Description
In response to recommendations by the AICPA Special Committee on Financial Reporting and the Association for
Investment Management and Research, the FASB recently invited comment regarding the question, ``Given [ecient]
markets, would any disservice be done to the interests of individual investors by allowing professional investors access
to more extensive information?'' [AICPA (1996) Report of the Special Committee on Financial Reporting and the
Association for Investment Management and Research, New York, p. 22]. Research in psychology [e.g. Grin &
Tversky (1992) The weighing of evidence and the determinants of con®dence. Cognitive Psychology, 411±435] suggests
that less-informed investors may suer from over-con®dence and trade too aggressively given their information. This
paper reports on an experiment designed to address these issues. In the experiment, security values are determined by
the price/book ratios of actual ®rms, ``more-informed'' investors observe three value-relevant ®nancial ratios derived
from Value-Line reports, and ``less-informed'' investors observe only one of those signals. Even after market prices
have stabilized after many rounds of trading, less-informed investors systematically transfer wealth to more-informed
investors as a result of biased prices and overly aggressive trading. However, alerting less-informed investors to the
extent of their informational disadvantage eliminates these welfare losses.
Con®dence and the welfare of less-informed investors
Robert Bloom®eld, Robert Libby*, Mark W. Nelson
Johnson Graduate School of Management, Sage Hall, Cornell University, Ithaca, NY 14853-6201, USA
Abstract
In response to recommendations by the AICPA Special Committee on Financial Reporting and the Association for
Investment Management and Research, the FASB recently invited comment regarding the question, ``Given [ecient]
markets, would any disservice be done to the interests of individual investors by allowing professional investors access
to more extensive information?'' [AICPA (1996) Report of the Special Committee on Financial Reporting and the
Association for Investment Management and Research, New York, p. 22]. Research in psychology [e.g. Grin &
Tversky (1992) The weighing of evidence and the determinants of con®dence. Cognitive Psychology, 411±435] suggests
that less-informed investors may su?er from over-con®dence and trade too aggressively given their information. This
paper reports on an experiment designed to address these issues. In the experiment, security values are determined by
the price/book ratios of actual ®rms, ``more-informed'' investors observe three value-relevant ®nancial ratios derived
from Value-Line reports, and ``less-informed'' investors observe only one of those signals. Even after market prices
have stabilized after many rounds of trading, less-informed investors systematically transfer wealth to more-informed
investors as a result of biased prices and overly aggressive trading. However, alerting less-informed investors to the
extent of their informational disadvantage eliminates these welfare losses. The results thus suggest that providing
information to only professional investors could harm the welfare of less-informed investors if less-informed investors
are not aware of the extent of their informational disadvantage. # 1999 Elsevier Science Ltd. All rights reserved.
1. Introduction
Reducing the extent to which some investors are
at an informational disadvantage relative to others
has been an integral part of the SEC's mission to
ensure fair disclosure (Beaver, 1981b, p. 193; see
also Foster, 1986, p. 40; Hand & Beatty, 1992).
1
However, regulators periodically consider reg-
ulatory changes that may increase informational
inequities. For example, both the FASB and the
SEC have experimented with allowing ®rms to
disclose only summary ®nancial information to
investors who are unlikely to understand more
detailed disclosures (Lee & Morse, 1990). More
recently, a report by the AICPA Special Commit-
tee on Financial Reporting recommended that
``regulators consider whether it would be in the
interest of e?ective capital allocation for certain
users... to have access to more extensive informa-
tion'' that other users would not have (AICPA,
1996, p. 22).
2
0361-3682/99/$ - see front matter # 1999 Elsevier Science Ltd. All rights reserved.
PI I : S0361- 3682( 99) 00025- 2
Accounting, Organizations and Society 24 (1999) 623±647
www.elsevier.com/locate/aos
* Corresponding author. Tel.: +1-607-255-3348; fax: +1-
607-254-4590.
E-mail address: [email protected] (R. Libby)
1
Similarly, the Financial Accounting Standards Board notes
that ®nancial statement users' ``understanding of ®nancial
information and the way and extent to which they use and rely
on it also may vary greatly'' and that ``e?orts may be needed to
increase the understandability of ®nancial information'' to
reduce informational inequities (FASB, SFAC 1, { 36, 1980).
2
Such di?erential disclosure could allow ®rms to release
information to some investors that they would be unwilling to
disclose publicly, due to its cost or proprietary nature.
These views are partly motivated by academic
research in ecient markets theory. If less-informed
investors always trade at prices that fully re¯ect all
available information, they are ``price-protected,''
and will earn the same returns as more-informed
investors (Beaver, 1973, 1981a; Foster, 1986; Watts
&Zimmerman, 1986). These arguments are re¯ected
in the FASB's recent invitation to comment, which
remarks that allowing the di?erential disclosure
recommended by the AICPA Committee ``could
be considered in the context of the ecient markets
theory. Given such markets, would any disservice
be done to the interests of individual investors by
allowing professional investors access to more
extensive information?'' (AICPA, 1996, p. 22).
Recent laboratory and archival research sug-
gests that less-informed investors may not in fact
be price protected, because market prices may not
always be informationally ecient.
3
Less-informed
investors can still protect themselves in such an
environment by trading in a way that re¯ects the
adverse selection they face when trading against
more-informed investors (Lev, 1988). However,
this requires less-informed investors to understand
how the statistical reliability of their information
compares to that of other investors. Previous psy-
chology studies (e.g. Grin & Tversky, 1992;
Lichtenstein & Fischho?, 1977; Lichtenstein,
Fishho? & Phillips, 1982; Peterson & Pitz, 1986,
1988) suggest that such accurate assessments may
be dicult because less-informed individuals tend
to be over-con®dent in their knowledge and infor-
mation relative to more-informed individuals.
Relative over-con®dence of less-informed inves-
tors may lead them to revise their estimates of
value too much in response to that information
and to be too certain of their biased estimates. As
a result, they will tend to buy too aggressively when
they have relatively favorable information and sell
too aggressively when they have relatively unfa-
vorable information. If relative over-con®dence or
other factors also cause prices to be too high when
less-informed investors hold favorable informa-
tion and/or too low when they hold unfavorable
information, this trading activity would cause less-
informed investors to lose wealth by ``buying high
and selling low.'' Providing less-informed inves-
tors with explicit information (``guidance'') about
the relative reliability (i.e. explanatory power) of
their information might reduce their relative over-
con®dence, and thus, reduce these wealth losses.
This paper reports an experiment that investigates
these issues.
In the experiment, two more-informed investors
observe three informative signals, while two less-
informed investors observe only one of those sig-
nals. In addition to the amount of information
available to particular investors, we manipulate
whether or not investors receive guidance as to the
reliability of their information. Our guidance
communicates only the amount of variance
explained by the information held by each class of
investor (i.e. no information useful in determining
the value of the security is conveyed). Trading
takes place in a computerized specialist market
(Bloom®eld & Libby, 1996).
