Outcome knowledge and auditor judgment

MISSING IMAGE

Material Information

Title:
Outcome knowledge and auditor judgment
Physical Description:
viii, 219 leaves : ill. ; 28 cm.
Language:
English
Creator:
Maddocks, Pauline Merle, 1944-
Publication Date:

Subjects

Subjects / Keywords:
Auditing -- Psychological aspects -- Research   ( lcsh )
Judgment   ( lcsh )
Prediction (Psychology)   ( lcsh )
Uncertainty (Information theory)   ( lcsh )
Accounting thesis Ph. D
Dissertations, Academic -- Accounting -- UF
Genre:
bibliography   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1989.
Bibliography:
Includes bibliographical references (leaves 210-218)
Statement of Responsibility:
by Pauline Merle Maddocks.
General Note:
Typescript.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001500391
oclc - 21387337
notis - AHB3119
System ID:
AA00002127:00001

Full Text












OUTCOME


KNOWLEDGE


AUDITOR


JUDGMENT


PAULINE


MERLE


MADDOCKS


A DI
OF THE


SSERTATION PRESENTED TO THE GRADUATE SCHOOL
UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY


UNIVERSITY


OF FLORIDA


1989





























Copyright


1989


Pauline Merle Maddocks
















ACKNOWLEDGMENTS


Many


people


have


aided


in this


accomplishment.


must


foremost


acknowledge


chairman,


Rashad


-khalik.


continual


support,


guidance


encouragement


throughout


doctoral


program


have


been


major


factors


in its


completion.


comments


suggestions


Bill


Mess


committee


members


Michael


Bamber


Joseph


Alba


provided


insight


direction.


am grateful


to Peat


Marwick


Main


Co., especially


John


Willingham


.Tom


Weedn


, for


providing


subjects


this


study.


Without


this


ass


instance


funding


provided


Fisher


School


of Accounting,


this


research


project


would


have


required


considerably


more


time.


Several


friends


have


always


given


freely


their


time


expertise


must


also


recognized.


Dimitrios


Ghicas,


Richard


Tubbs,


Francis


Bush


have


each


in his


own


made


a difference


life


work.


am especially


grateful


love


encouragement


provided


mother


Hazel


Malbrough


Guillotte,


well


unswerving


belief


me by


entire


family.


Most


importantly,


though,


must


thank


son


John


_


i

















TABLE OF CONTENTS






ACKNOWLEDGMENTS ....................... ..iii


ABSTRACT


I I


CHAPTERS


1 INTRODUCTION


. . . . . 1


Background .....
Research Objective
Research Motivation


. . . .
a. a. *. *. *..


. . . .. 1
. .S .S .S. .S. . 4
. .S. . . .5. . 5


Organization of the Dissertation


OVERVIEW OF RELATED RESEARCH ...


Non-Accounting Studies ......
Early Studies .............
Demand Characteristics ....
Effect of Outcome Severity
Attempts to Eliminate the B:
Accounting Studies ..........
Summary . .


ias . . .
. S . . . . . .
*.5 .S .S .S.S S .S S .S S ..S. S .S .S .


RESEARCH PROBLEM AND HYPOTHESES


Introduction


Audit Judgment Implications ......
Research Hypotheses Development .........


*.. . . ..S S S
. . ..S. ..* S
*.. .a . ..S S


4 THE EXPERIMENT: DESIGN, TASK, INSTRUMENT, SUBJECTS,


AND ADMINISTRATION


Introduction .......
Audit Task Area ....
Participants .......
Experimental Design


* .S 5 . ..S. .5. ..S. ..5. .5. .5. S S
. . . . . . . . . S S S S S
. .5 . ..3. .*. . ..5. . . S S S


Administration of the Experiment ..................
Experimental Methods ... . . . .
Hypotheses 1, 4, 5, and 6 .. ... .. ...... .











Development of Lne Chemco
Design of the Tasks .

EXPERIMENTAL RESULTS ......


Introduction ......
Raw Data Results .
Preliminary Tests .
Tests of Hypothesis
Tests of Hypothesis
Tests of Hypothesis
Preliminary Tests


Main Hypothesis Tests
Additional Tests ...


Tests
Tests
Tests
Tests
Cue
Cue


of Hypothesis
of Hypothesis
of Hypothesis
of Hypothesis
Recall
Relevance ...


Case


. . .. ..a .*.. a. .a .a

. . . . a a a a aa a S.


1 .. .S .*. . .*. .. . . .
2 .. .* . . .9. . . .
3 . . . .a. .S .S. . . .
.................................


s . . . . . ..S. .S S S S .
. . ..S. a a .a . .S S . S S .
*. a .a . .a ..S. . ..S. ..S. .S S S .
. . . . ..S. .a. . . S S S *
. .a .a .* .S .S . ..S. .S. .S. .S. S S
*. . . . . . ..S S S S S .S S .


Additional Analyses ..
Subjects' Experience


. . . ..S. .S. S S S . . ..S S
. . . . . ..a. ..S. .S. .S S S S


Subjects' Difficulty in Assessing Viability
Subjects' Awareness of the Effect of Outcome
Knowledge .. . . . . . . .


6 CONCLUSION


Introduction . . . .
Research Objectives ..........
Discussion of Research Results
General Discussion ...........


....................... 104
. .. .. . 104
................. .. 105
. .. .. ... 110


APPENDICES


A GENERAL INSTRUCTIONS FOR PARTICIPANTS ............

B CHEMCO CASE DATA . . . . . .

C NINO TEST INSTRUMENT . . . . . . .

D NIFO TEST INSTRUMENT . . . . . . .

E NISO TEST INSTRUMENT . . . . . . .

F FINO TEST INSTRUMENT. . . .....

G FIFO TEST INSTRUMENT .. . . . . .












K HISO TEST INSTRUMENT


REFERENCES


.. .. .. .. *.. a. a a S S a.I. aa ..-


BIOGRAPHICAL SKETCH
















Abstract


of Dissertation Presented to


University


Requirements


of Florida


the Graduate


School


in Partial Fulfillment


for the Degree of Doctor


of Philosophy


OUTCOME KNOWLEDGE AND


AUDITOR JUDGMENT


Pauline Merle Maddocks


August,


Chairman:


1989


Rashad Abdel-khalik


Major


Department:


Accounting


This


study


investigated the effect


of outcome knowledge


on the auditor's


ability to evaluate


what


"knew"


or was


expected


to have


known prior to


receipt


of the outcome


knowledge.


Based upon prior research,


it was expected


that


hindsight


assessment


of what


foresight


judgment


would have


been,


would be affected by


outcome


information


that


was


available at


that


that


prior point


outcome knowledge


tim


would affect


e. It

the kin


was also expected

d of preoutcome


knowledge


information


recalled,


as well


as the


relevance


assigned


to preoutcome knowledge data.


A second set


of expectations


was


related to


ameliorating


effect


of auditor


involvement


with


company


data


prior to or

Additionally


subsequent


was


to the


expected


receipt

that au


of outcome


ditor


information.


confidence would


w


e


I










found


to be greatest


in situations


where


actual


outcomes


are


unexpected


or surprising.


This


research


was


motivated


assumption


that


hindsight


bias


in the


audit


setting


have


undesirable


effects


on many


unstructured


audit


judgment


tasks,


especially


auditor


unaware


effect


on contemporaneous


judgments.


confidence


example,


in one's


it might


ability


foster


to accurately


undeserved


render


level


"correct"


highly


correlated


with


actual


outcome)


judgments


subsequent


situations


inhibit


systematic


learning


from


experience.


Seven


hypotheses


were


developed


examined


within


laboratory


experiment


in which


variables--outcome


involvement--were


manipulated.


dependent


variables


were


judgment,


confidence,


number


cues


recalled,


and


cue


weightings.


five-part


test


instrument


was


completed


senior-level


auditors


from


a Big


Eight


accounting


firm.


Planned


comparisons,


analysis


variance,


multivariate


analysis


of variance


were


used


test


hypotheses.


research


findings


strongly


supported


presence


an outcome


effect


on auditor


judgment


cue


recall


and


generally


supported


"surprise"


effect--hindsight


bias


was


greatest


when


nature


outcome


was


unexpected.


Involvement


was


found


to be marginally


significant


e.,
















CHAPTER


INTRODUCTION



Background


Several


studies


over


past


decade


have


demonstrated


that


an individual's


knowledge


of outcome


information


effect


on memory.


This


effe


is demonstrated


overes


timation


[Fischhoff,


orted


of preoutcome


1975;


outcomes


Fischhoff


as having


knowledge


Beyth,


been


memory


1975],


inevitable


states


perception


[Slovic


Fischhoff,


1977],


overestimation


of what


others


knew


prior


rece


of outcome


knowledge


[Wood,


1978;


Baron


Hershey,


an event


1988].


s outcome


In spite


before


of one's


fact,


inability


learning


to predict


that


occurred


often


leads


one


to believe


that


"knew


it all


along"


These


individuals


findings


are


have


called


implications


on to disregard


task


information


which


in order


fairly


evaluate


what


another


individual)


knew


was


expect


to have


known


prior


receipt


of outcome


knnwl edae


(e .i .


whether


or not t


auditor


of a firm


that


..










also has


implications


for the acquisition


decision-making


skills.


one


cannot


accurately


remember


how much he knew before


he will


tend to overestimate


receipt

the sa


of outcome


lience of


information,


the preoutcome


information


(e.g.,


causality vis-a-vis


the outcome)


or her predictive abilities.


These


tendencies


undoubtedly


contribute to


overconfidence


judgment


frequent


finding of


literature.


of these


implications


represent


potential


problems


for the auditor.


Auditing


standards


indicate


that


the evaluation of


audit


evidence and the decision


that


sufficient


competent


evidence


has been

largely


collected to s

on professional


support


the audit


judgment.


This


opinion,

judgment


depend

is tied to


decision-making


skills


which are


acquired almost


exclusively


from professional


However, th

formulation


experience


e nature and/or timing


process may


[Waller and Felix,


of feedback


complicate and/or distor


1984a].

the opinion

t the


learning process


[Hogarth,


1980;


Waller and Felix,


1984b],


which in


turn may affect memory.


the auditor


unaware


the effect


of outcome knowledge,


remembered probabilities


be higher


for events believed


to have occurred and


lower


events


believed to


have not


occurred.


The auditor


cannot


improve deci


sion-making ability


if he


is overly


confident










unrealistic


impressions


as to his


or her prediction


abilities.


This overconfidence may also have


serious


implications


for the degree of reliance

or structural aids [Waller


the auditor will


and Felix,


1984b].


place


on decision


Research


[Arkes


et al.,


1986


revealed that


one of the many


problems


with over-confident


judges


is that


they tend to


underutilize decision aids--they


assume


their


experience,


knowledge, insight,

decision aid. Thus,


etc.

the


will


enable them to outperform the


"I-knew-it-all-along"


[Fischhoff,


1975]


attitude,


typical


of hindsight-biased


judges,


lead


to overconfidence,


utilization


which


of decision aids,


turn may


lead to under-


suboptimal


decision making over


time,


and a


decrease


in audit


efficiency and


effectiveness.


A few researchers


workings


have attempted


of hindsight bias


assimilation and schemata


theory.


to explain


terms of


the causes


memory


Assimilation of


outcome


information


was


assumed to


immediate and natural


result


the overwriting


of preoutcome memory


such


that


could not


be accessed


once


the outcome


information had been


assimilated.


The Wood


[1978]


study only partially supported


that


notion.


Two recent


papers


[Hasher et


al.,


1981;


Dellarosa and Bourne,


1984


also weakened that


line of


reasoning by demonstrating that


outcome ma


affect


-L


evidence


1










A recent


analysis


review


hindsight


literature


[Hawkins


Hastie


, 1987


discussed


numerous


cognitive


motivational


factors


that


have


been


credited


contributing


bias.


Hawkins


Hastie


concluded


that


most


like


sources


bias


are


selective


recall


preoutcome


evidence,


selective


evaluation


of evidence,


and/or


model


change.


These


findings


have


implications


development


of decision


aids


to reduce


effect


of outcome


knowledge


on judgments


audit


environment.


Research


Objective


objective


issues


auditor


this


investigating


judgment


research


effect

task


an auditing


to address


of outcome


above


knowledge


evaluating


factors


that


affect


bias


an auditing


environment.


Previous


research


be discussed


in Chapter


psychology


judgment


literature


two


public


hed


acc


counting


studies


support


existence


of outcome


knowledge


effects


judgment.


This


study


differs


from


existing


accounting


auditing


literature


in several


respects.


First,


although


most


published


research


in both


judgment


accounting


literature


supports


existence


of a hindsight


hia -


I I 1-n


nn lv


nubl i


shed


auditinac


studv


fBuchman,


19851


was










Second,


tasks


used


in many


of the


non-accounting


studies


were


somewhat


artificial


and did not


address


expertise of


subjects.


The accounting


and auditing


studies


have been


did utilize more complex tasks


weakened by methodological


their


flaws


results may


their test


instruments.


Research Motivation


important


to note that


hindsight bias does


refer to all


retrospective


increases


in the


probabilities


assigned


to events


[Hawkins


and Hastie,


1987] .


Outcome


knowledge often


reveals


information about


relationship


between


prior


circumstances


subsequent


outcomes.


With


this new


information,


the observed


outcome's


an individual's


occurrence


estimated probability

should increase.


