Journal or Applied Psychology
1984. Vol 69 No 3, 531-545
Copyright 1984 by the
American Psychological Association, Inc
When Feedback is Ignored: Disutility of Outcome Feedback
Jacob Jacoby
David Mazursky
Department of Marketing,
New York University
Department of Marketing,
Hebrew University, Jerusalem, Israel
Tracy Troutman
Alfred Kuss
Department of Marketing,
New York University
Institute for Marketing and Consumer Research,
The Free University, West Berlin, West Germany
Following Hammond, McClelland, and Mumpower (1980), two types of feedback
are conceptualized—outcome feedback and cognitive feedback The latter is here
hypothesized as having either predictive or explanatory value Using security analysts
participating in a security analysis decision simulation, the hypothesis that, in
contrast to poorer performing decision makers, better performing decision makers
are more likely to ignore outcome-only feedback was confirmed (r = —.48, p =
02). Implications for theory revision and future research are discussed
The concept of feedback has long occupied
a hallowed place in the literature of applied
psychology. Despite some evidence to the contrary (e.g., Einhorn & Hogarth, 1978; Hammond & Summers, 1972; Steinman, 1976),
both the human performance and organizational behavior/management literature suggest
that feedback can exert a positive effect on
performance (e.g., Ammons, 1956; Annett,
1969; Carver & Scheier, 1982; Greller, 1980;
Ilgen, Fisher, & Taylor, 1979, Ivancevich &
McMahon, 1982; Nadler, 1979)
The generally accepted feedback-performance relationship implies that, other things
(e.g., task-related experience) being equal, better performing individuals would be more
likely to access and use feedback information
This is consistent with the notion that "Feedback is central to the learning of expertise"
(Hogarth, 1981, p 202) There are, however,
reasons for arguing just the opposite, especially
in situations that involve complex, cogmtively
rich decision making. Providing the rationale
for this argument requires a more detailed
consideration of the concept of feedback itself.
This work was supported, in part, by funding from New
York University's Institute of Retail Management
The authors acknowledge with appreciation the assistance provided by their colleagues, Professors Fred Renwick
and Arnold Sametz, in arriving at the set of 26 fundamental
factors used in this investigation
Requests for reprints should be sent to Jacob Jacoby,
Department of Marketing, New York University, Washington Square, New York, New York 10003
The cybernetic theorist Wiener (1948, 1950,
p. 35) is usually credited with introducing the
concept of feedback to the behavioral sciences
and popularizing its use (cf. Nadler, 1979)
Although there might be minor disagreement,
most would concur with the following description provided by Ilgen et al. (1979, p. 351):
"At its most basic level, feedback is information received by an individual about his or
her past behavior (Annett, 1969). It provides
some information about the correctness, accuracy or adequacy of the response (Bourne,
1966)" In addition to the central notion of
"information regarding the accuracy of response," Ilgen et al. conjectured that feedback
also possesses other characteristics. Specifically, they referred to the "information value"
of feedback, which they said depends on "the
incremental increase in knowledge about performance that the feedback provides the recipient" (p 351).
The notions of "accuracy of response" and
"information value" find parallels m the Social
Judgment Theory concepts of "outcome feedback" and "cognitive feedback," respectively
(cf. Hammond, McClelland, & Mumpower,
1980, p. 228). Whereas outcome feedback is
information that describes the accuracy or
correctness of the response, cognitive feedback
represents information regarding the how and
why that underlies this accuracy
Diagnostic or "cognitive feedback" may also
be considered in terms of the concepts of predictive and explanatory value. Predictive value
531
532
JACOBY, MAZURSKY, TROUTMAN, AND K.USS
refers to the correlation between accurate in- maker) rather than static. The securities marformation regarding one's past performance ket and security analyst decision making seem
and the ability of this information to predict to satisfy these criteria. As Slovic (1969, p. 79)
future levels of performance. In contrast, ex- comments:
planatory value represents information reanalysis, whether by expert or novice, might be
garding why certain relationships occurred. Security
labeled "the information game" In no other realm are
This may or may not be accompanied by an such vast quantities of information from such diverse
ability to predict if (or when) these relation- sources brought to bear on so many important decisions
ships will occur in the future, that is, predictive Careful accumulation and skilled interpretation of this
and explanatory value are conceptually in- information is said to be the sine qua non of accurate
evaluation of securities
dependent. By way of example, a parent might
be able to answer a child's question as to why For this reason, security analyst decision makan eclipse of the sun has occurred but not be ing has been found quite useful for testing
able to predict when it will occur again (i.e., various psychological propositions (e.g.,
explanatory value). In contrast, the same par- Clarkson, 1962; Ebert & Kruse, 1978; Kozent might be able to demonstrate to his auto minsky, Kintsch, & Bourne, 1981; Slovic,
mechanic the existence of a predicted rela- 1969, 1972; Slovic, Fleissner, & Bauman,
tionship between stepping on the gas pedal 1972).
and hearing a strange rattle yet be unable to
The types of information believed to have
explain why this has occurred (i.e., predictive an impact on stock prices are generally clasvalue).
sified into one of three categories: market facHammond and Summers (1972), Ham- tors, industry-specific factors, and companymond, Summers, and Deane (1973) and unique factors. Market factors refer to the genSteinman (1976) found that when predictive eral economic and political background and
value is low, providing feedback after each trial include such world and national events as wars,
can be both misleading and detrimental to changes in governments, monetary policy, or
learning a functional relationship. This sug- short-term interest rates. Industry-specific
gests that, given a situation involving experi- factors refer to trends and developments afenced decision makers who have access to task fecting an industry as a whole (e.g., the impact
environment feedback that is neither predictive of Japanese exports on the U.S electronics
nor explanatory, the better performing decision and computer industry; deregulation of the
makers are more likely to be the ones who airline and trucking industries, etc ), as well
have learned to disregard this feedback. More as the activities of the various competing firms
specifically, a principal objective of the present within that industry. Company-unique factors
investigation is to test the hypothesis that task may be broken into two categories: those inenvironment feedback providing completely volving consideration of a specific company's
accurate response information (i.e., outcome financial statement (often referred to as "funfeedback) but containing neither predictive nor damental analysis") and those involving nonexplanatory (i.e., cognitive or information) financial matters, such as the quality of its
value is less likely to be used by better as com- management or whether its labor force is about
pared to poorer performing decision-makers to negotiate a new contract.
