the early 1970s as a bear market,with sentiment at a low level. However, a set of established, large, stable, con-sistently profitable stocks known as the "nifty fifty" enjoyed notably high val-uations. Brown (1991), Malkiel (1990), and Siegel (1998) each highlight thisepisode. Siegel writes, "All of these stocks had proven growth records, contin-ual increases in div ide nd s. . . and high ma rket capital izat ion" (p . 106). Note tha tthi s specula tive episode is a mirro r image of those described above (and below).Th at is, the bubbles associated w ith high sen time nt periods centered on small,young, unprofitable growth stocks, whereas the nifty fifty episode appears tobe a bubble in a set of firms with an opposite set of characteristics (old, large,and continuous ea rning s and dividend growth) tha t happe ned in a period of^lowsentiment .

The late 1970s through mid 1980s are described as a period of generallyhigh sentiment, perhaps associated with Reagan-era optimism. This periodwitnessed a series of speculative episodes. Dreman describes a bubble in gam-bling issues in 1977 and 1978. R itter (1984) stud ies th e hot-issue m ark et of1980, and finds greater initial retums on IPOs of natural resource start-upsthan on large, mature, profitable offerings. Of 1983, Malkiel (p. 74-75) writesthat "the high-technology new-issue boom of the first half of 1983 was an al-most perfect replica of the 196O's ep iso de s. , . The bubble app ears to have bu rstearly in the second half of 1 9 8 3 . . . the carnage in the small company and new-issue markets was truly catastrophic." Brown confirms this account. Of themid 1980s, Malkiel writes that "What electronics was to the 1960s, biotech-nology became to the 1980s new issue s of biotech com panies were eagerlygobbled up having positive sales and e arnin gs w as actually considered adrawback" (p. 77-79). But by 1987 and 1988, "market sentiment had changedfrom a n accep tance of an exciting st o ry .. . to a desire to stay closer to ea rth w ithlow-multiple stocks that actually pay dividends" (p. 79).

The late 1990s bubble in technology stocks is familiar. By all accounts, in-vestor sentiment was broadly high before the bubble started to burst in 2000.


8/37

1652 The Journal of Financethe bubble, while Asness et al. (2000) and Chan, Karceski, and Lakonishok(2000) were arguing even before the crash that late 1990s growth stockvaluations were difficult to ascribe to rationally expected earnings growth.Malkiel draws para llels to episodes in the 1960s, i970s, and 1980s, and Shiller(2000) draws para llels to the late 1920s. As in earlier speculative episodes th atoccurred in high sentiment periods, demand for dividend payers seems to havebeen low (New York Times, 1/6/1998). Ljungqvist and Wilhelm (2003) find that80% of the 1999 and 2000 IPO cohorts had negative earnings per share andthat the median age of 1999 IPOs was 4 years. This contrasts with an averageage of over 9 years just prior to the emergence of the bubble, and of over 12years by 2001 and 2002 (Ritter (2003)).

These anecdotes suggest some regular pa tterns in the effect of investor sen ti-ment on the cross-section. For instance, canonical extreme growth stocks seemto be especially prone to bubbles (and subsequen t crashes), consistent with theobservation that they are more appealing to speculators and optimists and atthe same time hard to arbitrage. The "nifty fifty" bubble is a notable excep-tion, but anecdotal accounts suggest that this bubble occurred during a periodof broadly low sentiment, so it may still be consistent with the cross-sectionalprediction that an increase in sentiment increases the relative price of thosestocks that are the most subjective to value and the hardest to arbitrage. Wenow tu rn to formal tests of this prediction.

III. Empirical Approach and DataA. Empirical Approach - \ . ' Theory and historical anecdote both suggest th at sentiment may cause sys-tematic p atte rns of mispricing. Because m ispricing is hard to identify directly,however, our approach is to look for systematic patterns of mispricing correc-tion. For example, a pattern in which returns on young and unprofitable growthfirms are (on average) especially low when beginning-of-period sentiment is es-timated to be high may represent the correction of a bubble in growth stocks.Specifically, to identify sentiment-driven changes in cross-sectional pre-

dictability patte rns, we need to control for two more basic effects, namely, thegeneric impact of investor sentiment on all stocks and the generic impact ofcharacteristics across all time periods. Thus, we organize our analysis looselyaround the following predictive specification:Et-i[Rit]=a + aiTt^i-\-h[^t-i+^Tt.iKit-i, (1)

where i indexes firms , t denotes time, x is a vector of characteristics, and T is aproxy for sentim ent. The coefficient a i picks up the generic effect of sentim ent,and the vector bi the generic effect of characteristics. Our interest centers onb2. The null is tha t b2 equals zero or, more precisely, that any nonzero effect isrational compensation for systematic risk. The alterna tive is that b2 is nonzero


9/37

Investor Sentiment and the Cross-Section of Stock Returns 1653B. Characteristics and Returns

The firm-level data are from the merged CRSP-Compustat database. Thesam ple includ es all common stock (share codes 10 and 11) between 1962 throu gh2001. Following Fama and French (1992), we match accounting data for fiscalyear-ends in calendar year ( - 1 to (monthly) re tu rn s from Ju ly t through June/ + 1, and we use their variable definitions when possible.Table I shows summary stat ist ics. Panel A summarizes returns variables.Following common practice, we define momentum, MOM, as the cumulativeraw return for the 11-month period from 12 through 2 months prior to theobservation retum. Because momentum is not mentioned as a sal ient charac-teristic in historical anecdote, and theory does not suggest a direct connectionbetween momentum and the difficulty of valuation or arbitrage, we use mo-mentum merely as a control variable to understand the independence of ourresults from known mispricing pattems.The remaining panels summarize the f irm and securi ty characterist ics thatwe consider. The previous sec tions' discus sions po int us directly to sev eral va ri-ables. To that list, we add a few more characteristics that, by introspection,seem likely to be salient to investors. O verall, we roughly group chara cteristicsas pe rta ini ng to firm size and age, profitability, d ividend s, asset tangibility, an dgrowth opportunities and/or distress.Size and age characteristics include market equity, ME, from June of yeart, measured as price times shares outstanding from CRSP. We match ME tomo nthly ret u m s from July of year t through Ju ne of year t + I. Firm age, Ageyis the num ber of yea rs since the firm's first ap peara nce on CRSP, m easu red tothe neares t monthj and Sigma is the standard deviation of monthly retumsover the 12 months ending in June of year t. If there are at least nine retumsavailable to estimate it , Sigma is then matched to monthly returns from Julyof yea r t through Jun e of year t + 1. While historical anecdote does not identifystock volatility itself as a salient characteristic, prior work argues that it islikely to be a good proxy for th e difficulty of both v alua tion and arb itra ge .Profitability characteristics include the retum on equity, E-\-IBE, which ispositive for profitable firms and zero for unprofitable firms. Earnings {E) isincome before extraordinary items (Item 18) plus income statement deferredtaxes (Item 50) minus preferred dividends (Item 19), if earnings are positive;book equity {BE) is shareholders equity (Item 60) plus balance sheet deferredtaxes (Item 35). The profitability dummy variable E > 0 takes the value onefor prof itable firms and zero for unpro fitable firms.Dividend characteristics include dividends to equity, DIBE, which is divi-dends per share at the ex date (Item 26) t imes Co mp ustat shares outstand ing(Item 25) divided by book equity. The dividend payer dummy D > Q takes thevalu e one for firms w ith positive divide nds p er sh are by the ex da te. The declinenoted by Fam a and French (2001) in the percen tage of firms th at pay dividendsis apparent.


10/37

1654 The Journal of FinanceIIili

B a "a .2 = '5 3

P ^ e ^

M ^ " ^ ^ ^

" ;s ii.