Hindsight bias,


on the other


hand,


a projection


new


knowledge


into


past;


is a denial


that


anything new has


been


learned from the


outcome


information.


Thus,


subjects


who are provided outcome


individuals


information


who become aware of


in experiments,


outcomes


well


in day-to-day


life,


typically


claim


they


"knew


it all along"


[Fischhoff,


1975] .


Both


academic and professional auditing


literatures


emphasize t


- -


importance of professional


judgment,


which










Hogarth


[1981],


have


demonstrated the


complexities


learning through experience,


especially with regard to


relatively


infrequent


problem situations


(i.e.,


rare


event


problem)


Many


important


auditing problems


do occur


infrequently,


and little


is known about


and what


auditors


have


learned with


respect


to those


situations.


Equally


interesting and not


known


what


auditors


think


they


have


learned from such situations.


If outcome


knowledge about


such infrequent


along"


events


frame


leads


of mind,


auditor to assume an


he may underestimate


"I knew


how much


uncertainty accompanies


such


situations


overestimate his


or her


ability to handle


similar


situations


future.


theoretical motivation


for this


study


determine i

susceptible


f auditor training


to the effect


renders


of outcome


the auditor

knowledge.


less

Practical


motivations are


related


to auditor training.


If the auditor


susceptible as


those


subjects


in previous


studies,


auditor training programs may be


auditor's awareness of how much


needed to


uncertainty


increase


surrounds most


contemporaneous


problems.


Oraanizationu


of the Dissertation


Chapter


"Overview of Related Literature,


" discusses


nreviou S


research and helos


Provide a


background


fnr this


YV


VI










chapters


formulate


results of previous


the research hypotheses to be


research studies to


investigated.


Chapter


"The Experiment,


" describes


auditing


environment


in which


the hypotheses


were


tested and the


development


test


instrument.


It also addresses the


methodologies used to


test


the hypotheses


and addresses


choice of participants.


Chapter


"Experimental Results,


presents the


results


of the


study


Chapter


"Conclusion,

contribution


" discusses


those


limitations,


results


as well as


possible extensions of


this


study.
















CHAPTER


OVERVIEW OF RELATED


LITERATURE


Non-Accountina Studies


Early Studies


deleterious


initially


effects


demonstrated in


of outcome


a seminal


knowledge


article by


were


Fischhoff


[1975]1


and labeled hindsight bias.


In his


1975


study,


Fischhoff


compared foresight


and hindsight


judgments


several

brief (


experiments


150 word)


in which student


passage of


subjects


a historical


were


or clinic


given a

1 event


for which


four possible outcomes


were


provided.


Each


subject


was


then asked to estimate


likelihood of


occurrence of


each of


four possible outcomes


and to evaluate the


relevance of


subjects


each datum in


completed the


the event


tasks


description.


without


Some of


outcome


knowledge

asked to


Others were told


rate the


likelihood of


"true"


outcome and then


four possible outcomes.


Each possible outcome


subjects.


was


Fischhoff


given


true outcome


found that


reporting


some


an outcome's


nrnnrr'r nn yr 1 m Ctt


dnnihl rd


\ + Q ner n iv


lk1 Eio 1of










second


experiment


in this


study,


he provided


different


group


of student


subjects


with


same


stimulus


materials


used


first


experiment.


only


difference


was


that


subjects


received


outcome


information


were


explicitly


asked


res


pond


they


would


have


they


known


out come"


[1975,


Again,


effect


outcome


knowledge


was


significant;


subjects


were


"either


unaware


perce


of outcome


options


knowl


aware


edge


, they


having


are


an eff


unable


on their


to ignore


resc


that


effect"


[197


, p.


Fischhoff's


third


experiment


study


also


used


same


stimulus


materials


first


experiment.


However,


unlike


previous


experiments,


this


one


asked


subjects


to respond


they


assumed


other


students,


known


true


outcome,


respo


nded.


Fischhoff


was


interested


in investigating


seco


nd-guessing


phenomenon.


He found


that


those


subjects


with


outcome


knowledge


expe


cted


other


subjects


to have


seen


in foresight


what


they only


knew


as a result


of hindsight


information.


Additionally,


throughout


three


experiments


relevance


attributed


datum


was


highly


dependent


on which


outcome,


any,


subject


was


told


to be


"true"


Most


of the


subsequent


hindsight


bias


studies


continued


to use


student


subjects


historical


scenarios


[Fischhoff,










[Fischhoff,


1977;


Wood,


1978;


Hasher et


al.,


1981]


investigate the bias.


tasks


were


very


similar


several


studies.


some


instances


subjects


were asked to assign outcome


probabilities to various


historical


events


that


either


not


occurred


(e.g.,


outcome


results of President


Nixon's


visits


to China and Russia


[Fischhoff


and Beyth,


1975])


whose outcomes


were not


well


known


(e.g.,


the outcome of the


British-Gurka


war


[Fischhoff,


1975]


After the events


occurred or after outcome


feedback


(i.e.,


correct,


false or


information


concerning the outcome)


had been provided,


subjects


were asked to predict


have been had


they not


what


the outcome


their probabilities


knowledge


would


and/or to


recall


what


their original


predictions


had been.


Preoutcome


and postoutcome mean responses were compared.


results


indicated


responses


outcome knowledge


in that


affected the


those outcomes provided to


subjects'


subjects


were


judged


as more


likely to occur;


the bias


effect


was


stronger


surprising


(i.e.,


unexpected)


outcomes;


subjects'


recall


of preoutcome cues and


their


evaluation


of the


relevance of particular pieces


of preoutcome


information


was


related to


the outcome


received.


Demand Characteristics










time delays on


estimates


recall


and assessed the


of subjects'


potential


original


effect


probability


of certain


characteristics


from earlier


studies.


previous


studies,


the effect


appeared


to be robust


and not


an artifact


the experimental


procedures


used.


A few


studies


utilized more


realistic


tasks


(e.g.


Slovic and Fischhoff


[1977])


and experienced subj


ects


(e.g.,


Arkes et


[1981])


latter


study the


subjects


were


experts


task


(i.e.,


(i.e.,


physicians)


render


required


a second opinion as


to complete a


diagnostic


to the diagnosis


of a


of symptoms).


researchers


hypothesized


that


practicing physicians,


trained to make


diagnoses


based


upon


symptoms,


would be


less


likely to exhibit


the bias2


They


discovered otherwise;


the more


unexpected the diagnosis


(i.e.,


outcome)


provided the physician


subjects,


the greater


the bias


that


was


exhibited.


researchers


concluded


that


the doctors


the hindsight


group tried


to make


sense out


what


they


knew


happened rather than objectively analyzing the


symptoms


provided in


case materials.


Wood's


[1978


results


revealed that


subjects


who made


preoutcome


judgments


exhibited less bias


than subjects


had not made


such


judgments when


both


groups


were asked,


after the outcome was


known,


what


their preoutcome


judgments


were or would have been.


study also addressed the










vulnerability of the


interpretation by


effect


having


to a


some of


demand characteristic


subjects estimate


foresight


responses of their peers.


Wood


theorized that


subjects'


concern


with self-presentation


or desire


to please


the experimenter


(i.e,


"look


good"


might


be a


factor


driving the bias.


this


were


true,


subjects


would have


no reason


to exaggerate


what


others


knew prior to outcome


knowledge.


findings


revealed otherwise;


Fischhoff's


1975


study,


subjects


second-guessed their peers


expected them to


know


in foresight


what


they


(i.e.,


subjects)


only


knew


in hindsight.


Effect


of Outcome Severity


Although most


of the


studies that


investigated the


effect


of outcome


information


utilized


outcomes that


were


neutral


from the perspective of


subject,


other studies


examined


the effect


of outcome


knowledge


in situations


which the


outcome was


non-neutral


Mitchell


and Kalb


[1981]


hypothesized that both outcome


outcome would affect


knowledge and


a supervisor's


valence of


evaluation


subordinate's performance.


researchers


had 55


student


nurses


read a brief


description


of two


incidents


of poor


performance by


subjects


hypothetical


also received an


subordinates.


outcome


Half the nurse-


(either benign










negative


manipulation


were


patient


cardiac


arrest


patient


fell out


of bed.


outcomes


benign


manipulation


were


patient


health


continued


improve


patient


fall


ofi bed.


results


supported


those


of other


hind


sight


studies


tha .


increased


outcome


subjects


knowledge


' perceptions


whether


of the


negative


or benign


inevitability


that


out comrs.


More


interest


ting


though,


was


finding


that


effect


was


greater


with


negative


outcomes


than


with


benign


out comes.


This


finding


was


supported


in a later


study


Baron


decision


Hershey


maker's


1988


decisions


in whi


were


ch subjects'


tempered


evaluations


favorableness


decision


outcome.


Outcome


earliest


severity


studies


was


to reveal


also


a factor


a hindsight


one


bias.


In a study


actual: y


designed


to expl


ore


attribution


of responsibility,


Walstei


[1967]


subjects


(high


school


students


college


undergraduates


participate


in a scenario


in which


deci


sion


been


made


to buy


a house.


Each


subject


was


provided


information


as to


amount


of financial


gain


oss


that


result


ed from


purchase


of the


house


asked


to rate


confident


was


that


he would


have


anti


cipated


rer orted


financial


outcome.


While


conclusions


as to


a a.


e.,


___ _











confidence


that


they would have


anticipated the


reported


outcome


increased monotonically


as a function


severity


financial


consequences.


Thus


both


the direction and


the magnitude of the


outcome affected


the degree of hindsight


bias


demonstrated.


Attempts to


Eliminate


the Bias


Fischhoff


and Beyth


[1975]


investigated possible demand


characteristics


from a


different


perspective.


They used a


within-subject

questionnaire


design and asked


completed two


probabilities which


gave


their


subjects


weeks earlier,


then"


[1975,


to redo a


"giving the


same


This


manipulation was used to encourage


subjects to view the


hindsight


one of


task


their


test


current


of recall


knowledge.


or reconstruction,


However,


and not


in other


studies,


the hindsight bias


effect


was


reduced.


Telling the


subjects


about


the bias and/or asking them


to work


carefully


[Fischhoff,


1977;


Wood,


1978],


were also


unsuccessf

partially


in eliminating the bias.


successful


to consider why


Fischhoff


in reducing the bias by


an outcome might


[Slovic and Fischhoff,


1977;


Davies,


[1977]


forcing


S


have occurred and

1987] were able t


was


subjects

others


O


reduce the


size


of the effect


by asking their


subjects


alternative outcomes might


occur.










outcome


information.


In fact,


if subjects


were


told


that


outcome


information


was


wrong,


hindsight


bias


effect


was


actually


eliminated.


Hasher


et al.


concluded


that


"highly


unusual


circumstances"


[1981


, p.


subj


ects


can


ignore


outcome


information


recall


their


original


decisions


Another


study


[Davies


, 1987


compared


three


potential


techniques


reduce


bias.


results


indicated


that


subjects


were


owed


to review


notes


made


during


initial


reading


task


cases


exhibited


less


bias


than


control


subjects,


that


those


subjects


were


not


allowed


review


their


notes


exhibited


as much


bias


as the


control


subjects.


These


results


imply


that


foresightfull


encoding


elaboration"


1987,


alone


insufficient


reducing


the bias


effect.


Another


experiment


revealed


that


subj


ects


were


allowed


to review


their


notes


exhibited


much


bias


as those


generated


reasons


to why/how


each


outcomes


could


have


occurred)


in hindsight,


prior


making


their


judgments.


Although


both


groups


exhibited


less


bias


than


control


group,


implication


that


while


reviewing


foresightful


documentation


help


reduce


bias,


hindsight


alternative


generation


outcomes


might


reasons


have


as to how


occurred


and/or


be equally


effective.









Accounting


Studies


earliest


study


in the


accounting


literature,


Buchman


1985]


used


students


their


last


semester


masters


program)


in a bankruptcy


prediction


experiment.


research


was


motivated


third


parties'


financial


statement


users)


use


of outcome


information


was


intended


auditors)


to generalize


In a 2x2


particular


between-subjects


group


research


e.g.,


design


examined


failed


four


firm


hypotheses,


using


experimental


financial


case.


Only


statements


hindsight


bias


hypothesis


was


supported;


reporting


bankruptcy


increased


perceived


likelihood


that


it would


happen.


results


were


not


significant


"surprise "


scenario


where


type


of opinion


received


firm


was


manipulated


(i.e.,


qualified


versus


unqualifi


ed) ,


or as to the


relevance


cues


extent


of hindsight


bias


exhibited


subjects.


bias


In fact,


exhibited


results


manipulation


indicated


expe


there


cted


was


to produce


greater


less


bias


case


where


a qualified


opinion


been


issued)


In a working


paper


based


upon


dissertation,


Helleloid


[1985


used


a 2x3


split-plot


factorial


design,


tax


professionals


program),


(accountants


enrolled


six hypothetical


cases


a MBA


two


taxation


experiments


.e.,


.e.,










planning


problem


(foresight


group)


versus


a compliance


problem


was


(hindsight


tax-planning


group),


while


(foresight)


second


or tax


advocacy


expe


(hind


riment


sight


situation.


Neither


experiment


revealed


evidence


of hindsight


bias.


most


recent


accounting


study


Brown


Solomon,


1987


examined


effect


outcome


information


managerial


performance


evaluation


decisions.


