The Effect of Influence Type and Performance Outcomes on Attitude Toward the Influencer

8/7/2019 The Effect of Influence Type and Performance Outcomes on Attitude Toward the Influencer

1/16

LISA K. SCHEER and LOUIS W . S TE RN *

The authors demonstrate that a target's attitude toward an influencer is affectedby both (1) the influence type used by the influencer to achieve the target's com-pliance and (2) the performance outcomes that result from the behavior adoptedby the target in compliance with that influence. Before performance outcomes areknown, the target's satisfaction and trust are strongly affected by the type of in-fluence exercised; more dominating influence types result in less positive attitude.When outcomes of compliance become evident, however, favorable outcomes ap-pear to ameliorate negative attitudes, whereas unfavorable outcomes seem to un-dermine positive attitudes. These findings indicate the significance of performance

outcomes for understanding the ramifications of successful influence.

The Effect of Influence Type and PerformanceOutcomes on Attitude Toward the Influencer

In commercial exchange, when one party attempts toinfluence another to take speciflc actions (e.g., purchasea product, erect a display, increase space in a catalog),a dynamic ensues that can change the course and contentof their relationship. First, the target's attitude towardthe influencer is affected by the way in which the influ-ence attempt is presented. Second, the target decideswhether to comply and enact the requested behavior. Ifcompliance is selected, then, third, perfonnance out-comes result from the action undertaken and, fourth, thetarget's prevailing attitude is reinforced or altered, de-pending on the nature of those outcomes.Though an infinite number of extensions could be addedto this simple "model" of the influence process, the maineffects depiction identifies some extremely important is-sues that can best be resolved through empirical researchusing experimental designs. Marketing channel researchhas relied primarily on cross-sectional surveys to inves-tigate the issue that has been a central theme for channelmanagement over the past decadethe attitudinal ef-fects of influence attempts (e.g., Frazier, Gill, and Kale1989; Frazier and Rody 1991; Frazier and Summe rs 1986;

*Lisa K. Scheer is Assistant Professor of Marketing, College ofBusiness and Publ ic Administ rat ion, Universi ty of Missour i -Colum-bia. Louis W. Stern is the John D. Gray Distinguished Professor ofMarket ing, J. L. Kel logg Graduate School of Management , Nor th-western University.

Gaski 1986; Gaski and Nevin 1985; John 1984; Kale 19Lusch and Brown 1982). Though some laboratory (eAnand and Stem 1985; Dwyer and Walker 19McAlister , Bazerman, and Fader 1986; Stem , Stem thand Craig 1973) and field (Anand 1987; Keith, Jacksand Crosby 1990) experiments have been conducted, oKeith, Jackson, and Crosby (1990) have directly dressed attitudinal effects of channel influence. Moover, to the best of our knowledge, the ramificationsthe performance outcomes of compliance have never bconfronted by marketing scholars. In fact, with the ception of the work by Shaw and Condelli (1986), issue has not been addressed in an y discipline. Its glect is very surprising, because improved outcomes the ultimate objective of influence attempts.

We addressed both issues by using an experimendesign in a laboratory setting. To examine the flrst, borrowed from the marketing channels literature in veloping an influence framework that is generalizaacross all power resources. The central thought is tthe allocation of every resource can be contingent compliance or noncontingent to encourage complianeach resource also can be used positively or negativein a rewarding or punishing manner. The way in whinfluence attempts are constructed will directly affect target's attitude toward the influencer.To address the second key issue, we manipulated performance outcomes derived as a result of compliawith an influence attempt and gauged the effect on tarattitude. Most particularly, we wanted to flnd out whet128

Journal of Marketing ResearchVol. XXIX (February 1992), 12


2/16

FECT OF INFLUENCE TYPE AND PERFORMANCE OUT COM ES 129

e. Hence, the primary objective of our research is to

type of influence exercised and thegenerated by the behavior adopted

THEORYIn marketing channel relationships, the study of one(T ) has beenpotential power has focused on "bases of power." ST when T perceives that S has expertise,T's be-

T (e.g., French and Raven 1959; Gaski 1986;Nevin 1985; Wilkinson 1979). Research onor attempted influence typicallyeither the assistances-punishments frameworke.g ., Gaski 1986; Gaski and Nevin 1985; Hunt and Nevin1974; Lusch 1976; Lusch and Brown 1982) or the influ-razier, Gill, and Kale 1989; Frazier and Rody 1991;razier and Sum mers 1984, 1986; Kale 1986). Other re-searchers have focused on successful influence, exam-ining the extent to which 7"s compliance was motivatedby threatened punishments, promised rewards, 5's ex-pertise, information provided by 5, T's obligations, orr ' s attraction to 5 (e.g ., Busch 1980; John 1 984).

In our study, we focused on the mechanisms throughwhich 5 may attempt to influence T and the effects ofsuccessful influence on their relationship. We present acontingency-valence conceptualization of influence ex-ercise that is consistent with, and encompasses, the pre-ceding research streams (Stem and Scheer 1992). Unlikethe power base framework that tends to confound powerresources with the manner in which they are used, ourcategorization explicitly separates the two. Unlike theassistances-punishments framework that centers on thevalence of influence, our categorization acknowledgesthat whether power resources are used contingently ornoncontingently can also affect the relationship. Our re-search runs parallel to that on influence strategies in thata positive (negative) resource exercise need not alwaysbe presented by using a "promise" ("threat") influencestrategy.Power Resources

