8/10/2019 easton1999.pdf

1/15

1999 American Accounting ^s oc iat ionAccounting HorizonsVol, 3 No, 4December 1999pp- 399-412

COMMENTARY

Peter D Easton

Peter D Easton is a Professor at Ohio State University and aProfessorial ellow at the University of Melbourne

Security Returns and the ValueRelevance of ccounting Data

INTRO U TION

Empirical market-based accounting research seeks evidence of the value relevanceof accounting data via ana iysis of the relation between these data and various m arketvariables. The decade of the 1990s witnessed an increasing use of price per share andmarket rate of retu rn as the market variables of interes t. The focus of this commenttiryis on regressions of price per share on the levels of various financial statement data(referred to herein as price-levels regressions ) and regressions of re turns on changesin these financial statem ent variables (referred to herein as return s regressions ).The main points in the comm entary Eire: Studies that use returns of the fiscal period as the market metric provide evidence

regarding the role of accounting data as a summary of events that have affectedfirms over the reporting period. In contrast, studies of the m arke t response du ringa very short interval around the time of the announcement of the accounting dataexamine the role of these data in providing information to investors about eventsthat may affect their perceptions of the firm.

An argument for returns as the market m etric: Theoretical models that form a foundation for the use of security prices as the mar-

ket metric also provide a foundation for using rate of return as the market metric. Returns regressions may be used as the basis for tests of hypotheses regarding

the timeliness of the reporting of value changes in fin ncied statements.'

^ The concept of timeline ss oUhe accounting summary is the exte nt to which the value change as reportedin the financial statements is contemporaneous with the change in market value. This difFers from theconcept of timeliness of accounting information that relates to the investors' use of the information insetting prices.

These comments are based on my presentation at the American Accounting Association Doctoral Consor-tium in Lake Tahoe in Jun e 1999 and a t the A merican Accounting Association/Taiwan A ccounting Associa-tion First G lobalization Conference in Taipei in July 1999 1 tha nk Keji Chen Kirk Philipich Terry Shevlin

8/10/2019 easton1999.pdf

2/15

^^^ Accounting Horizons December 1999

Regressions th at use price as the dependent variable suffer from potentially serious scale problems. An obvious means of overcoming the scale or per share efTectis to rely on the returns specification.

Insights from the scatter plot o f returns and ea rnings: Recent increases in the power and availability of software fac ilitate detailed inspection of the huge da ta sets often encountered in market-based accounting research providing evidence beyond that from regression analyses.

A scatter-plot of returns and earnings shows distinct nonhnearities in the relationbetween these variables and provides insights regarding recent studies of these nonlinearities (including Hayn 1995; Basu 1997; Burgstahler and Dichev 1997).

CCOUNTING D T S SUMM RY OF EVEN TS

Ball and Brown (1968) and Beaver (1968) show tha t net income is value relevant in

the following senses. Beaver (1968) shows tha t the market reac ts with increased trading volume and increased price variability in the week of the earnings announcem entBall and Brown (1968) show that earnings increases (decreases) are associated (on average) with positive (negative) abnormal returns over the 12 months prior to the earnings announcementin short, the unexpected component of earn ings tends to have thsame sign as unexpected price changes. These seminal stud ies and much of the work inthe subsequent two decades emphasized the relation between the new information inearnings and e ither the marke t reaction to this information (as in Beaver 1968) or theassociation of this new information and the unexpected or abnormal component of returns (as in Ball and Brown 1968). This information perspective, which has continuedto influence research methods in the last decade, may be described as an investor, auser, or a finance perspective that views accounting as a source of information for use(either ac tual or potential) in investment decisions.

Studies such as Beaver (1968) that focus on the marke t response at the date of theannouncement of the accounting data (that is, the market metric is measured over arelatively short intervala few days or a few hours) examine the role of accountingdata in providing information to the market about events that may affect investorsperceptions of the firm ^ In contrast, the association between unexpected or abnormalreturns and unexpected earnings in Ball and Brown (1968) provides evidence of therole of accounting as a summary of the unexpected events th at have affected the firmover the 12-month period prior to the earnings announcement.

