A model of reputation building and destruction

A Model of Reputation Building and Destruction Paul Herbig JACKSONVILLE

STATE UNIVERSITY

John Milewicz JACKSONVILLE

STATE UNIVERSITY

Jim Golden JACKSONVILLE

STATE UNIVERSITY

Previous studies

of a firm’s reputation have either confirmed its existence as on influence agent or described in general terms its effects upon other attnbutes [quality, price, advertising, etc.) IhIs study attempts to quantify

Reputation and Credibility: An Overview

the reputation concept through a competitive evaluation. The competitive credibility model of reputation building is formulated and tested through snnulation. Results are reported and discussed. J BUSN RES 1994. 31.23-31

Reputation is the estimation of the consistency over time of an attribute of an entity (Herbig and Milewicz, 1993; Milewicz and Herbig, 1993). This estimation is based upon the entity’s willingness and ability to repeatedly perform an activity in a similar fashion. Reputation is a multidimensional construct; a

R

firm can have numerous reputations-a reputation for quality, one for marketing, one for product innovation, etc. All these

eputation

and credibility

Examples include whether made by a manufacturer’s

are concepts

familiar

to us all.

to believe the product claims advertising, credit check/

verification for a new account, or whether to believe delivery dates or claims made by a vendor. The value of a firm’s overall reputation

is easily seen in its relationship

to a firm’s revenues.

As a firm’s reputation improves, so does its sales (Shapiro, 1982). A firm with a good overall reputation owns a valuable asset“goodwill? Furthermore, a firm’s favorable reputation may well translate into more credible ads (Goldberg and Hartwick, 1990), better received brand names, more recognizable corporate logos, and higher customer loyalty. Brand names can often be repositories for a firm’s reputation; high quality performance on one product can often be transferred to another product via the brand name (Moorthy, 1985; Wernerfelt, 1988).

Objectives and Contribution The objective of this study is to quantify the relationship between reputation and credibility and to quantitatively describe the different effects of credibility transactions upon the reputation and credibility

of the signaling

firm. This study provides

a quantitative dimension to these two constructs and shows their dramatically different effects upon a firm’s performance and the powerful disincentives they provide to a firm thinking about risking (milking) its hard-earned reputation for short-term This is achieved by a simulation in pricing reputation.

gains.

Addresscorrespondenceto: Paul Herbig, Department of Marketing and Management, College of Commerce and Business Administration, Jacksonville State University, Jacksonwile, AL 36265. This article was selected for The Best Paper Award of the Southwestern Marketing Association 1993 Annual Conference. Journal of Business Research 31, 23-31 (1994) Q Elsevier Science Inc., 655 Avenue of the Americas, New York, NY 10010

dimensions together for the firm, which and corporate

can also yield an overall (global) reputation is usually perceived in the brand names

logos. Reputation

(either of an attribute or globally)

is an aggregate composite of all previous transactions over the life of the entity, a historical notion, and requires consistency of an entity’s actions over a prolonged time for its formation. A firm’s perceived reputation lowers if it repeatedly fails to fulfill its stated intentions or market signals. This consequential loss of its perceived reputation prevents the firm from signaling effectively as its signal will then be given little attention by its competitors. A firm, then, has considerable incentive to work hard to establish a credible reputation. Companies often use reputation as a means of predicting the actions of competitors. Reputation, though, is an imperfect attribute because there is always a time lag effect: companies must continually adjust reputation after the latest period (Shapiro, 1982). The critical factor in the reputation lag is the time frame concerned (and costs) of learning (information flows).

with the speed

Reputation accounts for strong intertemporal linkages along a sequence of otherwise independent situations and in its presence causes quite different behavior from what otherwise would occur in its absence. The ‘chain store paradox” illustrates this phenomenon well:

(Selden,

1978)

An entrepreneur owns a chain of drug stores. Another entrepreneur wishes to open his own chain. The established businessperson gets to. choose whether to play tough (bash the new entrant to forbid his entry into the marketplace) or soft (acquiesce and allow the other to enter the marketplace). The payoff matrix of this decision is shown [(x,y) refers to (current entrepreneur payoff [xl, new entrant payoff Ly])]. In an isolated

situation,

the equilibrium

is at the minimax

MN

0148.2963/94/$7.00

24

J Bum

Res 1994:31:23-31

I? Herbig et al.