Upper


level


students


in business


were


asked


assume


role


ass


istant


controller


to evaluate


a capital


budgeting


proposal.


planned


comparison


research


design


used


between-subjects


independent


variable


es--outcome


involvement.


outcome


factor


four


levels,


two


which


contained


a nested


manipulation


to investigate


whether


or not


evaluate's


response


ibility


outcome


(from


evaluator's


perspective)


influenced


effect


of outcome


information


evaluator's


judgement.


second


independent


studies


variable


[Fischhoff,


one


1977;


shown


Wood,


to be


1978])


significant


was


in earlier


evaluator's


involvement


with


decision.


Prior


involvement


was


operationalized


having


subject


evaluate


case


data


decide


among


other


things)


whether


or not


proposal


outcome


was


to be


favorable


or unfavorable.


prior


involvement"


simply


meant


that the


subj


was


expli


city










dependent


variable


was the evaluator's


assessment


of the


decision.


The experimental


results


revealed significant


differences


between


foresight


and hindsight


performance


evaluations


when


evaluator


(i.e.,


subject)


had no


prior


involvement


with


proposal,


evaluate


"relatively


higher


res


ponsibility


anticipating the


outcome" [

situations


1987,


(e.g


575]


when an


However,


since


evaluate's


in certain


decision processes


are


not


observable)


use of


outcome


information may not


represent


bias,


Brown


and Solomon


concluded


that


their


results


"appear to


support


least


a partial bias


interpretation"


[1987,


575]


Summary


As mentioned above,


Helleloid


[1985]


did not


find


evidence of the bias


results may


among


have been


subjects


(tax professionals)


driven by the nature of


task


and the expertise and


training


of his


subjects.


The task was


similar to a real-world advocacy


situation


in which the


emphasizes the need on


part


tax adviser to


heavily


weigh


taxpayer's priors


vis-a-vis


the outcome


information


available.


Also,


manipulations of hindsight


foresight


were


unusual


that both


groups


actually received










test


of hindsight


entirely


bias,


unexpected.


finding


Additionally,


in his


no effect


second


not


experiment


there


was


no control


group;


was


more


a test


differential


effects


between


favorable


unfavorable


outcomes.


Buchman


[1985


found


that


reporting


bankruptcy


increased


perceived


likelihood


that


would


happen,


that


effect


was


stronger


surprise


scenario


(where


unqualified


opinion


had


been


issued)


experimental


case,


however,


consisted


condensed


financial


statements


an actual


firm


that


ed for


bankruptcy


six months


after


issuance


firm'


s annual


report.


data


may


have


been


so diagnostic


of a fail


ed firm


that


overshadowed


kind


of opinion


ssued.


Also,


other


research


[Mutchler,


1984;


Mutchler,


1985]


shown


that


going-concern


audit


opinion


be redundant


some


degree.


this


case,


"opinion


issued"


may


have


been


an effective


manipulation


testing


surprise


hypothesis.


Brown


Solomon


1987]


also


found


evidence


effect


though,


outcome


they


influence


bias


information


identified


judgment.


factors


an accounting


task.


More


that


importantly,


appear


However,


their


wording


in the


prior


involvement


manipulation


which


their


cautious


finding


of hindsight


bias


based)


have


_










basic


events


subject's


hindsight


in hindsight


lack


effect


are


i.e.,


affected


awareness


of the


probability


outcome


effect


ratings


knowledge),


outcome


knowl


edge


on hindsight,


change


in the


predictive


importance


preoutcome


cues


once


outcome


information


available.


It also


appears


that


experience,


task


demands,


involvement


with


decision


prior


learning


outcome,


one


s responsibility


or control


over


factors


that


influenced


outcome,


and/or


surprisingnesss"


outcome


affect


degree


of hindsight


bias


exhibited.
















CHAPTER 3


RESEARCH PROBLEM AND


HYPOTHESES


Introduction


Published research has demonstrated that


hindsight


bias


effect


robust


over


a variety


of subject


populations,


experimental


designs,


task


contents and response


instructions.


This


robustness


is due


in part


to the


generally adaptive nature of


the bias1


and the benefits


offers

biases,


strategies


(e.g.,


favorable


persists


self-presentation).


in part because


to reduce mental


effort


humans

[Simon,


Like many


adopt

1955]


other


processing

and have


both


limited information processing capacity


and limited


memory


capacity


[Newell and Simon,


1972] .


Given


these


limitations,


appears


important


that


individuals


sensitive


to new information and


change


[Einhorn


and Hogarth,


1988] .


Although


term


"bias"


generally


carries


pejorative


implications,


reasonable


postoutcome


an individual


situations


to reconstruct


it may not


a prior,


always be


less-


informed state of mind;


appears


least


efficient


to omit


-I


~V










obsolete or


seemingly


irrelevant


data


(i.e.,


data


that


do not


support


the known outcome)


Audit


Judament


Implications


However,


while


possibly adaptive


some day-to-day


situations,


setting.


hindsight


It may


bias may not be


fact have


innocuous


undesirable effects


audit


in many


unstructured audit


judgment


tasks,


especially


if the auditor


unaware of


its presence and/or


Although numerous


studies


have


effect


investigated


judgment.

the effect


outcome


there


knowledge


little basis


judgment


(i.e.,


generalizing


hindsight bias),

from them to the audit


setting.


Few


studies


used realistic tasks2


and even


fewer


used experienced subjects


and,


even among those,


results


are


contradictory.


It may be


that high involvement


and/or


expertise enable


auditor to not


only


recall


the audit


records


and client


data,


to also avoid the biased recall


and/or


biased interpretation


of such data


that


often


exhibited in


behaviors


laboratory


(i.e.,


studies


biased recall


with student


and/or


subjects.


interpretation of


These

cues)


appear to be a


contributing


factor


in hindsight bias.


example,


Helleloid' s


[1985]


tax professionals,


performing a


realistic task,


did not


exhibit


the bias.










the other hand,


it may be


that


expertise


is actually


conducive


to the bias


effect because


the expert's


greater


knowledge base may provide


cognitive activity


a wider


in hindsight"


scope


[Davies,


for the

1987,


"biasing


which


could,


turn,


lead


to favoring


evidence


support


reported outcome.


found evidence of


example,


outcome


the Arkes et


knowledge effects


study


among


[1981]


experts


(i.e.,


physicians)


asked to


render a


"second opinion"


on a


medical


diagnosis


a set


symptoms.


There are also motivational


factors that may


contribute


to the effect


, especially


among


experts.


Fischhoff


[1975]


suggested


to believe,


is quite


that


flattering to believe,


we would have


known all along wh,


lead others

at we could


only


know with


outcome


knowledge,


that


that


we possess


hindsightful


foresight"


[1975,


298]


This


concern


with


self-presentation


(i.e.,


the need to


feel


competent


or the


need to make a


good


impression)4


may encourage auditors


anchor on


outcome knowledge.


Expertise may also


foster


overconfidence.


Danos


et al.


[1984]


reported evidence


indicating that


experts


(i.e.,


bond


raters evaluating the effect


of management


forecasts on an


initial


rating)


tend to be more


confident


their


judgments


3However,


as note


d by Shanteau


[1987],


in the majority


of the










than


are


novic


es.


As noted


in Chapter


overconfidence


fundamental


judgment


error


in prediction


"hindsight


babl


primary


source"


Hawkins


Hastie,


1987,


On the other


hand,


earlier


studies


(e.g.,


Goldberg


[1959]


and


Oskamp


[1965


produced


different


findings;


diagnostic


confidence


of experienced


psychologists


was


less


than


that


those


with


less


experience.


that


expertise


leads


to a clearer


understanding


of the


degree


complexity


that


often


underlies


many


diagnostic


tasks.


potential


effects


of outcome


knowledge


may


also


extend


1981;


judgments


Brown


made


Solomon,


third


1987;


parties


Baron


[Mitchell


Hershey,


Kalb,


1988] .


third


party


individuals


.e.


, bankers,


stockholders,


auditors


serving


as expert


witnesses,


etc.)


are


also


subject


bias,


the major


behavior,


replicated in


most


studies


with


student


subjects,


tend


ency


to second-


guess


another


individual


s or group's


deci


sions.


example,


customer


a creditor


unexpected


incurs


business


sses


failure,


as a result


attempt


anal


preoutcome


to the


known


financial


outcome.


data


a result,


to construct


data


a causal


supporting


business


s failure


be assigned


more


relevance


than


was


originally


ass


signed


auditor


prior


occurrence


~










the business


failure.


an auditor,


in the


role of


expert


witness


for the


court,


shares these


tendencies,


result


may be undeserved financial


and reputational


losses


for the


failed firm's


auditor


if the court's


ruling


is more strongly


influenced by actual


outcome


information


(e.g.,


failure


firm),


rather than by an


objective evaluation


of the


audit


procedures


used.


Outcome


should not be


deciding


factor


as to whether


or not a


preceding


course of


action


was


"good"


[Einhorn and


Hogarth,


1987] .


Rather,


the appropriateness


of actions


time


they were


implemented should be more


relevant


[Gibbins,


1984] .


However,


the evaluation of


quality


decision i

actions or


s difficult because many


decisions


important


long period of time


outcomes

(i.e.,


follow

the


delayed feedback problem)


In such


cases,


decision


quality


should depend upon


of the decision


such

the


things


time


as the


wisdom or consistency


implemented,


without


reference


to the outcome


[Gibbons,


1984].


Whereas


Helleloid' s


[1985]


subjects


did not


exhibit


bias,


subjects


two other


accounting


studies


[Buchman,


a review of


hindsight


research,


Arkes


1981]


wrote,


"there


always


enough


outlandish diagnoses...


evidence


Given


to nourish all but


enough data,


many


the most


diagnoses


can


appear


obvious"


[1981,


326]


7This


view


congruence


with


formal


decision


theories










1985;


Brown and Solomon,


1987]


did.


With only these


accounting


studies


(some of which may


suffer


from


methodological


problems),


no real


theory of hindsight bias


[Hawkins


and Hastie,


1987],


likelihood of


increased


auditor responsibility


a result


issuance of ten


new auditing


statements


since early


1988


and the


litigious


auditing environment


[Collins,


1985],


important


investigate


the effect


of outcome knowledge


within an


auditing


setting using experienced auditors


as subjects.


results


of such an


investigation may have


implications


auditor training,


use of


decision aids,


the extent


documentation


required in


audit


workpapers,


the manner


which


workpapers are


reviewed and/or the use


of auditors


expert


witnesses


court


cases.


follows


that


hindsight bias


generally adaptive,


training


experience of the auditor may not be sufficient


its

"the


effect

very


the audit


setting.


outcome knowledge


which


According to

gives us the


to mitigate


Fischhoff,

feeling that


understand what


the past


was


about may prevent


us from


learning


anything about


[1982,


343]


Research Hypotheses


Devel moment


This


research may be


viewed


as descriptive.


As with


other biases and heuristics,


there


is no theory


of hindsight










to whether


or not


auditors


are


affected


outcome


knowledge


nor


an investigation


to determine


what


factors


might


influence


effect


of outcome


knowledge


judgment


effect


bias


on the


evaluation


preoutcome


evidence.


If the

information,


auditor


previous


susceptible


research


effects


suggests


several


of outcome


behaviors


that


be exhibited.


First,


outcomes


may


appear


more


expected


in hindsight.


fundamental


hypothes


then


that


auditors


are


subj


to hindsight


this


will


serve


a base-line


prediction


(see,


example,


Brown


Solomon


1987])


hypothesized


prior


Specifically,


that


knowledge


auditor


state


will


audit


s judgment


be biased


setting


of his or another's

information received


subsequent


to that


prior


point


in time.


The
that
will


with


auditor
outcome


an auditor


outcome


as having


without


knowledge


een


more


outcome


will


probably


judge
e than


knowledge.


Second,


although


most


hindsight


bias


studies


used


tasks


that


were


outcome


neutral


see


earlier


discussion


prior


resea


audit


judgment


implications)


a few


studies


[Walster,


1967;


Mitchell


Kalb,


1981


found


valence


of the


outcome


(e.g


benign


as oppose


d to negative,


or financial


gain


opposed


to finan


cial


loss)


to affect


degree


to whi


ch the


bias


effect


exhibited.


audit


environment


some


outcomes


(e.g.,


successful


clients)


are










preferred


over


others.


This


preference


factor


may,


turn,


affect


extent


to which


bias


is exhibited


auditors.


other


lawsuits


hand


against


, as


a result


auditors


of the


involved


with


increasing


firms


frequency


have


failed,


auditors


feel


that


should


be able


accurately


measure


viability


therefore


influenced


more


"bad


news"


outcomes


(e.g.,


subsequent


bankruptcy)


than


ood news"


outcomes


significant


improvement


of a trouble


ed firm)


Auditors


will
bias
info


not


receiving
exhibit t


auditors


failure


same


rece


giving


outcome
amount o
success


information
hindsight


outcome


rmation.


third


knowledge


hypothesis


on auditor


investigates


confidence.


effect


If outcome


of outcome


knowledge


encourages


individuals


to believe


that


they


"knew


it all


along,


" it


follows


that


outcome


knowledge


should


also


affect


or contribute


to overconfidence.


Because


outcome


knowledge


makes


events


seem


more


certain


predictable


than


they


actually


were


prior


receipt


outcome,


auditor


overestimate


or her


judgmental


ability


preoutcome-knowledge


situations


thus


express


greater


confidence


in his


or her


judgment.