More recent studies have tended to move away from the information-content per-spective and to focus more clearly on the view tha t fmancial statem ents a re a sum maryof the events th at have affected the firm over the fiscal period for which the report hasbeen prepared. This perspective is similar to that in much earher studies such as Patonand Littleton (1940) and Edwards and Bell (1961). Empirical studies that adopt thisperspective require a benchmark against which to evaluate the effectiveness of theaccounting summ ary. Since the events th at have affected the firm over the fiscal periodare captured in change in firm value (or returns), market returns are the obviousbenchmark.

Easton et al. (1992) argue that there are two reasons why earnings will not be aperfect summ ary of events of the corresponding return interval: 1) value-relevant events- Other prototypical representations of the information-content perspective are Patell and Wolfson

8/10/2019 easton1999.pdf

3/15

Security Returns and the Value Relevance of ccounting Data 401

observed by the m ark et (a nd therefo re captu red in re tur ns ) in a prior period may affectaccounting earnings of the current period, and (2) value-relevant events observed by themarket in the current period may not be repori;ed in accounting earnings of the currentperiod. In short, acco unting rep orts th e effects of economic even ts with a lag. Easton et al.(1999) show tha t in a return-e arnin gs regression, the omitted variable th at arises due toaccounting recognition lag is perfectly negatively correlated with the included variable(accounting earnings). The effect of this omitted variable is to bias the estimate of theearnin gs coefficient (and, implicitly, th e regression R toward zeroceteris pa ribus a lowerearnings coefficient and/or a lower R suggests th at earn ings are a poorer sum m ary of theeven ts th at hav e affected re tu rn s of th e fiscal period.^ The effect of thi s lag is tha t th e R''from a regression of re tu rn s on eeirnings will be less tha n 1.

At any point in t ime, price reflects all returns (that is, changes in market value)since the firm came into existence, while book value represents all accounting mea-sur es of change in value (ea rnin gs) du rin g this period. Book value will reflect th e cum u-lative effect of accoun ting rep orti ng lagsome of the va lue-rele van t eve nts observed bythe market (and therefore captured in returns) in early years will be included in ac-counting earnings of later years, hut some will remain unrecorded in book value. Theeffect of this accounting reporting lag in the price-levels regression is similar to theeffect in the returns regressionthe R will be less tha n 1.

Most of the following discussion adopts the perspective that accounting data maybe viewed as a summary of the events that have affected the firm although some com-m ents will be m ade regard ing the information-content perspective.

S I M P L E M O D E L

As a starting point, consider a firm with two types of assetsthose for which wewould (as a first approximation or best guess ) use book value as the basis for determinin gmarket value and those for which we would use earnings as the basis for determiningvalue. The following simple model provides the valuation of this firm.' '

Consider the following rationale for using book value as the basis for determiningmarket value. Today's market value (per share) represents entit lements to a flow, orseries, of expected dividends. Likewise, the accounting book value (per share) repre-sents the accou ntant 's mea sure of the firm's resources and comm itments th at togetherwill determine the expected dividend flow per share. Assume (initially) that the bookvalue p er sh are of firm j at tim e t B. ) perfectly re cords the valu e of a sha re in th e sense

th at i t is equal to the m ark et value of the sha re (P.^). Th at is;

We can also rationalize a relation between m ark et value per share and earnin gs persh ar e. If firm i's ea rni ng s for period t rep res en t ear nin gs of each future period in

' T his effect of accou nting recognition lag is simila r to the effect of stale earnings in Kothari and Zimm ermanH995). However, since the emphasis in Kothari and Zimmerman (1995) is on obtaining an unbiasedestimate of the coefficients relating prices to earnings and returns to earnings, the bias in the returns

regression is seen as misspecification. In contrast, when the research emphasis is on the validity of earnings as a summary variable, the effect of this omitted variable (that is, the events that are not summa-rized by earnings) is precisely the focus of the investigation.