Oi intc t^00 in

^ rj" cc


11/37

Investor Sentiment and the Cross-Section of Stock Returns 1655The referee suggests that asset tangibility may proxy for the difficulty ofvaluation. Asset tangibility characteristics are measured by property, plantand equipment (Item 7) over assets, PPEIA, and research and developmentexpense over assets (Item 46), RDIA. One concern is the coverage of the R&Dvariable. We do not consider th is variab le prior to 1972, because the FinancialAccounting Standards Board did not require R&D to be expensed until 1974and Com pustat coverage prior to 1972 is very poor. Also, even in recent yearsless than half of the sample reports positive R&D.Characteristics indicating growth opportunities, distress, or both includebook-to-market equity, BEIME, whose elements are defined above. Extemalfinance, EFIA, is the change in assets (Item 6) minus the change in retainedearnings (Item 36) divided by asse ts. Sales growth (GS) is the change in net sales(Item 12)dividedby prior-year net sales. Sales growth GS/10 is the decile of the

firm 's sales growth in the prior year relative to NYSE firm s' decile breakpoints.As will become clear below, one must grasp the multidimensional na ture ofthe growth and d istress variables in order to understand how they interact w ithsentiment. In particular, book-to-market wears at least three hats: High valuesmay indicate dis tress ; low values may indicate high growth opportun ities; and,as a scaled-price variable, book-to-market is also a generic valuation indicatorthat varies with any source of mispricing or rational expected returns. Sim-ilarly, sales growth and extemal finance wear at least two hats: Low values(which are negative) may indicate distress, and high values may reflect growthopportunities. Further, to the extent th at market timing motives drive extem alfinance, EFIA also wears a th ird h at as a generic misvaluation indicator.All explanatory variables are W insorized each year a t the ir 0.5 and 99.5 per-centiies. Finally, in Panels C through F, the accounting data for fiscal yearsending in calendar year ( 1 are m atched to monthly re tum s from July of yeart through Ju ne of year t -\- 1.C. Investor Sentiment

Prior work suggests a number of proxies for sentiment to use as time-seriesconditioning variables. There are no definitive or uncontroversial measures,however. We therefore form a composite index of sentiment th at is based on thecommon variation in six underlying proxies for sen timent: the closed-end funddiscount, NYSE share tumover, the number and average first-day returns onIPOs, the equity share in new issues, and the dividend premium. The sen timentproxies are measured annually from 1962 to 2001. We first introduce eachproxy separately, and then discuss how they are formed into overall sentimentindexes.The closed-end fund discount, CEFD, is the average difference between thenet asset values (NAV) of closed-end stock fund shares and their market prices.Prior work suggests that CEFD is inversely re lated to sentim ent. Zweig (1973)uses it to forecast reversion in Dow Jones stocks, and Lee et al. (1991) arguethat sentiment is behind various features of closed-end fund discounts. We


12/37

1656 77ie Journal of Financethrough 1993 from Neal and Wheatley (1998), for 1994 through 1998 fromCDAAViesenberger, and for 1999 through 2001 from tum-of-the-year issues ofthe Wall Street Journal.

NYSE share turnover is based on the ratio of reported share volume to av-erage shares listed from the NYSE Fact Book. Baker and Stein (2004) suggestthat turnover, or more generally liquidity, can serve as a sentiment index: In amarket with short-sales constraints, irrational investors participate, and thusadd liquidity, only wben they are optimistic; hence, high liquidity is a symp-tom of overvaluation. Supporting this, Jones (2001) finds that high turnoverforecasts low market returns. Turnover displays an exponential, positive trendover our period and the May 1975 elimination of fixed commissions also has avisible effect. As a partial solution, we define TURN as the natural log of theraw turnover ratio, detrended by the 5-year moving average.

The IPO market is often viewed as sensitive to sentiment, with high first-day returns on IPOs cited as a measure of investor enthusiasm, and the lowidiosyncratic returns on IPOs often interpreted as a symptom of market timing(Stigler (1964), Ritter (1991)). We take the number of IPOs, NIPO, and theaverage first-day returns, RIPO, from Jay Ritter's website, which updates thesample in Ibbotson, Sindelar, and Ritter (1994).

The share of equity issues in total equity and debt issues is another measure offinancing activity that may capture sentiment. Baker and Wurgler (2000) findthat high values of the equity share predict low market returns. The equityshare is defined as gross equity issuance divided by gross equity plus grosslong-term debt issuance using data from the Federal Reserve Bulletin.^OUT sixth and last sentiment proxy is the dividend premium, p^-^^^ the logdifference of the average market-to-book ratios of payers and nonpayers. Bakerand Wurgler (2004) use this variable to proxy for relative investor demand fordividend-paying stocks. Given that payers are generally larger, more profitablefirms with weaker growth opportunities (Fama and French (2001)), the divi-dend premiiim may proxy for the relative demand for this correlated bundle ofcharacteristics.

Each sentiment proxy is likely to include a sentiment component as well asidiosyncratic, non-sentiment-related components. We use principal componentsanalysis to isolate the common component. Another issue in forming an indexis determining the relative timing of the variablesthat is, if they exhibit lead-lag relationships, some variables Tnay reflect a given shift in sentiment earlierthan others. For instance, Ibbotson and Jaffe (1975), Lowry and Schwert (2002),and Benveniste et al. (2003) find that IPO volume lags the first-day returns onIPOs. Perhaps sentiment is partly behind the high first-day returns, and thisattracts additional IPO volume with a lag. More generally, proxies that involvefirm supply responses (S and NIPO) can be expected to lag behind proxies

^ While they both reflect equ ity issues, the nu mb er of IPOs and the equity sha re have im po rtan tdifferences. The equity share includes seasoned offerings, predicts market returns, and scales bytotal external finance to isolate the composition of finance from the level. On the other hand, theIPO variables may better reflect demand for certain IPO-like regions of the cross-section that


13/37

Investor Sentiment and the Cross-Section of Stock Returns 1657th at are based directly on investor dem and or investor behavior (RIPO, p ^ - ^ ^ ^TURN, and CEFD).We form a composite index that captures the common component in the sixproxies and incorporates th e fact tha t some variab les tak e longer to reveal thesame sentiment.^ We start by estimating the first principal component of thesix proxies and their lags. This gives us a first-stage index with 12 loadings,one for each of the current and lagged proxies. We then compute the correla-tion between the first-stage index and the current and lagged values of eachof the proxies. Finally, we define SENTIMENT as the first principal compo-nent of the correlation matrix of six variableseach respective proxy's lead orlag, whichever has higher correlation with the first-stage indexrescaling thecoefficients so that the index has unit variance.

This procedure leads to a parsimon ious indexSENTIM ENTt = -0.24lCEFDt-^0.242TURN t-i-\-0.253NIPO i

-\-0.257RIPOt-i -h 0.112S, - 0.283P ,^7^ , (2)where each of the index components has first been standardized. The firstprincipal com ponent exp lains 49% of the sam ple v ariance, so we conclude th atone factor ca p tures much of the common variation . The correlation betwee n the12-term first-stage index and the SENTIMENT index is 0.95, suggesting th atlitt le information is lost in dropp ing the six term s w ith other time subsc ripts.T he SENTIMENT index has several appealing properties. First, each indi-vidual proxy enters with the expected sign. Second, all but one enters withthe expected timing; with the exception of CEFD, price and investor behaviorvariables lead firm supply variables. Third, the index irons out some extremeobservations. (The dividend premium and the first-day IPO retums reachedun pr eced ente d levels in 1999, so for the se p roxies to work as individu al p redic-tors in the full samp le, these levels m ust be matched exactly to extreme futurereturns.)