After many rounds of trading, prices in these
markets converge to a ``steady-state price'' at
which supply approximately equals demand. We
®nd that average investors' estimates of value are
too high (low) and the securities are over-priced
(under-priced) when relatively favorable (unfavor-
able) information is held by less-informed inves-
tors. Furthermore, in the absence of guidance
about signal reliability, the less-informed inves-
tors' estimates of value are higher (lower) than
those of the more-informed investors for these
over-priced (under-priced) securities. As a con-
sequence, less-informed investors show a strong
tendency to buy from (sell to) more-informed
investors at steady-state prices when securities are
over-priced (under-priced) in this condition.
Thus, less-informed investors e?ectively ``buy
high and sell low,'' systematically transferring
wealth to more-informed investors when guidance
is absent.
4
3
See, e.g. archival studies by Bernard and Thomas (1989,
1990), Hand (1990), Frankel and Lee (1998), Sloan (1996), and
Lee, Myers, and Swaminathan (in press), and experimental
studies by Bloom®eld (1996b), Bloom®eld and Libby (1996),
and Calegari and Fargher (1997).
4
Although such welfare losses would not be surprising in
early rounds of trading (e.g. see Plott & Sunder, 1982, 1988;
Forsythe & Lundholm, 1990), in an ecient market no such
losses would occur once the market reaches a steady-state
(equilibrium) price.
624 R. Bloom®eld et al. / Accounting, Organizations and Society 24 (1999) 623±647
As predicted, we also ®nd that providing less-
informed investors with guidance on signal relia-
bility reduces (and in fact eliminates) their welfare
losses at steady-state prices, by causing their esti-
mates of values and trading behavior at steady
state prices to become indistinguishable from
those of more-informed investors. Interestingly,
however, we also ®nd that guidance does not suf-
®ciently reduce the over-con®dence of the less-
informed investors in early rounds to eliminate the
overall biases in steady-state market prices Ð even
in the presence of guidance, steady state prices are
too high (low) when less-informed investors hold
favorable (unfavorable) information. This sug-
gests that the biases in steady-state prices are due
to general diculties of aggregating public and
private information in ®nancial markets, similar to
those discussed in Bloom®eld (1996a) and Bloom-
®eld and Libby (1996), which may be compounded
when the over-con®dent less-informed investors
trade aggressively in early rounds.
To clarify the role of con®dence in the above
results, a second group of subjects performed a
supplementary pencil and paper task to directly
compare the more- and less-informed investors'
con®dence and the impact of guidance on con-
®dence. In the absence of guidance, the less-
informed investors' con®dence was indistinguish-
able from the more-informed, indicating that the
less-informed were relatively over-con®dent. Gui-
dance signi®cantly decreased the con®dence of the
less-informed.
These results, combined with those of the
experiment, are consistent with our proposed
linkages between amount of information, gui-
dance, con®dence, and trading behavior. Welfare
losses by less-informed investors at steady-state
prices are caused not by the fact that less-informed
investors are at an informational disadvantage
(which they always are in our markets), but rather
by the tendency of less-informed investors to
underestimate the extent of that disadvantage (i.e.
to be relatively over-con®dent). This result sug-
gests that less-informed investors might be helped
by disclosures which act in a manner similar to
our guidance manipulation, simply making less-
informed investors more aware of their informa-
tional disadvantage. For example, even if less-
informed investors' lack the training or resources
to interpret footnotes, disclosing those footnotes
might improve their welfare by highlighting the
extent of their information disadvantage relative
to investors who have greater training and
resources. Conversely, proposals to provide only
summary ®nancial statements to less-informed
investors, and to better inform some investors
without disclosure to other investors, might harm
less-informed investors by concealing the extent of
their informational disadvantage.
The remainder of the paper is organized as fol-
lows. Section 2 provides background and hypoth-
eses. Sections 3 and 4 describe the method and
results. Section 5 discusses the implications of the
results and directions for further study.
2. Theory and hypotheses
2.1. Over-con®dence
Previous research in psychology suggests that
errors in assessing the reliability of information
are likely to take a consistent form: people with
little information will tend to be too con®dent,
relative to those with more information (Lichten-
stein & Fischho? 1977; Lichtenstein et al., 1982).
Two related sets of studies suggest a cause for
these ®ndings. First, Grin and Tversky (1992,
Study 1) provide evidence in a variety of contexts
that individuals rely on the balance of data for or
against a hypothesis, with insucient regard for
the reliability of the data. They suggest that this
leads less-informed individuals to be over-con-
®dent, particularly when relatively unreliable data
takes on extreme values. Second, Peterson and
Pitz (1986, 1988) provide evidence that possessing
a greater number of independent information sig-
nals increases the chance that the signals con¯ict,
and this con¯ict decreases both the extremity of
predictions and con®dence. Thus, while more
information may increase con®dence, con¯ict
among items of information can decrease con-
®dence. They also show less-informed individuals
to be particularly insensitive to the reliability of
their data when that reliability does not vary much
from one decision to another. This circumstance is
R. Bloom®eld et al. / Accounting, Organizations and Society 24 (1999) 623±647 625
likely to re¯ect actual investment settings. For
example, most individual investors probably
receive some information of low reliability for all
of the securities they trade.
2.2. Welfare e?ects
If the steady-state price is perfectly informa-
tionally ecient (i.e. it fully re¯ects all information
available to the market in the aggregate), less-
informed investors are ``price-protected'' and will
not lose any money as a result of their trading
strategy. However, previous laboratory research
indicates that prices tend to respond more strongly
to information that is held by more investors
(Bloom®eld, 1996a, 1996b; Bloom®eld & Libby,
1996; Forsythe & Lundholm, 1990; Lundholm,
1991; Plott & Sunder 1982, 1988). The informa-
tion that is held by more investors is likely to be
the information that less-informed investors pos-
sess (i.e. more-informed investors know what less-
informed investors know, and more). Conse-
quently, prices will tend to be higher when less-
informed investors hold a favorable item of infor-
mation than when less-informed investors hold an
unfavorable item of information, even though
more-informed investors know both items (so that
the information available to the market remains
constant).
5
The psychology studies discussed above suggest
that less-informed investors will be relatively more
over-con®dent than will more-informed investors.
This relative over-con®dence will cause their esti-
mates of value to be higher (lower) than those of
the more-informed investors when they hold rela-
tively favorable (unfavorable) information. They
will also be too certain that their biased estimates
are accurate. As a consequence of these di?erences
in estimates of value and their undue certainty,
less-informed investors will show a tendency to
buy from (sell to) more-informed investors at
steady-state prices when they hold relatively
favorable (unfavorable) information.
6
If, as sug-
gested above, securities are overpriced (under-
priced) when the less-informed investors hold
relatively favorable (unfavorable) information,
employing this trading strategy will cause the less-
informed to transfer wealth to more-informed inves-
tors as a result of ``buying high and selling low.''
Our ®rst hypothesis examines whether, in the
absence of guidance, less-informed investors
transfer wealth to more-informed investors at
steady state prices.
H1: Less-informed investors transfer wealth
to more-informed investors at steady-state
prices in the absence of guidance on signal
reliability.