Power is obtained through the possession and controlof resources that are valued by another party (Patchen1974; Tjosvold, Johnson, and Johnson 1984). Power re-sources, the raw material of influence attempts, includeflnancial resources, expertise, information, services, le-

gitimacy, or status (Dwyer, Schurr, and Oh 1987), aswell as 5's possession of attributes (e.g., attractiveness)or rightful claims on T (e.g., 7"s contractual obligations)that may motivate 7"s compliance. "The same under-lying resources can serve as the foundation for more thanone base of power" (Frazier 1984, p. 71). When at-tempting influence, 5 must select the resources to use aswell as the manner in which those resources are exer-cised. Exercising the same power resource(s) in differentways is expected to have different attitudinal effects.The Influence Attempt

Typically, it is claimed that reward and coercive powerdiffer from other bases of power in that 5 mediates someconsequences for T (e.g., Frazier 1984; John 1984; Ka-sulis and Spekman 1980; Raven and Kruglanski 1970).Hinkin and Schriesheim (1989), however, contend thateach power base is deflned best as the ability to admin-ister tangible (things) or intangible (feelings) conse-quences for a target. They defme legitimate power as"the ability to administer to another feelings of obliga-tion or responsibility" and referent power as "the abilityto administer to another feelings of personal acceptanceor approval" (p. 562). Similarly, we maintain that everyinfluence situation involves some influencer-controUedconsequences mediated by 5 and other outcomes derivedfrom the marketing environment. The crucial distin-guishing features of influence attempts are the valenceof 5's resource exercise and whether 5 exercises thoseresources contingently or noncontingently.A positive influence attempt involves reward, the be-stowal of consequences that T evaluates as desirable, or

relief, the withdrawal of consequences that T evaluatesas aversive; a negative influence attempt involves pun-ishment, the bestowal of aversive consequences, or pen-alty, the withdrawal of desirable consequences (Hinkinand Schriesheim 1989; W oods 19 74).'In a contingent influence attempt, 5 uses promises orthreats to signal explicitly that it mediates positive ornegative consequences that it will bestow or withholdcontingently after T's behavioral response. When 5 at-tempts contingent positive influence, 5 indicates that anavailable reward/relief will be provided only if T com-plies. In a contingent negative influence attempt, 5 linkspunishment/penalty with 7"s noncompliance. T's com-pliance decision may also be affected by its expectationsof potential performance outcomes and environmentalconsequences associated with compliance versus rejec-tion. In a noncontingent influence attempt, 5 again me-diates consequences for the target, but it bestows those

'It is possible for an influence attempt to contain no explicit ref-erence to desirable or aversive consequenc es, but often such explicitlyneutral influence carries an implicit valence. For example, when Smakes a suggestion, T is rewarded with the information that S per-ceives one of r ' s behavior options as preferable. If S exercises le-gitimate authority and demands 7"'s compliance, S bestows an aver-sive consequence on T, the obligation to comply.


3/16

130 JOURNAL OF AAARKETING RESEARCH, FEBRUARY 19consequences unilaterally in the hope that T will sub-sequently adopt the behavior sought by S. 5's resourceexercise takes place prior to 7's compliance (Baldwin1971; Harsanyi 1962). In a noncontingent positive in-fluence attempt, 5 solicits 7's compliance through theunconditional provision of reward/relief such as when Sunilaterally provides 7 with economic or noneconomicrewards (Frazier 1984; Frazier and Sheth 1985) or whenS noncontingently provides advice (suggestion, recom-mendation, warning, normative plea) or valued infor-mation (information exchange) about environmentalconsequences not mediated by S (Angelmar and Stem1978; Frazier 1984; Frazier and Sheth 1985; Tedeschi,Schlenker, and Bonoma 1973). S attempts noncontingentnegative influence when it unilaterally provides eco-nomic or noneconomic punishment/penalty without priorwarning in the hope of modifying 7's behavior (Frazier1984; Frazier and Sheth 1985).

5 controls the explicit content of its influence attempt,but the effect of an influence exercise depends on 7'sperceptions. Though 7 may sometimes perceive contin-gency where none is indicated or intended, an explicitcontingent influence attempt is expected to generatestronger perceived contingency than a noncontingent in-fluence attempt.^ Similarly, the verbal framing may af-fect 7's perception of the influence exercised and thuspotentially alter the effect of an influence exercise. In apositively framed influence attempt, the relative benefitsassociated with compliance are emphasized ("if youcomply"), whereas in a negatively framed attempt, therelative detriments associated with rejection are empha-sized ("if you don't comply").Effects of Influence Type at the Time Influence IsExercised

Channel power base research has consistently dem-onstrated that the use of coercive power results in morenegative target attitudes than the use of expert, referent,or information power (Busch 1980; Gaski 1986; John1984; Keith, Jackson, and Crosby 1990; Shaw and Con-delli 1986; Wilkinson 1979), but that finding can beviewed as evidence of a valence effect, a contingencyeffect, or both. Research using the assistances-punish-ments framework generally has found that the provisionof assistances is related positively to target satisfaction,whereas the threat or use of punishment has the oppositeeffect (Gaski 1986; Gaski and Nevin 1985 ; Hunt andNevin 1 974; Lusch 1976). Those studies demonstratevalence effects, as they examine rewards and punish-ments but not the manner in which those sanctions areprovided. All else being equal, positive influence at-tempts are expected to result in more positive target

^Personal or legal pleas may be used to imply contingency withoutexplicitly contingent promises or threats (Frazier and Sheth 1985), butsuch intentional, implicit contingency is beyond the scope of our re-search.