Table 1. Payoff Matnx

tion of other entitles of a specific firm. This creates the scenario whereby 90% of the model of reputation of a firm consists of

Decision of New Entrant to Enter Market Current

Entrepreneur

Enter

Do Not Enter

(x,y)

(x,y)

22 0.0

5,1 5.1

Soft Tough

credibility

(soft, in) and both parties gain 2; therefore the current entrepreneur will be better off playing soft. What is missing from the analysis and provides rationale for the observed market behavior is the concept of reputation. By playing tough in the earlier periods, the current entrepreneur will have established a reputation to deter future entrants Reputation, therefore, causes a different response in a dynamic situation than statically. Critical to the concept of reputation is credibility. Credibility is the believability of an entity’s intentions at a particular moment

in time. That is, credibility

is the trustworthiness

or

the extent of confidence in the source’s actually carrying out its intentions (Herbig and Milewicz, 1993; Milewicz and Herbig, 1993). Credibility is whether a company can be relied upon to do what it says it will do. Credibility is time-sensitive; the entity’s perceived

credibility

today can differ immensely

behavior of other firms and not the signaling firm. Only in the lower right-hand part of the model exist those signals and ultimate actions that can be controlled by the signaling firm The central module of the model is the relationship between

from

and reputation

and the feedback loops from the latest

Credibility Transaction to the module. A credibility transaction (CT) is the firm’s comparison between a competitor’s pronouncements or intentions and its true behavior or final actions. The credibility of a competitor increases if its actions agree with its statements and its credibility decreases if its actions and pronouncements

are inconsistent.

The four types

of credibility

transactions are: (1) true posmve (CTl: + + when a competitor indicates it will do something and subsequently does it), (2) true negative (CT4: - when a competitor indicates it will not do something and then does not), (3) a false positive (CT2: + - when

a competitor

indicates

it will do something

or take

some action and then reneges), and (4) a false negative (CT3: - + when a competitor says it will not do something and reverses itself by going ahead and doing it). A false positive or false negative CT is also called a mixed signal, that is signaling one intention then not fulfilling the signal. Note that CT1 and similar but opposite transactions as are CT2 and CT3 rying out an action). The feedback loop delivers dR (change in reputation transactions) and dC (change in credibility between

CT4 are (not carbetween transac-

its perceived credibility by the same firm on a previous or future date. A firm that fails to follow through loses its credibility; to regain credibility, it must again pay the high costs of reputation

tions) as the result of a particular CT. The CT module compares what a firm signals (through public announcements, intents, releases, etc.) to its actual performance. The results of this transaction (dC and dR) are fed back to impact both credibility

building.

and reputation

Once a reputation

is established,

the firm has ample

incentive to maintain that reputation. For example, a public accounting firm charges fees that reflect its reputation for accuracy in verifying client firms’ financial statements. Credibility would be lacking if the firm were to offer unverified statements. Loss of that credibility would mean fewer clients would be billed and hence less profits received. To summarize the constructs, credibility is the believability of the current intention; reputation is a historical notion based

in the central

module.

Note that positive

and

negative are all relative; a firm may view a drop in price negatively, whereas consumers may view such a move positively. Reputation building is related to the consistency of the outcomes: repeated similar outcomes strengthen a competitor’s reputation. A true CT can be either positive or negative, for in either case repeated consistency provides believability (in-

The relationship between reputation and credibility is modeled from a competitive perspective. Illustrated in Figure 1 is the competitive credibility model of reputanon. A firm bases its deci-

creased credibility) and heightened reputation (but in opposite directions). A firm can have a horrible reputation but be totally credible (as long as it is consistently bad!). Note the different effects: repeated true positive CTs lead to a higher positive reputation (for example, for quality or on time delivery) and a positive higher state of current believability credibility while repeated true negative CTs lead to a lower negative reputation (poor quality) but still a positive higher credibility (you know it is consistently going to be bad quality or always late). That is, repeated promises and follow through of what is judged a positive action leads to an outstanding reputation and supports the credibility of the firm, whereas repeated promises and follow through of what is judged a negative action leads to a poor

sion upon a competitor’s

reputation

upon the sum of the past behaviors

of the entity

Both credibil-

ity and reputation are dynamic in nature; both are prone to change over time and are a function of time. Reputation and credibility are states; reputation building and credibility establishment are processes.

The Model

past behavior

(among

other

factors)

to help determine the credibility of the competitor’s current signal, using reputation and credibility as a means of predicting the intents of its competitors. A firm’s reputation is the percep-

yet a high level of credibility.

On the other hand, a mixed signal decreases credibility and repeated mixed signals result in a total lack of credibility. Inconsistent mixed signals tend to erode a firm’s reputation; how-

Model of Reputation

Building and Destruction

25

J Busn Res 1994:31:23-31

,

dR

~

b

REPUTATION I,

Reaction to Marketing Signal

Formulation

Figure 1. Competlnve credibility model Abbreviations: CT, credibility rransacnon, dR. change m reputanon as a result of the credibility transaction; dC. change m credoblllty as a result of the crediblllty transaction

_--------------Pronouncement (Signal)

Signaling Firm I

I

t

1 Action I I ever, consistent mixed signals do eventually establish a reputation (of sorts). For example, if a competitor were to consistently

inconsistent reputation)

give a false positive,

The greater and more enduring the level of uncertainty, the faster the firnis reputation should approach zero and credibility ap-

fail to do something,

always promising

to

deliver but fails, the firm will establish a poor reputation. Or, conversely, if a competitor were to always provide superior products regardless of the signal shown, they will eventually develop a reputation for high quality. The CTs (the comparison of out-

mixed signals, reputation trends toward zero (“no” and credibility trends towards the unbelievable state.