One might object to equation (2) as a measure of sentiment on the groundsthat the principal components analysis cannot distinguish between a commonsentiment component and a common business cycle component. For instance,the n um ber of IPOs varies with the b usiness cycle in pa rt for entirely ra tionalreasons. We w ant to identify when the n um ber of IPO s is hig h for no good reason .We therefore construct a second index that explicitly removes business cyclevariation from each of the proxies prior to the princip al comp onents an alysis.Specifically, we regress each of the six raw proxies on growth in the indus-trial production index (Federal Reserve Statistical Release G.17), growth inconsumer durab les, nondu rables, and services (all from BEA National IncomeAccounts Table 2.10), and a dummy variable for NBER recessions. The residu-als from these regressions, labeled with a superscript -L, may be cleaner p roxiesfor investor sentiment. We form a n index of th e orthog ona lized p roxies followingthe sam e procedure as before. The resultin g index is


14/37

1658 The Journal of FinanceSENTIMENT,^ = -0.198CEFDi- + 0.225TURN,^_j +

Here, the first principal component explains 53% of the sample variance of theorthogonalized variables. Moreover, only the first eigenvalue is above 1.00. Interms of the signs and the timing of the components, SENTIMENT-^ retainsall of the appealing properties of SENTIMENT.Table II summarizes and correlates the sentiment measures, and Figure 1plots them. The figure shows immediately that orthogonalizing to macro vari-ables is a second-order issue. It does not qualitatively affect any componentof the index or the overall index (see Panel E). Indeed, Table II suggests thaton balance the orthogonalized proxies are slightly more correlated with eachother than are the raw proxies. If the raw variables were driven by commonmacroeconomic conditions (that we failed to remove through orthogonal ization)instead of common investor sentiment, one would expect the opposite. In anycase, to demonstrate robustness we present resu lts for both indexes in our mainanalysis.More importantly. Figure 1 shows that the sentiment measures roughly lineup with anecdotal accounts of fluctuations in sentiment. Most proxies pointto low sentiment in the first few years of the sample, after the 1961 crash ingrowth stocks. Specifically, the closed-end fund discount and dividend premiumare high, while turnover and equity issuance-related variables are low. Each

variable identifies a spike in sentim ent in 1968 and 1969, again m atching anec-dotal accounts. Sentiment then tails off until, by the mid 1970s, it is low by mostmeasures (recall that for turnover this is confounded by deregulation). The late1970s through mid 1980s sees generally rising sentiment, and, according tothe composite index, sentim ent has not dropped far below a medium level since1980. At the end of 1999, near the peak of the Internet bubble, sentiment is highby most proxies. Overall, SENTIMENT-^ is positive for the years 1968-1970,1972, 1979-1987, 1994, 1996-1997, and 1999-2001. This correspondence withanecdotal accounts seems to confirm that the measures capture the intendedvariation.There are other variables that one might reasonably wish to include in asentiment index. The main constraint is availability and consistent measure-ment over the 1962-2001 period. We have considered insider trading as a sen-timent m easure. Unfortunately, a consistent series does not appear to be avail-able for the whole sample period. However, Nejat Seyhun shared with us hismonthly series, which spans 1975 to 1994, on the fraction of public firms withnet insider buying (as plotted in Seyhun (1998, p. 117)). Lakonishok and Lee(2001) study a similar series. We average Seyhun's series across months toobtain an annual series. Over the overlapping 20-year period, insider buyinghas a significant negative correlation with both the raw and orthogonalized

sentiment indexes, and also con-elates with the six underlying components asexpected. I


15/37

Investor Sentiment and the Cross-Section of Stock Returns 1659

a ffl o ^"a!afla

tha

0)" Dc0)u

isu

e

aop

-oh

3-a 3

aev


16/37

1660 The Journal of FinancePanel A. Closed-end fund discount % Panel B. Turnover

IM I IMT 1971Panel D. Average first-day return

1M2 19B7 197! 1977 ISSI 1917

Panel E. Equity share in new issues Panel F. Dividend premium

Panel E. Sentiment index (SENTIMENT)

Figur e 1. Investor sent iment , 196^2001. The first panel shows the year-end, value-weightedaverage discount on closed-end mutual funds. The data on prices and net asset values (NAVs) comefrom Neal and Wheatley (1998) for 1962 through 1993, CDAAViesenberger for 1994 through 1998.and tum-of-the-year issues of the Wall Street Journal for 1999 through 2001. The second panelshows detrended log turnover. Turnover is the ratio of reported share volume to average shareslisted from the NYSE Fact Book. We detrend using the past 5-year average. The third panel showsthe annual number of initial public offerings. The fourth panel shows the average annual first-dayreturns of initial public ofFerings. Both series come from Jay Ritter, updating data analyzed inIhbotson, Sindelar, and Ritter (1994). The fifth panel shows gross annual equity issuance dividedby gross annual equity plus debt issuance from Baker and Wurgler (2000). The sixth panel showsthe year-end log ratio of the value-weighted average market-to-book ratios of payers and nonpayersfrom Baker and Wurgler (2004). The solid line (left axis) is raw data. We regress each measure on thegrowth in industrial production, the growth in durable, nondurable, and services consumption, thegrowth in employment, and a flag for NBER recessions. The dashed line (right axis) is the residualsfrom this regression. The .solid (dashed) line in the final panel is a first principal component index ofthe six raw (orthogonalized) measures. Both are standardized to have zero mean and unit variance.In the index, turnover, the average annual first-day return, and the dividend premium are lagged


17/37

Investor Sentiment and the Cross-Section of Stock Returns 1661IV. Empirical Tests

A. SortsTable III looks for conditional characteristics effects in a simple, nonpara-

metric way. We place each monthly return ohservation into a bin according tothe decile rank that a characteristic takes at the beginning of that month, andthen according to the level of SENTIMEN T^ at the end of the previous calen-da r year. To keep th e m ean ing of the deciles similar over time, we define thembased on NYSE firms. The trade-off is tha t th ere is not a uniform distrib utio n offirms across bins in any given month. We compute the equal-weighted averagemonthly retum for each bin and look for patterns. In particular, we identifytime-series changes in cross-sectional effects from the conditional difference ofaverage re tu rn s across deciles.The first rows of Table III show the effect of size, as measured by ME, con-ditional on sentiment. These rows reveal that the size effect of Banz (1981)appears in low sentiment periods only. Specifically, Table III shows that whenSENTIMENT^ is negative, returns average 2.37% per month for the bottomME decile and 0.92 for the top decile. A similar pattern is apparent when con-ditioning on CEFD (not reported ). A link betw een the size effect an d closed-endfund discounts is also noted by Sw am inath an (1996). This patt ern is consistentwith some long-known results. Namely, the size effect is essentially a Januaryeffect (Keim (1983), Blume and Stambaugh (1983)), and the January effect, intu rn , is stron ger after a period of low retu rn s (Rein ganu m (1983)), which is alsowh en sen tim en t is likely to be low.As an aside, note th at the av erage r etu rn s across the first two rows of Table IIIil lustrate that subsequent retums tend to be higher, across most of the cross-section, when sentiment is low. This is consistent with prior results that theequity share and tumover, for example, forecast market returns. More gen-erally, it supports our premise that sentiment has broad effects, and so theexistence of richer pa tte m s w ithin the cross-section is not su rprising .The conditional cross-sectional effect of Age is striking. In general, in-vestors appear to demand young stocks when SENTIMENT^ is positive andprefer older stocks when sentiment is negative. For example, when senti-

ment is pessimistic, top-decile Age firms return 0.54% per month less t hanbottom-decile Age firms. However, they retum 0.85% more when sent imentis optimistic. When sentiment is positive, the effect is concentrated in thevery youngest stocks, which are recent IPOs; when it is negative, the con-trast is between the bottom and top several deciles of age. Overall, there isa nearly monotonic effect in the conditional difference of returns. This re-sult is intriguing because Age has no unconditional effect.^" The strong con-ditional effects, of opposite sign, average out across high and low sentimentperiods. .'"This conclusion is in seeming contrast to Barry and Brown's (1984J evidence of an uncondi-tional negative period-of-listing effect; however, their sample excludes stocks listed for fewer than