attitude toward the influencer than negative influeattempts.However, rewards may result in negative effects silar to those associated with threats and contingent pishments (Balsam and Bondy 1983). Though some edence indicates that reward power is similar to "noncoercive" pow er bases (Busch 1 980; Gaski 19Gaski and Nevin 1985; Wilkinson 197 9), other studdemonstrate similarity between reward and coercive pww(John 1984; Keith , Jacks on, and Crosby 199 0; Shaw aCondelli 1986). Both promises and threats have befound to have a negative effect on 7's attitude (Broand Frazier 1978; Frazier, Gill, and Kale 1989; Frazand Rody 19 91; Frazier and Summers 1986). Circustances in which positive influence has been found have negative attitudinal effects often involve the ctingent exercise of positive influence. When a contingreward is provided, 7's intrinsic motivation is likelybe undermined by the extemal explanation for the havior (Bem 1967; Staw 1976), and reactance may oc(Brehm 1966). Noncontingent rewards are less likely thcontingent rewards to undermine intrinsic motivatiinterest, and positive attitudes (Deci and Ryan 191987). If we assume that positive influence positivaffects 7's attitude, negative attitudinal effects of ctingent rewEirds must be caused by the contingency the influence. Hence, contingent influence is expecto result in more negative target attitude toward the fluencer than noncontingent influence.

Any contingent influence exercise will employ a pitive frame ("if you comply"), a negative frame ("if ydon' t comply"), or both. Typically, the framing and lence of the resource exercise match, but what happewhen framing and valence do not match? Kahneman aTversky (1979), discussing decision making under certainty, note that framing can alter the reference poithe baseline from which potential gains and losses assessed. A positive, compliance-based frame is likto foster 7's perception of a positive sanction, that the relative gain associated with compliance. A net ive, rejection-based frame, in contrast, is likely to pmote 7's perception of a negative sanction, that is, relative loss associated with noncompliance. All else beiequal, positive framing is expected to result in a mpositive attitude toward the influencer than negatframing.The influence type exercised is also expected to aff7's autonomy and motivation for compliance. Autoomy is the freedom to be one's own boss and to maindependent decisions (Schul, Little, and Pride 1985,16); it "connotes an inner endorsement of one's actiothe sense that they emanate from oneself and are onown" (Deci and Ryan 1987, p. 1025). Motivation compliance involves 7's beliefs about its reasons for egaging in the behavior sought by S (e.g., Busch 19John 1984). Though many motivations can be identifi(e.g., information, legal obligations, performance f)ectations, etc.), we focus on partner-contingent mo


4/16

EFFECT OF INF LUEN CE TYPE A N D PERFORAAANCE OU TCO ME S 131vation for compliance, that is, 7's admission that its be-havior was motivated by contingencies controlled by S.When 7's autonomy is low, partner-contingent motiva-tion for com pliance is likely to be strong. Howe ver, evenwhen 7 has strong autonomy, its compliance may stillbe motivated by influencer-controlled contingencies.

Kale (1986, p. 392) speculates that "low pressure" in-fluence promotes 7's belief that it is acting autono-mously. Noncontingent influence enhances 7's auton-omy by making a wide range of reactions feasible. Positivesanctions may be interpreted by 7 in alternative waysthat will be perceived to infringe on 7's autonomy invarying degrees. Alternate interpretations of a reward mayinclude, for example, a bribe or incentive to make anunpalatable choice more acceptable, a well-deservedpayment for past activities, or a bonus added to an al-ready preferable option. Contingent influence and neg-ative sanctions are more likely to be perceived as pres-suring 7 to adopt the advocated behavior, therebyinfringing on 7's autonomy. Similarly, more contingentand more negative influence will stimulate greater part-ner-contingent motivation for compliance because thesource of the compliance behavior is more clearly theinfluencer. More intrinsic motivations are discounted whenan extrinsic source is present to explain behavior. Hence,more contingent and more negative influence are hy-pothesized to result in lower autonomy and greater part-ner-contingent motivation for compliance. As negativeframing is expected to promote perceptions similar tothose due to negative influence, negatively framed in-fluence is expected to have similar effects.

H,: When the perfonnance outcomes of compliance areunknown, the target experiences less positive atti-tude toward the influencer when(a) more contingent influence is exercised,(b) more negative influence is exercised, and(c) contingent influence is framed negatively ratherthan positively.H2: The target experiences lower autonomy when(a) more contingent influence is exercised,(b) more negative influence is exercised, and(c) contingent influence is framed negatively ratherthan positively.H3: The target experiences stronger partner-contingentmotivation for compliance when(a) more contingent influence is exercised,(b) more negative influence is exercised, and(c) contingent influence is framed negatively rather

than positively.Effects When the Performance Outcom es ResultingFrom Compliance Are Known

The conclusion drawn from previous reseiirch, that morecontingent and more negative influence result in morenegative attitudes, seems very logical. However, a chan-nel member's attitudes are affected also by its perfor-mance outcomes (Frazier 1983). Successful influence re-sults in the modification of 7's behavior, whichsubsequently may result in two types of outcomes for

the targetconsequences bestowed by 5 and 7 's per-formance outcomes. 7 's performance outcomes are ex-temal consequences that result from the behavior adoptedby 7 in compliance with the influence. For example, if7 is influenced to adopt a new product, perfonnance out-comes would include the sales and profits derived fromthat product. Various channel studies have examined theantecedents of channel performance (e.g., Gaski andNevin 1985; Lusch 1976), but none have examined theeffects of the performance outcomes 7 receives as a re-sult of its compliance. Given the retrospective reportsused in field research, 7's attitudinal reactions could beexplained by the nature of the infiuence, 7 's perfor-mance outcomes, or both. Therefore the effects of theinfiuence type exercised must be disentangled from theeffects of the performance outcomes resulting from com-pliance. At the time influence is exercised, more con-tingent and more negative influence are expected to re-sult in more negative attitude toward the infiuencer, butwhen performance outcomes resulting from 7's compli-ance are realized, that initial attitude is expected to bemodified by those outcomes.