8/10/2019 easton1999.pdf

4/15

^^^ Accounting Horizons /December 1999

perpetuity, and these earnings are paid out as dividends in the period in which theyaccrue, then , cum-dividend price per sha re (P., d.,) is a multiple of earnings per share(X^). T hat is:

P, - fd , = ( l + r X (2)where r. is the expected ra te of retu rn on share s of firm i.

Now assume tha t a proportion k of the asse ts of the firm (human capital and otheintangibles may be examples of these assets) may be valued using model (2) and aproportion 1 - k) of the firm s assets may be valued using model (1) (property, plantand equipment m ay be examples of such asse ts). Firm value may be determined as:

Pn: = ( l - k ) B , + k [ ( l + f J ) X , - d J . (3)

Ohlson (1995) provides a rigorous foundation for equation (3) in a dynam ic uncertain environment that relies on the clean surplus assumption and the Miller andModigliani (1961) propositions. In this framework, variables other than book valueearnings and dividends play a role in valuation. These variables are captured by thscalar v^ and the valuation relation is:

P.t = 1 - k )B , -f- k [ 1 -f f;) X^ - d J av^^. (3a)

Ohlson s (1995) work is cited as the theoretical foundation for many recent studies othe relation between price, book value (and components of book value), and earning(and components of earnings). These studies are based on variations of the followinglinear (price-levels) regression:^

An important contribution of Ohlson s (1995) work is that it forms a framework forunderstanding the relation between prices and accounting da ta and a basis for interpreting estimates of the regression coefficients a^,, a,, and a^ For example, (a) the modelprovides a valuation role for other information and dividendsa nonzero intercept a^suggests that the average incremental explanatory power for prices (over book valueand earnings) is nonzero, b) the coefficient a^ on book value is negatively related to thepersistence in abnormal earnings so th at a higher coefficient on book value implies thatthese earn ings are less persistent, and , (c) the coefficient a^ on earnings is positivelyrelated to this persistence and negatively related to the expected rate of return.^ Notehowever, that Ohlson (1995) is only a starting point for understanding the relationbetween prices and accounting da ta. O ther theoretical papers (including, Feltham andOhlson 1995; Ohlson and Zhang 1998; Zhang 1999) provide critical additional insights.For example, Feltham and Ohlson (1995) form a framework for understanding the roleof conservative accounting, while Zhang (1999) shows, inter alia tha t with growth andconservative accounting, the coefficient on book value may, indeed, be negative.

Although Ohlson (1995) provides obvious motivation for price-levels regressionsthe model also provides motivation for returns regressions. Taking first differences in

The effect of trea ting dividends as part of the ran dom er ror term is unclea r. Hand a nd L and sm an (1998)address this issue in the context of a price-levels regression. The effect of other information, v , is ana-

8/10/2019 easton1999.pdf

5/15

Security Returns and the Value Relevance of ccounting D ata 403

equation (3a), invoking the clean surplus assumption, dividing through by beginning ofperiod price and rearrang ing terms yields the re turns relation:

ret^, = 1 - k ) [X ,/P^^J + kCl+O IAX^,/P.,.,] + k[d^,_,/P^,_J + oAv^ (5)

where ret is the ra te of return on investmen t in firm n for the fiscal period t - 1 to t.Equation (5) may be viewed as the theoretical basis of the re turns regression:

The important point from this discussion is tha t regressions based on equations (4and 6) are both motivated by the same theoretical foundation. The theory predicts tha ta^ = p = 0, a( = pj, and a = pg. This raises two issues. Are there any questions that cashould be addressed via an analysis of price levels (equation [4]) that can not/should not beanswered by an analysis of returns (equation [61)and vice versa? In view of the fact thatinferences fi om studies based on regression (4) often differ from inferences from studiesbased on regression (6), which of these two regression models is better specified?