Real-world decision makers operating in a
There seems to be increasing agreement that
task environment that was both cognitively "market factors have very little to do with incomplex and dynamic were considered best vestment performance" (Hagaman & Jensen,
for testing the hypothesis. This is because it 1977, p 64). According to King (1966) and
is further anticipated (though not tested by Blume (1971), general market factors explain
the present investigation) that the hypothesized only about 30% of the total variation in the
relationship is more likely to be manifested price of a particular company's stock. Further,
with task environments that are relatively industry-specific factors seem to explain percomplex (i.e., ones in which the task environ- haps another 12% of this variance (King,
ment is rich in information) rather than sim- 1966). "The implication of these findings is
ple, and are dynamic (i.e., actively changing that 50% of the variance in stock prices is due
independent of the actions of the decision to factors unique to the specific firm" (Tersine
533
WHEN FEEDBACK IS IGNORED
&Celec, 1976, p. 32). Accordingly, it becomes
important to learn just which types of company-unique factors are most useful, since
analysis of these fundamental factors seems to
represent the key to rational investment decisions Regarding the two categories of company-unique factors, Bernstein (1975) comments:
change stocks to test the hypothesis that completely accurate outcome feedback from the
task environment that contains neither predictive nor explanatory value is less likely to
be used by better as compared to poorer performing security analysts.
Method
Even assuming that psychological and other (I e , market
and industry-specific) factors do account for as much as
50% of price changes, those who abandon fundamental
analysis on the basis of such an argument abandon control
over the other 50% of their investment decisions In fact,
they surrender much more because the fundamental factors
affecting a business entity lend themselves to a much more
detailed, disciplined and systematic analysis than do some
of the extreme factors affecting the price-earnings ratio
Thus fundamental analysis is a necessary ingredient of a
rational and prudent investment decision (p 60)
Sample
Subjects were recruited through mailed invitations to
professional security analysts working for different investment firms in the Wall Street area These letters indicated that research was being undertaken on security
analyst decision making and that this research would take
the form of a competition The core of the letter reads as
follows
The purpose of this letter is to invite you to participate
in this competition The winner will receive a $500
prize and a press release identifying the winner will be
distributed to the appropriate media
Briefly, each competitor will be confronted with the
same set of eight common stocks, representing eight
consumer oriented firms listed in the NYSE, and be
given access to 26 different types of fundamental factors
regarding each of these stocks The task will be to identify
Based on the above review, fundamental
factors can thus be assumed to account for at
least 30% to 35% of the variance in security
analyst decision making (see Figure 1). Accordingly, this study used actual fundamental
factor information for New York Stock Ex-
30% CA.
/
/
NATIONAL AND
INTERNATIONAL
EVENTS
'
1
2
3
4
5
/
/
/
/
/
10-12% CA.
INDUSTRY-SPECIFIC
FACTORS
/
'
'
COMPANY-UNIQUE
FACTORS
58-60% CA.
1. FUNDAMENTAL
FACTORS
(30-35%)
2 OTHER FACTORS
(E G , IMPRESSIONS OF
MANAGEMENT QUALITY)
(20-25%)
FUNDAMENTAL FACTORS
v
\
\
\
\
\
\
P/E RATIO FOR LAST 12 MOS
%PR CHG LAST 3 MOS
PRICE LAST MONTH
LAST EARN TREND (UP, DOWN, NO CHANGE)
EARN /COM SHR 12 MOS LAST QTR
6
7
8
9
10
ADJTD ANNL EARN $/COM SHR , PAST 4 YRS
INT'MEARN LASTREPTDJ/COM SHR RECENT REPORT
INT'M EARN PREV YR $/COM SHR
YLD ON IND1CTD 12 MOS OF DIV
EARN/COM SHR 12 MOS ENDING NEXT TO LAST QTR
11
12
13
14
15
$ NET PR CHG (LAST 9 M O S )
PR RANGE CURRT YR HI LO
LONG TERM DEBT
TOTAL CURRENT ASSETS
TOTAL CURRENT LIABILITIES
16
17
18
19
20
PR RANGE LAST 10 YRS HI LO
LAST DIV TRND (UP, DOWN, NO CHANGE)
CASH AND EQUIVALENTS
DIV ($ PER COM SHR , LAST YR ANN'L RATE)
NO OF COMMON SHARES
21
22
23
24
25
NO OFINST HOLDINGS
NO OF SHRS HELD BY INST
CASH DIV /COM SHR LAST YR ANN'L RATE
NO OF PREFERRED SHARES
CASH DIV/COM SHR THIS AND LAST CLNDR YRS
\26
DATE OF EX DIV
Figure 1 Elements incorporated into security analyst decision making
534
JACOBY, MAZURSKY, TROUTMAN, AND KUSS
the "best buy" from among the set of eight stocks, not
once, but for each of four different time periods spaced
90 days apart
. The analyst whose four decisions
produce the highest cumulative percentage growth over
the four test periods will be declared the winner
The $500 prize and promised press coverage were both
used to instill a degree of realistic motivation The industry
is accustomed to providing recognition in the form of
annually published and widely disseminated lists of top
performing analysts Furthermore, the $500 prize meant
that there was something of consequence riding on the
decision Hence, each participant could be expected to
treat the task seriously, thereby satisfying the "consequentiahty" desideratum (cf Jams & Mann, 1977, p 69)
More than twice as many analysts returned response
forms volunteering to participate as the number that actually participated The invitation letters were mailed near
the end of an extended period of sluggish market activity
(July, 1982) By the time most were phoned to schedule
a test interview, the market experienced a dramatic resurgence in activity and many analysts who had earlier
replied that they would participate then chose to decline
Second, because analysts had to be tested individually,
scheduling problems made it inconvenient for others who
had expressed a willingness to participate to actually do
so Third, the study was computer-administered and several
computer malfunctions, both between subjects and during
the time a subject was actually participating, created further
decreases in sample size Notwithstanding these problems,
the obtained sample of 17 is still considerably larger than
that employed by either Clarkson (1962), Slovic (1969),
or Ebert and Kruse (1978) and is only one less than that
employed by Slovic et al (1972)
The sample consisted of 3 women and 14 men Seven
were between 21 and 30 years old, another 7 were between
31 and 40, and 3 were over 40 Their careers as professional
security analysts ranged from 1 5 to 17 years (M = 6 9
years, SD = 5 7), and the amount of time spent working
with their present employer ranged from (L months to 6
years (M = 2 3 years, SD = 1 8) Fifteen of the analysts
had master's degrees (14 M B A s and I M S ) , two held
bachelor's degrees Two analysts declined to respond to an
item regarding the income they derived from their activities
as professional security analysts Another 8 indicated that
their income was below $75,000 a year, 7 indicated that
their income was above this amount
Test Setting
Testing took place in a specially equipped test facility
at NYU's Graduate School of Business Administration,
immediately adjacent to the American Stock Exchange,
a site conveniently situated for most firms in the Wall
Street area
Task and Instructions
Each analyst participated in a behavioral process (BP)
simulation A general overview of the BP approach was
provided in Jacoby, Chestnut, Weigl, and Fisher (1976)
More recent papers (e g, Chestnut & Jacoby, 1982, Jacoby,
