Earnings, Book Values, and

Dividends in Equity Valuation

(OHLSON, 1995)
∞ ~
Pt = ∑ R −f τ Et [d t +τ ] (A1)
τ =1

Pt = market value (or price) of equity at date t.

dt= net dividends at date t.
Rf = the risk-free rate plus one.
Et [.] = the expectation operator conditional on the date
t information.
yt = (net) book value at date t.
xt = earnings (net income) for the period (t— 1,t).
Labeling of x, and y, is obviously arbitrary and


gratuitous unless the model exploits structural

attributes inherent in accounting. Of interest are
at least two closely related attributes. F First, the
change in book value between two dates equals
eamings minus dividends, that is, the model
imposes the clean surplus relation. Second,
dividends reduce current book value, but not
current eamings.
  To formalize these two aspects of owners' equity
accounting, we introduce the following
mathematical restrictions:


 yt-1 =yt + dt – xt (A2a)



 and

∂yt
= −1
∂d t


 (A2b)
∂xt
=0
∂d t



  Though (A2b) does not follow from (A2a), (A2b) is
consistent with (A2a) in the sense that


∂yt −1 ∂yt 
∂d t ∂xt
= + −
∂d t ∂d t

∂d t ∂d t


0 = −1 + 1 − 0 

 We distinguish between (A2a) and (A2b) because many

conclusions depend only on (A2a). One can apply the
clean surplus relation (A2a) to express Pt, in terms in
terms of future (expected) earnings and book values in
lieu of the sequence of dividends in the PVED
formula.
Define:


 x ≡ xt − ( R f − 1) yt −1
a
t
earnings and book values in lieu of the sequence of


Using this d ≡ x − y +
a
 expression
t t R f y t −d1t+1 ,dt-2 , … in the
tot
replace
PVED formula yelds the equation


 (1)
∞ ~a

Pt = yt + ∑ R Et [ x t +τ ]
−τ
f
τ =1
 ~a
 provide that Et [ x t +τ ]
 →0
τ →∞ Rf
 as


 We assume that the last regularity condition in

satisfied. That the clean surplus equation
implies equivalence of equation (l) and PVED
has long been known in the accounting
literature. See, for example, Edwards and Bell
(1961), and Peasnell (I981), (1982).

  We will refer to x a as abnormal earnings. The
t
terminology is motivated by the concept that “normal”
earnings should relate to the “normal” retum on the
capital invested at the beginning of the period, that is,
netbook value at date t-1 multiplied by the interest rate.
Thus one interprets as eamings minus a charge for
a the use of capital.
xt 

A positive indicates a “profitable” period since the book

rate of return, xt +1
yt
exceeds the firm’s cost of capital, R f −1.
Relation (1) has a straightforward and intuitively appealing interpretation: a
firm's value equals its book value adjusted for the present value of
anticipated abnormal eamings. In other words, the future profitability as
measured by the present value of the anticipated abnonnal earnings
sequence sequence reconciles the difference between market and book
values.
While relation (l) may appeal to one's intuition, its equivalence to. PVED
depends only on relatively trite algebra. As Peasnell (1982) notes, this
formula is peculiar because one interprets it by referring to accounting
concepts, yet the formula works regardless of the accounting principles
that measure book values and eamings. Accounting constructs eyond the
clean surplus restriction are irrelevant, and (1) does not even rely on
assumption (A2b). The third and final assumption concemsthe time-series
behavior of abnormal earnings. Since any analysis of the valuation
function generally depends critically on various aspects of this
assumption, it demands careful elaboration An analytically simple linear
model formulates the information dynamics. Two variables enter the
specification: abnormal earnings, , and information other than
abnormal earnings, .
xta
vt
 ~a
Assume (A3)
{x t }τ ≥1
satisfies the sthocastic process
~a ~
(2a) xτ +1 = ωx + vt + ε 1t +1
a
t

~ ~
(2b)
vτ +1 = + γvt + ε 2t +1
(3) yt = x + R f yt −1 − d t
a
t
~
(4)
Et [ x t +1 ] = ( R f − 1) yt + ωx + vt
a
t

(5) Pt = y t +α x + α 2 vt
t
1 a
where,

α1 = ω /( R f − ω ) ≥ 0
α 2 = R f /( R f − ω )( R f − γ ) > 0
Straightforward manipulations yield:
(6)
~ ~ ~ ~
( P t +1 + d t +1 ) / Pt = R f + (1 + α1 ) ε 1t +1 / Pt + α 2 ε 2t +1 / Pt

(7)
Pt = k (ϕxt − d t ) + (1 − k ) yt + α 2 vt
where

ϕ = R f /( R f − 1)
k = ( R f − 1)α1 = ( R f − 1)ω /( R f − ω )
 To make the point we simply (w.l.o.g) by
putting vt=0. As the first special, let
 ω = k = 1.
 (8) Pt = ϕxt − d t


 (9) x~ = R x −( R − 1)d + ε ~
t +1 f t f t 1t +1


 (10)
Pt = yt


 (11) ~ ~

x t +1 = ( R f − 1) yt + ε 1t +1



 To make the point we simply (w.l.o.g) by
putting vt=0. As the first special, let
 ω = k = 1.
 (8) Pt = ϕxt − d t


 (9) x~ = R x −( R − 1)d + ε ~
t +1 f t f t 1t +1


 (10)
Pt = yt


 (11) ~ ~

x t +1 = ( R f − 1) yt + ε 1t +1



 Earnings, Book Values, and
Dividends in Equity Valuation
(OHLSON, 1995)
As is well-known, PVED and CSR imply the RIV model:
o introduce the RIV model, assume the following:

∞ ~
Pt = bt + ∑ R Et ( x t +τ )
−τ

where τ =1residual (or abnormal) income

defines
(eamings).
xta ≡ In
xt −fact,
r bt − 1 one can make the slightly stronger
statement that, given the clean surplus relation CSR,
PVED implies RIV, and conversely. This equivalence has
been much noted in recent literature, including DHS.
 Earnings, Book Values, and
Dividends in Equity Valuation
(OHLSON, 1995)
Having established RIV, it appears reasonable that one next
imposes a time-series stochastic process related to residual
income in lieu of dividends. The particularly simple first-order
auto-regressive (AR(1)) process suggests itself. Suppose that

~a ~
x t +1 ≡ ωxta + ε t +1
in which case

ω
Pt = (2),
The derivation that yields bt +given RIV xand
a
t (1), is of course
R −ω
elementary.
 Earnings, Book Values, and
Dividends in Equity Valuation
(OHLSON, 1995)
The EBD model adds no significant analytical complications. It extends the
simple AR(1) dynamic by introducing information other than current residual
income. Such information influences forecasts of subsequent residual
incomes. A scalar variable v, represents "other information", and two
stochastic dynamic equations specify the evolution of :

~a ~
x t +1 ≡ ωxta + ε t +1

ω
Pt = bt + a
xt
R −ω