REPUTATION

come to signal) influences credibility, but consistent of outcome alone is the final determinant of reputation. Figure 2 shows the credibility transactions and their projected influences on credibility and reputation. To achieve credibility a competitor must first develop a reputation (Bell, 1984); it usually takes many CTs before a reputation can be established. Credibility is, therefore, dependent upon reputation as well as the just prior CT. Credibility influences reputation only through the final outcome: promised quality must be delivered (a true CT) to build a positive reputation (Fitz-

Unbelievable

Believable + CREDIBILITY

gerald, 1988). We speculate the more consecutive positive CTs that occur the greater a firm’s positive reputation; whereas the more consecutive negative CTs that occur, the greater a firm’s negative reputation. In either case, the firm’s credibility should increase. Conversely,

as a result of uncertainty

resulting

from repeated

Figure 2. Relationship between reputation and credibility

26

J Bum Res

i? Herbig et al.

1994:31:23-31

Figure 3. Projected effects of different CTs upon preach the unbelievable state. As credibility tends towards unbelievable, reputation erodes and trends toward zero. Figure 3 shows the postulated relationship between reputation and credibility The absolute magnitudes of the changes (dR and dC) should move in the same direction and be approximately the same magnitude. tive credibility

This relationship

is a key part of the competi-

and

credibility

nal? What will be the effect of a second or third inconsistent mixed signal upon a firm’s reputation and credibility. The first mixed signal could be interpreted by ones competitors or customers as an act of nature and beyond a firms control. However, a second mixed signal, especially a second consecutive mixed signal should dramatically influence negatively a firms reputation

model.

reputanon

and credibility.

As the mixed

signals

occur

closer

together, the greater the effect on a firms reputation the second mixed signal should have. The more mixed signals that occur

Hypotheses to confirm reputation and credibility both in their individual relationships to CT’s as well as their relationship to each other.

consecutively, the quicker the reputation should be driven to zero and credibility to the unbelievable state. Thus, the following hypothesis appears to follow:

It is also necessary to verify the speculated differences seen between them; i.e., the historical attribute indicative of reputation but not of credibility. Hence the following hypotheses appear

H5: Mixed signals will have larger negative effects on reputation and credibility if spaced closer together than further apart.

How well does the model hold up under

testing? It is necessary

reasonable: Hla:

Reputation

Hlb:

transactions. Credibility is a function

H2:

prior reputation. A firm’s credibility

H3:

is a function

of the sum of the credibility of the current

is positively

CT and the just

related to its reputation.

The absolute value of the change of reputation and the absolute value of the change in credibility are positively related.

The four credibility

transactions

should

have different

(dR) (dC)

and credibility

but with different

magnitudes;

the

false positive CT should have a greater absolute magnitude than the false negative CT. The different CTs should also effect influence in different directions and magnitude. Hence the following hypotheses appear reasonable: H4a:

Credibility transactions are not equal on their effect on dR and dC. H4b: CTs can be rank-ordered in terms of their effect on the absolute magnitude of both dC and dR: False Posxive CT2

> False Negmve CT3

> True Posmve CT1

Procedure A market simulation One of the strengths

was used to view reputation formation. of simulation is that the rules of the game

are completely specified, the strategy set is known to all players, and an equilibrium predicts not only what is expected to happen but also the anticipated

ef-

fects A true positive CT should increase a firms reputation and its credibility. A true negative CT should decrease a firms reputation but increase its credibility. Inconsistent usage of the two mixed signals or false CTs should both result in a decrease in reputation

Method

> True Negatwe. CT4

What are the effects of a mixed signal? How long does it take to reestablish a firms reputation after delivering a mixed sig-

consequences

of any deviation

from

the optimal strategies of the players. Twenty-four graduate student teams were used in the experiment. The use of student subjects as a convenience sample was a major consideration in the research design in order to maximize homogeneity and internal validity Students were arbitrarily assigned to teams, and treatments were randomly assigned to teams. Teams were used to emulate the multiple buying influences found in the industrial decision-making process. Sixteen game turns were held. The experiment involved airline industry pricing on a single route. The airline industry is a particularly good choice because pricing is highly dynamic in nature (daily or even hourly rate changes can easily be made in response to competitor actions), competitive response can be quick, competition is well defined and has bordered product lines (routes) that are different and distinguishable from other products (preferred for ease of measurement purposes), competitors are well known with usually no more than three or four competitors on any route (a manageable number to simulate), a history on any one competitor

is well established

(a known

track record),

and the out-

J Busn Res 1994:31:23-31

Model of Reputation Building and Destruction

comes (profits) are typically exists on the profit profiles

not perfectly known (uncertainty for each competitor).