1977, Jacoby, Chestnut, Hoyer, Sheluga, & Donahue, 1978,
Major, 1980, Sheluga, Jaccard, & Jacoby, 1979) will guide
the reader to more than 20 investigations that have employed this approach
The analysts were tested individually Upon arriving at
the test facility, analysts were asked to read the following
brief descriptions of the simulation and the rules for the
competition
The purpose of this task is to examine how professional security analysts make decisions as to which common stocks represent a good purchase
In a short while, you will be asked to decide which
one of eight stocks represents the "best buy" For purposes of this study, the best buy is denned as that stock
most likely to show the greatest percentage of growth in
price per share over the next 90 day period
The eight stocks represent major U.S retailing firms
listed on the New York Stock Exchange during the years
1969-1970 This time frame coincides with the first
term of the Nixon presidency, the Vietnam conflict, and
was generally a period characterized by economic expansion and growth So that your knowledge regarding
the present condition of these firms does not influence
your decision-making, we have given them fictitious
names However, all the information to which you will
have access is authentic and is the actual information
that was available for these stocks during the 1969-1970
period
You will be able to acquire information on any of
26 different fundamental factors for each stock A list
of these factors is provided at the back page of these
instructions While we recognize that professional analysts often integrate yet other kinds of information into
their decision making, for purposes of this task, you are
asked to arrive at a decision based only on the 26 fundamental factors that are available
Thus, you will have access to 208 different items of
information—that is, 26 items of information for each
of 8 stocks (=208 total items of information) All this
information is stored in the computer Since it would
be unreasonable to expect you to keep all the information
in your head while you're arriving at your decision, a
set of recording sheets has been provided so that, if you
wish, you can make notes
You will actually be making four different "best buy"
decisions The first will be for October 1969 Once you
have made this decision, the computer will update the
available information by three months (so that they
reflect the data existing as of January 1970), and you
will be asked to arrive at a second "best buy" decision
regarding these same eight stocks After you have reached
this second decision, the information will again be updated by three months to reflect April 1970 data This
cycle is repeated for a total of four periods In all, you
will be asked to make four "best buy" decisions regarding
the same set of 8 stocks, beginning with October 1969
and ending with July 1970
Throughout, there are five basic rules you need to
bear in mind as you go about obtaining information
from the computer These are
1 You may obtain information only for the 3-month
period provided That is, you will not be able to go
backward to access information from an earlier period,
or forward to access information for a future date
2 Within each period, you are free to examine up
to 104 of the 208 items of information available There
535
WHEN FEEDBACK IS IGNORED
is no "correct" or "right" amount of information to
look at You may choose to look at 2 items, 20 items,
or 100 items, It's all up to you
3 You may examine the available information in
any order you want
4 You may examine an item of information more
than once That is, you may return to look at information
you have seen before Simply instruct the computer to
display that item again
5. You are free to make as many or as few notes on
the recording sheets as you like
So that you can become familiar with how to operate
the computer, we will begin with a short practice task
involving three stocks taken from the airline industry
Although it was estimated that each session would require I'h to 2'h hours, no time limit was imposed The
amount of time actually used ranged from V/i to 6V2 hours
The 104-item limit noted in the second rule was necessitated by the storage limitations of the computer, as described below
Information Environment
For each test period, the information environment confronting the analyst consisted of 208 separate items of
information—26 fundamental factors for each of eight
companies To eliminate the possibility of extraneous confounds, all the stocks were drawn from the same industry
To reduce the possibility that the respondents' memories
would exert an effect on their information accessing, the
data were taken from the period spanning October 1969
through July 1970 The only information that was not
authentic was the names of the firms; the letters J, K, L,
M, N, Q, R, and S were used instead of firm names The
names of the firms actually involved were Lane Bryant,
Gimbels, Hughes and Hatcher, R H Macy, J C Penney,
Sears, Roebuck and Co, F W Woolworth, and Zayres
The fundamental factor information regarding each of
these companies was taken from the "Monthly Stock Digest," which was published by Data Digests (1969-1970)
and mailed monthly by Merrill Lynch to its clients This
information was taken from the October 1969, January
1970, April 1970, and July 1970 issues It should be noted
that test Periods 1, 2, and 4 reflected a relatively stable
to "bullish" market, whereas Period 3 reflected a "bearish"
market, that is, all eight stocks decreased in value during
this period The 26 fundamental factors that were used
were selected after both a review of the literature and
consultation with knowledgeable colleagues revealed that
these 26 factors were rather comprehensive in terms of
covering factors typically considered by security analysts
(see Figure 1) Note, also, that one of the factors,"% price
change over the past three months," was a precise statement
of the performance criterion, that is, the analysts were
instructed to select "that stock most likely to show the
greatest percentage of growth in price per share over the
next 90 day period "
All information was stored in a Cromemco Z-2H microcomputer This enabled the order of presentation of
both the stock and the fundamental factors to be randomized across respondents, that is, though each respondent received the same eight stocks and 26 fundamental
factors, these were presented in different orders for each
respondent To avoid confusing the respondents, the order
remained constant for each respondent across the four
separate test periods for that respondent
Procedure
Respondents commenced reaching their "best buy" decision for Period I by communicating with the computer
via a light pen attached to a color video monitor After
an initial series of person-machine interactions to review
the task instructions and rules, two "menu" lists were
displayed—one containing the names of eight stocks, and
the other being a list of 26 fundamental factors—and the
analyst was asked to indicate which type of fundamental
factor information he or she wanted to see for which security The analyst was permitted to access only a single
item of information at a time, for example, "price/earnings" information for Company J Accessing of this information was accomplished by touching the screen twice
with the light pen—once to identify the stock for which
the information was being requested and a second time
to identify the fundamental factor—and the desired information appeared on the screen almost instantaneously.