We examine wealth transfers by constructing
markets in which two more-informed investors
hold all three value-relevant signals and two less-
informed investors hold only one of those signals.
For each security traded, di?erent groups of
investors trade the same security in settings that
are identical except for which of the three signals
the less-informed investors hold. Because the
information available to the more-informed inves-
tors is the same (and therefore the total informa-
tion available to the market is the same), the
informationally ecient price is the same in both
settings. However, the favorability of the signal
held by the less-informed investors di?ers depend-
ing on which signal the less-informed investor
holds. While the informationally ecient price is
the same in both settings, as we note above, we
expect prices to be biased in the direction of the
signal held by the less-informed investors. We also
expect the less-informed investors' estimates of
value and trading to be more biased than that of
the more-informed. As a consequence, they will
transfer wealth to the more-informed by buying
over-priced and selling under-priced securities to
the more-informed.
Presumably, wealth transfers could be arising at
every point throughout the trading process. We
5
Stronger reactions to public information than to private
information do not preclude under-reactions to public infor-
mation. A market could under-react to all information avail-
able, but under-react less to public information.
6
If less-informed and more-informed investors were equally
over-con®dent, both would trade more aggressively than they
would otherwise and the e?ects of over-con®dence would o?set.
626 R. Bloom®eld et al. / Accounting, Organizations and Society 24 (1999) 623±647
focus only on wealth transfers that occur after
many rounds of trading have led the market to a
``steady-state'' price at which supply equals
demand, because these prices should represent the
most informationally ecient prices, and therefore
should price-protect less-informed investors to the
greatest extent. Also, previous research has
already established that uninformed investors earn
less than better-informed investors before infor-
mation is impounded in market prices as com-
pletely as it can be (Forsythe & Lundholm, 1990;
Plott & Sunder, 1982). If less-informed investors
are unable to protect themselves from welfare los-
ses at prices that are as informationally ecient as
possible in our market setting, it seems unlikely
that they would avoid losses before that point.
The studies cited above have found no sig-
ni®cant welfare di?erences between informed and
uninformed investors once prices reach steady-
state levels. For example, Plott and Sunder (1982)
®nd that closing prices are so informationally e-
cient that all traders earn approximately the same
amount from trade. Experienced subjects in For-
sythe and Lundholm (1990) reveal a similar result.
However, the simplicity of the information environ-
ments in these studies make it dicult for psycho-
logical biases such as over-con®dence to a?ect
prices and welfare. For example, in many of these
experiments less-informed investors are entirely
uninformed, which would discourage any attempts
by them to trade on their own information. Also,
more-informed investors in those markets receive
information that is so informative and objective
that there is little room for them to exercise their
judgement in interpreting it. [For example, sub-
jects in most of the markets in Plott & Sunder,
Forsythe & Lundholm, and Lundholm (1991)
learn the true value with certainty, learn that the
security has one of two values with certainty, or
receive no information at all about security value.]
One of the goals of this paper is to explore psy-
chological biases that arise when investors must
exercise judgement in interpreting their information.
2.3. Guidance
Understanding the psychological cause of poor
decision performance facilitates identi®cation of a
decision aid that is designed to improve perfor-
mance (Bonner, Libby & Nelson, 1996). If overly-
aggressive trading behavior by less-informed
investors is in fact driven by the relative over-
con®dence of less-informed investors, then helping
these investors assess the reliability of information
should diminish their welfare losses. Two ®ndings
in the psychology literature suggest a simple and
e?ective form of guidance that should reduce the
con®dence of less-informed investors. First, a
large number of studies have examined the e?ec-
tiveness of di?erent forms of feedback in multiple
cue probability learning tasks. This literature is
aimed at determining what feedback conditions
promote or deter learning in tasks similar to our
trading settings, where participants predict uncer-
tain outcomes based on cues with di?erent
reliabilities [see Libby (1981, Chapter 2) for a
description of typical studies and results]. Many of
these studies compare the e?ectiveness of ``out-
come'' feedback (where subjects are given the true
value after each case) with ``task-properties'' feed-
back (where the relative reliabilities of the cues are
provided). In general, learning and performance
are far superior in the task-properties-feedback
condition.
7
Second, while the probability judgment
literature has generally found over-con®dence to be
dicult to debias, a recent study by Tra®mow and
Sniezek (1994) found that providing subjects with
information about their general knowledge level (a
simple form of task properties feedback) reduces
subjects' con®dence, and thus subjects' over-con-
®dence. These ®ndings suggest that providing less-
informed investors with guidance about the rela-
tive reliability of their information may reduce the
relative over-con®dence of these investors com-
pared to that of more-informed investors. This, in
turn, should reduce the bias in their value esti-
mates and the aggressiveness of less-informed
investors' trading decisions, and reduce their
wealth transfers to the more-informed investors.
If the trading bias predicted by H1 is in fact
driven by the relative over-con®dence of less-
informed investors, then providing guidance
7
See Bonner and Walker (1994) for similar ®ndings that
show the advantage of explanatory feedback over outcome
feedback.
R. Bloom®eld et al. / Accounting, Organizations and Society 24 (1999) 623±647 627
concerning the reliability of their information
should diminish less-informed investors' welfare
losses. To address this issue, we manipulate whether
or not less-informed investors are given information
about the reliability of the signals that are available
to the market. If trading bias is driven by the relative
over-con®dence of less-informed investors, then gui-
dance should make those investors less con®dent,
and reduce this bias.
H2: Providing less-informed investors with
guidance on signal reliability reduces wealth
transfers fromless-informed investors to more-
informed investors at steady-state prices.
We also manipulate whether or not more-informed
investors are given guidance on signal reliability.
However, we make no hypothesis about the e?ect of
guidance on more-informed investors, because that
e?ect depends on whether or not more-informed
investors who lack guidance are over-con®dent. If
more-informed investors tend to be over-con®dent
in the absence of guidance, then providing guidance
to more-informed investors will decrease their
over-con®dence. This in turn may increase wealth
transfers by increasing the over-con®dence of less-
informed investors relative to the con®dence of
more-informed investors. On the other hand, if
more-informed investors tend to be undercon®dent
in the absence of guidance, providing guidance to
more-informed investors may increase their con-
®dence and decrease wealth transfers by decreasing
the over-con®dence of less-informed investors rela-
tive to the con®dence of more-informed investors.
Finally, more-informed investors who lack gui-
dance may accurately assess the reliability of the
information they hold, yielding no guidance e?ects.
We cannot discriminate between these predictions,
so we state no hypotheses about the e?ect of pro-
viding guidance to more-informed investors.
3. Method
3.1. Subjects
All subjects (``investors'') were MBA students at
Cornell's Johnson Graduate School of Manage-
ment who had participated in at least one prior
experimental market.
3.2. Security values and information
In our markets, investors make judgments about
securities whose values are derived from values of
real-world securities, given ®nancial information
derived from ``Value-Line'' analyses of those
securities. Speci®cally, security values are deter-
mined by each ®rm's price/book ratio, and inves-
tors are given information on the ®rm's return-on-
equity, growth in return-on-equity (as projected
by Value-Line), and book value per share.