Outcomes have been found to affect attitudes throughattributions (Russell and McA uley 1986; W einer, R us-sell, and Lerman 1979). In their research on infiuence,Shaw and Condelli (1986) found that targets are morelikely to have internal attributions for favorable than forunfavorable outcomes that result from compliance. Manystudies (e .g. , Bettman and Weitz 1983; Schoene man etal. 1986; Weiner 1974) provide evidence for the self-serving bias in attribution, the tendency for one to attri-bute success to internal causes and failure to extemalcauses (Bradley 1978). More favorable relationship out-comes therefore are expected to result in weaker partnerattributions. The self-serving bias may not always occur,however (Anand and Stem 1985). Though 7 may be pre-disposed to take responsibility for favorable performanceoutcomes and to find excuses for unfavorable outcomes,the relationship will be strongest when a firm receivesfavorable performance outcomes an d attributes thoseoutcomes to its partner.The contingent use of coercion and reward has beenfound to result in stronger infiuencer attributions than thenoncontingent exercise of referent, expert, and infor-mation infiuence (Litman-Adizes, Fontaine, and Raven1978; Shaw and Condelli 1986). When noncontingentinfiuence is exercised, 7 is expected to be more inclined

to take credit for favorable outcomes and less likely toblame 5 for unfavorable outcomes. When more contin-gent infiuence is exercised, however, 7 is expected tobe more aware of 5's role and thus more likely to give5 credit or blame. Hence, infiuence type and outcomesare posited to have an interactive effect on 7's attitudetoward the infiuencer.Outcomes also affect attitudes directly; favorable out-comes are generally associated with positive attitudes,whereas unfavorable outcomes are associated with neg-ative attitudes (Russell and McAuley 1986; Weiner,


5/16

132 JOURNAL OF AAARKETING RESEARCH, FEBRUARY 19Russell, and Lerman 1979). Shaw and Condelli (1986)found that more favorable outcomes of compliance in-crease 7"'s attraction to S, regardless of the influence typeinvolved. Therefore, in spite of whatever influence typeeffects or influence by outcome interactions are found,more positive target attitude is expected when favorablerather than unfavorable performance outcomes are ob-tained.

H4: The target has stronger partner attributions when(a) more unfavorable perfonnance outcomes are re-ceived and(b) more contingent influence is exercised.H5: When the perfonnance outcomes of compliance areknown, the type of influence exercised and the per-formance outcomes received have an interactive ef-fect on the target's attitude toward the influencer.Specifically, when favorable (unfavorable) perfor-mance outcomes result from compliance, the targetexperiences more (less) positive attitude toward theinfluencer when more contingent influence is exer-cised.H^: The target experiences more positive attitude towardthe influencer when more favorable performanceoutcomes are received as a result of compliance.METHOD

In ongoing relationships, attitudes have been shapedby prior influence episodes, performance outcomes re-ceived, and many odier factors. Given our objective ofdisentangling influence type and performance outcomes,internal and construct validity are of paramount concern.We conducted laboratory research because it enabled usto manipulate influence type and performance outcomesand isolate their effects for exam ination. Our experimentwas designed to separate the type of influence exercisedfrom the performance outcomes received, test the re-search hypotheses, and further develop a theoretical modelfor later testing in the field.Subjects

A total of 233 MBA students enrolled at a midwestemUnited States university completed the experiment. A pilottest with 32 executives enrolled in executive educationprograms was used to develop materials. In a separateprestudy with 74 MBA students and 26 executives assubjects, no significant differences were found betweenMBA students and executives in the key dependent vari-ables of interest, which suggests that MBAs at this uni-versity, most of whom have a minimum of several years'work experience, are able to approximate executives' re-sponses to the issues examined in the experiment.Procedure

The subjects were assigned randomly to the 12 treat-ment conditions in a 4 (influence type) x 3 (p erfonnan ceoutcomes) factorial design. The first page of the exper-iment booklet described the subject's role as the mar-keting manager for a company facing a strategic deci-sion. Subjects were instructed to relate how they would

actually react if they experienced a similar situatioBackground information was provided:You are the Marketing Manager for DiagnostixDistribution Company, a distributor of medical di-agnostic equipment. Diagnostix distributes a varietyof products for 15 manufacturers. It is your respon-

sibility to select the products that Diagnostix sells.Two firms, Image Manufacturing and MDS Inc..have independently develop)ed competing versionsof a new, innovative diagnostic m achine. Image hasformally offered distribution rights for its new prod-uct, B IOSCAN, to D iagnostix. M DS sells its equip-ment via its own salesforce and does not use dis-tributors. It will take at least 12 months for anothercompany to be ready to go to market with a com-petitive product.It was also noted that approximately 20 independent dtributors sell Image equipment in North America. Dianostix has been selling Image products for 10 years, athose products currently generate 25 % of Dia gno stigross profit. The contingent penalty condition includthe additional sentence: "The last 2 years Diagnostix hbeen honored as one of only 4 'Premium Image Distriutor s ' . " The background information then reminded tsubject, "As Marketing Manager, you must deciwhether or not Diagnostix should carry BIOSCAN."