After acquiring this first item of information, the computer inquired whether the analyst wanted to make a best
buy recommendation at that point or wanted to acquire
additional information If the analyst replied that he or
she wanted to acquire information, the two lists were displayed again, and the analyst used the light pen to indicate
which item of information he or she next wished to consider Each analyst continued in this way until the point
at which he or she felt ready to make a "best buy" recommendation This, too, was indicated via the light pen
After the analyst arrived at a "best buy" decision for Period
1. the information for the eight securities was automatically
updated by three months and the analyst proceeded m a
similar manner to arrive at "best buy" decisions for Periods
2, 3, and 4 As a result of having worked through the brief
practice task, which served as a warm-up exercise, none
of the analysts evidenced any difficulty with this procedure
Following completion of all four periods, the analyst
completed a brief questionnaire, which inquired about the
task, collected certain demographic and employment information, and included the following manipulation check
as the very first question "Did you think you knew the
identity of any of the stocks''" In all there were (8 stocks X
17 analysts) 136 opportunities for the analysts to correctly
guess the identities of the test stocks Only one of the
analysts responded "yes," and was directed to answer the
following question "If yes, please indicate the name of
the firms that you think were involved by writing the
names of these firms next to the corresponding letter"
This analyst wrote in the name of only one firm, which
was correct This manipulation check thus revealed that
the tactic of using letters to camouflage the identities of
the stocks worked as intended
Results
The procedures described above generate
voluminous quantities of data, much of which
is relevant for answering other questions. A
detailed report of much of these data is being
536
JACOBY, MAZURSKY, TROUTMAN, AND KUSS
Table 1
Depth of Search (Total Sample)
Information accessed
Total no of items acquired (with
redundancies)
Range
M
Mdn
Total no of different items
M
Mdn
Overall
Period 1
Period 2
Period 3
Period 4
4-105
42.6
38 5
4-97
62 5
60 0
10-96
35 6
28.5
14-105
34.5
315
6-98
38 1
32 0
41.1
36 0
59 3
59 5
34 5
26 0
33 8
30 5
36 8
315
197
173
28 5
28 6
166
12 5
16 3
147
177
15 1
79
80
78
80
80
80
80
80
76
80
79
75
11 4
10 5
65
70
71
65
66
55
65 9
60 0
66 7
70 8
66.3
46 4
59 5
58 6
73 6
71.6
% of full S X F matrix accessed
M
Mdn
No of different securities considered
M
Mdn
No of different factors considered
M
Mdn
% of submatnx accessed
M
Mdn
provided elsewhere (Jacoby, Kuss, Troutman,
& Mazursky, 1984). Except for providing a
suitable context by briefly summarizing some
of the findings bearing on the depth of search,
The present report focuses exclusively on the
data relating to feedback.
Background
Overall depth of search Table 1 summarizes the descriptive data regarding the amount
of information acquired and provides perspective across all four test periods. Several
findings are noteworthy. First, on average and
across all four periods, analysts accessed 42.6
items of information per test period. (This
mean decreases to 41.1 when accessing of a
previously accessed information is excluded.)
In other words, the analysts accessed less than
20% of all the available information.
Second, m most cases the respondents considered information for each of the eight stocks
at least once. However, on average, they considered fewer than 8 of the 26 fundamental
factors that were available {M = 7.9; Mdn =
7.5). This latter finding compares quite favorably to those reported by Slovic (1969).
Third, dividing the mean number of stocks
considered (7.9) into the mean number of
items considered (411) indicated that, on av-
erage, the analysts paid attention to only five
items of information for each of the stocks
that they considered.
Fourth, the 26 X 8 information environment represents an investigator-oriented perspective on depth of search. However, not all
the available information is necessarily useful,
worthwhile, or even meaningful to all respondents Hence, it is also insightful to adopt a
respondent-oriented perspective and ask: If attention were limited to only those stocks and
factors considered by the analyst at least once,
then what percentage of the information from
that submatrix was accessed? Using this perspective, search seems to have been relatively
extensive. On average, 66% of each analyst's
own submatnx was accessed, indicating a tendency for the analyst to compare each of the
stocks along a subset of similar factors.
Comparing depth of search across test periods The breakdown for periods shows relatively extensive search in the first period
(mean number of items accessed = 62.5), followed by a sharp and substantial decline for
the next three periods (means = 35.6, 34.5,
and 38.1, respectively). This trend may be explained as follows.
Several of the fundamental factors represented annual information that did not change
throughout the four test periods (or at least,
537
WHEN FEEDBACK IS IGNORED
for two successive periods). Thus, once accessed, several analysts tended not to reaccess
this information in subsequent periods. Moreover, although Periods 1, 2, and 4 reflected
either a steady or increasing market, the third
period represented a sharp decline in stock
prices. Consequently, analysts may have engaged in more extensive information acquisition in the fourth period as a result of obtaining negative feedback on their preceding
(third period) performance.
The general trend (of a sharp decline in
depth of search after the first period) is also
consistent with traditional learning theory. Although it could be assumed that the participating analysts were unfamiliar with all the
stocks during the first period, their increased
familiarity resulted in more efficient decision
processes in subsequent periods. The slight
deviation during the fourth period can be explained by sharp decline in stock prices at the
end of the third period
Operationahzing Performance
The performance criterion used to distinguish between the better and poorer analysts
was based directly on the increases in price
per share of each security, calculated separately
for each of the four test periods The analyst
whose four choices produced the greatest net
yield was declared the winner. The yields that
were actually possible ranged from +33.1%
to -76.3%.
Several interesting findings emerge from a
consideration of this performance index. First,
had they invested money in a real-world situation as they recommended in the task, only
two of the analysts would have managed to
make money over the 12-month period; the
other 15 would have lost money—seven of
them would have lost 40% to 55% of the original value of their stocks. Much of this loss
was experienced during Period 3, which was
characterized by a sharp market decline. In
all fairness to the participants, analysts operating in a real-world environment are able
to more closely monitor the day-to-day performance of their stocks and would not wait
up to 3 months to divest themselves of a sinking security. If calculation of the performance
index is based only on Periods 1, 2, and 4,
then the numbers are reversed, with 15 out of
Table 2
Comparing Analyst Performance to a
Baseline Random Choice Strategy
No of
successes
(r)
Pr (r/p = 1/8)
Expected
no of
analysts
(Pr X 17)
4
3
2
1
0
.0002
0068
0718
3350
5862
003
120
122
5 68
9 97
Total
17 00
Obtained
no of
analysts
0
0
1
10
6
17
Now Pr = performance
17 analysts having positive indices (range =
+35 4 to - 7 . 2 ; M = +14.25).
Second, performance was not related to
several key characteristics Across the entire
sample of 17 analysts, performance correlated
negatively and nonsigmficantly with age (r =
—.22; p = .40) and tenure with present employer (r = - . 3 7 ; p= .14) and correlated negligibly with either tenure as an analyst (r =
.03; p = .92) or income (r = .04; p = .89).
Suggested by one anonymous reviewer, additional perspective on performance is provided by comparing the actual performance
of the analysts against a random choice strategy
where the probability of picking the winning
stock in each period is p = 1/8. Therefore, for
each analyst, the probability of getting r successes in four trials can be computed using
the binomial expansion. As can be seen from
Table 2, though the obtained distribution reflected slightly better than chance performance, this difference was not significant (chisquare = 5.03; ns).
Operationahzing Feedback
Several important points need to be made
regarding the manner in which feedback was
operationalized. First, feedback information
was made available as part of the external information environment. More specifically, it
was one of the 26 fundamental factors (namely,
"perceived price change over the last three
months") that could be accessed by the analysts.