We attempt to present investors with a valua-
tion task similar to that facing investors in real-
world exchanges, while still ensuring that the
informative signals presented to investors are easy
to understand, can explain an economically sig-
ni®cant portion of variation in security values, and
possess a minimal amount of colinearity and
interdependence. We achieve these goals by deriv-
ing security values from price/book ratios and
deriving signals from common accounting ratios.
In this way, we are able to use extant theory and
empirical work (Bernard, 1994; Frankel & Lee,
1998) to develop a small number of signals that
explain a majority of variation in price/book ratios.
8
To obtain our securities and signals, we ®rst
construct a database which includes all 291
8
Using information derived from real-world sources (as
opposed to more objective information such as random num-
bers) allows investors to determine the meaning and statistical
reliability of their signals by relying on their own knowledge
and experience. These are the circumstances in which over-
con®dence tends to be observed. A real-world valuation task
should also increase the motivation of our subjects, and allow
the study to provide some educational bene®t to the subjects,
who are MBA students. These advantages are diminished to the
extent that the subjects are unfamiliar with the nature of price/
book ratios, and to the extent that the signals we use di?er
from the naturally occurring signals reported in Value-Line.
However, we have designed our experiment so that these pos-
sibilities cannot account for welfare losses by the less-informed
investors. In particular, welfare losses require the less-informed
investors to be more over-con®dent than the more-informed.
While unfamiliarity with the information signals could cause all
subjects to be over- or under-con®dent, it is not clear how it
could cause a greater di?erence in the con®dence levels of the
two groups.
628 R. Bloom®eld et al. / Accounting, Organizations and Society 24 (1999) 623±647
manufacturing ®rms which were pro®table in 1993
and covered by Value-Line reports. We rank these
®rms according to their price/book ratios (the
most recent market price available prior to pub-
lication of the Value-Line report divided by the
net book value per share). The percentile rank of
each ®rm (denoted PRICE/BOOK) is then used as
the value of the associated security. For example,
a ®rm with a price/book ratio greater than 24% of
all ®rms in the database is given a value of 24
francs (a ®ctitious currency ultimately converted
to cash). The ®rst signal providing information
about PRICE/BOOK is an ROE score which
re¯ects the ®rm's relative return-on-equity (net
income before extraordinary items divided by net
book value) relative to other ®rms in the popula-
tion. The second signal is a GROWTH score,
which re¯ects the rate at which Value-Line pro-
jected return-on-equity for the ®rm to grow from
1992 to 1994. The third signal is a BOOK score,
which re¯ects the relative book value for each
share of the ®rm's stock. We also use a simple
orthogonalization technique to ensure that the
signals are largely independent, and present all
scores on an 11-point scale. This orthogonaliza-
tion should increase the e?ectiveness of our
manipulation, because it insures that less-
informed investors are signi®cantly less informed
than more-informed investors.
9
Appendix A con-
tains instructions to subjects which describe the
construction of the securities, including the ortho-
gonalization technique, in detail.
Panel A of Table 1 provides the results of a lin-
ear regression of PRICE/BOOK onto ROE,
BOOK and GROWTH. Consistent with empirical
research, all three of the signals are positively and
statistically signi®cantly related to PRICE/BOOK.
Overall, the three signals explain 60.7% of the
variation in security value. Panel B of Table 1
shows the ten securities used in the experiment.
The securities were selected randomly and have
values throughout the 100-franc scale and signals
with realizations throughout their 11-point scales.
3.3. Trading
Subjects trade in computerized specialist mar-
kets similar to those used by Bloom®eld (1996b)
and Bloom®eld and Libby (1996). These markets
are based on the New York Stock Exchange, in
which a specialist quotes prices at which investors
can trade, the specialist ®lls all orders investors
tender at those prices, and the specialist changes
those prices in response to investors' previous trades.
In our laboratory markets, each group of four
investors (which we call a cohort) always trades
together. Each of the 10 securities (a claim on a
terminal dividend) they trade is denominated in
``francs.'' Trading is a batch process Ð all four
investors in a cohort enter their buy or sell orders
before any investor chooses their next action.
Trading round refers to each time all investors in
the cohort enter buy and sell orders for shares of a
security. Each security is traded for 15 rounds
before trading begins in the next security. We refer
to the sequence of 15 rounds of trading for a single
security as a market.
Before learning the market price for each round
of trading, investors enter their best estimates of
the security's value. These estimates allow us to
test whether trading losses are driven by biased
estimates of value. Investors then enter orders to
buy (sell) from 0 to 10 shares if their estimate is
above (below) the market price, but cannot trade
if their estimate is exactly equal to the market
price. After each round, investors learn the aggre-
gate numbers of buy and sell orders at the pre-
vious price. The specialist ®lls all orders.
Our specialist market allows shortselling, so the
number of shares bought and sold at a given price
is not in¯uenced by either wealth constraints or
how shares are distributed among investors.
Ganguly, Kagel and Moser (1994) show how
short-selling limitations can allow biases to persist,
because investors who wish to exploit high prices
cannot do so unless they own shares. The lack of
shortselling restrictions (beyond the maximum 10
share trade/round) allows investors in our markets
to exploit price errors in any direction, regardless
9
If the single signals possessed by less-informed investors
were highly correlated with the three signals held by more-
informed investors, there would be little di?erence in the
amount of explanatory power of the information held by less-
and more-informed investors, and consequently a weak treat-
ment e?ect.
R. Bloom®eld et al. / Accounting, Organizations and Society 24 (1999) 623±647 629
of their holdings, so prices in our markets provide
a less noisy re¯ection of investors' value estimates
and their con®dence in those estimates.
10
Trading for a security always begins at 50
francs. To set the price for the next round, the
computer revises the price upward (downward) if
there are more (fewer) buy orders than sell orders,
with the magnitude of movement an increasing
function of the imbalance.
11
Under this price revi-
sion rule, prices should not settle at a level which
di?ers substantially from investors' estimates of
value Ð if the price is too high (low), investors will
drive it downward (upward) until they can achieve
no further gains. Eventually, the market settles into
a narrow range of prices at which average supply
roughly equals average demand. We call the aver-
age of these prices the ``steady-state'' price. Trad-
ing continues for 15 rounds, which we expected
would be long enough to allow several rounds of
trading after the steady-state price is reached.
3.4. Instructions
The ®rst 20 minutes of each experimental ses-
sion are spent reviewing two sets of instructions.