18/37

1662 The Journal of Finance

^ 1

= s s -r

assi4 - 2 -S

2--.

a S2 E

S ^

p [ > c q p ^ n 0 5 D O o q q c pd d d d d d d d - n d d d

o o oM ID ^ ^ C3 O ^ ^ C^ ^ IO IO OO CO IO N ^ 00 ^ IO t~ ^ 00 rH CC ^ IO OJ "^OS O ^ C^ *H c^ " f 1.H Ol ^ ^ *' ^^ O O ^ lO CC ^ O ^ "^ ' ^ O IO lO ^ ^ 'i"

'-' "-f O O *-< O

O O -H O o o o

O -H O

o o oto (D O O5 O) O V30 W Tl; C^ lO Ci C-;o d c - i - H - ^ - - d d - - ' r^ .-* i n d COd q- l or.; d c-l1 ^ 30. J lOd "1 '^ d

l O O J I O T f C ^ C O i - l t D l O O m MM I O N ^ m o q o M p C C W Q Od e 4 s - ) ' - ; - H - H d d - J dI I I

I ^ ^ ^ ^ n j { ^ i ^ ^


19/37

Investor Sentiment and the Cross-Section of Stock Returns 1663Puiel B, Age PMKlC.ToulfUtk

PMMI E Dividends Pmet F. PPE/A

ftmetG-RD/A

ii

Figure 2. Two-way sorts: Future retum s by sentiment index and firm characteristics,1963-2001. For each month, we form 10 portfolios according to the NYSE breakpoints of firm size(.ME), age, total risk, earnings-book ratio for profitable firms (E/BE), dividend-book ratio for payers(D/BEl fixed assets (PPE/A), research and development (RD/A), hook-to-market ratio (BE/ME),external finance over assets (EF/A), and sa les growth IGS), We also calculate portfolio re tur ns forunprofitahle. nonpaying, zero-PP&E, and zero-R&D firms. The solid bars are r et um s following pos-itive SENTIMENT^ periods, and the clear bars are returns following negative sentiment periods.The dashed line is the average across both periods and the solid line is the difference, SENTI-MENT-^ is positive for 1968-1970. 1972. 1979-1987, 1994, 1996-1997, and 1999-2001 (returnsend in 20 01, so th e las t value used is 2000).

The next rows of Table III indicate that the cross-sectional effect of retumvolatility is conditional on sentiment in the hypothesized manner. In particular,high Sigma stocks appear to be out of favor when sentiment is low, as they earnreturns of 2.41% per month over the next year. However, just as with Age,the cross-sectional effect of Sigma fully reverses in low sentiment conditions.Loosely speaking, when sentim ent is high, "riskier" stocks earn lower retums.When sentiment is low, they earn higher retu ms. A natu ral interpretation isthat highly volatile stocks are, like young stocks, relatively hard to value andrelatively hard to arbitrage, making them especially prone to fluctuations insentiment.Figure 2 shows the results of Table III graphically. Panel C, for example,shows the unconditional average m onthly re turns across Sigma deciles (dashedline), which is essentially flat; the average monthly retum in high sentiment


20/37

1664 The Journal of Financedeciles; and the difference in conditional returns (solid line). The solid linesummarizes the difference in the relationship between Sigma and future re-turns across the two regimes and clearly illustrates t hat the future retu rns onhigh Sigma stocks are more sensitive to sentiment.The next rows examine profitability and dividends. For average investors,perhaps the most salient comparisons are simply those between profitable andunprofitable (E < 0) firms and payers and nonpayers (D 0). These contrastsare in the extreme right column, where we average returns across profitable(paying) firms and compare them to unprofitable (nonpaying) firms. Thesecharacteristics again display intriguing conditional sign-flip patterns. Whensentiment is positive, monthly retu rns over the next year are 0.61% higher onprofitahle than unprofitable firms and 0.75% higher on payers than nonpayers.When it is negative, however, re turn s are 0.95%' per month lower on profitablefirms and 0.89% lower on payers. The left column shows that these patternsare driven mostly by conditional variation in the returns of unprofitable andnonpaying firms, although there are also some differences across levels of div-idend payments and profitability. Again, this is consistent with unprofitable,nonpaying firms being generally harder both to value and to arbitrage, thusexposing them more to sentiment fluctuations.

The next two rows look at asset tangibility characteristics under the notiontha t firms with less tangible assets may bemore difficult to value. The pa tternshere are not so strong, but there is a suggestion th at firms w ith more intangibleassets, as m easured by less PPE/A, are m ore sensitive to fluctuations in senti-ment. (This pattern is only apparent within firms that report positive PPE/A.)The clearest pattern in RD/A is a modest unconditional effect in which higherRD/A firms earn higher returns .The remaining variablesbook-to-market, external finance, and salesgrowthalso display intriguing patterns. Most simply, running across rows,one can see that each of them has some unconditional explanatory power. Fu-ture retu rns are generally higher for high BE/ME stocks, low EF/A stocks, andlow GS decile stocks. The EF/A result is reminiscent of Loughran and Ritter(1995) and Spiess and Affleck-Graves (1995, 1999), while the GS result is sug-gested in Lakonishok, Shleifer, and Vishny (1994).A closer look reveals that after controlling for these unconditional effects, aconditional pattern emerges. Specifically, there is a U-shaped pattern in theconditional difference. Consider the GS variable. The difference in re turn s onbottom-decile GS firms is 1.79% per m onth. Forfifth-decilefirms, he differ-ence is only 0.26% per month. But for tenth-decile firms, the difference isagain large, -1.64% per month. U-shaped patterns also appear in the condi-tional difference row for BE/ME and EF/A. The solid lines in Panels H-J ofFigure 2 show these "frowns" graphically. The figure illustrates why one mustcontrol for the strong unconditional effects in these variables in order to see theconditional effects.Thus, in all three of these growth and d istress variables, firms with extremevalues react more to sentiment than firms with middle values. What does


21/37

Investor Sentiment and the Cross-Section of Stock Returns 1665the U reflect? It reflects the multidimensional nature of the growth and dis-tress variables. Consider GS. High-GS firms include high-flying growth firms,low-GS firms are often distressed firms with shrinking sales, and middle-GSfirm s are steady, slow-growthfirms.Thus, relative tofirms n the middle deciles,firms with extreme va lues of GS are harde r to value, and perhaps to arbitrage,and thus may be more sensitive to sentiment. Put differently, firms with ex-treme values of GS are likely to seem riskier, in a salient sense, than firmsin the middle. The same explanation may help to explain the U-shaped pat-terns in the conditional difference row of EFIA and BEIME. There again, lowEFIAfirmsand high BEIMEfirms nclude distressed firms, high EFIA and lowBEIME firms include high-flyers, and the middle deciles tend to be populatedby the most "stable" firms.In unreported re sults, we sort retum s not just on positive and negative valuesof SENTIMENT^ but also on > 1 and < 1 standard deviation values. Not sur-prisingly, conditioning on more extreme values of sentiment leads to strongerresults. We take more formal account of the continuous natu re of the sentimentindexes in the next subsection. Also, for brevity, we omit sorts on SENTIMENT(the nonorthogonalized version), which give similar results. We present resultsfor both indexes in the next section. Finally, we have also sorted returns onpositive and negative SENTIMENT^, where positive and negative are definedrelative to a 10-year average. By requiring a 10-year history of sentiment, oneloses a little more than one-quarter of the sample. The resu lts are qualitativelyidentical to those in Table III, although slightly weaker except for Age, whichis slightly stronger.

f

B. Predictive Regressions for Long-Short PortfoliosAnother way to look for conditional characteristics effects is to use sentim entto forecast equal-weighted portfolios that are long on stocks with high values ofa characteristic and short on stocks with low values. Above we see that the aver-age payer, for example, earn s higher re tum s than the average nonpayer whensentiment is high, so sentiment seems likely to forecast a long-short portfolioformed on dividend payment. But a regression approach allows us to conduct

formal significance tests, incorporate the continuous nature of the sentimentindexes, and determine which characteristics have conditional effects that aredistinct from well-known unconditional effects.Table IV starts by plotting the average monthly r etu m s on various long-shortportfolios over time. The first several rows show that, not surprisingly, long-short portfolios formed on size (SMB), age, volatility, profitability, dividendpayment, and (to a lesser extent) tangibility are typically highly correlated.Thus, a good question, which we address in subsequent tables, is whether theresults from the sorts are all part of the same pa ttem or are somewhat distinct.This question is also relevant given that our portfolios a re equal-weighted. Bycontrolling for SMB in portfolio forecasting regressions, we can examine theextent to which the conditional predictability pa tterns are independent of size.