Second, making feedback information
available is qualitatively different from insuring
538
JACOBY, MAZURSKY, TROUTMAN, AND KUSS
that feedback information is actually provided
to the decision maker. Although the real world
reflects both types of situations, virtually all
prior research has involved providing information to the decision maker—either after the
response has been made or, in some instances,
before the decision maker even begins to interact with the task environment (see the discussion of "feed forward" in Hammond et al.,
1980, p. 229) Some might contend that simply
making information available does not constitute feedback. Others, however, could contend that providing feedback information also
fails to satisfy the classic definition. For example, Nadler (1979, p. 310), in evaluating
Wiener's (1950, p. 35) classic definition, observed: "Thus, feedback is information about
the actual performance or actions of a system
used to control the future actions of a system"
(italics added). In other words, unless said information is then used to control future actions, simply providing feedback information
is insufficient for defining feedback in the classic sense. For present purposes, we rely on a
looser definition and justify this by observing
that many real-world situations may be characterized as ones that simply contain feedback
information as part of the external environment. However, as is discussed below, the distinction between "information availability"
and "information provision" has substantial
implications for conceptualizing and modeling
the feedback process
would have tied such an hypothetical analyst
with the 15th ranked analyst in our investigation. As Casey (1979, p. 91) pointed out:
"It is an axiom in security analysis that past
performance is not a guarantee of future performance."
Performance and the Accessing of
Feedback Information
Table 3 provides a detailed breakdown, for
each analyst and each security, of whether and,
if so, just when feedback information was accessed during each of the periods (2, 3 and 4)
for which feedback is relevant. This table is
explained as follows. The 17 analysts are arranged in order of their net performance across
all four test periods, from the best performing
analyst on the left to the worst performing
analyst on the right. The block of rows under
the label "Period 2" refers to the information
that the analysts accessed during Period 2 while
in the process of arriving at their stock selection
decisions. How each security (labeled from J
to S) actually performed during the previous
90-day period is indicated under the heading
"Feedback." The row labeled "Total" indicates
the total number of items of information acquired by each of the analysts during that period. Blank cells indicate that no feedback information was accessed by that analyst during
that period for that security. The numbers in
cells that contain a number indicates the order
in which that item of feedback information
was accessed. For example, the fourth-bestperforming security analyst accessed a total
of 56 items of information before making his
choice during Period 2. He did not access any
feedback information until doing so for Stock
N on the 17th item. He then went on to consider feedback information for stocks Q, K,
S, R, M, J, and L, in that order. From the
25th through 56th item in his search, he looked
at other types of fundamental factor information. Note that the mne cells containing
multiple entries (e.g., the fifth-best-performing
analyst for stock S during Period 4) indicate
that the analyst accessed that particular feedback information more than once.
Third, as is required for testing the hypothesis, the feedback information was completely accurate m regard to the correctness
of the response. By accessing this information,
the decision maker could tell exactly how well
each security had performed and, in turn, how
well he or she had performed
Fourth, also as is required for testing the
hypothesis, the feedback information was not
explanatory; it provided no information as to
why the obtained outcome actually occurred.
Finally, the feedback information also failed
to be predictive. That is, had a decision maker
employed the simple choice heuristic of accessing only feedback information and then
based a decision on just this information, the
net performance index (based on predicting
Inspection of Table 3 reveals several interperformance from Period 1 to Period 2, from esting findings. First, not all the analysts chose
Period 2 to Period 3, and from Period 3 to to access feedback information. Five of the
Period 4) would have been —41.7%. This analysts, including the very best and very worst
Table 3
Extent and Sequence of Accessing Feedback Information Across All Securities Within Each Period
Analyst
Feedback
(%)
Stock
+9 6
0
-5 8
-5 4
+ 12 5
-2 9
0
-5 3
Total
J
K
L
M
N
Q
R
S
+4 2
+8 3
-125
+8 6
+6 2
-1 5
-2 4
0
Total
J
K
L
M
N
Q
R
S
-122
-9 0
-35 7
-58 0
-29 4
-136
-37 5
-33 3
J
K
L
M
N
Q
R
S
Total
Better performing
1
2
96
26
52
23
19
24
22
17*
18
21
20
56
37
7
3
8
6
1
2
5"
4
40
54
62
7
4
8"
6
2
3
1
5
55
30
32
27
28
29*
33
31
25, 26
59
5
9
6
11
8
7*
12
10
105
98
21
12
10
3
26
28
23
24
25
29
27
22
30
Period 2
5
3
2
4a
6
7
1
25
37
25
Period 3
5
4
3
1
6"
7
2
8
26
19
25
Period 4
5
3"
2
4
6
7
1
32
" Denotes the stock selected by the analyst as the "best buy" for that period
40
1
7
5
6
2°, y
21
2
7
4"
6
1
8
3
5
31
1
7
4"
6
2
8
3
8
5
26
23
18
14
12
16
4
10
6
8"
48
6
4
7
1
5
2
3*
33
6"
4
7
1
5
2
3
32
11
17"
18
18
4
3"
1
7
6
2
5
14
4
3"
1
7
6
2
5
16
12
13
7
5
6"
10
2
8
9
3
10
4
3
6*
8
1
5
7
2
17
14
15
40
43
39
12
5
29
21
33"
17
8
25
34
41"
42
38
37
44
23
44
19
36
43
13
5,36
30
20"
32
16
8
25
24
43
36
5
3
2'
4, 11
6,8
9
1
10
55
58"
54
56
57
53
52
59
31
64
18
17"
39
42
38
40"
41
37
Poorer performing
16
17
14
4", 33a
27
23.30
29
18, 19
7, 8, 12
25
33
11
10"
13
12
8
9
29
31
1
_
6
—
7
2
3
30
32
2
5
6
1"
8
7
3
4
24
35
540
JACOBY, MAZURSKY, TROUTMAN, AND KUSS
performing analysts (Numbers 1 and 17), did
not consider any feedback information whatsoever. Second, those that did access feedback
did not do so consistently. More specifically,
although five of the analysts accessed all the
available feedback information for all eight
stocks across all three periods, the remaining
seven analysts accessed some feedback for
some of the securities some of the time but
not for other securities at other times.
In terms of the hypothesis, the patterns seem
clear. Given three periods and eight securities,
each analyst could have acquired up to 24
different items of feedback. Yet, of the seven
best performing analysts, four accessed no
feedback information whatsoever and only one
accessed all the available feedback information In contrast, of the seven worst performing
security analysts, only one accessed no feedback information at all and two accessed all
the available feedback information.
Moreover, within a period, when one of the
top seven performing analysts accessed feedback for one security, he or she did so for all
eight securities. In contrast, on five different
occasions (out of 7 analysts X 3 periods = 21
such occasions), when one of the seven poorest
performing analysts accessed feedback information for a security, he or she apparently did
not feel compelled to access feedback information for all the other securities. This tendency for better performing decision makers
to engage m more systematic accessing is likely
not limited to the accessing of feedback information but probably reflects a tendency toward predecision information accessing in
general (cf. Kahneman & Tversky, 1979; Kozminskyetal., 198 l;Sternberg& Powell, 1983).