The ®rst set of instructions explains the computa-
tion of gains and losses from buying and selling,
the computation of gains from accurate estimates
of security value, the trading process, the mechan-
ism by which prices are adjusted after each round
of trading, and the conversion of winnings from
Table 1
Relation of signals to security value.
a
Panel A
b
: estimation of regression model
ROE
GROWTH
BOOK
R
2
À0.06 0.058 0.020 0.035 60.7%
(À2.06) (17.21) (5.75) (9.99)
p=0.04 p
In response to recommendations by the AICPA Special Committee on Financial Reporting and the Association for
Investment Management and Research, the FASB recently invited comment regarding the question, ``Given [ecient]
markets, would any disservice be done to the interests of individual investors by allowing professional investors access
to more extensive information?'' [AICPA (1996) Report of the Special Committee on Financial Reporting and the
Association for Investment Management and Research, New York, p. 22]. Research in psychology [e.g. Grin &
Tversky (1992) The weighing of evidence and the determinants of con®dence. Cognitive Psychology, 411±435] suggests
that less-informed investors may suer from over-con®dence and trade too aggressively given their information. This
paper reports on an experiment designed to address these issues. In the experiment, security values are determined by
the price/book ratios of actual ®rms, ``more-informed'' investors observe three value-relevant ®nancial ratios derived
from Value-Line reports, and ``less-informed'' investors observe only one of those signals. Even after market prices
have stabilized after many rounds of trading, less-informed investors systematically transfer wealth to more-informed
investors as a result of biased prices and overly aggressive trading. However, alerting less-informed investors to the
extent of their informational disadvantage eliminates these welfare losses.
Con®dence and the welfare of less-informed investors
Robert Bloom®eld, Robert Libby*, Mark W. Nelson
Johnson Graduate School of Management, Sage Hall, Cornell University, Ithaca, NY 14853-6201, USA
Abstract
In response to recommendations by the AICPA Special Committee on Financial Reporting and the Association for
Investment Management and Research, the FASB recently invited comment regarding the question, ``Given [ecient]
markets, would any disservice be done to the interests of individual investors by allowing professional investors access
to more extensive information?'' [AICPA (1996) Report of the Special Committee on Financial Reporting and the
Association for Investment Management and Research, New York, p. 22]. Research in psychology [e.g. Grin &
Tversky (1992) The weighing of evidence and the determinants of con®dence. Cognitive Psychology, 411±435] suggests
that less-informed investors may su?er from over-con®dence and trade too aggressively given their information. This
paper reports on an experiment designed to address these issues. In the experiment, security values are determined by
the price/book ratios of actual ®rms, ``more-informed'' investors observe three value-relevant ®nancial ratios derived
from Value-Line reports, and ``less-informed'' investors observe only one of those signals. Even after market prices
have stabilized after many rounds of trading, less-informed investors systematically transfer wealth to more-informed
investors as a result of biased prices and overly aggressive trading. However, alerting less-informed investors to the
extent of their informational disadvantage eliminates these welfare losses. The results thus suggest that providing
information to only professional investors could harm the welfare of less-informed investors if less-informed investors
are not aware of the extent of their informational disadvantage. # 1999 Elsevier Science Ltd. All rights reserved.
1. Introduction
Reducing the extent to which some investors are
at an informational disadvantage relative to others
has been an integral part of the SEC's mission to
ensure fair disclosure (Beaver, 1981b, p. 193; see
also Foster, 1986, p. 40; Hand & Beatty, 1992).
1
However, regulators periodically consider reg-
ulatory changes that may increase informational
inequities. For example, both the FASB and the
SEC have experimented with allowing ®rms to
disclose only summary ®nancial information to
investors who are unlikely to understand more
detailed disclosures (Lee & Morse, 1990). More
recently, a report by the AICPA Special Commit-
tee on Financial Reporting recommended that
``regulators consider whether it would be in the
interest of e?ective capital allocation for certain
users... to have access to more extensive informa-
tion'' that other users would not have (AICPA,
1996, p. 22).
2
0361-3682/99/$ - see front matter # 1999 Elsevier Science Ltd. All rights reserved.
PI I : S0361- 3682( 99) 00025- 2
Accounting, Organizations and Society 24 (1999) 623±647
www.elsevier.com/locate/aos
* Corresponding author. Tel.: +1-607-255-3348; fax: +1-
607-254-4590.
E-mail address: [email protected] (R. Libby)
1
Similarly, the Financial Accounting Standards Board notes
that ®nancial statement users' ``understanding of ®nancial
information and the way and extent to which they use and rely
on it also may vary greatly'' and that ``e?orts may be needed to
increase the understandability of ®nancial information'' to
reduce informational inequities (FASB, SFAC 1, { 36, 1980).
2
Such di?erential disclosure could allow ®rms to release
information to some investors that they would be unwilling to
disclose publicly, due to its cost or proprietary nature.
These views are partly motivated by academic
research in ecient markets theory. If less-informed
investors always trade at prices that fully re¯ect all
available information, they are ``price-protected,''
and will earn the same returns as more-informed
investors (Beaver, 1973, 1981a; Foster, 1986; Watts
&Zimmerman, 1986). These arguments are re¯ected
in the FASB's recent invitation to comment, which
remarks that allowing the di?erential disclosure
recommended by the AICPA Committee ``could
be considered in the context of the ecient markets
theory. Given such markets, would any disservice
be done to the interests of individual investors by
allowing professional investors access to more
extensive information?'' (AICPA, 1996, p. 22).
Recent laboratory and archival research sug-
gests that less-informed investors may not in fact
be price protected, because market prices may not
always be informationally ecient.
3
Less-informed
investors can still protect themselves in such an
environment by trading in a way that re¯ects the
adverse selection they face when trading against
more-informed investors (Lev, 1988). However,
this requires less-informed investors to understand
how the statistical reliability of their information
compares to that of other investors. Previous psy-
chology studies (e.g. Grin & Tversky, 1992;
Lichtenstein & Fischho?, 1977; Lichtenstein,
Fishho? & Phillips, 1982; Peterson & Pitz, 1986,
1988) suggest that such accurate assessments may
be dicult because less-informed individuals tend
to be over-con®dent in their knowledge and infor-
mation relative to more-informed individuals.
Relative over-con®dence of less-informed inves-
tors may lead them to revise their estimates of
value too much in response to that information
and to be too certain of their biased estimates. As
a result, they will tend to buy too aggressively when
they have relatively favorable information and sell
too aggressively when they have relatively unfa-
vorable information. If relative over-con®dence or
other factors also cause prices to be too high when
less-informed investors hold favorable informa-
tion and/or too low when they hold unfavorable
information, this trading activity would cause less-
informed investors to lose wealth by ``buying high
and selling low.'' Providing less-informed inves-
tors with explicit information (``guidance'') about
the relative reliability (i.e. explanatory power) of
their information might reduce their relative over-
con®dence, and thus, reduce these wealth losses.
This paper reports an experiment that investigates
these issues.
In the experiment, two more-informed investors
observe three informative signals, while two less-
informed investors observe only one of those sig-
nals. In addition to the amount of information
available to particular investors, we manipulate
whether or not investors receive guidance as to the
reliability of their information. Our guidance
communicates only the amount of variance
explained by the information held by each class of
investor (i.e. no information useful in determining
the value of the security is conveyed). Trading
takes place in a computerized specialist market
(Bloom®eld & Libby, 1996).