22/37


c V.

5 -fl ^ S

O

a. .2r

8

d dI 1

oo

ooI-i

d

t oodSj id

d16t od

COt od1Sato

O CO C- to CO


23/37

Investor Sentiment and the Cross-Section of Stock Returns 1667In the last several rows of Table IV, we break the growth and distress vari-ables into "high m inus medium " and "m edium m inus low" portfolios. In th e caseof th e GS varia ble, for examp le, the se portfolios are h ighly negatively correlatedwith each other, at - 0 . 6 3 , indicating th at h igh and low GS firms actually move

together relative to middle GS firms. Likewise, the correlation between "highminus medium" and "medium minus low" EFIA is -0.60. Thus, simple "highm inus low" analy ses of these variab les would om it crucial aspects of the cross-section.The question is whether sentiment can predict the various long-short port-folios analyzed in Table IV. We run regressions of the type^^

Rxu=mgb.t - Rx,,=L,>w,t =c + dSENTIMEN Tt^i -\- ,,. (4)The dependent variable is the monthly retum on a long-short portfolio, suchas SMB, and the monthly retums from January through December of t areregressed on the sentiment index that prevailed at the end of the prior year.We also distinguish novel predictability effects from well-known comovementusing the m ultivariate regression

Rxu^High.t - Rxu=Low.t = c -F dSENTIMEN Tt-x + fiRMK Tt -H sSMBt-F hHMLt + mUMDt -\- uu. (5)

The variable RMRF is the excess retum of the value-weighted market overthe risk-free rate. The variable UMD is the return on high-momentum stocksminus the retum on low-momentum stocks, where momentum is measuredover m onths t- 1 2 , -2 ] . As described in Fama and French (1993), SMB is theretu rn on portfolios of small and big ME stocks that is sep ara te from ret ur ns onHML, where HML is cons tructed to isolate the difference betw een h igh an d lowBEIME portfolios. 2 We exclude SM B and HML from the right side when theyare the portfolios being forecast. Standard errors are bootstrapped to correctfor the bias induced if the autoco rrelated se ntim ent index has innovations th atare correlated with innovations in portfolio retums, as in Stambaugh (1999).Table V shows the results. The results provide formal support to our pre-liminary impressions from the sorts. In particular, the first panel shows thatwhen sentiment is high, returns on small, young, and high volatility firms arerelatively low over the coming year. The coefficient on sentiment diminishesonce we control for RMR F, SMB, HM L, and UMD, but in most cases the signif-icance of the predictive effect does not depend on including or excluding thesecontrols. In term s of m ag nitu des , the coefficient for predicting SMB, for exam-ple, indicates that a one-unit increase in sentiment (which equals a one-SDincrease, because the indexes are standardized) is associated with a -0.40%lower monthly retum on the small minus large portfolio.

" Intuitively, in terms of equation (1), this a mo unts to a regression of (bi AX -|- h-^T,., AX) on sen-timen t proxies T, ,, where AX is the difference between "high" and "low" levels of a c hara cteristic.


24/37

1668 The Journal o f FinanceTable V also shows th at the coefficients on SENTIMENT and SENTIMENT^are very similar.Keep in mind that the coefficients on SENTIMENT ^ are essen-tially the same as one would find from regressing long-short portfolio returns

directly on a raw sentiment index and controls for contemporaneous macroeco-nomic conditionsthat is, regressing X on Z and using the residuals to predictY is equivalent to regressing 7 on X and Z. The similarity of the results onSENTIMENT and SENTIMENT^ thus suggests that macroeconomic condi-tions play a minor role.For profitability and dividend payment, we run regressions to predict thedifference between the profitable and paying portfolios and the unprofitableand nonpaying portfolios, respectively, because the sorts suggest that these arelikely to capture the main contrasts. The results show that sentiment indeedhas significant predictive power for these portfolios, with higher sentiment fore-

casting relatively higher returns on payers and profitable firms. The patternsare litt le affected by controlling for RMRF, SMB, HML, and UMD.As wefindwith the sorts, the tangibility characteristics do not exhibit strongconditional effects. Sentiment does have marginal predictive power for thePPE/A portfolio, with high sentiment associated with relatively low future re-turns on low PPE/A stocks, but this disappears after controlling for RMRF,SMB, HML, and UMD. The coefficients on the RD/A portfolio forecasts are notconsistent in sign or magnitude.Also as wefindwith the sorts, the "growth and distress" variables do not havesimple monotonic relationships with sentiment. Panel D shows tha t sentimentdoes not predict simple high minus low portfolios formed on any of BE/ME,EF/A, or GS. However, Panels E and F show that when the multidimensionalnature of these variables is incorporated, there is much stronger evidence ofpredictive power. We separate extreme growth opportunities effects from dis-tress effects by constructing High, Medium, and Low portfolios based on thetop three, middle four, and bottom three NYSE decile breakpoints, respectively.The results show that when sentiment is high, subsequent returns on bothlow and high sales growth firms a re low relative to re turn s on medium growthfirms. This illustrates the U-shaped pattern in Table III in a different way,and shows that it is statistically significant. An equally significant U-shapedpattern is apparent with externalfinance;when sentimen t is high, subsequen treturns on both low and high external finance firms are low relative to moretypical firms. In the case of BE/ME, however, although sentiment predicts thehigh minus medium and medium minus low portfolios with opposite signs,neither coefficient is reliably significant. This matches our inferences from thesorts, where we see that the U-shaped pa ttern in the conditional difference forBE/ME is somewhat weaker than for EF/A and GS.Equations (4) and (5) offer a simple framework in which to address somerobustness issues. To test whether the results are driven by an overall trend,we include a post-1982 dummy in the regressions, with no change in inferencesfrom those in the last column of Table V. Also, the resu lts are slightly stronger


25/37

Investor Sentim ent and the Cross-Sec tion of Stock Returns 1669

3IC .2

IIO o

c

no 5c 6a) aCO

I - - S3 1^ O (Ua 5 p in 5

'M * ^ f lj tJ C

Q - s gilli CO M Tj;o d d

" j ; lo asd d d

Tp iq qd d i-


26/37

1670 Journal of Financett"5.

3"a

Vi.2

a

6IB

B'a


27/37

Investor Sentiment and the Cross-Section of Stock Returns 1671indicates that tax-motivated trading and associated fluctuations in liquidityaround the tum of the year do not drive the main results. Further, our port-folios are equal-weighted. As mentioned previously, the purpose of this is th attheory predicts tha t small firm s will be most affected by sentiment, and hencevalue weighting will obscure the relevant patterns. Yet by sorting on charac-teristics tha t are correlated with size, as several of our characteristics are, andthen equal-weighting these characteristics portfolios, one worries that we arejust picking up the size effect once again. By controlling for SMB in portfolioforecasting regressions, we can see that several of the conditional predictabilitypa tte rns are distinguishable from size, though as theory predicts the predictivecoefficient is attenuated . Finally, while we omit the results for brevity, the sixindividual sentiment components generally predict the portfolio returns withthe expected sign. The number of IPOs and the closed-end fund discount offerthe best individual performance, followed by the equity share , turnover, the av-erage first-day retum, and the dividend premium. (Those results are reportedin the NBER working paper version of this paper.)