Outcome
confident
ability.


knowl


edge


in his


will


increase


preoutcome


auditor'


knowledge


judgment


g.,










when


he actively


evaluates


data


prior


learning


outcome)


Without


this


preoutcome


or fore


sight


involvement,


outcome

salient)


focal


around


which


piece


auditor's


information


memory


i.e.,


more


preoutcome


information


be restructured.


Involvement


with


preoutcome


information


interrupt


tendency


toward


biased


recall


and/or


biased


interpretation,


which


turn


reduce


a hindsight


bias


effect.


Auditor


foresight


involvement


will


decrease


hindsight


bias.


Fifth,


even


if foresight


strategies


prior


involvement


with


data)


are


effective


in reducing


hindsight


bias,


they


have


limited


applicability


from


practical


viewpoint.


Often,


unexpectedness


surprisingnesss"


original


an outcome


data


that


prompts


or an examination


a reexamination


memory


of the


original


data.


In such


situations,


preoutcome


or foresight


techniques


to reduce


bias


are


at best


limited,


more


likely,


unavailable.


such


situations


only


hind


sight


strategies


be available.


fifth


postoutcome


examines


hypothesis


strategy


whether


investigates


is effective


effect


whether


in reducing


of outcome


a hindsight


bias.


knowledge


diminished


auditor


explicitly


considers


after


having


g.,


I










tendency


toward


biased


recall


and/or


biased


interpretation


original


data.


Auditor


hindsight


involvement


will


decrease


hindsight


bias.


Sixth,


independent


whether


either


foresight


and/or


hindsight


strategies


are


or are


not


effective


in reducing


bias,


hindsight


state


gies


have


wider


applicability.


Therefore,


is desirable


to determine


if hind


sight


strategies

Hypothesis


Hypothesis


are


least


compares


with


as effective


foresight


hindsight


foresight


strategy


strategy


strategies.


examined


examined


Hypothesis


to investigate


this


point.


Auditors
strategy


utilize


do not


auditors


a hind


exhibit
utilize


sight
same


involvement


amount


a foresight


of bias


involvement


strategy.


Seventh,


outcome


interpretation


cues.


knowledge


affect


An auditor


attach


recall


more


and/or


relevance


and/or


more


causal


attribution


those


items


of evidence


that

would


support

have w


rithou


known

t the


outcome

outcome


than


knowledge.


* another

Outcome


auditor

knowledge


encourage


increase


a causality-directed


availability


search


of supporting


cue


cues


recall


memory


[Dellarosa


Bourne,


1984] .


In effect,


may


drive


search


evaluation


cues.


example,


Arkes


and Harkness


[1980]


found


that


when


physicians


assessed










consistent


with


final


diagnoses,


were


falsely


remembered


as having


been


presented.


Additionally,


supporting


evidence


searched


evaluated


first,


implication


that


disconfirming


evidence


may


search


later.


Einhorn


Hogarth


1981]


suggest


that


disconfirming


evidence


toward


decision


process


implication


that


be underweighted


"sufficient


and/or


(although


ignored.


necessa


explanations"


will


readily


accepted
3^~ ~~ / TlT


as predictive


[Hogarth,


1980,


106]


result


in the


auditor


attaching


great


a predictive


value


to such


cues


when


they


occur


in subsequent


judgment


situations.


Auditors


will


recall


more


cues


supporting


known


outcome


than


cues


supporting


alternative


achieved)


outcome.


Auditors


with


outcome


knowledge


will


attach


more
than


weight


will


cues


audit


supp


ors


orting


not


a known


have


outcome


outcome


knowledge.
















CHAPTER


EXPERIMENT


DESIGN


, TASK,


INSTRUMENT


SUBJECTS


, AND


ADMINISTRATION


Introduction


A laboratory


experiment


consisting


four


experimental


tasks


a debriefing


task


was


conducted


test


research


outlines


hypotheses


expe


eve


riment


loped


Chapter


designed


test


This


those


chapter


hypotheses.


experimental


design,


task


choice


of subjects


discuss


development


research


instrument


administration


of the


experiment


is also


described.


Experimental


results


are


examined


in Chapter


Audit


Task


Area


Although


many


aspects


audit


process


may


conducive


to the


exhibition


of hindsight


bias


on the


part


auditor,


this


study


will


investigate


presence


prediction


of business


failure.


popular


press


, according


to former


AICPA


President


Philip


Chen


often


equates


busin


ess


failure


with


audit










underlying business or economic


conditions,


not because of


poor auditing"


[1985,


be accurate,


reality


that


stockholders


want


the auditor to


reveal more about


likely


success


or failure of their


investments.


Two recent


studies


[St.


Pierre and Anderson,


1984;


Palmrose,


pursue


1987]


legal


reveal


remedies


willingness


against auditors


of stockholders


case


business


failure.


Although Palmrose's


[1987]


analysis


revealed


that


less


than a


fourth


of failed firms


become


involved in auditor


Pierre and Anderson

approximately half


litigation,


[1984]


results

lawsuits


also


that


supported the St.


indicated


involving


auditors


defendant s


since


19601


followed bankruptcy filings


and/or


significant


losses by the


plaintiffs.


Pierre and


Anderson


discovered that


there was a surprise


factor


in many


cases,


with bankruptcy


or a significant


loss occurring


after


several


years of profitable operations.


This


represents the


type of


situation


(i.e.,


a situation


with an


unexpected outcome)


in which outcome effects


have often been


demonstrated


e.g.


Wood


[1978]).


Although


the business


failure


prediction


task has been


investigated in numerous


studies,


issuance of


("The Auditor


s Consideration


of an Entity's


Ability to


Continue


as a Going Concern")


(AICPA,


1987a


SAS


4


1


v










("Reports


on Audited Financial Statements")


[AICPA,


1987b],


concern


auditor


responsibility,


as well


the nature


task


itself,


support


use


in a study


investigating


the dysfunctional


effect


of outcome


knowledge.


Additionally,


in a list


steps


covering the entire audit


process,


business


viability


decision


was


ranked first


in difficulty


and fourth


1987] .


in criticalness


Therefore,


378 auditors


recent adoption


[Chow et


of SAS


al.,


and SAS


increase


the workload of the auditor


(who


already


under tremendous


budget


pressure


in a


competitive market)


while the additional


responsibility may require the auditing


profession to reevaluate


current


audit


procedures


training.


Other


aspects


this


task,


such as


judgmental


nature and delayed feedback also


suggest


that


outcome effects


may be present.


Additionally,


is a


task


that


is conducive


use of


decision aids.


If auditors'


judgments


situations


where a


firm's


viability


is uncertain


are biased


by their


(i.e.,


auditors')


knowledge of


the audit


information of


other


failed firms,


auditor training may need


to include decision aids


that ameliorate


tendency.


Otherwise,


overestimate his


the auditor


is unaware of


ability to make


such


the bias,


decisions,


he may

may


inappropriately weigh


various pieces of


evidence,


and in









Participants


use of


students as


subjects


or participants


behavioral


accounting


and auditing


research has,


in the


past,


been


commonplace.


There are obvious


advantages of


surrogation,


such as


greater


availability and lower


costs.


fact,


early


studies


in accounting


[Dyckman,


1966;


Mock,


1969;


Hofstedt,


1972]


revealed a


strong relationship between


student


decisions and


those


of practicing


experts.


Other


studies


[Copeland et


al.,


1973;


Abdel-khalik,


1974]


concluded


that


students are not


reliable


surrogates


experts and


that


there do not


appear to be predictable


conditions


establishing the adequacy


of students


[Dickhaut


et al.,


1972] .


Other


studies,


using


students


subjects


in auditing


research,


have


also yielded


conflicting results.


Using


essentially the


same


internal


control


task,


Ashton


and Kramer


[1980]


and Hamilton and Wright


[1982]


concluded that


there


were no significant


correlations


between experience and


such


factors


as consensus,


cue


weighting,


self-insight,


consistency.


However,


Krogstad et


[1984]


found that


their


auditor


subjects


displayed


greater


consensus,


consistency,


and self-insight


than


their


student


subjects


an audit


materiality task.


While


results


n early research on


imoac


C.. L


_ j VI A


L- A-


r j










fact,


significant


[Weber,


1980;


Libby


, 1985;


Frederick and


Libby,


1986] .


These


research results


have eroded


support


use of


students


surrogates


auditors


in research


studies


[Krogstad et


al.,


1984] .


Abdolmohammadi


and Wright


[1987]


concluded that


students are especially


dubious


surrogates


in audit


tasks


that


are


unstructured or


semi-


structured.


research hypotheses


this


study were tested


within


context


of a laboratory experiment


using practicing


auditors.


One


hundred seventeen auditors


from the offices of


a Big Eight


while


accounting firm participated


in attendance at


this


a firm-sponsored training


experiment

session


during the


summer of


1988.


Exenrimental Desian


This experiment


resembles


the Buchman


[1985]


study


that


subjects


were asked to evaluate


information about


firm and


assess


firm's


viability for the


subsequent


period.


Unlike


Buchman


study,


this


experiment


used


practicing auditors as


subjects


and investigated


the effect


strategies to reduce


illustrates

experiment.


the bias


the between-subject


Additional


effect.

design s


information about


Figure


.chemat ic


4-1

for this


the design


will be


presented later


this


chapter.








37


0
H
LO 4

66)-



H 0 rO E

n
CO) >1
-d 0
-o H H
0r 0) *-H -
H as U -
CO NH U0 0fH
< s-H > ?
Hw w~ s-





H 04 u < i
4- C C






6) -H G
(U -H
0 0 rd 4P
o' 5









N 6 CC

f^(0 0 0)1 (1
.3 ft -d k









0 0 0 rl

^ 4 2) N H a) 6) -H 0 Nd




-" -- (
I-z- 0 P 0 0 C
c, c c; U






bH CO bJ (
x! u a> r -









cr, 0) 0 43P
N-'- -H~ c3) 4 C N
H- 5.4 0 0)
I O4 NI O -H -H4 +- 4)

rl ----- CO FiC ) N X
H CJ) Q) S ^
CO N X NtY' 0 CO
,o MD 0 0)^4
Elr .W 4-J -r- l0 Qa -Ii










I- -I a-H C)

0) 4 0 5


OH a >-H na )
6)6 > C- 6 )
^ w^~c co u- o>

p, k-H














-- U M 0 O (d
ON N) C3 (U- H
|C/ FC( 4H -nJL +j-r -n m
I NrC C 4->.0 .0 .0
.g L 9^ < 3 0
c* A^ A<- ^^ CO cO cO
^^^^^3^-.k r:^^^^^^^ M^_ _^ _^_^^ _









Administration


of the


Experiment


researcher


administered


test


instrument


participants


projects


during


during


time


a week-long


period


training


aside


session.


such


Although


research


there


were


nine


experimental


groups,


subjects


were


actually


divided

training


into


five


site.


groups


Group


size


range


space

d from


limitations


to 25


subjects.


Prior


commencement


experiment


subjects


were


assured


that


their


identity


would


remain


anonymous.


Five


training


session


leaders


assisted


distribution


collection


of the


test


instrument.


These


five


individuals


participate


experiment


During


experiment


researcher


moved


from


room


room


to monitor


participants


to be available


questions.


Each


subject


first


read


of General


Instructions


see


Appendix


then


Chemco


Case


data


(see


Appendix


This


was


identified


as Task


Those


subjects


Foresight


Involvement


treatment


groups


also


received


paper


upon


which


they


were


encouraged


to record


information


they


considered


helpful


in making


a viability


decision


about


Chemco.


After


this


task


was


completed,


Chemco


Case


data


was


placed


in a large


envelope


subject


and


not


used


remaining


tasks


the experiment.


Each


subject


then


___










completed


three


additional


tasks


(Tasks


debriefing


questionnaire


(Task


Each


task


was


completed


placed


large


envelope


with


Chemco


Case


before


next


task


was


begun.


subjects


were


not


allowed


(except


Foresight


Involvement


manipulations4


to refer


back


previously


completed


task


or to


Chemco


Case.


research


five


training


leaders


observed


participants


carefully


to make


certain


that


they


remove


Chemco


Case


from


envelope


time


after


completion


of Task


completion


experiment


researcher


(again


with


assistance


of the


five


training


leaders


collected


test


materials,


thanked


subjects


their


participation


discussed


experimental


tasks


with


some


parti


cipants.


ExPerimental


Methods


between-subjects


variables,


each


having


three


levels


were


manipulated


treatment


variable


to test

e was O


hypotheses.


outcome ;


was


primary


manipulated


as part


Task


experiment


Subjects


received


either


Outcome


or were


provided


information


that


firm


failed


subsequent


year


Failure


Outcome


group)


been


successful


in overcomincr


nrbnhlems


(i .


I r


e.,


J-J.LiL S-JL


1 1 l


.










judgment.


No Outcome


subjects


received


information


about


subsequent


performance


firm


prior


to making


their


viability


judgments,


they


thus


served


as a control


group.


Involvement


was


second


treatment


variable.


Subjects


rec


eived


either


No Involvement,


Foresight


Involvement


Hindsight


Involvement


treatment.


Foresight


Involvement


subjects


were


encouraged


to record


notes


while


reading


background


data


firm


whose


viability


they


were


evaluate;


this


was


part


Task


those


subjects.


Hindsight


Involvement


Involvement


subjects


were


not


allowed


to record


information


while


evaluating


case


data.