In all, the seven top performing analysts
accessed 48 items of feedback information out
of a possible (3 periods X 7 analysts X 8 securities =) 168 such items, or 28.6%. In contrast, the seven poorest analysts accessed a total
of 115 items of feedback information, or
68.5%. Applying Ferguson's (1966, pp. 204206) test for the significance of the difference
between uncorrelated proportions yielded a
chi-square of 117 (p < 0.0001). However, the
fact that the middle three analysts accessed all
the available feedback information suggests
that a more appropriate test would be a correlation, based on the entire sample, between
the proportion (from 0 to 24) of feedback in-
formation accessed and the performance criterion. Across all 17 analysts, the correlation
between the proportion of outcome-only feedback accessed and decision quality is —.48 (p =
.02). Based on these findings, the hypothesis
that task environment feedback that provides
only accurate outcome feedback but that contains neither predictive nor explanatory value
(i.e., cognitive feedback) will be less used by
better peformmg decision makers is considered
confirmed.
Additional perspective is provided by Table
4, which considers the order in which feedback
information was accessed on a secunty-by-secunty basis. The two numbers appearing under
each analyst should be read as follows. The
second number indicates the total number of
information items accessed by that analyst for
that security during that period; the first number indicates the order in which feedback information was first accessed from among all
the information accessed for that security during that period. For example, the fourth best
performing analyst accessed a total of seven
items for information regarding stock J during
Period 2, with feedback information being accessed third out of these seven items.
Because different analysts engaged m different amounts of search, the following procedure was employed to enable aggregation
and comparisons to be made across analysts,
securities, and periods. For each analyst who
considered feedback information, the order in
which that information was accessed relative
to the other items of information accessed for
that security was converted to a 100-pomt
scale, where 1 indicated highest priority and
100 indicated lowest priority. This was accomplished by dividing the rank order by the total
number of items accessed. To illustrate, take
the example noted above m which the fourth
best performing analyst accessed seven items
of information for stock J during Period 2,
with feedback information being accessed
third out of this set. One seventh is equivalent
to .14 and three sevenths equals .42; this is
therefore the "accessing priority score" assigned to the fourth analyst on Period 2 for
Stock J. Had the analyst accessed 12 items of
information for that security, with feedback
being the ninth fundamental factor considered,
the accessing priority score would have
been .75.
Table 4
Extent and Sequence of Accessing Feedback Information for Each Security Within Each Period
Analyst
Stock
Better performing
1
2
.
3
10
12
11
13
14
Mean of
accessing
Poorer performing priority
15
16
17
indices (%)
Period 2
J
K
L
M
N
Q
R
3,7
3,6
3,8
3,7
3,7
3,7
3,7
3,7
S
3,3
3,3
7, 7
3,3
3,4
3,3
3,3
4,4
1,4
1.3
1,3
1,3
1,2
1,3
1,3
1,4
,2
,2
,2
,3
,6
,2
,3
,1
2,6
2,5
1, 4
2,7
3,8
2,6
3, 6
2,6
3,3
3,3
, 1
,I
,2
, 1
, 1
, 1
, 1
,2
5,5
5,5
5,5
5,5
7, 7
5,5
7, 7
5,5
3,4
5,5
4,4
4,4
4,5
4,4
3,4
4,4
5,5
5,5
5,5
5,5
5,5
5,5
6,6
7,7
4,4
5,6
4,4
2,5
2,4
3,5
3,4
4,4
,3
—
,3
,6
,4
,4
,4
7, 7
8,9
7, 7
10, 11
8,8
7,8
7,7
7, 7
4,4
3,7
2,2
4,5
2,2
4,5
3,6
2,2
,2
,2
,4
2,3
2,9
1,2
2,3
2,3
2,3
M
66
72
72
60
63
70
67
72
68
Period 3
J
K
L
M
N
Q
R
S
—
—
—
—
—
—
—
—
5,5
1, 2
1,2
1,2
1,2
1,4
1,2
1,2
—
—
—
—
—
—
—
—
1,3
1,3
1,4
1,3
1,3
1,3
1,4
1,3
,5
,5
,5
,5
,5
,5
,5
,5
,5
,3
,4
,4
,4
,3
,5
,3
,4
,4
,3
,4
,5
,4
,4
,5
,2
,2
,3
,2
,2
,1
,1
,1
1,
1,
1,
1,
1,
1,
1,
1,
1
1
10
1
1
1
1
1
—
3, 4
7, 11
3, 3
,3
M
58
57
43
48
47
52
58
59
53
Period 4
J
K
L
M
N
Q
R
1, 5
1,7
1, 10
1,5
1,9
1,6
1,7
1,6
S
M
4, 7
3,5
4,8
4,8
4,8
5,9
4,6
5,8
1,4
1, 4
1,2
1,3
1,3
1,4
1,3
1,3
,3
,3
,3
,2
,4
,3
,3
,2
,4
,4
,4
,4
,4
,4
,4
,4
,1
,2
,2
,2
,4
, !
,2
,2
1,6
1, 2
1,2
1, 11
1, 5
1,3
1,4
1, 2
1,2
,6
,4
,2
,2
,2
56
44
42
43
43
50
44
54
47
55
i
a
542
JACOBY, MAZURSKY, TROUTMAN, AND KUSS
Table 5
Changes in Search Content (i e, Factors Accessed) Across Successive Periods
No of properties accessed in periods
Analyst
1—2*
1—2"
1—2C
2—3
2-3
2—3
1-4
3—4
3—4
Five best performing
Five worst performing
27"
17"
36
15"
3
35
28
7
9
11
8
37
32
9
4
5
14
32
* To be read as number of factors accessed by a given analyst in both Periods 1 and 2, summed across all five analysts
b
To be read as number of factors accessed by a given analyst in Period 1, but not Period 2, summed across all five
analysts
c
To be read as number of factors not accessed by a given analyst in Period 1 which were accessed in Period 2, summed
across all five analysts
" Cell entries are to be read as follows The five best performing analysts accessed information on a combined total
of 59 factors in Periods 1 and 2. Of these, 27 factors that had been accessed in Period 1 were also accessed in period
2, 17 factors that had been accessed m Period 1 were not reaccessed in Period 2, and 15 factors not accessed in Period
1 were accessed in Period 2
The right-hand column of Table 4 contains
the aggregated accessing priority scores for
feedback information for Periods 2, 3, and 4.