After many rounds of trading, prices in these
markets converge to a ``steady-state price'' at
which supply approximately equals demand. We
®nd that average investors' estimates of value are
too high (low) and the securities are over-priced
(under-priced) when relatively favorable (unfavor-
able) information is held by less-informed inves-
tors. Furthermore, in the absence of guidance
about signal reliability, the less-informed inves-
tors' estimates of value are higher (lower) than
those of the more-informed investors for these
over-priced (under-priced) securities. As a con-
sequence, less-informed investors show a strong
tendency to buy from (sell to) more-informed
investors at steady-state prices when securities are
over-priced (under-priced) in this condition.
Thus, less-informed investors e?ectively ``buy
high and sell low,'' systematically transferring
wealth to more-informed investors when guidance
is absent.
4
3
See, e.g. archival studies by Bernard and Thomas (1989,
1990), Hand (1990), Frankel and Lee (1998), Sloan (1996), and
Lee, Myers, and Swaminathan (in press), and experimental
studies by Bloom®eld (1996b), Bloom®eld and Libby (1996),
and Calegari and Fargher (1997).
4
Although such welfare losses would not be surprising in
early rounds of trading (e.g. see Plott & Sunder, 1982, 1988;
Forsythe & Lundholm, 1990), in an ecient market no such
losses would occur once the market reaches a steady-state
(equilibrium) price.
624 R. Bloom®eld et al. / Accounting, Organizations and Society 24 (1999) 623±647
As predicted, we also ®nd that providing less-
informed investors with guidance on signal relia-
bility reduces (and in fact eliminates) their welfare
losses at steady-state prices, by causing their esti-
mates of values and trading behavior at steady
state prices to become indistinguishable from
those of more-informed investors. Interestingly,
however, we also ®nd that guidance does not suf-
®ciently reduce the over-con®dence of the less-
informed investors in early rounds to eliminate the
overall biases in steady-state market prices Ð even
in the presence of guidance, steady state prices are
too high (low) when less-informed investors hold
favorable (unfavorable) information. This sug-
gests that the biases in steady-state prices are due
to general diculties of aggregating public and
private information in ®nancial markets, similar to
those discussed in Bloom®eld (1996a) and Bloom-
®eld and Libby (1996), which may be compounded
when the over-con®dent less-informed investors
trade aggressively in early rounds.
To clarify the role of con®dence in the above
results, a second group of subjects performed a
supplementary pencil and paper task to directly
compare the more- and less-informed investors'
con®dence and the impact of guidance on con-
®dence. In the absence of guidance, the less-
informed investors' con®dence was indistinguish-
able from the more-informed, indicating that the
less-informed were relatively over-con®dent. Gui-
dance signi®cantly decreased the con®dence of the
less-informed.
These results, combined with those of the
experiment, are consistent with our proposed
linkages between amount of information, gui-
dance, con®dence, and trading behavior. Welfare
losses by less-informed investors at steady-state
prices are caused not by the fact that less-informed
investors are at an informational disadvantage
(which they always are in our markets), but rather
by the tendency of less-informed investors to
underestimate the extent of that disadvantage (i.e.
to be relatively over-con®dent). This result sug-
gests that less-informed investors might be helped
by disclosures which act in a manner similar to
our guidance manipulation, simply making less-
informed investors more aware of their informa-
tional disadvantage. For example, even if less-
informed investors' lack the training or resources
to interpret footnotes, disclosing those footnotes
might improve their welfare by highlighting the
extent of their information disadvantage relative
to investors who have greater training and
resources. Conversely, proposals to provide only
summary ®nancial statements to less-informed
investors, and to better inform some investors
without disclosure to other investors, might harm
less-informed investors by concealing the extent of
their informational disadvantage.
The remainder of the paper is organized as fol-
lows. Section 2 provides background and hypoth-
eses. Sections 3 and 4 describe the method and
results. Section 5 discusses the implications of the
results and directions for further study.
2. Theory and hypotheses
2.1. Over-con®dence
Previous research in psychology suggests that
errors in assessing the reliability of information
are likely to take a consistent form: people with
little information will tend to be too con®dent,
relative to those with more information (Lichten-
stein & Fischho? 1977; Lichtenstein et al., 1982).
Two related sets of studies suggest a cause for
these ®ndings. First, Grin and Tversky (1992,
Study 1) provide evidence in a variety of contexts
that individuals rely on the balance of data for or
against a hypothesis, with insucient regard for
the reliability of the data. They suggest that this
leads less-informed individuals to be over-con-
®dent, particularly when relatively unreliable data
takes on extreme values. Second, Peterson and
Pitz (1986, 1988) provide evidence that possessing
a greater number of independent information sig-
nals increases the chance that the signals con¯ict,
and this con¯ict decreases both the extremity of
predictions and con®dence. Thus, while more
information may increase con®dence, con¯ict
among items of information can decrease con-
®dence. They also show less-informed individuals
to be particularly insensitive to the reliability of
their data when that reliability does not vary much
from one decision to another. This circumstance is
R. Bloom®eld et al. / Accounting, Organizations and Society 24 (1999) 623±647 625
likely to re¯ect actual investment settings. For
example, most individual investors probably
receive some information of low reliability for all
of the securities they trade.
2.2. Welfare e?ects
If the steady-state price is perfectly informa-
tionally ecient (i.e. it fully re¯ects all information
available to the market in the aggregate), less-
informed investors are ``price-protected'' and will
not lose any money as a result of their trading
strategy. However, previous laboratory research
indicates that prices tend to respond more strongly
to information that is held by more investors
(Bloom®eld, 1996a, 1996b; Bloom®eld & Libby,
1996; Forsythe & Lundholm, 1990; Lundholm,
1991; Plott & Sunder 1982, 1988). The informa-
tion that is held by more investors is likely to be
the information that less-informed investors pos-
sess (i.e. more-informed investors know what less-
informed investors know, and more). Conse-
quently, prices will tend to be higher when less-
informed investors hold a favorable item of infor-
mation than when less-informed investors hold an
unfavorable item of information, even though
more-informed investors know both items (so that
the information available to the market remains
constant).
5
The psychology studies discussed above suggest
that less-informed investors will be relatively more
over-con®dent than will more-informed investors.
This relative over-con®dence will cause their esti-
mates of value to be higher (lower) than those of
the more-informed investors when they hold rela-
tively favorable (unfavorable) information. They
will also be too certain that their biased estimates
are accurate. As a consequence of these di?erences
in estimates of value and their undue certainty,
less-informed investors will show a tendency to
buy from (sell to) more-informed investors at
steady-state prices when they hold relatively
favorable (unfavorable) information.
6
If, as sug-
gested above, securities are overpriced (under-
priced) when the less-informed investors hold
relatively favorable (unfavorable) information,
employing this trading strategy will cause the less-
informed to transfer wealth to more-informed inves-
tors as a result of ``buying high and selling low.''
Our ®rst hypothesis examines whether, in the
absence of guidance, less-informed investors
transfer wealth to more-informed investors at
steady state prices.
H1: Less-informed investors transfer wealth
to more-informed investors at steady-state
prices in the absence of guidance on signal
reliability.