Hindsight


Involvement


subjects


were


required


list


as many


adverse


factors


mitigating


factors


they


could


recall


.e.


without


referring


back


case)


This


was


part


outcome


either


of Task


those


information


Failure


was


Outcome


subjects


was


provided


or Success


performed


subject


Outcome


was


group)5


before


viability


judgment


was


made.


in the


Outcome


manipulation


discussed


previously,


subjects


Involvement


group


illustrates


served


overall


a control


design


used


group.


this


Figure


study


acronyms


to be


used


to describe the


subjects


in each


cell.


after











Appendices


subjects


through


in each


K contain


nine


actual


treatment


tasks


groups.


O I


FAILURE


SUCCESS


NO NINO NIFO NISO


FORESIGHT FINO FIFO FISO


HINDSIGHT HINO HIFO HISO


FIGURE
Experimental


4-2
Groups


Subjects


were


randomly


assi


gned


one


of the


nine


possible


Outc


ome-Involvement


cells.


To eliminate


problem


of differential


carryover


effects


[Keppel,


1982],


to avoid


inherent


problems


of within-subject


designs


Pany


Reckers,


1987;


Joyce


Biddle,


1981],


to minimize


time


required


to complete


experimental


instrument,


each


subject

subject


received

analyses


only

were


one

used


treatment c

to analyze


combination.


Between-


data.


Hyvotheses


Hvnnth-h s o


T -I -'


S/" 1 1 1 -


1' cr.- 1 4{


h1 trnn t h


trim nth+ a Q o o


TC M E


ii


E 1I


C'










compared


the efficacy


the two


Involvement


manipulations to


one


another.


subjects'


Task


responses


(i.e.,


their probability


judgments


Hypothesis


the viability of


The precise


firm)


wording of


were


used to test


the elicitation


of the


judgment


response


chapter.


task will be discussed later


also Appendices C through K.


ANOVA


this


(one-way


and

data


two-way)


and post


Hypothesis


hoc tests were


utilized to analyze


Planned comparisons


were


used to test


Hypotheses


Hypothesis


Planned comparisons


were


used


to investigate


whether


the nature of the outcome


information


(i.e.,


Failure as


opposed


to Success)


differentially affected


the bias


Subjects


' responses


to Task


(i.e.,


the viability


judgment)


were used


to test


this hypothesis.


The wording


of the


Failure and


Success


discussed later


manipulations


in this


chapter.


this task will be


also Appendices C


through K.


Hypothesis 3


Hypothesis


investigated


knowledge on the auditor's


the effect


confidence


of outcome


in his


viability


judgment.


The dependent


variable was


each


subject's


response


w










chapter.


Two-way


one-way


ANOVAs


were


used


investigate


this


hypothesis.


Hypotheses


7a and


Hypotheses


7a and


7b investigated


effect


of outcome


knowledge


on cue


recall


cue


relevance.


Subjects'


responses


to Task


(i.e.,


cue


recall


task)


Task


cue


weighting


task)


were


used


to investigate


effect


of outcome


information


these


factors.


Planned


comparisons


dependent


were


variable


used


was


each


to investigate


subject'


Hypothesis


number


failure


cues


minus


or her


number


success


cues.


These


cues


were


obtained


from


Task


in which


subjects


were


asked


list


information


could


recall


about


Chemco


from


original


case


data.


precise


wording


of this


task


will


be discussed


later


this


chapter


can


also


found


in Append


ices


through


actual


procedure


used


to code


classify


these


responses


will


be discussed


Chapter


A two-way


MANOVA


(multivariate


analysis


of variance)


was


used

each


Failure


test


subject's


:ypothesis

ratings


Outcome


7b.

of 17


Success


dependent


Outcome


groups)


variables

case of


cues


were


the


from


Chemco


case


This


data


was


obtained


from


Task


in which


.e










discussed


later


in this


chapter.


also


Task


Appendices


Additional


though


Tests


In addition


above


tests


follow-up


tests


initial


result


several


tests


analyses


were


done


responses


to the debriefing


questionnaire


administered


expe


riment


.e.


Task


These


analyses


their


results


will


be discussed


in the


next


chapter.


DeveloDment


Experimental


Instrument


experimental


instrument


consisted


of information


about


a problem


firm


i.e.,


one


experiencing


financial


problems),


four


experimental


tasks,


a debriefing


questionnaire.


Development


of the


Chemco


Case


case


used


this


study


i.e.,


Chemco


Corporation)


was


carefully


constructed


to resemble


a chemical


manufacturing


firm


with


a potential


viability


problem.


included


one


page


of background


information


narrative


form),


financial


highlights,


comparative


financial


statements,


ratio


data


both


firm


industry


see


Appendix


In order


to create


such


a case


it was ne


!cessarv


to first


I


w










last

firms


years,


in order


have examined the


assess


financial


the predictability


information

of business


failure


(e.g.,Beaver


[1966],


Altman


[1968],


Deakin


[1972],


Libby


[1975],


McKee


1976]


, Ohison


[1980],


Casey


[1980],


Zavgren


[1985],


and Lau


[1987])


Typically,


these


studies


utilized discriminant


anal


ysis


(univariate


and multivariate)


or conditional


probability models


(e.g.,


logit model)


identify


financial


variables and


their predictive


accuracy


distinguishing


between bankrupt


and nonbankrupt


firms.


Several


firm.


one


of the


For example,


that


studies


McKee


was experiencing


developed a


[1976]


leas


composite


defined a pr

t one of the


problem


oblem firm

following:


. Bankruptcy
. Receivership
. Significant


deficit


in retained earnings


four


successive


years


Substantial


loss


. Reorganization
shareholders


for four


in which
or reduce


consecutive


the creditors


their


years
either become


claims


Disposing of


or discontinuing a major product


under


adverse


loss


conditions


Prior to


guidance


the adoption


the presence of


of SAS


34 provided


going-concern


uncertainties.


Although


did not


purport


to provide


an all-inclusive


list


events


that


would or


should signal


going-concern problems,


suggest


that


following were examples of


"contrary


'%%r.nH.


I


1 1


I r


rl *










information"


that might


indicate


viability problems


[AICPA,


1981,


section 340.04]:


. Recurring operating
. Recurring working c
. Recurring negative
. Adverse key ratios


ses


capital


deficiencies


cash flows


from operations


Loan


defaults


Arrearages


in dividends


As part


of a study of


auditor


decision making,


Kida


[1980]


composed a


list


items that


prior


research had


suggested


audit


were


partners


indicative of


a problem company and asked


(via a mailed questionnaire)


to evaluate each


item.


Based on


usable


responses


from 27


partners,


Kida


classified a


problem firm


(i.e.,


one having


going-concern


problems)


if it


was


likely that


firm would experience one


or more of the


following


events


within


year:


Enter receivership


Enter


reorganization pr


Inability to meet


oceedings


interest


payments


. Experience it,
losses


third


consecutive


year of


substantial


Liquidation
Experience


of its


assets


its third consecutive


year


significant


deficit


More


recently,


Kida


[1984a]


had 20 audit


partners


managers


evaluate a


list


of 20


information


items


about


a firm


to whether


each of


items


provided support


failure,


provided support


support


continued operations,


or provided


for both failure and continued operations.


Unlike


most


of the previous


research,


this


study utilized a


larger











TABLE


Ratio


Rankings by


Auditors


VARIABLE


RATING


Significant

If needed,


current 1

additional


oss f

debt


rom operations


capital


will be difficult


to obtain


Discussions with management


liability


indicate


from litigation


that


a material


likely


Discussions


with management


indicate


that


new


legislation may make


difficult


to market


firm's major products


one


Management


indicates


losing a major


that


there


is a


good chance of


customer


competence


of the


firm'


management


been


questioned by outside observers


company


than


Management
that


significantly


the average


and labor
labor will


less


firm in


indicate that


working
industry


there


capital


is a chance


strike


firm is arrears


in dividend payments to preferred


stockholders


Firm's market


share


is below average


for the


industry


SOURCE:


Kida


[1984a]


Appendix A,


339.


Mutchler


[1983,


1984,


1985,


1986]


continued the


investigation


of the


definition


of a problem firm in a


comprehensive


study that


participants were


16 audit


expanded


upon


partners


(two


the Kida


from each


studies.


of the Big


_____~~_____I_~______~~___ -----~-. ------- --ICII1 -----I--1III----1-CI-l--
-----~~~-----~~-------~~-- -----~~~-------~-~-~----~~11---~11-~












TABLE


Problem


Company


VARIABLES


Criteria


RATING


SCALE


as Ranked


Auditors


FREQUENCIES


*MEAN


RANKING


Enter

Enter


Receivership

Reorganization


Inability


erest


to Meet


Due


Third


Year


Substantial


Losses


Third


Year


Deficit


Negative


Worth


Going


cern


Audit


Report
Year


Previous


Second

Second


Year

Year


Deficit


Loss


Negative
Capital


Working


Negative
From 0


First

First


Cash


Flow


operations


Year

Year


Deficit


Loss


Liquidation


of Assets


= Ranking


a 5-point


sca


importance


identifying


a problem


company


SOURCE


: Mutchler


1984


Table


~--~~~~------~~~---~-~I--- -~~~I-----~~~I------~~~___~~______~I~
"~--~~~"--~~~~~'----~~~---------------










and means


each of


items


that might,


from an


auditor's perspective,


indicative of


a problem


firm.


In another part


the questionnaire Mutchler provided


the auditors


with a


list


ratios


(also derived from the


interviews and previous

them according to how i


research)


important


and had the auditors


they were


rank


the going


concern audit


report


decision.


Table


summarizes


auditors'


responses.


Both the Kida


[1984a]


the Mutchler


[1983]


choices as


to what


information


identifies


a problem firm were based upon


their


analyses


of previous


research and the


questionnaire


responses


of experienced audit


partners


and managers.


Together,


their


choices


represent


a well-documented means


identifying a problem firm or


creating the


financial


background for a hypothetical problem


firm.


This


study used a


combination


problem firm


characteristics


and ratios


identified by


Kida


Table


Mutchler


(Tables


4-3),


and SAS No.


and the


list


of problem firms


selected


from the Disclosure


II Database


the Mutchler


study


[1983]


to first


identify


a problem firm.


Tables


4-5 present


the characteristics and ratios


used to


identify a problem firm for purposes


of this


study.











TABLE


Ratio


Rankings


Auditors


~~- I-'~~~~-----~-~~~~~- ---


VARIABLE S


RATING


SCALE


FREQUENCIES


MEAN


RANKING


Cash


Flow


Total D

Current


ebt


Assets


Current


Net


Worth


Total


**Total


Debt


Debt


Total


Total
Total


Net


Assets


Assets


Income
Sales


Cash
Total


Working


Assets


Capital


Total


Assets


Income


Total


Current


Assets


Assets


Total


Net


Assets


Sales


Total


Assets


Ranked


from


most


important


least


important


to the


going-


concern


audit


report


deci


sion


Cash


Flow


= Working


Capital


from


Operations











TABLE


Problem


Firm


Characteristics


SAS


VARIABLE


Kida
1980)


Kida


Mutchler


{1984a)


Entering


receivership


Reorganization


Arrearages
Inability


Three


Three


in dividends


to meet


payments
consecutive
substantial
consecutive
significant


interest


years
losses


years o
deficits


Liquidation
Significant


assets


current


OSS


from
Difficulty


operations
in obtaining


additional


debt


capital


Likelihood


of material


liability
litigation


Difficulty
major


new


from


in marketing
product due


legislation


High


probability


losing


a major


customer


Competence


ment


of firm's
questioned


manage-


outside


observers


Working


capital


less


than


industry


average


High


probability
strike


of a labor


Negative net
Going-concern


worth


audit


report


in previous


year


Second y
Second y
Negative
Negative


ear
ear


deficit


oss


working capital
cash flow from


operations


First


year


deficit











TABLE


Ratios


Used


Identify


Problem


Firms


BEAVER


DEAKIN


MUTCHLER


VARIABLE


ALTMAN


KIDA


THIS


STUDY


Cash


Flow
Total


T. debt/
WC / T.
Current


Total


Debt


Assets
. Assets


Assets
Ratio


Retained


Earn


Assets


EBIT


Assets


MV equity


BV T.


debt


assets


Curren
Quick


Assets


Assets


assets


assets


Cash


Assets


ck Assets


Current


Cash


Current


Current
Assets


Liabilities


ck Assets


Working
Cash /


Sales


Capital
Sales


Net


Worth


Debt


Liabilities


EBT


Assets


Sales


. Assets
. Assets


firm


ected


was


Balchem


Corporation,


company


prince


ipally


engaged


in the


manufacture


sales


of specialty


chemi


cals


Mutc


hler


study


1983]


correctly


classified


Balchem


a problem


firm


that


not


rec


eived


going


-concern


opinion


It


was


also


firm


that


been


classified


Balchem


a problem


s financial


firm


report


Kida


s typifi


ed the


study

kind


1980]


of problem-firm


__










negative


signals.


The overall


impression gleaned


from the


annual


reports


was one


a troubled firm,


necessarily


one


of a failing firm.


Actual Balchem financial


statement


data and ratios


from


1982-1984


were


reviewed and


condensed


to create


the original


Chemco


Case


that


was


used in


this


study.


Industry averages


firms


with


same


four-digit


SIC number were


included


as part


case


case


was


materials


then reviewed by two


comparative purposes.


former auditors


(average experience of


five


years)


and,


based on


their


comments,


slightly modified


improve


readability.


Neither


auditor viewed


Chemco


failing firm and both


indicated


that most


auditors


would consider


to be


viable.