EXPLAINING RETURNS PROVIDES EVIDENCE OF TIMELINESS

Since the focus of retu rns studies is on the events tha t have affected prices ovethe re turn interval, they address the question of the timeliness of the accounting summary: is the value change as reported in the financial statements contemporaneouwith the change in market value? Note that the return interval may be of any length(for example, a qu arter in Warfield and W ild [1992] to 10 years in Easton et al. [1992]).If the retu rn interval is short and the financial statemen ts report the change in valuwithin this short interval, the summary provided by the accounting data is particularly timely.

As an illustration of this point, consider a study that focuses on the valuation role ofan adjustment to book value (such as asset revaluations in the U.K. and Australia, ochanges in the liability for post-retirement benefits under Statement of Financial Accounting Standards No. 106 in the U.S.). The effect of these adjustments is to changthe alignment of the market s m easure of value and the accountant s record of value othe balance sheet at the time that the adjustment is recorded. A researcher could in-clude the amount of the adjustm ent in a price-levels regression and design tes ts baseon the significance and th e magnitude of the estimate of the coefficient on the adjustment. Alternatively, we may use a returns specification with tests based on the estimate

of the coefficient on changes in the balance shee t amounts (in our example, the chang

Equ ation (5) sugge sts th at lagged dividends should be an add itional explanatory variahle in regressio(6). For prac tical purp oses , hoth d.j i/P,i_, and, in p artic ular, the cross-sectional variation in this varia bleare suificiently small to have little effect on the estimates of the coefficients on earnings and eamingchanges. Omission of this variable may, however, lead to a positive intercept.

* The choice of cited stu dies (particu larly those for which I am an a utho r) in the rem ainin g sections reflecmy famiharity with the details therein. There are many other studies that could have, equally approprately, been used as illustrations.

For examp le, Ba rth an d Clinch (1998) regress price per sha re on book value per shar e, the balance in thasset revaluation reserve per share , earnings per sha re, and the per-share increment to the asset revaluation reserve. Indicatio ns of value relevance of asBet revalu ation s a re obtained via test s of significance the estimates of the coefficients on the balance in the asset revaluation reserve and on the increment tthis reserve. Sim ilarly, Cham bers et al. (1999) exam ine th e effect on th e R^ from a regression of price pshare on book value and earnin gs of adjusting book value by subtracting the book value of property plan

8/10/2019 easton1999.pdf

6/15

'^^'^ Accounting Horizons /December 1999

in the asset revaluat ion reserve or the change in the l iabi l i ty for post-ret i remenbenefits).^ The se latte r tes ts provide evidence beyond th a t provided by the price-levelregressions they provide evidence regarding the timeh ne ss of the adjustm ents.

T H E E F F E C T S O F S C A LEStatistical associations between price and the explanatory variables in price-levels

regressions such a s equ ation (4) may be a spuriou s effect of scale. In genera l, large (smallfirms w ill hav e a large (small) total m ark et valu e, large (small) book value and large (smallnet income. Additionally, many other variables for these large (small) firms will also belarge (small) s th at a regression of m ark et va lue on firm a ttrib ute s will lead to coefficientsth at may cap ture no mo re th an scale effects. Expre ssing all variab les on a per-s hare basiwill not overcome thi s scale effect. The a rgum en t is as follows.