Next was an internal product evaluation from Dianostix's purchasing department. BIOSCAN scored highon some dimensions and the competing product scorhigher on others, but overall the products had roughequivalent, fairly high ratings. It was also reported: "Dto the innovative and unique nature of this equipme5-year sales and profitability projections are uncertaiSome experts think that this equipment will revolutioize the industry, but other experts are skeptical."A letter from the distribution manager of Image folowed. It began:Diagnostix has been a valued partner in the dis-tribution of Image equipment. We at Image wantour distributors to carry our new diagnostic machineBIOSCAN, a technological breakthrough with tre-mendous sales potential. Concurrent with the intro-duction of BIOSCAN, we have revised our distri-bution program. This new program includes thefollowing key elements: . . . .

The body of the letter then presented the appropriate ifiuence induction.Influence Attempts

Noncontingent reward, contingent reward (positiveframed), negatively framed contingent reward, and cotingent penalty were examined. '

'Be cau se rewards are considered more justifiable and socially ceptable than punishments, we focused on reward and penal ty. Ncontingent penalty was omitted to conserve experimental resources the other three influence types.


6/16

EFFECT OF INFLUENCE TYPE AND PERFORMANCE OUT COM ES 133

The noncontingent reward influence attempt stated thatDiagnostixhad been selected as a Premium Image Distributor, would be one of the four Premium Distributors speciallyfeatured in Image advertising campaigns, andwould receive this marketing support wbetber or not

Diagnostix distributed BIOSCAN.The contingent reward influence attempt stated that Diag-nostixwould be selected as a Premium Image Distributor if itagreed to sell BIOSCAN,w ould be one of the four Premium Distributors speciallyfeatured in Image advertising campaigns if it carriedBIOSCAN, andwould receive that marketing support only if it distrib-uted BIOSCAN,The contingent reward (negatively framed) influence at-tempt stated that Diagnostixwould not be selected as a Premium Image Distributor

if it refused to sell BIOSCAN,would not be one of the four Premium Distributors spe-cially featured in Image advertising campaigns if it didnot carry BIOSCAN, andwould not receive that marketing support if it did notdistribute BIOSCAN.The contingent penalty influence attempt stated that all Pre-mium Distributors must carry BIOSCAN and that Diagnos-tixwould not be retained as a Premium Image Distributorif it refused to sell BIOSCAN,w ould no long er be one of the four Premium Distributorsspecially featured in Image advertising campaigns if itdid not carry BIOSCAN, andwould lose Image's marketing support if it did not dis-tribute BIOSCAN.The same set of power resources was involved in eachinfluence attempt; only the nature of the influence ex-ercise and the conditions under which Diagnostix wouldreceive the resources differed. Each letter concluded:

Because our competitor is ready to enter the mar-ket, we must finalize our BIOSCAN distributionplans. We at Image hope that you will become partof our BIOSCAN network. Does Diagnostix wantto distribute BIOSCAN?Given the mental role-playing in the experiment, one

can question whether the subjects were sufficiently in-volved to internalize the manipulations and provide validresponses to the measures. The approach we adopted wasmodeled to a large extent after that used in psychologyexperiments studying attributions or influence, in whichsubjects are asked to offer their own perspectives abouta hypothetical situation (e.g., Litman-Adizes, Fontaine,and Raven 1978; McG raw 1987; Russell and McAuley1986; Schm idt and Weiner 19 88; We iner, Russell, andLerman 1979) or to speculate about a hypothetical thirdparty's reactions to a given scenario (e.g., Forgas, Bower,

and Moylen 1990; Hamilton et al. 1990; Shaw and Con-delli 1986). Subjects in our experiment, however, werenot simply given information about a situation in whichsuccessful influence was a/a/f accompli. Instead, theywere given a role and the responsibility to decide whetheror not to comply. Subjects were required to commitDiagnostix to distributing the product by indicating "YES,Diagnostix will dis tr ibute Image's BIOSCAN" or "NO,Diagnostix will not dis tr ibute Image's BIOSCAN." Ap-proximately 89% of the subjects indicated Diagnostixshould distribute BIO SCA N. After indicating their de-cisions, subjects responded to perceived contingency,autonomy, and expectation items and then received theperformance outcome manipulation.Performance Outcomes

Favorable and unfavorable outcomes were examined,as was a condition in which outcomes were not speci-fied. Subjects in the favorable and unfavorable outcomeconditions were informed that 5 years had passed sinceDiagnostix had chosen to distribute BIOSCAN. Perfor-mance outcomes resulting from Diagnostix's distributionof BIOSCAN were described in relation to benchmarksfor comparison. The favorable (unfavorable) outcomeinformation stated:1. Research indicates that customer satisfaction with BIOS-CAN is grea ter (lower) than with other products carriedby Diagnostix.2. BIOSCAN's introductory sales performance is better(worse) than that achieved by most new products. Orig-inally BIOSCAN sales were projected to increase 20%per year for the next 3 years. Now , however, BIOSCANsales are expected to increase 35% (5%) per year.3. This year Diagnostix's BIOSCAN distribution is fore-

casted to result in a net profit that is approximately dou-ble (half) the projected profit.Subjects randomly assigned to the unspecified outcomecondition received no information about BIOSCAN per-formance outcomes; after completing the first battery ofmeasures, they were simply instructed to proceed to thesecond battery. The experiment concluded with attitude,attribution, motivation, and literal manipulation checki tems. All measures except attributions were assessed on7-point agree-disagree Likert scales.Measures

Thoug h r ' s gene ral attitude toward the influencer maybe of interest in itself, marketing researchers usually haveexamined specific attitudinal variables. We examinedsatisfaction, trust, and commitment. Satisfaction has beendefined as the overall approval of and positive affect to-

''The proportion of subjects deciding in favor of BIOSCAN was;noncontingent reward 53/59, contingent reward 55/58, negativelyframed contingent reward 47/57, and contingent penalty 53/59. Onlycontingent reward and negatively framed contingent reward resultedin different compliance rates ( = 2. 11 , p = .019, one-tailed).