Each of the 24 teams was provided with a prearranged

se-

quence of treatments. One group was given positive CTs (+ +) for the first eight periods to allow a positive reputation to be built. A second group was given negative CTs (-) for the first eight periods to allow a negative reputation to be established. Transactions for the third group were totally randomized for the first eight periods by throwing a die before choosing the CT For the duration of the experiment, all the teams were the recipient

of a specified

pattern

of CTs.

Measurement After having allowed reputations to be established (after eight periods), all the teams were given a mixed signal during turn nine. This allowed measurement on the effects of mixed signals on the dependent variables. The teams were then given customized patterns according to the manipulation desired (the number of turns before another false signal or the effect of multiple consecutive false actions). The teams receiving positive reputation building during the first set of turns were given negative signals during the last half of the game and vice versa. This provided balance to the experiment and gave all teams equal opportunity to “win”- to garner the highest cumulative profitability A master game matrix (of treatment versus team for which period) was created before the experiment commenced and adhered to throughout the duration of the experiment. The payoff matrix was a simple 2 by 2 designed to have similar characteristics as the classical prisoner’s dilemma. If both your firm and the competing firm (Zeta) chose to price high, your firm would have received profits of $250 for that period. If both your firm and the competing firm chose to price low, your firm

27

receiving a mixed signal) to remain conservative and price low indefinitely. These results confirm many of initial expected responses. The two key constructs were credibility and reputation. A firm would make a pricing decision (high or low price) based upon the other firms (Zeta’s) previous turn comments and actions (CT) and the payoff matrix. Afterwards, the players were asked to evaluate two interval scales. Credibility was rated by the teams by evaluating the believability of its competitors comments or actions: Zeta Believability Rating (ZBR). At one extreme was - 100 representing “don’t believe a word, cannot trust, totally unbelievable? At the other extreme was +lOO representing ‘totally believable and unquestioned” The middle of the scale was the 0 point representing “neutral, do not know, not enough information? Zeta was rated every turn on players’ believability of Zeta comments

or actions.

The reputation construct was also evaluated every turn by each firm by completing a similar interval scale, the Zeta Reputation Index (ZRI). Zeta was also rated every term by the players’ on its pricing reputation. The reputation construct was based upon a determination by the players on whether it was perceived to be a high-price or low-price firm. At one extreme was - 100 “low price oriented!’ At the other extreme was + 100 “high price tendency: cant ascertain, When

period

At 0 in the middle of the scale was “don’t know, not enough information decisions

were issued

by the firms,

the rat-

ings were noted for the previous turn (ZBR and ZRI for period 10 are compared to Zeta CT for period 9). Data, therefore, tended to lag one turn. This was done because credibility and reputa tion is a lag function of the just prior CT,

hxntives

and Debriefing

would receive profits of $50. If your firm chose to price low while Zeta priced high, your firm would have received profits of $150. Vice versa, if you happened to chose to price high while Zeta underpriced your route, your profits for that period would be $0. Thus, the payoff for both to cooperate is greater ($250)

Enough incentives were given to motivate participants to compete (it was surmised that being exempt from the final was powerful enough incentive). The reward structure was expanded to provide part of the class grade as a reward for participation

than the payoff for either to disagree ($150) but the risk of pricing high while the other firm priced low is a major concern ($0). Even though a firm could enjoy high profits by playing

by the game administrator. cern as upon investigation same learning curve.

high if its opponents also play high, the threat of its opponent undercutting its price and resulting in a $0 payoff will naturally cause the firm to choose the safety (low price) choice if there is any doubt whatsoever about its competitor’s sincerity; the profit potential is not as high but the downside is minimized (50). Firms will gravitate towards this position only when they have great confidence (credibility) of the other player’s position (reputation). This will occur owing to the potential cost to the firm of an opponent defectingthe firms payoff falling from $250 to $0. It was expected that a series of true positive credibility transactions would be necessary before the playing firm took the greater risk (and consequential greater payoff) of selecting the high price. Equilibriums were reached after the fifth or sixth turn. A propensity was seen for all firms (after

in this study. A learning

To eliminate

effect of about two game turns was noted This was not thought to be a conall the teams had approximately the

any end-game

actions

(that is, illogical actions

that might have resulted if the players knew the duration of the game), the last turn was not declared until after the surveys were turned in on the last turn. After the full 16 turns with only a 17th to allow final indices to be calculated, the administrator debriefed the playing teams, surveyed them about the exercise via a written questionnaire, and interviewed samples on whether they had surmised the intent of the experiment and suggestions on improving future experiments. Results were then tabulated, awards given to the winning teams, and a presentation of the findings were given to the participants. Post-experimental interviews and surveys indicated that the subjects were unaware of the experiment’s purpose. The cover story was that the simulation was a study of competitive situa-