M anagem ent h as discretion over the num ber of sha res outstanding. They may chooseto split their firm's stock, to offer stock dividends and/or to un de rta ke a reve rse stock splitThese sp lits could conceivably be used to change th e price of sh are s without changing theeconomic characteristics of the firm Arguably, therefore, the magnitude or scale of thedep end ent varia ble in price-levels regressions reflects no more tha n th e choice by m anag ement of the number of shares outstanding. This management choice will also affect thescale of the per-share measure of many firm attributes. It follows that a regression ofsha re price on the firm attr ibu tes w ill lead to coefficients th at ma y simply cap ture th e facttha t, at th e per-s har e level, all variables h ave the s am e scale and scale differs across firms

Eas ton (1998) shows th at in cross-sectional price-levels regre ssion s whe re price pe rshare of firm i is matched with book value per dollar of market value for firm j andearnings per dollar of market value for firm k, the regression statistics are strikinglysimilar to the s tatistic s fi-om a regression wh ere all of the se va riables are ta ke n fromthe sam e firm. Although this is not unequivocal evidence tha t the resu lts of price-levelsregres sions a re due only to scale effects, the evidence does sugg est th at caution shouldbe exercised w hen in terpr eting th e re sults of these regressions.

Since the ra te of re tu rn is a scale-fi-ee va riab le th a t ce teris paribus is not affected bythe m an ag er s' choice of nu m be r of sh are s outs tand ing, a n obvious me ans of overcomingthe scale or per-share effect is to rely on the ret ur ns specification. Furthe rm ore , Plosserand Schw ert (1978), Schw ert (1981) and Christie (1987) sugg est th at time-differencinga misspecified p rice-levels mo del can ge ne rate a well-specified model in th e differences.That is, the returns specification not only has the advantage that the scale effects areremoved , but the im plicit time-differencing also improv es the m ode ls' specification. Th us ,unless there are compelling reasons for focusing on the levels relation, the alternativere tu rn s specification should be d^ '

For example, Easton et al. (1993) and Barth and Clinch (1998) regress returns on price-deflated earningsthe price-deflated increment to the asset revaluation reserve, price-deflated change in earnings, and theprice-deflated change in the increment to the asset revaluation reserve. Indications of value relevance ofasset revaluations are obtained via tests of significance of the estimates of the coefficients on the incrementto the asset revaluation reserve and the change in this increment. Similarly, Chambers et al, (1999) exam-ine the effect on t he ^ from a regression of retu rns on deflated e arning s an d deflated chan ge in earnings ofadjusting earnings by adding back depreciation and subtracting capital expenditures.

Easton and Somm ers (1999) suggest m eans of overcoming the scale problem th at do not necessitate the useof retu rns models. The essence of the ir idea is tha t when estim ating levels relations such as regression (4),each fitted value is scaled by its own price. For example, rath er tha n m inimizing the m ean squared difler-

8/10/2019 easton1999.pdf

7/15

Security Returns and the Value Relevance of Accounting Data 6

A CLOSER LOOK AT THE RETUR NS EA RNINGS RELATION

The vast majority of price-levels studies and retu rns studies use the same (linear)relation to describe the entire sample of observations. ^ The m odeling of the regressionrelations in section three of this commentary suggests that observations should begrouped into subsam ples with sim ilar attr ibu tes . For example, the simple model (devel-oped rigorously by Ohlson [1995]) relying on the intuition that firms may (as a firstapproximation) be treated as if they have two types of assetsthose for which we woulduse book value as the basis for determining market value and those for which we woulduse earnings as the basis for determining valueleads to the prediction that for somefirms the weight on earnings will be high, while for others it will be low. Two notablestudies that group observations according to a priori expectations about the form andstrength of the relation between returns and earnings are Hayn 1995) and Basu (1997).I will briefly discuss these studies and then show how the points th at they have madevia careful economic reasoning are also very evident from simple plots of the da ta.

Hayn (1995) provides several reasons why both the estimate of the return-earningscoefficient and the regression ^ will be lower for firms reporting a loss than for firmsreporting profits. These reasons include: (1) because shareholders have a liquidationoption, losses are not expected to pe rpetuate, and (2) the (related) trans itory na ture oflosses will result in a lower coefficient estimate. In simple pooled cross-section andtime-series regressions of returns on earnings (deflated by beginning-of-period price),Hayn 1995) finds that th e ^ for 14,512 loss firms is 0.0 percent and the estimate of theslope coefficient is 0.01, while the R= for 61,366 firms that reported a profit is 16.9percent and the estim ate of the slope coefficient is 2.62.