We examine wealth transfers by constructing
markets in which two more-informed investors
hold all three value-relevant signals and two less-
informed investors hold only one of those signals.
For each security traded, di?erent groups of
investors trade the same security in settings that
are identical except for which of the three signals
the less-informed investors hold. Because the
information available to the more-informed inves-
tors is the same (and therefore the total informa-
tion available to the market is the same), the
informationally ecient price is the same in both
settings. However, the favorability of the signal
held by the less-informed investors di?ers depend-
ing on which signal the less-informed investor
holds. While the informationally ecient price is
the same in both settings, as we note above, we
expect prices to be biased in the direction of the
signal held by the less-informed investors. We also
expect the less-informed investors' estimates of
value and trading to be more biased than that of
the more-informed. As a consequence, they will
transfer wealth to the more-informed by buying
over-priced and selling under-priced securities to
the more-informed.
Presumably, wealth transfers could be arising at
every point throughout the trading process. We
5
Stronger reactions to public information than to private
information do not preclude under-reactions to public infor-
mation. A market could under-react to all information avail-
able, but under-react less to public information.
6
If less-informed and more-informed investors were equally
over-con®dent, both would trade more aggressively than they
would otherwise and the e?ects of over-con®dence would o?set.
626 R. Bloom®eld et al. / Accounting, Organizations and Society 24 (1999) 623±647
focus only on wealth transfers that occur after
many rounds of trading have led the market to a
``steady-state'' price at which supply equals
demand, because these prices should represent the
most informationally ecient prices, and therefore
should price-protect less-informed investors to the
greatest extent. Also, previous research has
already established that uninformed investors earn
less than better-informed investors before infor-
mation is impounded in market prices as com-
pletely as it can be (Forsythe & Lundholm, 1990;
Plott & Sunder, 1982). If less-informed investors
are unable to protect themselves from welfare los-
ses at prices that are as informationally ecient as
possible in our market setting, it seems unlikely
that they would avoid losses before that point.
The studies cited above have found no sig-
ni®cant welfare di?erences between informed and
uninformed investors once prices reach steady-
state levels. For example, Plott and Sunder (1982)
®nd that closing prices are so informationally e-
cient that all traders earn approximately the same
amount from trade. Experienced subjects in For-
sythe and Lundholm (1990) reveal a similar result.
However, the simplicity of the information environ-
ments in these studies make it dicult for psycho-
logical biases such as over-con®dence to a?ect
prices and welfare. For example, in many of these
experiments less-informed investors are entirely
uninformed, which would discourage any attempts
by them to trade on their own information. Also,
more-informed investors in those markets receive
information that is so informative and objective
that there is little room for them to exercise their
judgement in interpreting it. [For example, sub-
jects in most of the markets in Plott & Sunder,
Forsythe & Lundholm, and Lundholm (1991)
learn the true value with certainty, learn that the
security has one of two values with certainty, or
receive no information at all about security value.]
One of the goals of this paper is to explore psy-
chological biases that arise when investors must
exercise judgement in interpreting their information.
2.3. Guidance
Understanding the psychological cause of poor
decision performance facilitates identi®cation of a
decision aid that is designed to improve perfor-
mance (Bonner, Libby & Nelson, 1996). If overly-
aggressive trading behavior by less-informed
investors is in fact driven by the relative over-
con®dence of less-informed investors, then helping
these investors assess the reliability of information
should diminish their welfare losses. Two ®ndings
in the psychology literature suggest a simple and
e?ective form of guidance that should reduce the
con®dence of less-informed investors. First, a
large number of studies have examined the e?ec-
tiveness of di?erent forms of feedback in multiple
cue probability learning tasks. This literature is
aimed at determining what feedback conditions
promote or deter learning in tasks similar to our
trading settings, where participants predict uncer-
tain outcomes based on cues with di?erent
reliabilities [see Libby (1981, Chapter 2) for a
description of typical studies and results]. Many of
these studies compare the e?ectiveness of ``out-
come'' feedback (where subjects are given the true
value after each case) with ``task-properties'' feed-
back (where the relative reliabilities of the cues are
provided). In general, learning and performance
are far superior in the task-properties-feedback
condition.
7
Second, while the probability judgment
literature has generally found over-con®dence to be
dicult to debias, a recent study by Tra®mow and
Sniezek (1994) found that providing subjects with
information about their general knowledge level (a
simple form of task properties feedback) reduces
subjects' con®dence, and thus subjects' over-con-
®dence. These ®ndings suggest that providing less-
informed investors with guidance about the rela-
tive reliability of their information may reduce the
relative over-con®dence of these investors com-
pared to that of more-informed investors. This, in
turn, should reduce the bias in their value esti-
mates and the aggressiveness of less-informed
investors' trading decisions, and reduce their
wealth transfers to the more-informed investors.
If the trading bias predicted by H1 is in fact
driven by the relative over-con®dence of less-
informed investors, then providing guidance
7
See Bonner and Walker (1994) for similar ®ndings that
show the advantage of explanatory feedback over outcome
feedback.
R. Bloom®eld et al. / Accounting, Organizations and Society 24 (1999) 623±647 627
concerning the reliability of their information
should diminish less-informed investors' welfare
losses. To address this issue, we manipulate whether
or not less-informed investors are given information
about the reliability of the signals that are available
to the market. If trading bias is driven by the relative
over-con®dence of less-informed investors, then gui-
dance should make those investors less con®dent,
and reduce this bias.
H2: Providing less-informed investors with
guidance on signal reliability reduces wealth
transfers fromless-informed investors to more-
informed investors at steady-state prices.
We also manipulate whether or not more-informed
investors are given guidance on signal reliability.
However, we make no hypothesis about the e?ect of
guidance on more-informed investors, because that
e?ect depends on whether or not more-informed
investors who lack guidance are over-con®dent. If
more-informed investors tend to be over-con®dent
in the absence of guidance, then providing guidance
to more-informed investors will decrease their
over-con®dence. This in turn may increase wealth
transfers by increasing the over-con®dence of less-
informed investors relative to the con®dence of
more-informed investors. On the other hand, if
more-informed investors tend to be undercon®dent
in the absence of guidance, providing guidance to
more-informed investors may increase their con-
®dence and decrease wealth transfers by decreasing
the over-con®dence of less-informed investors rela-
tive to the con®dence of more-informed investors.
Finally, more-informed investors who lack gui-
dance may accurately assess the reliability of the
information they hold, yielding no guidance e?ects.
We cannot discriminate between these predictions,
so we state no hypotheses about the e?ect of pro-
viding guidance to more-informed investors.
3. Method
3.1. Subjects
All subjects (``investors'') were MBA students at
Cornell's Johnson Graduate School of Manage-
ment who had participated in at least one prior
experimental market.
3.2. Security values and information
In our markets, investors make judgments about
securities whose values are derived from values of
real-world securities, given ®nancial information
derived from ``Value-Line'' analyses of those
securities. Speci®cally, security values are deter-
mined by each ®rm's price/book ratio, and inves-
tors are given information on the ®rm's return-on-
equity, growth in return-on-equity (as projected
by Value-Line), and book value per share.