Design


the Tasks


Task


Task


entailed reading the


four-page Chemco


Case.


Additionally,


those


subjects


the Foresight


Involvement


treatment


groups


(i.e.,


FINO,


FIFO,


and FISO)


were encouraged to record notes


(on paper


included with


case)


about


firm while


reading the


case.


Task


Task


asked


subject


assess


Chemco's


viability


1987 at


the end of


1986.


Each


was


instructed to


Express
between


your judgment
0% and 100%,


in t
where


erms of


a probability


indicates


Chemco


not


likely to continue


as a going


concern









1987


indicates


Chemco


highly


likely


to continue


as a going


concern


1987.


Additionally,


each


subject


also


indicated


degree


confidence


that


judgment.


latter


response


was


measured


following


scale:


not
confident

0


extremely
confident


Those


FIFO,


subj


HIFO,


ects


NIFO)


in the


read


Failure


a brief


Outcome


statement


groups


orior


requests


viability


judgment


that,


in effect,


revealed


that


Chemco


filed


protection


under


Chapter


of the


Bankruptcy


Code


1987


Likewise,


those


subjects


Success


Outcome


groups


FISO


, HISO,


NISO)


read


a brief


statement


orior


to the


request


viability


judgment


that,


in effect,


said


emco


had


been


successful


in disposing


failing


divis


1987


Task


also


included


Hindsight


Involvement


manipulation.


Those


subjects


HINO,


HIFO,


HISO


groups


were


told


Based


Case


ONLY


data


upon
that


your


was


rec


Chemco


available


the 1986


audit,


indicate


those


factors


that


I I -.


1flr0 r


100%


.e ,


.e.


l m ^ /-


"P r


I


II4









cause


survival
factors),

reassure


to doubt


in 1987


Chemco's


(i.e.,


and those

you about


continued existence


adverse


that

Chemco's
in 1987


(i.e.,


mitigating


factors)


This


task preceded the


requests


for the


viability


judgment


and confidence


assessment.


A primary research


concern


with


viability


judgment


task


was


the potential


effect


framing


[Tversky and


Kahneman, 1981].

and Reckers, 1981;


Several

Waller


studies [I

and Felix,


loriarity,

1984b; F


1979;


icchiut


Shultz

e, 1984;


Stock and Watson,


1984]


have


concluded


that


task


presentation


mode has

psychology


an effect


on decision making.


[Einhorn and Hogarth,


1981]


Researchers

and auditing


in both

[Holt,


1987]


have provided evidence


as to how alternative


representations


of questions,


problems


and hypotheses may


lead to systematic variations


subjects'


responses.


example,


Kida


[1984b]


found


that


initial


framing


of a hypothesis


had an


impact


types


information a


group of


auditor


subjects


considered relevant


their


evaluation


of the


viability


of a firm.


The auditors


had been


provided with a


case


consisting


description


information


going-concern


items pointed to


items auditors


decisions)


consider


of a firm.


failure and the


important


One-half


other


half


when making


case


implied continued










evaluate


case


from


perspective


of whether


or not


firm


would


fail


within


years


while


other


half


of the


subjects


that


were


firm


asked


would


eva


remain


luate


case


viable


from


at least


perspective


two


more


years.


As hypothesized,


framing


effect


was


significant.


to this


concern


with


framing


effects,


Task


subjects


to determine


probability


that


firm


would


fail


would


or to determine


successful.


Instead,


probability


Task


that


cited


firm


subject


viability


judgment


on a sca


le of 0%


100%,


with


0% defined


"not


to remain


viable"


100%


"very


likely


remain


viable


Task


Task


was


a recall


task


that


each


subject


List
about


below


Chemco.


information


Your


can


include


rec


actual


numerical


data


.g.,


the balance


in an account


or the


amount


data


evaluations


or a spec
"healthy"
the firm


ific


ratio)


g.,


ratio),


qualitative


, or qualitative
an evaluation o


competence


manageme


subjects


were


not


allowed


to refer


back


Chemco


Case


before


beginning


this


task.


Although


subjects


rece


ived


same


version


of this


task,


those


subjects


Hindsight


Involvement


groups


.e.


, HINO,


HIFO,


HISO)


were


instructed


that


could


need


repeat


items


listed


as part


of Task


Task


In Task


subject


was


provided


a list


of 17


g.,










receiving


outcome


informational


from


Chemco


Case


asked


"indicate


important


each


of the


following


was


your


assessment,


viability


end of 1986,


circling


appropriate


of Chemco


number


s 1987


each


item.


These


responses


were


measured


on the


following


bipolar


scale


not
important

0


extremely
important


cue


list


included


ratios


provided


four-page


Chemco


Case,


seven


characteristics


that


have


been


used


assess


viability


previous


studies,


Failure


Outcome


Success


Outcome


treatment


groups)


outcome


provided.


subj


ects


within


each


Outcome


treatment


group


rece


ived


same


vers


this


task.


Task


Task


was


a questionnaire


designed


to elicit


background


as to the


information


helpfulness


about


of SASs


subj


ects


34 and


their


was


also


opinions


used


ascertain


whether


or not


those


part


cipants


Failure


Success


Outcome


treatment


groups


were


aware


of the


effect


outcome


through


information


K contain


on their


questions


judgments.


asked


Appendices


subjects.
















CHAPTER 5

EXPERIMENTAL RESULTS



Introduction


The discussion


in this


chapter


first


addresses


tests


formal hypotheses


and then


comments on the


additional


information


obtained


from the


responses to the Debriefing


Questionnaire.


SPSSx


statistical


software package


was


used to perform all

Planned comparisons,


statistical analyses

analysis of variance


for this

(ANOVA),


study.

and


multivariate analysis


of variance


(MANOVA)


were


used


examine the hypotheses


developed in


Chapter


Raw Data Results


raw data


providing the data


collected and the


can be


treatment


classified as


groups


indicated below:


Participant's


recorded notes


information about


the Chemco case


- Foresight


Involvement


subjects


Participant's


recall


of data


that


could be


classified as mitigating


or adverse


- Hindsight


Involvement


subjects.


(Recorded in an


annn-Pndedr










Participant's


viability


judgment


- all


subjects.


probability


between


100%.


Participant's


confidence


in his


viability


judgment


subjects.


(Response on a


ten-point bipolar


scale anchored on


"not


confident"


"extremely


confident"


Participant's


recall


of data


from the Chemco Case


subjects.


(Open-ended notes


recorded


on paper


provided with


task


instrument.


Participant's evaluation of


importance of


specific cue


s from the


case


to their viability


judgment


- all


subjects.


(Response on a


seven-point


bi-polar rating


scale anchored on


"not


important"


"extremely


important"


each


of 17


or 18


cues.


Participant's thoughts as


to the


relationship


between each


of the


following


and the


viability


judgment


- all


subjects


(measured


on a ten-point


bipolar


scale.


perceived responsibility


in viability


judgments,


the helpfulness of


34 and SAS


the amount


training provided by the


firm,


amount


of guidance


provided by the


firm's


audit


manuals,










difficulty


assessing


viability


this


experiment


Months


of audit


Position


experience


firm


- all


subjects


subjects


Responses


three


open-ended


questions


- all


subjects


Participant


s awareness


effect


out come


knowledge


in this


experiment


- Failure


Success


Outcome


subjects


above


actual


raw


data


are


reported


with


discussion


of the


research


results


hypotheses


tested


debriefing


questionnaire.


Preliminary


Tests


A preliminary


analysis


viability


judgment


responses


was


conducted


to ascertain


whether


or not


there


was


effect


group


differences.


participants


been


divided


into


five


groups


(due


space


limitations)


training


leaders


prior


to the


arrival


researcher.


This


divi


sion


was


not


random;


was,


in fact,


based


upon


training


schedule


participants.


Although


there


was


reason


to believe


that


there


would


a group


effect


groups


been


determined,


a one-way


ANOVA


with


Group


independent


variable


Judgment












TABLE 5-1
Effect of GROUP Partitioning on JUDGMENT

-------------------------------------------------
------------------------------------------------


SOURCE


d.f.


MEAN SQUARE


Between Groups
Within Groups


271.9447


.6544


.6250


.5900


difference


was


found


(F=.6544,


p=.625)


Accordingly,


judgment


responses


for the subjects


were


collapsed


across


five Groups in testing the hypotheses in this study.


COUNT


=> approximately


.4 occurrences)


MIDPOINT


*******************************


********** #
J~ kji3:l ~kf ;J t~ f3(k ( ; kk 4


********************


*********************


*****#
* **


7**#


*******#*******************


#***










A second group of preliminary tests was


undertaken


ascertain


if the assumptions underlying the


ANOVA model


were


valid for the data


(i.e.,


judgment


responses)


For


fixed effects model


(such


one


in this


study)


lack


normality,


unless


is extreme,


is not


an important matter


[Neter et


al.,


1985,


624]


In general,


F test


robust


and-leaf


such


situations.


display were


A frequency histogram and a


constructed


to visually


stem-


check


extreme nonnormality


(see Figures


and 5-2)


Neither


figure


raised any alarm of


extreme nonnormality.


000000035
00
00000000000025
0000000000000005555


0000000000005


5555555555555559


00000000000000000055555558
0000000055555
0


FIGURE


Stem and Leaf Display


JUDGMENT


A third concern


was


the error terms


the possibility


study.


of nonnormality of


way to detect


non-


normality


through


use of


a normal


probability plot


residuals.


If residuals


are normally distributed the










suggests


that


the normality assumption has not been


significantly violated.


1.00


+---------+---------+-- ------- +----- ----+


EXPECTED


1.00


FIGURE


Normal


Probability


Plot


(Standardized Residuals)


JUDGMENT


Figure


represents


another method of


testing the


normality assumption


-- a histogram of


standardized


residuals.


It also


suggests that


the normality assumption


appears


valid in this


Tests


study


homogeneity


of variances were also conducted.


Results


from both


the Cochran


(C=.4383,


.151)


and the










residuals.


Additionally,


since


the use of


equal


cell


sizes


also tends


to minimize the


effects of


unequal


variances


on inferences


with the F


distribution


[Neter


et al.,


1985,


624],


unequal


variances


were not


expected


pose


a problem


in this


study. 1


EXPECTED
N


6.42


14.66
15.49
14.66


*********#
*A**ww


Case


=> Normal


Curve


FIGURE


Histogram


(Standardized Residuals)


JUDGMENT


1However,


since


homoscedasticity present


a serious threat


=> 1


*****&#**
********-*****
*****f**k******~******


v --











+-----+-----+-----+ -----+-----+-------+ +


SYMBOLS


Max


OUT+


+-----+--1--+-----+-----+-----++


+3 OUT


PREDICTED


FIGURE


Standardized


Scatterplot


JUDGMENT


Residuals


Tests


of Hypothesis


As discussed


in Chapters


Hypothesis


concerns


effect


of outcome


knowledge


auditor's


assessment


of what


viability


judgment


would


have


been


without


that


outcome


knowledge.


Outcome


knowl


edge


was


manipulated


between


subjects


Outcome,


levels.


as one


Failure


was


of the


Outcome,


hypothesiz


independent


Success


ed that the


variables,


Outcome


subject's


with


three


assessment


w v










Table


presents


descriptive


statistics


viability


Judgment


Outcome


Involvement.


Involvement


Outcome


were


principal


treatment


variables


in the


study


Each


cell


labeled


with


treatment


acronym


, NIFO),


contains


judgment


responses


, the


mean


judgment,


the standard


deviation


judgment


responses


that


cell.


overall


judgment


mean


subjects


was


.7521.


Results


two-way


fixed


effects


ANOVA


conducted


test


Hypothesis


are


presented


in Table


5-3.


There


was


interaction


effect


.355


.840


Involvement


was


significant


(F=1


a significant


.627,

: eff


p=.201)


Outcome


on subjects'


knowledge,


viability


however,


Judgments


(F=14


.917,


.000


significant


effect


of Outcome


in this


study


allows


judgment


rejection


responses


across


null


levels


hypothesis


supports


of equal


results


previous


studies.


A set


of follow-up


analyses


was


conducted


using


only


responses


of the


subjects


No Involvement


cells


NINO,


NIFO,


NISO)


to completely


remove


potential


Involvement


confounding.


The result


s (see


Table


supported


initial


anal


.6152,


.0371


Outcome


continued


to show


a significant


effect


judgment.


Therefore,


e auditors v


were


told


,w Chemco ha


, p=


.e.,











TABLE


Cell Judgments, Means, and Standard Deviations
------------------------------------------------


NO
INVOLVEMENT


FORESIGHT
INVOLVEMENT


HINDSIGHT
INVOLVEMENT


NO
OUTCOME


NINO


70

71.5385%
(8. 9872)
n=13


FINO


.6923%


(20.8782)
n=13


HINO


.8462%


(18.4700)
n=13


FAILURE
OUTCOME


NIFO


65

60.0000%
19.1485)


n=13
FIFO


.8462%
.4531)


n=13
HIFO


75

56.9231%
17.5046)


n=13


SUCCESS
OUTCOME


NISO


85%
75
90
100
55
50


76. 0000
16.9706


n=13
FISO


7692%


(14.4115)
n=13


HISO


95

81.1538%


10.237


n=13


69.1795%
(15.8476)


63.7692%
(21.2819)


70.3077%
(18.4750)












TABLE


ANOVA


Results


- JUDGMENT


INVOLVEMENT


OUTCOME


SOURCE


d.f.