To test whether the trend of increased priority
given to accessing feedback over the periods
was statistically significant, a one-way analysis
of variance (ANOVA) was employed. The test
yielded an F ratio of 31.8 (p = .0001), indicating a significant change in the accessing
priority given to feedback information across
the three periods. A Scheffe test was then applied to determine whether the stages of accessing information feedback differed between
adjacent periods. The results show that the
priority that security analysts assigned to feedback information in the second period was
significantly lower than the priority that they
assigned to feedback in the third and fourth
periods (p = .05). However, despite the fact
that the fourth period reflects earlier accessing
of feedback information, the difference between the third and fourth periods is not significant.
Learning From Experience
A related set of questions concerns information accessing of the other 25 fundamental
factors across the four periods, particularly as
this relates to learning from experience (Einhorn, 1980; Einhorn & Hogarth, 1981). Given
that the analysts focused on only a limited
number of the available factors (see Table 1),
one might expect that, as the number of periods increased, there would be greater con-
vergence or overlap in the factors considered
from one period to the next, and decreased
attention to factors not considered during the
previous period.
To examine this question, the data for the
five best and five worst performing analysts
were sorted into three categories (see Table 5).
The first category included all instances in
which at least one item of information was
accessed on the same factor in two successive
periods. The second category included all cases
in which information accessed on a given factor m one period was not reaccessed on that
factor in the subsequent period. The third category included all instances in which a factor
that was not considered in one period was
accessed in the immediately succeeding period.
A one-way ANOVA applied to these data revealed that, though present, neither the trend
toward increasing convergence (27, 35, 37) nor
the trend of decreasing attention paid to different factors (15, 11, 5) was significant for
the five better performing analysts. However,
for the five poorer performing analysts, although the degree of overlap remained roughly
the same across periods, there was a significant,
F(2, 13) = 5.3, p = .03, increase in the attention paid to new factors (3, 8, 14). These data
suggest that the better performing analysts
might have been more effective learners. Alternatively, these data may also be interpreted
as indicating that the poorer performing analysts were also learning—as evidenced by the
fact that, m view of then- poor performance,
they continued looking at other factors.
WHEN FEEDBACK IS IGNORED
543
to the others with whom they were competing.
Yet as Hogarth (1981, p. 210) pointed out:
Two basic findings are suggested by this in- "An important ecological dimension missing
vestigation. First, given an environment that from most research on judgment and choice
permits decision makers to be selective in is that decisions are often made in competitive
choosing the information that they will and and other social situations.... In competitive
will not consider, not everyone chooses to ac- situations optimal responses are not necessary
cess feedback information—at least not when for survival. Instead, responses only have to
such information possesses only "outcome be better than those of competitors . . . (Einvalue" and fails to also possess either predictive horn, 1980; Hammond, 1972)." Though the
and/or explanatory value. Second, as pre- real-world security analyst's task environment
dicted, better performing decision makers are does not reflect direct access to relative feedthe ones who are less likely to access such back, it is clear that future research needs to
"outcome only" feedback information than be devoted to this issue.
are poorer performing decision makers.
This study also has other implications. First,
These findings need to be considered in both in terms of developing theory and conterms of several limitations. First, the sample ducting research, it seems obvious that adconsisted of 17 volunteer analysts. Much ditional consideration needs to be devoted to
greater confidence would be obtained from a the diagnostic (that is predictive and explanatory) nature of feedback information. To the
larger and more representative sample.
More importantly, the present investigation extent possible, prior research needs to be rewas also limited by the fact that respondents viewed to determine the type of feedback that
were able to access feedback information on had actually been provided to the respondent.
only three occasions (that is, during Periods It may be that outcome feedback has little
2, 3, and 4). This was due to the amount of utility for cognitive (as opposed to motor) tasks
time required for the task. As previously noted, and that only predictive and/or explanatory
the analysts devoted an average of 2'/2 hours feedback would be useful. Hence, the general
to the task, and one analyst actually spent more feedback-performance relationship would
than 6 hours. Real-world security analysts, as have to be qualified.
well as other types of decision makers, generally
Second, it is also clear that theorists need
have access to much greater numbers of feed- to revise the generally accepted view that the
back occasions. Future research would do well impact of feedback on performance is virtually
to examine the impact of a greater number of always positive. Sufficient evidence is beginfeedback trials.
ning to accumulate (Emhorn & Hogarth, 1978;
Third, the task that was employed did not Hammond & Summers, 1972; Steinman,
faithfully model real-world security analyst 1976; this study) to indicate that it depends
decision making in at least two key respects. on the type of feedback that is involved. In
First, though it required that the analyst make particular, outcome feedback may be partica "buy" decision, it did not also permit a "sell" ularly disfunctional in a complex, dynamic
response; second, it did not adopt a "portfolio" environment
approach, that is, the buying and selling of
Another implication is that models of feedtwo or more securities at one time. Note, how- back that implicitly assume that the decision
ever, that this limitation did not detract from maker will be given feedback information need
our ability to test the hypothesized relation- to be revised in order to accommodate the
ship.
active search for and accessing of feedback.
It should also be noted that the present in- When the source of feedback is the task envestigation provided what might be termed vironment itself, especially a complex task en"absolute" rather than "relative" feedback. vironment, the individual must actively engage
That is, the feedback that was available in- in the search for and accessing of this infordicated to the respondents how well they had mation. Without such prior accessing, the inperformed in selecting the best out of the eight dividual never will have the opportunity to
available stocks. It did not, however, provide "perceive" (qua receive and interpret, cf. Ilgen
any indication as to how well they did relative et al., 1979) this feedback.
Discussion
544
JACOBY, MAZURSKY, TROUTMAN, AND KUSS
Finally, most theories of human judgment
and decision making actually fail to study
feedback (cf. Hammond et al., 1980, p. 228).
As but one example, after acknowledging that
feedback/learning constituted one of three key
subprocesses of decision making, Payne's
(1982, p. 386) integrative review continued.
"The feedback/learning processes, though assumed to interact with information acquisition
and evaluation, will not be stressed "
Yet the notion of feedback is not only central
to our understanding of such phenomena as
information evaluation, learning, and decision
making, but to our understanding of all of
human behavior. Consider the following:
The human information processing system that has
evolved to cope with the environment is characterized by
essentially sequential processing, limited memory, selective
perception, and reliance on cognitive simplification mechanisms (I e , heuristics) Central to these means is the role
played by feedback (Hogarth, 1981, p 199)
All behavior involves strong feedback effects, whether
one is considering spinal reflexes or self-actualization
Feedback is such an all-pervasive and fundamental aspect
of behavior that it is as invisible as the air we breathe
Quite literally it is behavior—we know nothing of our own
behavior but the feedback effects of our own outputs (Powers, 1973, p 351)
Given the apparently pivotal role of feedback
in all human behavior, greater conceptual and
empirical attention to this phenomenon than
has heretofore been the case appears warranted.