28

J Bum

Res 1994:31:23-31

P Herbig et al.

tlons. Interviews indicated subjects thought the purpose of the game was a study of “negotiations” or “competitive strate&‘Sev-

Table 2. Results of Regression Analysis of Reputanon as a Function of Credibility Transaction

eral indicated

Predictor Variable

that it “tested the effect that information

our decision-making process:’ real intent helps in assessing

had on

The fact that no one guessed validity of the results.

the

ctl

Limitations Reputation building occurs only in lengthy games: a game consisting of only two to four periods will have insufficient time for reputation building to occur (Selden, 1978). Typically, 20. period games will certainly allow sufficient time to encounter reputation building phenomena, and a minimum of eight periods is usually

required

(Sobel,

1985; Kreps

and Wilson,

ct2 ct3 ct4 Constant aS~gn~ficanr ‘xl

-

Std. Error

Regression Coeffkient 11978 -54 361 25.741 -5.095 - 15 077 at the a -

true positwe.

01

cr2 -

of B

t-Statistic

R*

Std. Error

f-Value

2.91” 8.96” 9 03” 2 93 11.25

4.12 -6 07” 2 85” 1.74 - 1.34

037”

52.10

2559

level false pmuve,

cr3 -

lake

negatwe,

ct4

-

true

negarwe

1982a).

Camerer and Weigelt (1988) observed reputation building in games of only eight turns. The duration of the game held in

transience distinguishes reputation from credibility. It is, therefore, necessary to establish that the summation model does not

this experiment therefore non building to occur.

perform

was of sufficient

duration

for reputa-

If finite, the game must also be sufficiently long and its expiration uncertain to prohibit end-game strategies from occurring. The number of periods remaining is critical; an end-game strategy tends to be played if the game is finite, the duration of the game known, and a relatively small number of periods remain to be played in the game. With 16 turns, sufficient time was given to establish and measure reputation. And with the control of uncertainty of when the end was coming, it is felt no end-game

strategies

One of the limttations

were played. comes from the simulations

game. Al-

though easy to run with a limited set of results, simulanons often are a poor reflector of reality. Concerns also exist about commitment and importance of this experiment in the minds of the student

subjects

In the real world,

competition

is always

well under credibility

and that the single-period

model

does not perform well with reputation as the dependent variable. Only if all four factors are significant will the hypothesis be considered confirmed. To verify reputation

works under the summation

model (hy-

pothesis la), the cumulative number of each of the CTs were regressed upon reputation. For example, if at the end of the seventh turn, SIX true positives and one false positive have occurred, the reputation score would be compared to 6 CT1 and 1 CT2. The results can be seen in Table 2. Reputation is highly significant when regressed upon cumulative CTs (F = 25, R* = ,386, p < ,001). Note the signs and magnitudes of the CTs are as expected - CT1 and CT3 are positive in sign whereas CT2 and CT4 are negative m sign. The constant coefficient term is not significant as expected. A similar regression using credibility was run; its results are shown in Table 3. The results using credi-

on the mind of the firm, whereas in this experiment, the attention given it by the student is uncertain. Because the company existence is often uppermost to its managers, competition is a life and death activity to them, whereas the game cannot im-

bility show a decisive lack of significance throughout. With an R* of .O and an F of 2, coefficients that do not significantly differ from zero, and an overall insignificant probability, this is sufficient evidence to disconfirm credibility as a function of the

pose the same importance to the student subjects. This concern could cause validity and reliability problems and must be

summation model. To define credibility

guarded against. Another concern is timeliness-previous actions are often forgotten or misread if the time lag is great. The “noise lag” in the intervening period might preclude drawing clear-cut conclusions (Riley, 1985).

esis lb), a regression

Despite

these limitations,

the intent of the study was an ex-

ploration test various hypotheses on reputation and credibility. The results for many of the hypotheses were very strongly confirmed, indicating that real-world testing of the model should verify the laboratory results.