We attempt to present investors with a valua-
tion task similar to that facing investors in real-
world exchanges, while still ensuring that the
informative signals presented to investors are easy
to understand, can explain an economically sig-
ni®cant portion of variation in security values, and
possess a minimal amount of colinearity and
interdependence. We achieve these goals by deriv-
ing security values from price/book ratios and
deriving signals from common accounting ratios.
In this way, we are able to use extant theory and
empirical work (Bernard, 1994; Frankel & Lee,
1998) to develop a small number of signals that
explain a majority of variation in price/book ratios.
8
To obtain our securities and signals, we ®rst
construct a database which includes all 291
8
Using information derived from real-world sources (as
opposed to more objective information such as random num-
bers) allows investors to determine the meaning and statistical
reliability of their signals by relying on their own knowledge
and experience. These are the circumstances in which over-
con®dence tends to be observed. A real-world valuation task
should also increase the motivation of our subjects, and allow
the study to provide some educational bene®t to the subjects,
who are MBA students. These advantages are diminished to the
extent that the subjects are unfamiliar with the nature of price/
book ratios, and to the extent that the signals we use di?er
from the naturally occurring signals reported in Value-Line.
However, we have designed our experiment so that these pos-
sibilities cannot account for welfare losses by the less-informed
investors. In particular, welfare losses require the less-informed
investors to be more over-con®dent than the more-informed.
While unfamiliarity with the information signals could cause all
subjects to be over- or under-con®dent, it is not clear how it
could cause a greater di?erence in the con®dence levels of the
two groups.
628 R. Bloom®eld et al. / Accounting, Organizations and Society 24 (1999) 623±647
manufacturing ®rms which were pro®table in 1993
and covered by Value-Line reports. We rank these
®rms according to their price/book ratios (the
most recent market price available prior to pub-
lication of the Value-Line report divided by the
net book value per share). The percentile rank of
each ®rm (denoted PRICE/BOOK) is then used as
the value of the associated security. For example,
a ®rm with a price/book ratio greater than 24% of
all ®rms in the database is given a value of 24
francs (a ®ctitious currency ultimately converted
to cash). The ®rst signal providing information
about PRICE/BOOK is an ROE score which
re¯ects the ®rm's relative return-on-equity (net
income before extraordinary items divided by net
book value) relative to other ®rms in the popula-
tion. The second signal is a GROWTH score,
which re¯ects the rate at which Value-Line pro-
jected return-on-equity for the ®rm to grow from
1992 to 1994. The third signal is a BOOK score,
which re¯ects the relative book value for each
share of the ®rm's stock. We also use a simple
orthogonalization technique to ensure that the
signals are largely independent, and present all
scores on an 11-point scale. This orthogonaliza-
tion should increase the e?ectiveness of our
manipulation, because it insures that less-
informed investors are signi®cantly less informed
than more-informed investors.
9
Appendix A con-
tains instructions to subjects which describe the
construction of the securities, including the ortho-
gonalization technique, in detail.
Panel A of Table 1 provides the results of a lin-
ear regression of PRICE/BOOK onto ROE,
BOOK and GROWTH. Consistent with empirical
research, all three of the signals are positively and
statistically signi®cantly related to PRICE/BOOK.
Overall, the three signals explain 60.7% of the
variation in security value. Panel B of Table 1
shows the ten securities used in the experiment.
The securities were selected randomly and have
values throughout the 100-franc scale and signals
with realizations throughout their 11-point scales.
3.3. Trading
Subjects trade in computerized specialist mar-
kets similar to those used by Bloom®eld (1996b)
and Bloom®eld and Libby (1996). These markets
are based on the New York Stock Exchange, in
which a specialist quotes prices at which investors
can trade, the specialist ®lls all orders investors
tender at those prices, and the specialist changes
those prices in response to investors' previous trades.
In our laboratory markets, each group of four
investors (which we call a cohort) always trades
together. Each of the 10 securities (a claim on a
terminal dividend) they trade is denominated in
``francs.'' Trading is a batch process Ð all four
investors in a cohort enter their buy or sell orders
before any investor chooses their next action.
Trading round refers to each time all investors in
the cohort enter buy and sell orders for shares of a
security. Each security is traded for 15 rounds
before trading begins in the next security. We refer
to the sequence of 15 rounds of trading for a single
security as a market.
Before learning the market price for each round
of trading, investors enter their best estimates of
the security's value. These estimates allow us to
test whether trading losses are driven by biased
estimates of value. Investors then enter orders to
buy (sell) from 0 to 10 shares if their estimate is
above (below) the market price, but cannot trade
if their estimate is exactly equal to the market
price. After each round, investors learn the aggre-
gate numbers of buy and sell orders at the pre-
vious price. The specialist ®lls all orders.
Our specialist market allows shortselling, so the
number of shares bought and sold at a given price
is not in¯uenced by either wealth constraints or
how shares are distributed among investors.
Ganguly, Kagel and Moser (1994) show how
short-selling limitations can allow biases to persist,
because investors who wish to exploit high prices
cannot do so unless they own shares. The lack of
shortselling restrictions (beyond the maximum 10
share trade/round) allows investors in our markets
to exploit price errors in any direction, regardless
9
If the single signals possessed by less-informed investors
were highly correlated with the three signals held by more-
informed investors, there would be little di?erence in the
amount of explanatory power of the information held by less-
and more-informed investors, and consequently a weak treat-
ment e?ect.
R. Bloom®eld et al. / Accounting, Organizations and Society 24 (1999) 623±647 629
of their holdings, so prices in our markets provide
a less noisy re¯ection of investors' value estimates
and their con®dence in those estimates.
10
Trading for a security always begins at 50
francs. To set the price for the next round, the
computer revises the price upward (downward) if
there are more (fewer) buy orders than sell orders,
with the magnitude of movement an increasing
function of the imbalance.
11
Under this price revi-
sion rule, prices should not settle at a level which
di?ers substantially from investors' estimates of
value Ð if the price is too high (low), investors will
drive it downward (upward) until they can achieve
no further gains. Eventually, the market settles into
a narrow range of prices at which average supply
roughly equals average demand. We call the aver-
age of these prices the ``steady-state'' price. Trad-
ing continues for 15 rounds, which we expected
would be long enough to allow several rounds of
trading after the steady-state price is reached.
3.4. Instructions
The ®rst 20 minutes of each experimental ses-
sion are spent reviewing two sets of instructions.
The ®rst set of instructions explains the computa-
tion of gains and losses from buying and selling,
the computation of gains from accurate estimates
of security value, the trading process, the mechan-
ism by which prices are adjusted after each round
of trading, and the conversion of winnings from
Table 1
Relation of signals to security value.
a
Panel A
b
: estimation of regression model
ROE
GROWTH
BOOK
R
2
À0.06 0.058 0.020 0.035 60.7%
(À2.06) (17.21) (5.75) (9.99)
p=0.04 p