In summary, the regressions essentially confirm the significance of the pat-tems suggested in the sorts. When sentiment is high, future retums are rela-tively low for small firm s, the youngest firms , firm s with volatile stock returns,unprofitable firms, non-dividend-paying firms, high growth firms, and dis-tressed firms. And vice-versa. In general, the results support predictions thatsentiment has stronger effects on stocks that are hard to value and hard toarbitrage.C. A Brief Look at Earlier Data

Reliable accounting information, especially on the so rts offirmsmost affectedby sentiment, is not easy to obtain for the pre-Compustat era. Some of oursentiment proxies also are not available. However, using CRSP data, we canperform a reduced set of tests over a longer period. Specifically, we form asentiment index from 1935 to 2001 using the firs t principal component of CEFD ,S, and TURN,where TURN is lagged relative to the others, in the spirit ofequation (2).'^ We also orthogonalize these sentiment proxies with respect toconsumption growth variables and NBER recessions (industrial production isnot available over the full period) to form an index in the spirit of equation (3).We use these indexes to forecast the return on SMB and long-short portfoliosformed on Age, Sigma, and dividend payer status.The results are in Table VI. With the exception of the Age portfolio, forwhich the results are not significant, the results from the full 1935-2001 pe-riod and the "out-of-sample" 1935-1961 period are similar to those in more

""The closed-end fund discount is first available in 1933 from Neal and Wheatley (1998):"Wiesenberger's survey has published end-of-year fund prices and net asset values since 1943.Moreover, the first edition of the survey contains end-of-year data from 1933 to 1942." Turnoverand the equity share in new issues are available in earlier years. None of our inferences in PanelA of Table VI change w hen we use a longer samp le period an d a se ntim en t index based on these


28/37

1672 The Journal of FinanceTable VITime Series R egress ions of Portfolio Return s, 1935 to 2001

Regressions of long-short portfolio returns on lagged SENTIMENT, the market r isk premium(RMRF), the Fama-French factors {HML and SMB), and a mom entum factor {UMD).Rx,,^mgh.t -Rx,,=Lou.;t =c+dSENTlMENTt-\+pRMRFt+sSMBt+hHMLt+mUMDt + , .

The long-short portfolios are formed based on firm characteristics (X): firm size {ME), age, totalrisk {a), and dividends (D). High is defined as a firm in the top three NYSE deciles, and low isdefined as a firm in the bottom three NYSE deciles. The sentiment index is the first principalcomponent of the closed-end fund discount {CEFD), the equity sh are (S), and t he lag of detren dedlog turnove r (TURN). Average monthly returns are matched to SENTIMENT from the previousyear-end. SENTIMENT^ index is based on six sentiment proxies that have been orthogonalizedto growth in in dustrial production, the growth in durab le, nondurab le, and services consumption,the growth in employment, and a flag for NBER recessions; the com ponents of SENTIMENT arenot orthogonalized. The first and third sets of columns show univariate regression results, whilethe second and the fourth columns include RMRF, SMB, HML, an d UMD as control variables.SMB (HMD is not included as a control variable when SMB (HMD is the dependent variable.Bootstrapped p-values are in brackets.SENTIMENTt-i

SENTIMENT,.d pid)

Con trolling forRMRF, SMB,HML, UMDd p( d ) p ( d )

Controlling forRMRF, SMB,HML, UMD

d p(d)Panel A: 1935-2001

MEA^eaD

M EA genD

SM BHigh-LowHigh-Low> 0 - = 0

SM BHigh-LowHigh-Low> 0 - = 0

- 0 . 30.2- 1 . 00.9

- 0 . 3- 0 . 1- 0 , 80 .9

10.03]i0.18]10.00!10.00)

[0.05][0.41]10.01][0.01]

- 0 .20 .1- 0 . 40.5

Panel B:- 0 . 3- 0 . 1- 0 . 8

0 .9

[0.071[0.38[[0.00][0.01]1935-1961

(0.05][0.41][0.01][0.01]

- 0 . 30.2- 0 . 80 .7

- 0 . 1- 0 . 1- 0 . 40.5

[0.04]10.101[O.OOI[0 .001

[ 0 . 3 3 ][ 0 . 0 4 ]10.14]to. 101

- 0 . 20 .1- 0 . 40 .4

- 0 . 1- 0 . 1- 0 . 40.5

[0.07[[0.26[[0.001[O.Oll

[0.33][0.05[[0.16[[0.11]

recent data.^** One possibility for the insignificant results on the A ge portfoliois that we measure age as the number of months for which CRSP data areavailable. Anecdotal evidence suggests that in these early data, there are fewertruly "young" firms listing on the NYSE. In contrast, in recent years, manygenuinely young IPOs start trading on Nasdaq, so our way of measuring agemay be more meaningful.The longer time series make it possible to conduct an out-of-sample test. Inunreported results, we compare the in-sample reduction in root mean squared' For a more detailed look at ea rlier da ta, see Gruber (1966). He documents changes in the cross-sectional det erm inan ts of stock prices between 1951 and 1963, and argues that they are connected


29/37

Investor Sentiment and the Cross-Section of Stock Returns 1673error (RMSE) in Table VI to the reduction in RMSE that an investor mightsee using only past data. The results suggest that a substantial fraction of theportfolio predictability would have been "knowable" in advance. The exceptionsare the.^e portfolio and the SMB portfolio in the post-1980 period (for whichthe in-sample predictive power for SMB is also modest). A table is available onrequest.Together, the longer-sample results and the out-of-sample exercise rule outthe possibility tha t a spurious correlation is behind the main resu lts. The factthat there are at least several fluctuations in sentiment, and the fact that thecross-sectional pattems tend to work in the predicted directions, cast furtherdoubt on that notion.D. Systematic Risk

At face value, the conditional characteristics effects seem unlikely to becompensation for systematic risk. Among other considerations, the indexSENTIMENT-^ is orthogonalized to macroeconomic conditions; the pattemsmatch predictions about where sentiment should matter most; and the pattem sline up with anecdotal accounts of bubbles and crashes. Intuitively, the sys-tematic risk explanation requires that older, profitable, less volatile, dividend-paying firms often require higher returns than younger, unprofitable, morevolatile, nonpaying firms , and are recognized as riskier in the relevant sense bythe marginal investor. While this proposition already seems counterintuitive,we attem pt to rule it out more rigorously.

Systematic risk explanations come in two basic flavors. One is that the sys-tematic risks (beta loadings) of stocks with certain characteristics vary withthe sentiment proxies, despite our effort to isolate them from macroeconomicconditions. We investigate this directly in Table VII, where we ask whethersentiment coincides with time-variation in market betas in a way that couldat least qualitatively reconcile the earlier results with a conditional CAPM.Specifically, we predict re tu m s on the characteristics portfoliosRx,^igh,t - Rxu=Low,t = c + dSENTlMENTt-i

+ ^{e +fSENTIMENT,_^)RMRFt + u^. (6)The time-varying betas story predicts that the composite coefficient fif, reportedin Table VII, has the same sign as the estimates ofd inTable V. However, it tu rnsout that when the coefficient fif is significant, it is typically of the wrong sign.We obtain similar results when we replace RMRF by aggregate consumptiongrowth. A table is available upon request.The second systematic risk story keeps stocks' be tas fixed, but allows the riskpremium to vary with sentiment, which means tha t the difference in requiredretums between the high and low beta stocks varies in proportion. However,this story runs into trouble with the simple fact that the predicted effect ofseveral characteristics varies not jus t in magnitude over time, but also in sign.It would seem then that the bulk of the resu lts do not reflect compensation for


30/37

1674 The Journal of FinanceTable VIIConditional Market Betas, 1963-2001

Regressions of long-short portfolio returns on the market r isk premium (RMRF) and the marketrisk premium interacted with SENTIMENT.