Results Hypothesis 1 To verify Hypothesis 1, that reputation and credibility are as speculated from the competitor credibility model- reputation is a function of the summation of the CTs and credibility is a function of the latest CT and the just prior reputation-it is necessary not only to verify the model but also to test the thesis that

was computed. reputation was ity score would credibility are

as per the single-period

model (hypoth-

of the latest CT and just prior

reputation

For example if on the seventh turn, the latest 70 and the CT was a false positive, the credibilbe compared to 70 and 1 CT2. The results for shown in Table 4. The credibility regression is

Table 3. Results of Regression Analysis of Credibhty Function of Credlbhty Transaction

as a

Predictor

Regression

Variable

Coefficient

Std. Error of B

t-Statistic

R*

Std. Error

f-Value

ctl ct2 ct3 ct4 Constant

-0.71 -12.19 -12.45 -0.97 1784

3 70 11.38 11.47 3.73 14.30

-0 19 -1.07 -1.08 -0.26 1.25

0.03

66.20

2.14

No SlgnlfIcant ctl

-

results

rrue posm”e,

ct2

-

false posmve,

ct3

-

false negatwe,

ct4

-

true neganve

Model of Reputation

Building and Destruction

120,

Table 4. Results of Regression Analysis Using Credibility as a Function of Reputation and Credibility Transaction Predictor Variable

Regression Coefficient

ctl

Std. Error of B

12.49 -48.67 -71.64 17.88 0.17 9.55

ct2 ct3 ct4 Reputation Constant

4.37= -5.25” -8.29” 6.47“ 2.31” 0.903

2.86 9.26 8.64 2.76 007 10 57

Std. Error

f-Value

0 31d 48.67

16.29

t-Statistic

29

J Bum Res 1994:31:23-31

RZ

60-m

-20

!

I

I

I

I

I

I

-20

0

20

40

60

60

100

rep Each dot represents

“Indraw s~gmhcance at the a - 02 level Ctl - true pomve. ct2 - false posn,ve; ct3 - false negawe,

a pax

1 0

revised

and does not stgnify quanttties

ct4 - true negative

Figure 4. Reputation versus credibility. (F = 16.29, R2 = .33, p < ,001). The signs and

very powerful

magnitudes are again as expected: CT1 and CT4 are positive whereas CT2 and CT3 are negative. The constant coefficient as predicted does not exist. The coefficient of reputation is small but positive at ,169. All these factors were expected given the model. A similar regression using reputation is shown These statistics (F = 1.5, p < ,174, nonsignificant

in Table 5. R2 = ,017)

disconfirm reputation as a function of the single-period model. These four results clearly show the expected result of a particular credibility transaction upon a firm’s reputation and credibility and clearly competitive

confirm

credibility

the first set of hypotheses

and the

model.

was that the absolute value of both variables would change with approximately the same magnitudethe same event would influence the constructs with approximately the same impact (magnitude). The coefficient was ,614 thus indicating a high equivalence between dC and dR and the changes to both major constructs were of approximately same magnitude. Results (F = 100, p < 0.001, R2 = 0.371) confirm the hypothesis.

Hypothesis 4 The fourth hypothesis

concerned

actions upon reputation, &NOVA for these factors

the effect of credibility

trans-

dR, credibility, and dC. One-way versus CTs were run. Highly signifi-

abso-

cant results were seen thus indicating differences between the CTs for all four factors (F > 11 and p < 0.001 for all factors). But where were the differences? The means of dR and dC (as

lute reputation. Figure 4 shows the results of regressing the modified data. The coefficient was 0.85 showing considerable equivalence between the two constructs. The constant was marginally

differential effects were desired) for each separate CT were calculated then compared via t-test to determine if there was any significant differences (or if their difference was significant from

significant. The modified constructs move in the same direction and had similar adjusted magnitudes. The results (F = 192,

zero). Different

p < 0.001, R2 = 0.533) confirm

structs.

Hypothesis 2 Hypothesis

2 was tested

by regressing

credibility

upon

the hypothesis.

Hypothesis 3 Hypothesis 3 was tested by regressing the absolute value of dC (change between periods of credibility) upon the absolute value of dR (change

between

periods

of reputation).

The hypothesis

Regression Coefficient

Std. Error

of B

t-Statistic

Std.

ctl ct2 ct3 ct4 Constant

-0.65 -2.98 -4.35 0.39 2.86

2.65 8.17 8.24 2.67 10.28

-0.24 -0.36 0.53 0.15 0.145

Error f-Value

R* 0.017

47.39

1.562

No s~gndicant results

ctl - true posmve. ct2 - false pomve,

ct3 - false negmve;

have differing

CT produced

effects upon the con-

a positive

change for both

reputation and credibility. A true negative produced a positive change for credibility but a negative change for reputation. For mixed signals, dC and dR are negative for credibility. The false negative dR has a positive ent from 0. As reputation

sign but it is not significantly differis associated with consistency, one

possible explanation is that by consistently showing a false negative (saying negative but showing positive), then reputation is given a positive boost. The absolute magnitude of the true positive is greater than that for the true negative. The absolute magnitude of the false positive is greater than that for the false nega-

Table 5. Results of Regression Analysis of Reputation as a Function of Prior Credibility Predictor Variable