7/16

134 JOURNAL OF AAARKETING RESEARCH, FEBRUARY 19ward another party (Anderson and Narus 1984; Gaskiand Nevin 1985; Rusbu lt, Johnson, and Morrow 1986;Schul, Little, and Pride 1985). Trust is the belief thatone's partner can be relied on to fulfill its future obli-gations and to behave in a manner that will serve thefirm's needs and long-term interests (Anderson and Na-rus 1990; Dw yer, Sch urr, and Oh 1987; Larzelere andHuston 1980). Commitment is a party's intention to con-tinue a relationship (Lund 1985; Mic haels, Aco ck, andEdw ards 1986) and thus represents an implicit or explicitpledge of relational continuity between exchange part-ners (Dwyer, Schurr, and Oh 1987).

Autonomy involves a perception of independence,freedom of action, and lack of outside interference (Schul,Little, and Pride 1985). Autonomy reflects the subject'sbelief that its decision about BIOSCAN was made vol-untarily, freely, and without pressure. Partner-contin-gent motivation for compliance is the extent to wh ich thesubject's compliance behavior was motivated by partner-controlled contingencies. Items similar to John's (1984)motivation for compliance operationalization of rewardand coercive influence were used. Partner attributionsreflect the extent to which the firm perceives its partneras responsible for performance outcomes it receives. At-tributions concern causes of outcomes; motivations forcompliance concem reasons for behavior. A firm in amarketing relationship may attribute causality to itself,its partner, or causes extemal to the relationship (Holtz-worth-Munroe and Jacobson 1985). Because our prelim-inary studies had indicated that subjects attributed out-comes to both intemal and partner loci if given theopportunity, subjects in the main experiment were notgiven that option. Attributions were assessed by two itemsrequiring a forced choice among Diagnostix, Image, orneither (e.g., Hamilton et al. 1990) and two items on 7-point scales anchored by Diagnostix and Image with amidpoint of "neither."

Influence type can affect attitudes toward the influen-cer and about the behavior undertaken. Attitude aboutBIOSCAN measured the subject's evaluation of BIOS-CAN and willingness to make the same decision again.No systematic differences in expectations were antici-pated, but expectations about BIOSCAN were measuredto detennine whether there were systematic differencesacross treatment conditions. Anderson and Narus (1984,1990) demonstrated that outcomes given comparison levelcan affect a channel member's cooperation and satisfac-tion. Because of our emphasis on perfonnance out-comes, different expectations generated by influence typecould mask treatment effects or provide rival explana-tions for some of the hypothesized effects..Manipulation checks assessed the subject's percep-tions and literal recognition of the manipulations. Anyinfluence attempt will be either explicitly contingent ornoncontingent, but the target's perceived contingency alsocan be assessed. Perceptions of positive and negativecontingencies were uncontaminated by the performanceoutcome manipulation as they were obtained in the first

battery. Literal contingency and performance outcomanipulation checks were measured at the end of experiment.RESULTS

Data analysis was conducted in three stages. First, dependent variable manipulations were evaluated and construct validity of dependent variable measuremscales was assessed by principal components factor anysis with varimax rotation and standard reliability checSecond, hypotheses were tested by analysis of varianand correlation analysis. Third, a posteriori cell mecomparisons were conducted by the Tukey-hsd produre at the .05 significance level to investigate hypoesis rejection.Manipulation Checks

Analyses of variance indicated that subjects in noncontingent and contingent conditions were signcantly different on perceived contingency (E,204 = 69p < .001) and the literal contingency manipulation che(^1,192 = 346 .61 , p < .001). On the outcome maniplation check, subjects receiving favorable outcomes hhigher values than those receiving unfavorable outcom(^1.135 = 989.08, p < .001); subjects receiving no ocome information had lower values than those in the vorable outcome condition (F, ,25 = 274.43, p < .00and higher values than those in the unfavorable outcocondition (F, ,24 = 38 1.3 8, p < .001 ).Scale Construction

In a factor analysis of the items measured in the fibattery, autonomy and expectation factors emerged (Tble 1). Factor analysis of the attitude, attribution, amotivation items resulted in one factor composed of bosatisfaction and trust items and four factors apparenrepresenting attitude about BIOSCAN, partner-contgent motivation, partner attr ibutions, and commitm(Table 2). Given the strong positive correlation expecbetween satisfaction and trust, it is not surprising to fithose items loading on a common factor. There is a cldistinction between satisfaction and trust, however, satisfaction assesses affect toward the influencer whertrust examines beliefs about the influencer and its tions. Consequently, separate satisfaction and trust vaables were retained. Subsequent individual factor anyses confinned the unidimensionality of all scales wthe exception of that for the partner-contingent motivtion items, which generated two factors. Separate scawere formed, representing negative and positive movations. Items were averaged to create the variables usin hypothesis testing. For the purpose of our resear

'One could argue that oblique rather than orthogonal rotation is mappropriate. In this case, the two rotations provide similar pattemsfactor loadings.