References
Ammons, R B (1956) Effects of knowledge of performance A survey and tentative theoretical formation
Journal of General Psychology, 54, 279-299
Annett, J. (1969) Feedback and human behaviour The
effects of results, incentives and reinforcement on learning
and performance Baltimore, Penguin
Bernstein, L A (1975) In defense of fundamental investment analysis Financial Analysts Journal, JanuaryFebruary, 57-61
Blume, M E (1971) On the assessment of risk Journal
of Finance, March 1-10
Bourne, L E , Jr (1966) Comments on Professor I M
Bilodeau's paper In E A Bilodeau (Ed), Acquisition of
skill (pp 297-313) Academic Press
Carver, C S., & Scheier, M F (1982) Control theory A
useful conceptual framework for personality-social,
clinical, and health psychology Psychological Bulletin,
92, 111-135
Casey, D R (1979) Crisis investing Opportunities and
profits in the coming great depression New\brk Simon
and Schuster (Pocket Books Division)
Chestnut, R. W, & Jacoby, J (1977) Consumer infor-
mation processing. Emerging theory and findings In A
Woodside, P D Bennett, & J N Sheth (Eds.), Foundations of consumer and industrial buying behavior (pp
119-133). New York Elsevier, North-Holland
Chestnut, R W, & Jacoby, J (1982) Behavioral process
research' Concept and application in consumer decision
making In G R Ungson & D N Braunstein (Eds),
Decision making An interdisciplinary perspective (pp
232-248) Boston, MA Kent
Ciarkson, G P E (1962) Portfolio selection A simulation
of trust investment Englewood Cliffs, NJ Prentice-Hall
Ebert, R J., & Kruse, T E (1978) Bootstrapping the
security analyst. Journal of Applied Psychology, 63(1),
110-119.
Einhorn, H J (1980) Learning from experience and suboptimal rules in decision making In T Wallsten (Ed.)
Cognitive processes in choice and decision behaviour
(pp 1-20) Hillsdale, NJ Erlbaum.
Emhorn, H J , & Hogarth, R M (1978) Confidence m
judgement Persistence of the illusion of validity Psychological Review, 85, 395-416
Einhorn, H J., & Hogarth, R M (1981) Behavioral decision theory Processes ofjudgement and choice Annual
Review of Psychology, 32, 52-88
Ferguson, G A (1966). Statistical analysis in psychology
and education N Y- McGraw Hill
Greller, M M (1980) Evaluation of feedback sources as
a function of role and organizational level Journal of
Applied Psychology, 65, 24-27
Hagaman, T C , & Jensen, A E (1977) Investment value
and security analysis Financial Analysts Journal,
March-April, 63-70
Hammond, K R (1972) Inductive knowing In J R
Royce & W W Rozeboom (Eds.), The psychology of
knowing (pp 285-327) London Gordon & Breach
Hammond, K R , McClelland, G H , & Mumpower, J
(1980) Human judgment and decision making Theories,
methods, & procedures New \brk Praeger
Hammond, K R., & Summers, D A (1972) Cognitive
control Psychological Review, 79, 58-67
Hammond,K R,Summers,D A.,&Deane,D H.(1973)
Negative effects of outcome-feedback in multiple-cue
probability learning Organizational Behavior and Human Performance, 9, 30-34.
Hogarth, R M (1981) Beyond discrete biases. Functional
and dysfunctional aspects ofjudgement heuristics Psychological Bulletin, 90, 197-217
Ilgen, D R , Fisher, C D , & Taylor, M S. (1979) Consequences of individual feedback on behavior in organizations Journal of Applied Psychology, 64, 349-371
Ivancevich, J M , & McMahon, J T (1982) The effects
of goal setting, external feedback, and self-generated
feedback on outcome variables. A field experiment
Academy of Management Journal, 25, 359-372
Jacoby, J (1977) The emerging behavioral process technology in consumer decision making research In
W D. Perrault, Jr (Ed), Advances in Consumer Research, 4, 263-265.
Jacoby, J., Chestnut, R W., Hoyer, W, Sheluga, D A , &
Donahue, M J (1978) Psychometric characteristics of
behavior process data Preliminary findings on validity
and reliability In Keith Hunt (Ed ), Advances in Consumer Research, 5, 546-554
Jacoby, J , Chestnut, R W., Weigl, K C , & Fisher, W.
WHEN FEEDBACK IS IGNORED
(1976) Pre-purchase information acquisition Description of a process methodology, research paradigm, and
pilot investigation In B B Anderson (Ed), Advances
in Consumer Research, 3, 306-314
Jacoby, J , Kuss, A., Troutman, T, & Mazursky, D (1984)
Effective vs. ineffective strategies in security analyst decision making. Evidence for the concept of control schemas Manuscript submitted for publication
Jams, I L., & Mann, L (1977) Decision making New
York The Free Press
Kahneman, D , & Tversky, A (1979) Prospect theory
An analysis of decisions under risk Econometnca, 47,
263-291
King, B F (1966) Market and industry factors in stock
price behavior Journal of Business, January 139-190
Kozminsky, E , Kmtsch, W, & Bourne, L E Jr (1981)
Decision analysis with texts Information analysis and
schema acquisition Journal ofExperimental Psychology
General, 110, 363-380
Major, B N (1980) Information acquisition & attribution
processes Journal of Personality and Social Psychology,
39, 1010-1023
Monthly Stock Digest (1969-1970) New York Data Digests, Inc
Nadler,D A (1979) The effects of feedback on task group
behavior A review of the experimental research Organizational Behavior and Human Performance, 23,
309-338
Payne, J W (1982) Contingent decision behavior Psychological Bulletin, 92, 382-402
Powers, W T (1973) Feedback Beyond behaviorism Science, 179, 351-356
545
Sheluga,D A,Jaccard,J J ,& Jacoby, J.(1979) Processdescriptive methods and the estimation of preference
An integrative approach Journal of Consumer Research,
6, 166-176
Slovic, P (1969) Analyzing the expert judge Adescnptive
study of a stockbroker's decision process Journal of
Applied Psychology, 53, 255-263
Slovic.P (1972) Psychological study of human judgment:
Implications for investment decision making Journal
of Finance, 27, 779-799
Slovic, P , Fleissner, D., & Bauman, W S (1972, April).
Analyzing the use of information on investment decision
making A methodological proposal The Journal of
Business, April, 45(2), 283-301
Steinman, D O (1976) The effects of cognitive feedback
and task complexity in multiple cue probability learning
Organizational Behavior and Human Performance, 15,
168-179
Steinberg, R J , & Powell, J S (1983) Comprehending
verbal comprehension American Psychologist, 38, 878893
Tersine,R J,&Celec, S E (1976) Portfolio management
A practical approach Atlanta Economic Review, September/October, 28-33
Wiener, N (1948) Cybernetics On control and communication in the animal and machine New York Wiley.
Wiener, N (1950) The human use of human beings Cybernetics and society Boston Houghton-Mifflin
Received November 29, 1983
Revision received March 7, 1984