CTs did indeed

A true positive

ct4 - true negmve

tive. These results add further

confirmation

for the competitive

credibility model. Hypothesis 4b predicts the rank order of the CTS and is verified for dC but has a different order for dR with much smaller absolute magnitudes. The effect of the four CTs upon credibility appears as was speculated but the effects of the mixed signals on dR is not as great as those of true CTs. In fact, the differences are not significant for the two mixed signals for both dR and dC. One possible explanation is that reputation is more

J Bum Res 1994:31:23-31

30

f? Herbig et al.

forgiving: one mixed signal could be explained as an act of nature or beyond the ability of firm. A mixed signal, therefore, has an adverse effect on dR but of a smaller magnitude than with dC. Partial support for the hypothesis is therefore provided.

Hype thesis 5 The fifth hypothesis was tested by comparing means of the changes in reputation and credibility for the number and interval of mixed signals and testing for differences between these and non-mixed signal sets. Hypothesis 5 says the less the interval, the greater the response should be. In fact the change in credibility is greater the more the interval. Figure 5 (F = 37, p < 0.001, R2 = 0.336) provides a graphic illustration of this phenomenon. This seems to indicate that the longer the interval between mixed signals, the more confidence one has or regains in one’s opponent, and when a mixed signal does arrive, the greater its negative impact will be on one’s credibility. Because the effect on reputation is scattered, it appears that one event does not make or destroy a reputation; some negative impact is seen but certainly not clearly nor conclusively. The dynamics of the interval effect are more complex than speculated with negative changes in credibility increasing the greater the interval, while interval has minimal difference for change in reputation. According to the hypothesis, the greater the consecutive number of mixed signals, the greater the response should be. Significant differences were noted for change in credibility and a marginal difference noted for change in reputation. Multiple occurrences have greater negative effect on reputations. The response to the first defect (mixed signal) has the greatest impact on credibility It appears that subsequent mixed signals although still adverse are not as great- that is, the second shock is easier to absorb.

Conclusions and Strategic Implications The findings tend to support the competitive credibility model. The implication is a firm’s reputation and subsequent credibility are the result of the continuous process of credibility transactions. These transactions reinforce reputation and credibil-

100-m

.

intervals

Figure 5. dc versus number of intervals

ity. Credibility transactions, when they involved mixed (false) signals, are devastating to reputation and subsequent credibility. There is a high price to pay in lack of confidence in subsequent marketing signals for a substantial period of time. The results of the experiment clearly show the destructive power on reputation of a mixed signal. Research demonstrates that a mixed signal has three to five times the impact of a true signal. It, therefore, takes many applications of a true signal to restore a firm’s reputation to where it was before the mixed signal occurred. Managers must think twice about not living up to their commitments with the duration and magnitude of the penalty for not living up to their signaled intentions. The major managerial implication in this paper is a warning concerning the fragility of reputation. The recent emergence of the multi-billion dollar takeover or leveraged buyout (LBO) principally concerns companies with a number of well established brands. Because brands can be visualized as repositories of reputation, the temptation for the winning LB0 group is to milk the br+cl as a means to lower debt. A firm can milk its reputation for short-term gains. This strategy has its dangers. Reputation is a valuable commodity; its value is directly proportional to its fragility (Kreps and Wilson, 1982b). Once eroded it takes precious time to rebuild. A firm would want to use this strategy only if the potential rewards were considerably higher than the costs involved in its loss of reputation-the potential costs inherent in rebuilding its reputation. As the results of this experiment clearly indicate, managers can milk a product only infrequently. Aaker (1991) seconds this notion by warning managers of the dangers that short-term thinking can do to brands. Milking a “cash cow” or existing brand will only diminish brand equity. Equity once destroyed is difficult to replenish. Repeated mixed (false) signals will destroy reputation. And as reputation goes, profits seem to follow. Managers can not afford to rest on their laurels (or that of established brands) for customers are always asking the question “what have you done for me lately?” Constant attention to the product, service, and customers must be given to maintain an established reputation and maximize profitability over the long-term.

Future Research An interesting extension would be to measure the effects on credibility and reputation by successful and called bluffs. That is, how long does It take to establish credibility and a reputation, and how vulnerable are those; what does it take to lose them? Bluffing is also an interesting topic in the particular aspect of what value does it take to induce one to defect in a multiturn game when reputation is important. Bluffing should not occur often in games of complete information or public information where outcomes are known. But in incomplete games or where the other’s moves not necessarily published (as in a winning hand of poker with all other participants have folded), bluffing should have greater value and usage. When on& reputation is important and not to be easily squandered, what rewards or temptations must it take to induce one to defect?