Basu (1997) observes that the effect of conservatism in accounting is tha t bad newstends to be reported more quickly than good news. He uses returns of the fiscal year asan indicator of the net (bad vs. good) news. In simple pooled cross-section and time-series regressions of e rnings (deflated by beginning-of-period price) on returns Basufinds that the ^ for a sample of 17,790 firms with negative returns is 6.64 percent andthe estimate of the slope coefficient is 0.275, while R^for 25,531 firms with positivereturns is 2.09 percent and the estim ate of the slope coefficient is 0.059.

These two papers represent important recent developments in our understandingof the simple returns-earnings association. Consistent with the reasoning that moti-vates these papers, the results appear to show distinct nonlinearities in the returns-earnings and the earnings-returns relations. As far as I am aware, these are the first

papers to document the nonlinearity of the returns-earnings regression at zero earn-ings and the nonlinearity of the earnings-returns regression at zero returns. Interest-ingly, as I will attem pt to demonstrate, the ideas in these papers may have been devel-oped much earlier if we had paid more attention to the data.'^ In addition, detailedanalysis of the da ta provides insights that are not available from the regression analyses in these papers.

Fran cis and Schipper (1997) and Nwa eze (1998) are notable exceptions. Fran cis and Schipper (1997repeat their analyses on a sample of firms selected from high-technology industries and on a samplwhere the technological change is likely to have had a lesser effect on the financial statements. Nwaeze(1998) compares coefTicients from regressions (4) and (6) for samples of firms from the electric utilityindustry and from the manufacturing sector.I t ti th t h h ld b d i b b d tt i th d t b t i

8/10/2019 easton1999.pdf

8/15

8/10/2019 easton1999.pdf

9/15

Security Returns and the Value Relevance of Accounting Data 4 7

e

m

1rt (D

IS 5

GT

Re

o

'

- _

' " ' : ' ,

. . ,

- : ' . - ^ - : - J

: . . . - , ry^s-s -y^M

- . i'.'j'i 'O ' - ' , - : ' '

. . ' : ' . 'v;.'^.'';.^Vf'

' . . . ' ' ' ' j ^

; / : ' , :

- , ; . - '

. :

m m

I

CNJ

r

o

I

CVJ

"a

Bd

O

11

oa

2 Sb S

g Oj 0)

I S ^

S g ^ii .2 O

S S J

1 1 i ^ w

T3 . 3 - a.d u . (i> O I -B W O V CC to 03 O.

z y '

^ OJ

a c -

O rt >- -C

8/10/2019 easton1999.pdf

10/15

408 Accounting Horizons December 1999

OJ

b s c

w aj 0 ^ c u

o

o - C'E.S *

C bi C a C

F iS -

rt ^ J '&

.> T3

Ji

M ^ C

8/10/2019 easton1999.pdf

11/15

Security Returns and the Value Relevance of ccounting D ata 9

Oi

01

S

o

oa

tD

M

ie

fl

fl

ta

o

h

r*

Fo

e

vite

c

Tc

ino

a

3 v

t j

, .

3

V

floCOboflf l

,e

o a.

is

ae

wfl

3h J01u

oa

fl

a

a; ^oticfl

Sfl

rdvd

M

fsc

y

01

01

o

oc p

i

>

?

pu

flca

J3u

a

in

a

1. .

rc

p

a

fp

o

otinfl

5

gn

01

-o01

o

dv

a13w

rh

.201

x:tfi

8/10/2019 easton1999.pdf

12/15

41 Accounting Horizons December 1999

r OJ

o

O {'-' drx) s6utujB3 V

OJOJ

OS

o

be

l O

n

e

h]

nt

u

pt

1)I M

h

e

r

e

of

T3

nCO

nv

p

o

e

n

dv

~

CO

xs

TX U

J 3 * ^

bed.