8/16

EFFECT OF INFLUEN CE TYPE A N D PERFORAAANCE OU TC OM ES 135Table 1

PRINCIPAL COMPONENTS FACTOR AAATRIX FOR AUTONOMY AND EXPECTATIONS MEASURES'

AutonomyI was no t pressured to recommend that Diagnostix distribute BIOSCANMy decision was primarily motivated by the potential and quality of the product rather than Image's actionsMy decision about BIOSCAN was made voluntar i lyMy decision regarding BIOSCAN was no t made freelyExpectationsBIOSCAN is a product that most distributors would be eager to sellThere is a good chance that BIOSCAN wil l no t generate acceptable salesCarrying BIOSCAN would be a reasonable st rategy under almost any ci rcumstancesDistributors that carry BIOSCAN probably will no t make a profitDisregarding Image's new dist r ibut ion program, carrying BIOSCAN would be a good decisionEigenvaluesPercentag e of variance 33.2 21.1"Factor loadings greater than .5 are underlined.

.751

.641.850

.853

.290

.052

.038-.021.056

2.99

-.171.318.114.114

.646

.623

.645

.678

.7331.90

Table 2PRINCIPAL COMPONENTS FACTOR AAATRIX FOR ATTITUDE, AHRIBUTION, AND MOTIVATION MEASURES"

SatisfactionDiagnostix is no t satisfied with the actions of ImageDiagnostix has a satisfactory business relationship with ImageDiagnostix is unhappy with ImageDiagnostix likes the way Image conducts its relationship with DiagnostixTrustImage does no t give Diagnostix the respect is deservesImage is fair in its dealings with DiagnostixImage can be trustedImage is sincere in its dealings with DiagnostixCommitmentDiagnostix should no t end its relationship with ImageDiagnostix would be better off if it stopped distributing all Image productsAttitude about BtOSCANDiagnostix is satisfied with BIOSCANDiagnostix would be better off if it did no t sel l BIOSCANAdding BIOSCAN to the product l ine was a good decisionIf I could do it again, I would no t recommend that Diagnost ix distr ibute BIOSCA NThe decision to distribute BIOSCAN was the right choicePartner outcome attributionsImage is most responsible for BIOSCAN's sales and profit performanceBIOSCAN's sales performance is primarily due to Image's ability or skillBIOS CA N's sales performance is pr imar ily determined by Image's ef for tsBIOSCAN's sales perfonnance is caused by factors that are predominant ly underImage's controlPartner-con tingent m otivation for complianceDiagn ost ix 's decision to dist ribute BIOSCA N was mot ivated by the desi re to be .307 - . 1 7 9 .531 .054 - . 08 1more closely associated with ImageDiagnostix decided to carry BIOSCAN in order to keep Image satisfied*Diagnostix decided to carry BIOSCAN in order to receive benefits that ImageofferedDiagnostix agreed to carry BIOSCAN to avoid penalties that Image could imposeImage's ability to punish Diagnostix motivated Diagnostix to carry BIOSCANDiagnostix's relationship with Image would have been damaged if Diagnostix re-fused to distribute BIOS CANEigenvaluePercentage of variance'Factor loadings greater than .5 are underlined.Items loading on a separate factor in a subsequent factor analysis of motivation items. Reliability = .57.

66874972081975879075473322718232015928008819012304307706 0

.452

.286

.448

.142

.042

.197

.137

.131

.251

.181

.772

.815

.819

.730

.743-.003-.072-.069-.025

-.039-.160-.081-.208

-.235-.076-.040-.018

-.033-.012

.023

.021-.041-.084-.036-.011-.069.050.012

-.003.036

-.016-.029

.038-.001.048.029

.035

.143

-.092-.135-.062.059

-.004

.768

.811

.788

.811

-.168.168.046.042

.051

.058

.257

.302

.789

.739

-.157.102.146.311.368.164.060

-.062.023

-.274.111

-.405-.409-.335

7.7631.1

-.055-.002.024

-.031.102

3.1112.4

.791

.634

.748

.753

.662

2.6810.7

.065-.102.086

-.032-.040

1.907.6

-.073.070

-.129-.133.216

1.255.0


9/16

136 JOURNAL OF AAARKETING RESEARCH, FEBRUARY 19.60 was considered the minimal acceptable reliability.*With the exception of the positive partner-contingentmotivation measure, all scales were unidimensional andreliable. Only the negative partner-contingent motivationscale was used in hypotheses testing.Hypothesis Testing

Our theory pertains to successful influence, so testingwas conducted with the 208 subjects who complied. Be-cause cell sizes were unequal, the model comparisonprocedure for nonorthogonal analysis of variance wasemployed (Appelbaum and Cramer 1974). Cell means

'Nunnally (1967) initially suggested that minimum reliability of .50to .60 is sufficient for early stages of resea rch, but he later (1978)recommended that a minimum of .70 be sought.

of dependent variables are given in Table 3. Correlatiand reliability coefficients are provided in Table 4.Main effect tests indicate satisfaction was affected both in fluen ce typ e (/ 3.i96 = 16 .12 , p < .001) and ocome type (F2,i96 = 41.37, p < .001). Influence ty(^3.196 = 14.23, p < .001) and outcome type (^2,1969 .28 , p < .001) also affected trust. No effects on comitment are found. Attitude about BIOSCAN was fected strongly by outc om e type (^2,191 = 158 .64 , p .001) , but an influence type by outcome type interactiof small magnitude also was present (Fg 19, = 2 .35 , p .05). Influence type effects are found for autonomy (F3= 13 .12 ,p < .001) and negative partner-contingent mtivation (^3,94 = 13.92, p < .001). No significant fects on attributions among subjects receiving perfomance outcomes and no effects on exjjectations are founThough main effect and interaction tests provide useinfonnation, planned contrasts directly tested the h