The long-short portfolios are formed based on firm characteristics (X): firm size {ME), age, totalrisk (a), profitability {E), dividends (D), fixed assets (PPE), research and development (RD), book-to-market ratio (BE/ME), external finance over assets (EF/A), and sales growth decile (GS). Highis defined as a firm in the top three NYSE deciles, low is defined as a firm in the bottom threeNYSE deciles, and medium isdefined as a firm in themiddle four NYSE deciles. Monthly returnsare matched to SENTIMENT from the previous year-end. SENTIMENT^ index is based on sixsentiment proxies that have been orthogonalized togrowth in industrial production, tbe growth indurable, nondurable and services consumption, the growth in employment and a flag for NBERrecessions; the components of SENTIMENT are not orthogonalized. Heteroskedasticity-robustp-values are in brackets. A superscript "a" indicates a statistically significant (if that matchesthe sign of the retu rn predictability from Table V; "b" indicates a statistically significant pf thatdoes not match.

MEAgea

ED

PPE/ARD/A

BE/MEEF/AGS

BE/MEEF/AGS

BE/ME

SMBHigh-LowHigh-Low

SENTIMENTt^i0f

Panel A: Size, Age, and Risk-0 . 03^0.050.00

[0.48][0.19]10.98]Panel B; Profitability and Dividend Policy

>0-0- = 0

High-LowHigh-Low

-0 . 04- 0 . 0 1Panel C: Tangibility

-0 . 000.12''

[0.47][0.75]

tO.94][0.01]Panel D: Growth O pportunities and Distress

HM LHigh-LowHigh-Low]

Medium-LowHigh-MediiunHigh-Medium

High-Medium

-0 .10 ' '0.06''0.42[0.02][0.00][0.06]

Panel E : Growth Opportxmities-0 . 060.030.05''

Panel F: Distress

[0.05][0.131[0.02]

[0.00]

SENTIM0f

-0 . 02-0 . 070.03

- 0 . 0 0- 0 . 0 4

- 0 . 0 1

-0 . 12 ' '0.07''0.37

0.040.06''

ENT^_^t


31/37

Investor Sentiment and the Cross-Section of Stock Returns 1675E. Predictive Regressions for Earnings Announcement Returns

Our last test is whether there are conditional characteristics effects in theretums around earnings announcements. La Porta et al. (1997) find that lowbook-to-market stocks have lower average returns at earnings announcementsthan high book-to-market stocks, suggesting systematic errors in earnings ex-pectations. Likewise, if errors in earnings expectations account for some of ourresults , we might expect that the average earnings announcement retum onsm all, young, volatile, unprofitable, no npaying, extreme growth, and/or distres sstocks would tend to be inversely related to sentim ent.

This methodology, while appealing at first glance, has only limited power todetect how expectational errors affect our resu lts. Th at is , our results are drivenby the correlated correction of mispricing, bu t a firm's a nno unc em ent event re -turn picks up the expectational corrections that occur only to it alone, withinits own announcement window. An anecdote from Malkiel (1999) illustratesthe problem: "The music slowed drastically for the conglomerates on January19, 1968. On tha t day, the gran dda ddy of the cong lomerates, Litton Ind ustrie s,announced that earnings for the second quarter of that year would be sub-stantia lly less th an forecast the ann oun cem ent was greeted with disbeliefand shock. In the selling wave that followed, conglomerate stocks declined byroughly 40 p er ce n t. , ." (p. 67). So al though a study of announ cemen t event re-turns captures the corrective effect of Litton Industries' announcement on itsown stock, it picks up none of its broader effects, which appear to be importantto our main results. Nevertheless, an analysis of earnings announcements mayprovide a lower bound on the effect that sentiment-driven expectational errorshave on our results.We gath er quarterly ea rnings anno uncem ent d ates from the merged CRSP-Compustat file. These dates are available beginning in January 1971. Thequarterly earnings ann ouncem ent sam ple represents approximately 75% ofthe firm-quarters (firm-months) analyzed in the main tables, so coverage isfairly complete. For each firm-quarter observation, we compute the cumula-t ive abnormal r et u m over the value-weighted m arket index over trading days[ - 1 , +1] around the report date. We then construc t a qu arterly series of averageannouncement effects for each characteristic decile, and attempt to predict itwith the composite sentiment index, that is,

CARx,,=Deciie.t = c -^ dSENTIMENT^^ i + u,. (7)Table VIII reports the coefficient estimates for each characteristic decile usingthe orthogonalized sentiment index. The results for the raw index are verysimilar.

Pe rha ps th e m ost strikin g feature of Table VIII is th a t m ost coefficients arenegative, thus eamings announcement effects are in general lower followinghigh sentiment periods. A very crude comparison can be made between thecross-sectional patterns in Table VIII and those in Table III. In Table VIII, 12of the 104 coefficients are significant at the 5% level. In Table III, 9 of the 104


32/37


k- -J

C T-

^ Pc [


33/37

Investor Sentiment and the Cross-Section of Stock Returns 1677value. The intersection of the two tab les' "strong resu lts" is six cells, and thesigns of the effects are congruent in all cases.Overall, this suggests that some portion of the conditional characteristicseffects may reflect the correction of errors in earnings expectations. However,as noted above, this te st is not powerful and provides only a lower bound on thecontribution of expectational erro rs.

V. ConclusionIn classical finance theory, investor sentiment does not play any role in thecross-section of stock prices, realized retums, or expected retums. This paperchallenges tha t view. We use simple theoretical arguments, historical accountsof speculative episodes, and most importantly a set of novel empirical resultsto demonstrate that investor sentiment, broadly defined, has significant cross-sectional effects.Our main empirical finding is tha t the cross-section of future stock re tum sis conditional on beginning-of-period proxies for sentiment. The pattems arerich but intuitive. When sentiment is estimated to be high, stocks that areattractive to optimists and speculators and at the same time unattractive toarbitrageurs younger stocks, small stocks, unprofitable stocks, non-dividend-paying stocks, high volatility stocks, extreme growth stocks, and distressedstockstend to earn relatively low subsequent returns . Conditional on low sen-timent, however, these cross-sectional patte rns attenuate or completely reverse.

The most striking finding is that several firm characteristics that display nounconditional predictive power actually hide strong conditional patterns thatbecome visible only after conditioning on sentiment. We consider the classi-cal explanation that the results reflect compensation for systematic risks, butseveral aspects of the results are inconsistent with this explanation.The results suggest several avenues for future work. In corporate finance,a better understanding of sentiment may shed light on pattems in securityissuance and the supply of firm characteristics that seem to be conditionallyrelevant to share price. In asset pricing, the re sults suggest tha t descriptivelyaccurate models of prices and expected retu rns need to incorporate a prom inentrole for investor sentiment.