Model of Reputation

-7

tnanK two anonymous,aK rev iewers for their insightful comments n.n-nrti--r A” Arof, of the manuscript slgnific--+ _‘lIIL JU~CJL,ULL> “11_..i”,. y LLvLu,u.s 1

I

ne autnors

and

1

I

J Bum Res 1994:3 1:23-3 1

Building and Destruction

7.,n

31

Kreps, David M., and Wilson, Robert, Reputation and Imperfect Information. J. Economic Theory 27 (August 1982a): 253-279 Kreps, David M., and Wilson, Robert, Sequential Equilibria. Econometrica 50 (198213): 863-894.

References Aaker, David A., Managing Brand Equity, Free Press, New York. 1991. Bell, Chip, Building a Reputation for Training Effectiveness. Training and Development J. 38 (May 1989): 50-54. Camerer, Colin, and Weigelt, Keith, Experimental Tests of a Sequential Equilibrium Reputation Model. Econometrica 56 (January 1988): l-36 DeJong, Douglas V, Forsythe, Robert, and Lundholm, Russell J., Ripoffs, Lemons and Reputation Formation in Agency Relationships: A Laboratory Market Study. J. Finance 40 (1985): 809-820. Dholokia, Ruby Roy, and Sternthal, Brian, Highly Credible Sources: Persuasive Facilitators Persuasive Liabi1ities.J. Consumer Res. 3 (March 1977): 223-232. Fitzgerald, Thomas J., Understanding the Differences and Similarities between Services and Products to Exploit your Competitive Advantage. J. .%-vices Marketing 2 (Winter 1988): 25-30. Goldberg, Marvin E., and Hartwick, Jon, The Effects of Advertiser Reputanon and Extremity of Advertising Claim on Advertising Effectiveness. /. Consumer Res. 17 (September 1990): 172-185. Harmon, Robert R., and Corey, Kenneth, The Persuasive Effects of Source Credibility m Buy Lease Situations.]. Marketing l&s. 19 (May 1982): 255-260. Heil. Oliver, Explaining and Predlctmg Competitive Reaction: A Marketing Signaling Approach. doctoral dissertation, University of Pennsylvania, 1988a. Heil. Oliver, A Basic Marketing Signaling Framework to Explain and Predict the Magnitude of Competitive Reactions. working Paper, Indiana University, 198813. Heil, Oliver, and Robertson, Thomas S., Toward a Theory of Competitive Market Signaling: A Research Agenda. Strategic Management]. 12 (1991): 403-418. Herbig, Paul, and Milewicz, John, The Relationship of Reputation and Credibility to Brand Success. J. Consumer Marketing (in press). Krapfel, Robert E., Jr., An Advocacy Behavior Model of Organizational Buyers’ Vendor Choice. J. Marketmg 49 (Fall 1985): 51-59

Levitt, Theodore, Communications 31 (April 1967): 15-21.

and Industrial Selling.]. Marketing

Milgrom, Paul, and Roberts, John, Predation, Reputation and Entry Deterrence. 1. Economic Theory 27 (1982a): 280-312 Milgrom Paul, and Roberts, John, Limit Pricing and Entry Under Incomplete Information: An Equilibrium Analysis. Econometrica 50 (198213): 443-459. Milewicz, John, and Herbig, Paul, The Value of a Firm’s Reputation on the Success of its Sales Force. Marketing Intelligence and Planning (in press). Moorthy, K. Sridhar, Using Game Theory to Model Competition.J. Marketmg Res. 22 (August 1985): 262-282. Rogerson, William P., Reputation and Product Quality The BellJ. Economics 14 (Autumn 1983): 508-516 Rogerson, Wllham P., Repeated Moral Hazard. Econometrica 53 (1) (January 1985): 25-37. Scherer, F., Industrial Market Structure and Economic Petjormance, Rand McNally College Publishing Company, Chicago, IL. 1980, p 338. Selden, R., The Chain Store Paradox. Theory and Decision 9 (1978): 127-159. Shapiro, Carl, Premiums for High Quality Products as Returns to Reputations. Quarterly j. Economics 98 (November 1983): 659-679. Shapiro, Carl, Consumer Information, Product Quality and Seller Reputation. BellJ. Economics 13 (1982): 20-35. Sobel, Joel, A Theory of Credibility 557-573.

Roj. Economic Stud. 52 (1985):

Spence, A. Michael, Market Signaling: The Informational Structure ofJob Markets and Related Phenomena, Harvard University Press, Cambridge MA. 1974. Sternthal, Brian, Dholakia, Ruby, and Leavitt, Clark, The Persuasive Effect of Source Credibility: Tests of Cognitive Response.J. Consumer Res. 4 (March 1978): 252-260. Wernerfelt, Birger, Umbrella Branding as a Signal of New Product Quality: An Example of Signaling by Posting a Bond. RandJ. Economics 19 (3) (Autumn 1988): 458-467.