MEAN


SQUARE


Main


Effects


4368


INV


.214
.419


14.917


.000


.627


eractions


OUT


x INV


.060


Explained


1263


.355


.188


.840


.314


.000


Residual


.836


Total


TABLE


ANOVA


Results


- JUDGMENT


Control


Group


By OUTCOME
Only)


-------~-~~~ --~-- lr,,~


SOURCE


Between
Within


d.f.


MEAN


Groups
Groups


SQUARE


.6152


.0371


.1453


Total


viability


probability


they


would


have


assessed


1986


1987,


were not


given


responded


outcome


differently


information


than


those


about


auditors


Chemco


prior


======~ ==3== I==============r======;==5==== ==== ===-


_










objective


(i.e.,


untainted


outcome


knowledge)


assessment


what


judgment


they


have


made


1986.


Multiple


specific


procedures


relationships


were


among


conducted


judgment


to determine


responses


three


Outcome


groups.


first


analysis


included


judgment


responses


of all


subjects.


results


from


Tukey-HSD


Scheffe


procedures


are


shown


Table


results


subjects


indicate


in the


that


Success


judgment


Outcome


responses


No Outcome


those


treatment


conditions


are


significantly


different


from


one


another,


that


both


are


significantly


different


from


those


Failure


Outcome


subjects.


TABLE


Multiple


Comparison


SCHEFFE


Tests


PROCEDURE


Hypothesis


TUKEY


PROCEDURE


RANGES


.05)


MEAN

56.5897

69. 0256


.6410


=> pairs o
level


groups


significantly


different


N => No Outcome


groups


F => Failure


come


groups


========================================


would













A second


multiple


comparisons


analysis


was


conducted


with


only


those


responses


from


subjects


in the


involvement


control


group


(i.e.,


No Involvement


cells).


results


from


both


Tukey-HSD


Scheffe


procedures


are


shown


Table


These


results


indicate


that


when


Involvement


excluded


a treatment,


only


Failure


Outcome


Success


Outcome


judgments


are


significantly


different


from


one


another.


TABLE


Multiple
(No


Compari


sons


Involvement


Tests
Control


Hypothesis


Group


Only)


SCHEFFE


PROCEDURE


TUKEY


PROCEDURE


RANGES


.05)


MEAN

60.0000

71.5385

76. 0000


=> pairs o
level


groups


significantly


different


N => No Outcome


groups


F => Failure
S => Success


come


Outcome


groups
groups


F L - 4.- -


----- ------ --C-C-II----- -----------L----- -
111 1-1-11-------- --1)---- ----- ------ -C--IC1--l--- ----------)---II--- -


. .. 9 /-.


II


L I


A -- .- A


m & -


r


m A


. A __ %- -. % _


L


__










affected


outcome


knowledge,


that


nature


of the


outcome


determines


direction


of the


bias.


Tests


of Hypothesis


Hypothesis


predicted


that


nature


of the


outcome


information

hindsight b


would


ias


have


exhibit


an effect

d. That


on the


those


degree


subjects


amount


receiving


Failure


Outcome


manipulation


(i.e.,


news


that


firm


failed


in the


subsequent


year)


would


exhibit


same


amount


of bias


those


subjects


received


Success


Outcome


manipulation


.e.


news


that


firm


had


remained


viable


chapter,


in the


subsequent


these


year)


outcomes


are


As dis


not


cussed


qualitatively


previous


equal,


such


differentially


affect


amount


of hindsight


bias


exhibited.


clear


from


Table


that


mean


judgment


responses


groups


differed


across


between


Involvement


Failure


levels.


Success


In fact,


Outcome


statistical


analyses


of those


means


Ided


highly


significant


res


ults.


both


Involvement


control


groups


.605


.013)


across


three


Involvement


levels


(t=5


.464,


.000


difference


between


subjects


Success


judgments


Outcome


of the


subjects


Failure


was


Outcome


significant.


Hypothesis


however,


is concerned


with the


, p=










were


utilized


to determine


in effect,


one


outcome


produced


more


bias


than


other.


Accordingly,


null


hypothesis


tested


Ho2:


was


JMTFO


= JMTSO


where


JMTFO


JMTSO


refer


to the


mean


probability


judgments


those


adjusted


subjects


mean


Failure


judgments


Success


those


Outcome


subjects


groups,


in the


Outcome


control


groups.


To facilitate


analysis,


each


of the


Failure


Outcome


means


each


Success


Outcome


means


was


reduced


No Outcome


treatment.


mean


This,


same


in effect


level


reduced


Involvement


original


matrix


Table


to the


matrix


Table


5-7.