REFERENCESAghion, Philippe, and Jeremy Stein, 2004, Growth versus margins: The destahilizingconsequencesof giving the stock m arket w hat it wants, Working paper. H arvard University.Amihud. Yakov, and Haim Mendelson. 1986, Asset pricing and the bid-ask spread, Journal ofFinancial Economics 17, 223-249 .Asness, Clifford S., Jacques A. Friedman, Robert J. Krail, and John M. Liew, 2000, Style timing:

Value versus growth. Journal of Portfolio Man agement 26. 50-60,Baker. Malcolm, and Jeremy Stein, 2004. Market liquidity as a sentiment indicator. Journal ofFinancial Markets 7, 271-299,Baker, Malcolm, and Jeffrey Wurgier, 2000. The equity share in new issues and aggregate stockreturns. Journal of Finance 55, 2219-2257 .Baker, M alcolm, and Jeffrey Wu rgier. 2004, A catering the ory of dividen ds. Journal of Finance 59,


34/37

1678 , The Journal of FinanceBanz. Rolf, 1981. The relationship hetween retu rn and m arke t value of common stocks. Journal of

Financial Economics 9, 3-18.Ba rberis. Nicholas, and Andrei Shleifer, 2003, Style investing, Journal of Financial Econ omics 68,161-199.Barh eris, Nicholas, Andrei Shteifer, and Jeffrey W nrgler, 2005 , Como vement, Journal of FinancialEconomics 75, 283-318.Barry, Ch ristoph er B., and S tephen J. Brown, 1984, Differential information and the sm all firmeffect. Journal of Financial Econom ics 13, 283-29 4.Benv eniste, Lawrence M., Alexander P. Ljungqvist, William J. Wilhelm, Jr., and Xiaoyun Yu, 2003,Evidence of information spillovers in the production of investment bimking services. Journalof Finance 58, 577-608 .Blume, Marshall, and Robert Stam baug h. 1983, Biases in computed retu rns: An application to th esize effect. Journal of Financial Economics 12, 387-404 .Brown, Gregory W., and Michael T. ClifF, 2004, Investor sentiment and the near-term stock market.Journal of Empirical Finance 11, 1-27.Brown, John Dennis, 1991, 101 Years on Wall Street (Prentice Hall, Englewood Cliffs, NJ).Brown, Stephen J., William N. Goetzmann, Takato Hiraki, Noriyoshi Shiraishi, and MasahiroWatanabe, 2003, Investor sentiment in Japanese and U.S, daily mutual fund flows. Workingpaper, Yale University.Brunn ermeier, Mark us, and Lasse Pedersen, 2005, Predatory trading. Journal of Finance 6 0 , 1 8 2 5 -1863.Cam pbell, John Y . and Joh n H. Coch rane, 2000, Explaining the poor performance of consumption-based asset pricing models. Journal of Finance 55, 2863-2878.Ch an, Louis K.C., Jaso n K arceski, and Josef Lakon ishok. 2000, New parad igm or sam e old hype inequity investing? Financial Analysts Journal 56, 23-3 6.Cochrane, John H., 2003, Stocks as money: Convenience yield and the tech-stock bubb le, in W ilhamC. H un ter. George G. Kaufinan, and M ichael Pomerleano, eds.: Asset Price Bubbles (MM Press.

Cambridge).D'Avolio, Gene, 2002, T he m arke t for borrowing stock. Journal of Financial Econom ics 66,271-306.Daniel. Kent, and Sheridan Titman, 1997, Evidence on the characteristics of cross-sectional vari-ation in stock returns. Journal of Finance 46, 1739-1764.De Long, J. Bradford, Andrei Shleifer, Law rence H. Sum me rs, and Rob ert Waldm ann, 1990, Noisetrader risk in financial marke ts, Journal of Political Econom y 98, 703-738.Dreman, David, 1979, Contrarian Investment Strategy, Dreman Contrarian Group.Duffle, Darrell, Nicolae Garleanu, and Lasse H. Pedersen, 2002, Securities lending, shorting, andpricing, Journal of Financial Econ omics 66, 307-339.Elton, Edwin J., Martin J. Grub er, an d Jeffrey A. Busse, 1998, Do investors care abou t sen timent?.Journal of Business 71 , 477500.Fama, Eugene F.. and Kenn eth R. French , 1992, The cross-section of expected stock re tu rn s, Journalof Finance 4 7. 4 2 7 ^ 6 5 .Fama, Eugene F., and Kenneth R. French, 1993, Common risk factors in the returns on stocks andbonds, Journal of Financial Econ omics 33 , 3-56.Fama, Eugene F , and Kenn eth R. French, 2001, Disappearing dividends: Chang ing firm charac-teristics or lower propensity to pay? Journal of Financial Economics 60, 3-44.Geczy, Ch ristophe r C , David K. Mu sto, and Adam V. Reed, 2002, Stocks are special too: An a naly sisof the equity tending ma rket. Journal of Financial Economics 66, 241-269.Gomes, Joao , Leonid Ko gan, and Lu Zha ng, 2003, Equilibrium cross section of returns. Journal ofPolitical Econom y 111, 693-732.Graham, Benjamin, 1973, The Intelligent Investor. 4th ed. (Harper & Row, Am sterdam).Greenwood, Robin, and Na tha n Sosn er, 2003, Trade and the comovem ent of stock retu rn s: Evidencefrom Ja pan , W orking paper. Harvard University.

Cruber, Martin J., 1966, The determinants of common stock prices, Ph.D. dissertation, ColumbiaUniversity.


35/37

Investor Sentiment and the Cross-Section of Stock Returns 1679Ibbotson, Roger, and Jeffrey F, Jaffe. 1975, 'Hot issue' ma rkets . Journal of Finance 30, 1027-1042.Ibbotson, Roger. Jody Sindelar, and Jay Ritter, 1994, The market 's problems with the pricing ofinitial public offerings. Journal of Applied Corporate Finance 7. 66-74,Jones, Charles, 2001, A century of stock market liquidity and trading costs. Working paper,

Columbia University,Jones. Charles, and Owen Lamont, 2002, Short sale constraints and stock returns. Journal ofFinancial Economics 66. 207-239.Keim. Donald. 1983, Size-related anomalies and stock retum seasonality: Further empirical evi-dence. Journal of Financial Econo mics 12. 13-32.Kothari, S. P., and Jay Shanken. 1997. Book-to-market, dividend yield, and expected market re-turns: A t ime-series an alysis. Journal of Financial Economics 44. 169-203.Kindleberger, Charles. 2001 . Man ian, Panics, and Crashes (Wiley, New York),La Porta. Rafael, Josef Lakonishok, Andrei Shleifer, and Robert W, Vishny, 1997. Good news for

value stocks: Fu rther evidence on market efficiency. Journal of Finance 52, 859-874.Lakonishok, Josef, and Inmoo Lee. 2001, Are insider trades informative?. Review of FinancialStudies 14 , 79-111 .Lakonishok. Josef. Andrei Sh leifer. and Robert W. Vishny, 1994, Contrarian investment, extrapo-lation, and risk. Journal of Finance 49, 1541-1578,Lam ont. Owen A,, and Richard H, Thaler, 2003, Can the market add and subtract? Mispricing andtech stock carve -outs. Journal of Political Econom y 111, 227-26 8.Lancaster, Kelvin J.. 1966. A new approach to consumer theory. Journal of Political Economy 74.132-157,Lancaster, Kelvin J., 1971, Consumer Demand: A New Approach (Columbia UP, New York).Lee. Ch arles, Andrei Shleifer, and Richard H. Thaler, 1991. investor sentiment and the closed-endfund puzzle. Journal of Finance 46, 75-109.Lettau. Martin, and Sydney Ludvigson, 200 1, Resurrecting the (C)CAPM: A cross-sectional testwhen risk premia are t ime-varying. Journal of Political Econom y 109, 1238-1287.Ljungqvist. Alexander, and William Wilhehn, 2003, IPO pricing in the dot-com bub hie. Journal ofFinance 58,723-752.Lougbran. Tim, and Jay Ritter, 1995, The new issues puzzle. Journa! of Finance 50, 23-51.Lowry. Michelle, and G, William Schw ert, 2002. IPO ma rket cycles: Bubbles or sequ ential le arnin g?.Journal of Finance 57.1171-1200.Malkiei. Burton, 1990, A Random Walk Down W all Street (W.W. Norton. New York),Malkiel, B urton, 1999, A Random Walk Down

Documents

Investor Sentiment