Table 3CELL MEANS

DependentvariabteSample

Satisfaction

Trust

Commitment

Attitude about BIOSCAN

Partner outcomeattributionsExpectationsAutonomyPerceived contingencyNegative partner-contingent motivation

OutcomeconditionFavorableUnspecifiedUnfavorableAll outcomesFavorableUnspecifiedUnfavorableAll outcomesFavorableUnspecifiedUnfavorableAll outcomesFavorableUnspecifiedUnfavorableAll outcomesFavorableUnspecifiedUnfavorableAll outcomesFavorableUnfavorableAll outcomesAll outcomesAll outcomesAll outcomes

Noncontingentreward191519535.84"5.53"4.07'5.12"5.46"5.40"4.50"5.10"6.45"6.47"6.42"6.44"6.42"5.93""3.79"5.34"4.25"4.85"4.82"4.88"4.22"

3.64"

Contingentreward191917555.22""5.09"3.60'4.68"4.71"^5.05"'4.15""4.65"6.39"6.29"5.97"6.23"6.34"5.76"3.54'5.29"4.21"4.75"4.71"4.49"5.30"4.30"

Contingentreward(negativeframe)181316474.89"3.56'3.22'3.95'4.47"3.54"3.98"'"4.05'6.58"6.08"6.41"6.38"6.36"5.47"4.46"5.47"4.99"5.08"4.90"3.70"5.37"'4.54"

Contingentpenalty181619534.78''3.52'2.96'3.75'4.46"3.58"3.16'3.73'6.44"6.00'6.00"6.15"6.40"5.31"4.01'"5.22"4.45'5.03'4.73"3.41"5.80'5.35'

At tinfluen746371208

5.19"4.48"3.474.394.78"4.45"3.94"4.396.47"6.21"6.20"6.306.38"5.63"3.945.324.47"4.93"4.794.135.174.44

NOTE: For each dependent variable, two cell means in the same row or column, two influence means in the same row, and two outcommeans in the same column are significantly different (p < .05) only when they do not share a common superscript.


10/16

EFFECT OF INFLU ENC E TYPE A N D PERFORAAANCE OUT CO ME S 137

Table 4CORR ELATIONS A N D RELIABILITY VALUES"

SatisfactionTrustCommitmentAttitude about BIOSCANPartner outcome attributionsAutonomyNegative partner-contingentmotivationPerceived contingencyExpectations

SATIS.8 9.86*.32*.58*- . 0 5.49*- . 4 7 *

- . 2 8 *.26*

TRUST.86*.8 6.34*.38*.0 1.50*- . 4 6 *

- . 3 0 *.31*

COMMIT.2 9.32**.7 0.35*.1 4.0 5- . 0 7.0 1.26*

BIOATT.51*.41*.60*.8 8- . 1 4.1 5- . 1 1

- . 0 1.25*

ATTRIB.0 9.1 3

- . 1 0- . 0 6.7 9- . 0 2.0 2.0 5.0 0

AUTON.61*.53*.0 5.36**.0 9.8 0- . 7 8 *

- . 5 5 *.36*

NEGMOT- . 5 7 *- . 4 2 *- . 0 5- . 34**- . 1 4- . 7 6 *.8 9

.64*- . 3 6 *

PERCON- . 5 4 *- . 4 7 *- . 0 4- . 1 8- . 0 5- . 7 4 *.76*

.6 7- . 1 0

EXPECT.42*.36**.38**.74*- . 1 1.52*- . 4 8 *

- . 37**.7 1"Reliability values are reported on the d iagonal. Correlations above the diagonal are for those in the unspecified outcome condition. Correlationsfor the entire sample are given below the diagonal.*p < .001.**p < .01.

potheses. The contingency effect (contingent rewardcompared with noncontingent reward), the valence effect(contingent reward compared with contingent penalty),an d the framing effect (contingent reward compared withnegatively framed contingent reward) were examined.The outcome effect compared the favorable and unfa-vorable outcome conditions.H,. Tests of the hypotheses about attitudinal effectswhen outcomes are unknown were conducted on the re-sponses of subjects in the unspecified outcome condi-tion. Significant valence and framing effects on satis-faction and trust are found, but no contingency effect isfound. When outcomes were unknown, subjects influ-enced by contingent reward reported greater satisfaction(F,.33 = 25.94, p < .001) and greater trust (F,,33 = 21.46,p < .001) than those influenced by contingent penalty.Contingent reward also resulted in greater satisfaction(F , 30 = 14.33, p = .001) and greater trust (F, 30 = 13.86,p = .001) when it was positively framed than when itwas negatively framed. No effects on commitment arefound. Though H|a is rejected, Hib and H,,. are supportedfor satisfaction and trust.

H2. No contingency effect is found, but subjects in-fluenced by contingent reward reported greater auton-omy than those com plying w ith contingent penalty (F , ,05= 16.48 , p < .001) or negatively framed contingent re-ward (F,,99 = 8.89, p < .01). H2a is rejected, but Hjband Hjc are supported./ / j . There was a valence effect, as contingent penaltyresulted in greater negative partner-contingent motiva-tion for complia nce than contingent reward (F, ,00 =17.1 4,/7 < .001 ). There was also a contingency effect,as contingent reward influence generated greater nega-tive partner-contingent motivation than noncontingentreward (Fi,io2 = 1.23, p < .01). H^^ and H3b are sup-ported, but Hjc must be rejected as no framing effect isfound.H4. There were no influence effects on attributions.Unfavorable outcomes resulted in slightly stronger part-

ner attributions than favorable outcomes (t = 1.88, p

Documents

The Effect of Influence Type and Performance Outcomes on Attitude Toward the Influencer