TABLE


OUTCOME


Adjusted


JUDGMENT


Cell


Means


~~~--~~~~~---~~~~~~---~~~II--~ -II- --C- --- ----C--- ---~Ll


I (NO 0
FAILURE


OUTCOME -
OUTCOME)


I(NO 0
SUCCESS


OUTCOME -
OUTCOME)


NO INVOLVEMENT


11.5385


.4615


FORESIGHT

HINDSIGHT


INVOLVEMENT


INVOLVEMENT


.8461


13.0769


.9231


.3076


37.3077


25.8460










total


bias


exhibited


Failure


Outcome


Success


Outcome


groups.


Therefore,


this


analysis


hindsight


bias


measured


as the


sum


of the


absolute


differences


between


cell


means


Outcome


treatment


groups


those


Outcome


control


groups


Table


summarizes


results.


Ove r


Involvement


Outcome


groups


amount


of bias


exhibited


Failure


Outcome


subjects


was


significantly


different


from


that


of the


Success


Outcome


subjects


(t=2


.415,


.017


TABLE


Effect


HYPOTHESIS

Ho2


of Kind


CONTRAST
VALUE

11.4617


of OUTCOME


STANDARD


ERROR


.7461


on JUDGMENT


PROB


VALUE


d.f.


.415


2-tail)

.017


investigation


of Table


reveals


interesting


points.


first


that


nature


Outcome


manipulation


differentially


groups.


i.e.,


Failure


affected


Interestingly,


or Success)

subjects ir


Hindsight


appears


1the


to have


Involvement


Involvement


group


highest


overall


mean


lowest


between


Involvement


treatment


groups.


These


observations


will


addressed


further


discussion


of Hvnaot ,hei


__1~~___~______ 1----- --11~1- -----1 -----CII--l--- I~lll-----C--- --II-)_
-----~I~~---~~I-1---1~~-~__~_~~~___~__~_


J-J.


L- .


I I










than


Success


Outcome


subjects.


This


be due


in part


nature


bias


itself.


As discussed


earlier,


usually


surprise


or unexpected


outcome


that


prompts


a strong


hindsight


bias


effect.


this


experiment


average


judgment


those


subjects


control


group


was


.54%.


This


implies


that


those


subjects


receive


outcome


information


were


almost


certain


that


Chemco


would


be viable

baseline


1987.


If this


judgment


percentage


subjects


also


approximated


Failure


Outcome


groups,


subsequently


receipt


failed


of the


1987


information


typified


that


kind


Chemco


surprising


out come


that


often


precedes


hindsight


bias.


Tests


of Hypothesis


Hypothesis


predicted


that


outcome


knowledge


would


have


an affect


on the confidence


level


of the


subject


in his


preoutcome


knowledge


judgment


ability.


Prior


investigating


this


relationship,


several


preliminary


tests


were


perform


Preliminary


Tests


Table


contains


confidence


scores


of all


subjects


summary


statistics


this


data


treatment


cell.


A preliminary


analy


confidence


scores


was












TABLE


Cell


CONFIDENCE


Scores


, Means,


and


Standard


Deviations


==~~~~~~~~~ -==t I== ==-I---------------- ---~-I-


NO
INVOLVEMENT


FORESIGHT
INVOLVEMENT


HINDSIGHT
INVOLVEMENT


NO
OUTCOME


NINO


>.8462
.9871
n=13


FINO


i .6923
..7022
n=13


FAILURE
OUTCOME


NIFO


.3846
.3868
n=13


FIFO


.4615


n=1 l


SUCCESS
OUTCOME


NISO


.2308
.4806
n=13


FISO


.3846
.2090
n=1l


I 0 --,


HINO


.3077


HIFO


HISO


.5385


5.8205
(1. 3153)


.8461


(1.6189)


.2051


___











Although


there


was


no a priori


reason


to believe


that


there


would


be a Group


effect


subjects


been


assigned


room


locations,


a one


fixed


effect


ANOVA


with


Group


independent


variable


Confidence


dependent


assumption.


variable


Table


was


5-10


performed


summarizes


investigate


results


this


test.


As expected,


no significant


Group


difference


was


found


(F=1


.5490,


.193)


TABLE


Effect


of GROUP


Partitioning


on CONFIDENCE


==-"~~~~~ --~--- --- -- -,


SOURCE


d.f.


MEAN


SQUARE


Between
Within


Groups
Groups


.2342
.0879


.5490


.1930


A second


preliminary


test


was


conducted


to verify


normality


assumption.


raw


data


was


used


to construct


histogram


a normal


curve


was


then


fitted


to the


plot.


Visual


inspection


of Figure


reveals


that


normality


assumption


not


been


seriously


violated.


Tests


homogeneity


variance


were


also


conducted.


Results


from


both


Cochran


.2488


.892


and


, p"










COUNT


MIDPOINT


=> approximately


.8 occurrences)


****N*
**************


*************N******************


8.00
9.00


**************

*** N


FIGURE


Frequency


Histogram for CONFIDENCE


Main Hypothesis Tests


The hindsight bias


effect


often been referred to


"I knew


outcome knowledge


all along"


phenomenon.


in general might


follows


increase


one's


that


confidence


in what


one's


judgment might


have been


without


that


outcome


knowledge.


Accordingly,


in this


experiment,


those


subjects


received outcome


information


1987,


prior to


having to


estimate


what


their


judgment


would have been at


the end of


1986,


might be


expected to


be more


confident


that


judgment


than


would subjects


who did not


have outcome


knowledge.


However,

news" as


the nature of the


opposed to


outcome


"bad news")


knowledge


well


(i.e.,


several


"good

other


factors


e.g.,


one's


responsibility


for the outcome or


one's


responsibility


for foreseeing the outcome)


might


also affect


one's


level


of confidence.











previous


analyses,


there


was


interaction


effect


between


Involvement


Outc


ome


.089,


p=.986)


Involvement


effect


.853,


p=.429)


There


was


, though,


an Outcome


effect


(F=3


.301,


.041);


outcome


know


edge


effect


degree


of confidence


expressed


subjects


in their


judgment


decisions.


TABLE


5-11


ANOVA


Results


- CONFIDENCE


INVOLVEMENT


and


OUTCOME


- ====~====~'~-~-- --
--~-1- -~~-~~~~=ZI== =


SOURCE


d.f.


MEAN


SQUARE


Main


Effects


.983
.803


INV


.301
.853


.041
.429


Inte


actions


x INV


.188


Explained


.089


.291


.083


.986


.381


Residual

Total


.115

.127


However,


when


Tukey


Scheffe


tests


were


run


on this


same


of data,


results


indicated


(see


Table


5-12


that


Outcome


groups


were


statistically


different


from


one


another,


not


from


No Outcome


group


Therefore,


although


ANOVA


indicated


that


outcome


had


a significant


v w


w










Failure


or Success)


not


outcome


itself


that


lead


ANOVA


results.


TABLE


5-12


Multiple


Comparison


Tests


Hypothesis


---------------------------------------------------------------------------------------------------------------------------------------------


SCHEFFE


PROCEDURE


TUKEY


PROCEDURE


RANGES


.05)


MEAN

.5385

.9487

.3846


=> Pairs


groups


significantly


different


level


N => No Outcome
F => Failure Ou


=> Success


groups


come
come


group
group


appears


from


Tables


5-11


that


outcome


knowledge


affects


judgment


confidence


that


effect


be related


more


to the


nature


outcome


information


than


actual


presence


information.


this


expe


riment,


subjects


receiving


Failure


Outcome


not


exhibit


same


amount


of confidence


those


subjects


receiving


Success


Outcome.


In fact,


subjects


receiving


Failure


Outcome


were not


only


less


confident


than


those










less


confident


than


those


subjects


receiving no outcome


information at


all.


These


results may be due


in part


to a


factor mentioned


previously.


was


noted


that


the auditor


subjects


(based


upon


case


data provided for the


firm)


appeared to view


Chemco as a


viable company,


as evidenced by the


71.54% mean


judgment


control


group.


Accordingly,


it may be


that


those auditors


who were


the Failure Outcome


treatment


groups


were more


surprised by the outcome


received and as


result


expressed


less


confidence


in their


judgment


ability.


Additional


Tests


A final


of post


hoc analyses


was


conducted to


determine


what


other


factors might be


correlated


with


confidence.


Table


5-13 present


Pearson


correlation


coefficients between


confidence


and several


other variables.


the coefficients


are extremely


low,


indicating no


linear


relationship between


confidence and experien


training,


difficulty


of the viability


decision.


Additionally,


plots of


these variables


indicated no apparent


higher


order


relationships.


Tests


of Hypothesis


Hypothesis


examined the


effectiveness of


preoutcome











TABLE


5-13


CONFIDENCE


Correlation


Coefficients


TEXP


MAN


CONFIDENCE


.004

.249


.007

.179


.003

.283


.014

.098


.010

.142


TEXP


=> Audit


experience


=> Perceived


amount


(measured
of formal


month


training


received


from


firm


MAN


=> Perceived


amount


of assistance


provided


audit


manual


DIF


=> Perceived


difficulty


of viability


assessment


typical


audit


=> Perceived


this


difficulty


experimental


of viability
task


assessment


=> two-tailed


probability


notion


that


if a decision


maker


actively


involved


with


data


upon


which


a deci


sion


to be based,


Drior


making


decision


Prior


to receiving


outcome


information,


he will


less


likely


to be influenced


outcome


knowledge.


That


outcome


knowledge


will


have


less


judgment


recall


in situations


where


individual


not


make


a decision


prior


to the


receipt


outcome


knowledge,


determination


of the


deci


sion


he would


have


made.










Involvement


treatment


would


exhibit


an amount


of hindsight


bias


greater


than


or equal


to those


subjects


Involvement


(i.e.


, control


treatment


groups.


Hindsight


bias


was


measured


differences


Outcome


absolut


between


treatment


mean


cells


.e.


value


judgment


Failure


sum


of the

Outcom


of the


subjects in t

e and Success


Outcome


mean


judgment


those


subjects


control


treatment


cells


.e.


No Outcome)


Operationally,


null


hypothesis


tested


was


FINO-FIFO


FINO-FISO


NINO-NIFO


+ININO-NISO


where


FINO,


FIFO,


FISO


, NINO,


NIFO,


NISO


refer


to the


mean


probability


Judgments


of those


groups


Table


5-14


(which


is an extension


exception


of the


of Table


5-7)


column,


clarification.


entries


in the


With


table


represent


differences


between


cell


means


Table


column


entries


represent


amount


of hindsight


bias


exhibited


each


three


Involvement


groups.


was


expected


that


subjects


' involvement


with


data,


to receiving


Failure


or Success


Outcome


prior


to having


to estimate


what


their


viability


judgment


would


have


been


without


that


knowledge,


would


reduce


effect


outcome


Planned


comparisons


knowledge


were


on their


used


judgment.


to investigate


prior


_~ __












TABLE


5-14


OUTCOME


Adjusted


JUDGMENT


Cell


Means


--IC11---l----l-l-~~~~~~- -1__1--- -~I-- ---11-----


OUTCOME


FAILURE


INVOLVEMENT


OUTCOME)


OUTCOME


SUCCESS


11.5385


OUTCOME)


.4615


16.0000a


FORESIGHT
INVOLVEMENT


.8461


.0769


HINDSIGHT
INVOLVEMENT


.9231


.3076


24.2307c


a => Amount
b => Amount


of bias exhibited


of bias


Invo


exhibited


Fores


Ivement


ight


group.


Involvement


group.


c => Amount


of bias exhibited


Hindsight


group


Involvement


opposed


to ANOVA,


can


be substantially


affected


unequal


variances.


Accordingly,


tests


homogeneity


variances


were


conducted


as part


planned


comparisons


analyses.


homogeneity


assumption


appeared


to be valid


both


comparisons


(Cochran


C=.4480,


.111;


Bartlett-Box


.971,


p=.140)


pooled


variance


statistics


this


hypothesis


are


shown


Table


5-15.


Based


upon


analysis


null


hypothesis

level. In


cannot


this


be rejected


study,


(t=1


involvement


.338,

with


.092


case


data


prior


II male


9230b


pt

th










reduce


effect


outcome


knowledge.


subjects


Fores


ight


Involvement


groups


actually


exhibited


more


bias


their


judgments


than


subject


s in


No Involvement


control)


groups.


TABLE


5-15


Effect


of FORESIGHT


INVOLVEMENT


on JUDGMENT


CONTRAST


HYPOTHESIS


VALUE


9230


STANDARD


ERROR


.1744


(1-tail)

.092


VALUE


.338


d.f.

113


Tests


of HVypothesis


Hypothesis


involvement


.e.


examined


, Hindsight


effectiveness


Involvement)


a postoutcome


in reducing


hindsight


bias


Some


studies


.g.,


Hasher


et al.


1981],


Davies


1987])


have


found


that


sometimes


possible


reduce


knowledge


effect


outcome


situations.


knowledge


example,


postoutcome


outcome


knowledge


appeared


to have


ess


effect


on an individual's


response


when


was


asked


to either


recall


the previous


decision


situations


where


receipt


individual


outcome


knowl


did

edge,


not

the


make a

decis


decision


prior


he would


PROB


t










hindsight


bias


greater


than


or equal


to those


subjects


in the


No Involvement


, control)


treatment.


Hindsight


bias


measured


between


as the


mean


absolute


judgment


value


of the


sum


subjects


differences


Outcome


treatment


cells


mean


treatment


Failure


judgment


cells


.e.


of those


Outcome


subject


, No Outcome)


Success


s in


This


Outcome


control


same


measure


used


to investigate


effect


of Foresight


Involvement


Hypothesis


above.


actual


values


used


are


included


column


Table


5-14.


A planned


comparisons


approach


was


used


investigate


following


null


hypothesis:


HINO-HIFO


HINO-HISO


NINO-NIFO


+ NINO-NISO


where


HINO,


HIFO,


HISO


, NINO,


NIFO


HISO


refer


mean


probability


Judgments


ose


groups


see


Tables


15) .


Analyses


can


using


be serious


planned


affected


comparisons


unequal


, as opposed


variances.


to ANOVA,


Tests


homogeneity


planned


of variances


comparisons


were


analyses


conducted


as part


Hypothesis


of the


Those


test


results


are


identical


Hypothesis


Therefore,


mentioned


above,


homogeneity


assumption


valid


this


comparison.


.e.,











Drior


to having


to estimate


what


their


viability


judgment


would


have


been


without


that


knowledge,


would


reduce


effect


outcome


knowledge


on their


judgment.


summary


statistics


this


analysis


are


shown


Table


5-16.


TABLE


Effect


of HINDSIGHT


5-16


INVOLVEMENT


on JUDGMENT


CONTRAST


HYPOTHESIS


VALUE


.2307


STANDARD


ERROR


.1744


VALUE


PROB


11 mim-


----------~-m----=t
t


(1-tail

.057


.591


Based


results


above,


null


hypothesis


upon


rejected


level.


In this


study,


postoutcome


involvement


prior


to having


to approximate


what


would


one


have


viability


assessed


to be without


that


outcome


knowledge,


to significantly


reduce


appear


effect


outcome


knowledge.


subjects


Hinds


eight


Involvement


group


exhibit


significantly


less


s than


those


subjects


in the


control


No Involvement)


.e.,


groups.


of Hypothesis


Tests


Although


neither


Involvement


treatment


found


to be


was


statistical


significan


reduction


of h i nds i b t


h ~ -


,


CM r L fcI










Hindsight


Foresight


outcome


Involvement


Involvement


knowledge


treatment


treatment


judgment


e* .


as compared


in reducing the e

hindsight bias).


to that


i.ffect

The


null


hypothesis


was:


Ho6:


JMTH I


= JMTF I


where


JMTHI


JMTFI


refer


to the


mean


judgments


those


subjects


in the


Hindsight


Foresight


Involvement


manipulations,


adjusted


mean


judgments


those


subjects


in the


Involvement


Outcome


control


groups.


Operationally,


JMTHI


(HINO-HIFO)


-(NINO-NIFO)


(HINO-HISO)+ (NINO-


NISO)


JMTFI


(FINO-FIFO)


- (NINO-NIFO)


(FINO-FISO)+ (NINO-


NISO) .


Refer


to Table


actual


means


used


these


formulas.


adjusted


means


used


to test


Hypothesis


are


means


from


column


Table


5-17


which


an extension


of Tables


5-14.


in the


previous


analyses,


hindsight


bias


measured


absolute


difference


between


cell


means


of the


treatment


groups


control


groups.










.401,


p=.689)


Therefore,


be concluded


that


in this


study


two


Involvement


groups


.e.


, Foresight


Involvement


Hindsight


Involvement


exhibit


significantly


different


amounts


of bias.


TABLE


5-17


OUTCOME


INVOLVEMENT


Adjusted


JUDGMENT


Cell


Means


FAILURE
OUTCOME


SUCCESS
OUTCOME I


SUM


NO INVOLVEMENT


FORESIGHT


INVOLVEMENT


.6924


.6154


.3078


NO INVOLVEMENT


HINDSIGHT


INVOLVEMENT


.3846


.8461


.2307


TABLE


5-18


Comparison


of FORESIGHT


HINDSIGHT


Involvements


t


CONTRAST


STANDARD


PROB


HYPOTHESIS

Ho6


VALUE


.0771


ERROR


.1744


VALUE

.401


d.f

113


(2-tail)

.689


Tests


Cue


of Hypothesis


Recall


====~== =======r==;=========:=~=== 2==2=====~========-










cues


(i.e.,


information


about


Chemco


Case)


supporting


outcome


manipulation


they


rece


ived


than


cues


supporting


alternative


outcome.


More


specifically,


was


hypothesized


that


those


auditors


rece


giving


the Failure


(Succ


ess


Out come


manipulat ion


would


recall


more


failure


(success)


cues


than


would


those


auditors


received


Success


(Failure)


Outcome.


Operationally,


null


hypothesis


tested


was


Ho7a


: CSNIFO


C CSNISO


where

and N


NIFO


ISO


refers

the N


No Involvement


o Involvement


Success


Failure

Outcome4


Outcome


group


group


equals


mean


of the


differences


between


each


subject'


number


of failure


cues


success


cues.


That


(Fi
X(Fi


n


where


refer


number


failure


cues


success


cues,


respectively,


recalled


subj


ect


in cell


denominator


n is


equal


cell


size


each


cell)


Prior


to actually


computing


these


means


was


necessary


to classify


each


subject's


responses


on the


cue


recall


task


Task


into


three


categories


failure


cues


i.e.,


.e.,


- Si)


.e.,










those


which


would tend to


support


the bankruptcy


outcome),


success


cues


(i.e.,


those


which


would


tend to


support


viable outcome),


and neutral


cues


(i.e.,


cues


which


could


be classified as


failure or


success)


This was a


three-


step process.


first


step entailed listing and then


alphabetizing all subjects'


responses


without


reference


to cell membership.


Each


of the


alphabetized responses


was


then


coded as


either a


failure cue or


a success


cue or


neutral


cue.


last


step this


coded


list


responses


was


used


to analyze


each


subject's


responses


and assist


calculation


the number of his


or her


failure cues and


success


cues.


specific direction of this test


is due


to the


disproportionate number of


failure cues


the Chemco Case.


greater number


failure cues present


in the


case


was


expected


to result


in all


subjects


recalling more


failure


cues


than


success


cues,


regardless


treatment


cell.


expected,


subtracting each subject's


number of


success


cues


from his


or her number


failure


cues


resulted in a


cue


score that


score


was


indicated a


positive


for almost


subject's


recall


subjects;


favored


failure


a positive

e cues,


whereas


a negative cue


favored success


cues.


score


indicated a


Therefore,


the mean


subject's

of the


recall

13 cue


scores i


the NIFO cell,


because of


Failure Outcome










Table


5-19


includes


means


cells


Table


summarizes


results


of the


planned


comparison


conducted


to examine


this


hypothesis.


Homogeneity


variance


tests


were


also


conducted


as part


of the


analysis.


Results


of the


latter


C=.3741,


supported


.00;


homogeneity


Bartlett-Box


assumption


F=.069,


(Cochran's


.933


TABLE


5-19


Cell


Means


Recall


Task


OUTCOME


FAILURE
OUTCOME


SUCCESS
OUTCOME


NO INVOLVEMENT


FORESIGHT

HINDSIGHT


INVOLVEMENT


INVOLVEMENT


.46*


95*


mean


difference


score


TABLE


Effect


of OUTCOME


Involvement


on Cue


Control


Recall


Group


Only)


t


CONTRAST


STANDARD


PROB


HYPOTHESIS


VALUE


ERROR


VALUE


d.f.


-tail)


Ho7a


.3077


1.1663


2.836


.004


---lCC--- ----I*I ------ ------C1--- --CICI---
-~~~-~~-----~~~~~,,-~ -~--- --13"-- --- --- ------ ---)-------- ---I---


.004










results


analysis


were


highly


significant


(t=2


.836,


.004)


this


experiment,


those


subjects


receiving


Failure


Outcome


manipulation


recalled


significantly


more


failure


cues


than


success


cues


as compared


those


subjects


Success


Outcome


manipulation.


Planned


comparisons


analyses


were


also


utilized


compare


cue


difference


scores


subjects


in all


Involvement


groups


(i.e.,


subject


Homogeneity


variances


tests


were


included


as part


of these


analy


ses


Cochran


Table


s C=.3731,


5-21


pa. 770,


summarizes


Bartlett-Box


F=.202,


results


three


.817


planned


comparisons.


overall


interpretation


this


analyses


that


outcome


knowledge


appears


to affect


cue


TABLE


5-21


Effect


of OUTCOME


on Cue


Recall


- -- -===------------------------ ------- --------------~~------------------------------------------------------- ------------------------------------------------------
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------


STANDARD


CONTRAST

FO-SO=0


VALUE


.3077


ERROR

.5805


VALUE


.976


d.f.

114


PROB

.000


NO-FO=0


-0.7949


.5805


.174


NO-SO=0


.5128


NO => Mean


cue


.5805


difference


.606


scores


.010


No Outcome


cells


FO => Mean of tl


cue


difference


scores


Failure