4 ^

repo

SX

4

ngs

cto

o

boc

CO i

a

p

c

9in

ti

id

y

d

CO

t n

en

dv

be

iceci

a.

CO

8/10/2019 easton1999.pdf

13/15

Security Returns and the Value Relevance of Accounting Data 411

announcement of information, these association studies examine the effectiveness of ac-counting data as a summary of the events that have affected the firm to date or over thefiscal return) period. Since price-levels models, which may be motivated by the sametheoretical foundation as returns models, suffer from potentially serious scale problems,

the inferences from returns models are probably more reliable and should be used. Inaddition, returns regressions provide evidence regarding the timeliness of the reporting ofvalue changes in the financial statements. Theoretical models may be used to clarify ourthinking about the relations between market variables and accounting data and they shouldbe a key ingredient in the development of our hypotheses. We may, however, also gainuseful insights by paying close attention to pervasive patterns in the data.

R F R N S

Ball, R., and P. Brown. 1968. An empirical analysis of accounting income nu mb ers. Journal ofAccounting Research 6: 159-178.

Barth, M., and G. Clinch. 1998. Revalued financial, tan gible, and intangible asse ts: Associationwith share prices and non-market-based value estimates. Journal of Accounting Research36: 199-233.

Bas u, S. 1997. The conservatism principle and th e asymm etric timeliness of earn ing s. Journal ofAccounting and Economics 24: 1-37.

Beaver, W. 1968. The information content of annual earnings announcem ents. Journal of Accounting Research 6: 67-92 .

Brown, P. 1994. Capital m arkets-based research in accounting. Coopers Lyb rand, M elbourne,Austral ia .

Burgstahler, D., and I. Dichev. 1997. Earnings management to avoid earnings decreases andlosses. Journal of Accounting and Economics 24: 99-126 .

Cham bers, D., R. Jen nin gs , and K. Thom pson. 1999. Evidence on the usefulness of capital expendi-tures as an alternative m easure of depreciation. Reuiew ofAccounting Studies. 4:169-195.Ch ristie, A. 1987. On eross-sectional analy sis in accounting research. Journal ofAccounting and

Economics 9: 231-258.Dechow, P., A. Hut ton , and R. Sloan. 1999. An empirical assessment of the resid ual income

valuation model. Journal of Accounting and Economics 26: 1-34.Easton, P., and M. Zmijewski. 1989. Cross-sectional va riation in the stock market response to

accounting earnings announcements. Journal of Accounting and Economics 11: 117-141., and T. Harris. 1991. Earnings as an explanatory variables for re turns . Journal of Ac

counting Research 29: 119-142.-, and J. Ohlson. 1992. Aggregate accounting earnings can explain m ost of security

retu rns : The case of long retu rn intervals. Journal of Accounting and

Economics 15: 119-142. , P. Eddey, and T. Harris. 1993. An investigation of revaluations of tangible long-livedassets . Journal of Accounting Research 31 (Supplement): 1-38.

. 1998. Discussion of: Revalued financial, tang ible, and intang ible a ssets : Association w ithshare prices and non-market-hased value estimates. Journal of Accounting Research 36(Supplement): 235-247.

, P. Shroff, and G. Taylor. 1999. Perm ane nt and transitory earning s, accounting recogni-tion lag and the earnin gs coefficient. W orking paper, Ohio Sta te Univ ersity.

-, and G. Som mers. 1999. Estim ation of the relation between price and financial sta tem entdata . Working paper, Ohio State University.

Edwards E., and P. Bell. 1961 . he Theory and Measurem ent of Business Income. Berkeley, CA:Un iversity of California Pres s.

Feltham, G., and J. Ohlson. 1995. Valuation and clean surp lus accounting for operating andfinancial activities. Contemporary Accounting Research 11; 689-731.

8/10/2019 easton1999.pdf

14/15

8/10/2019 easton1999.pdf

15/15