RICHARD L. OLIVER*

As a result of a lack of empirical investigotion, the variance in salesmen's

performance attributable to motivational constructs has not been estimated.

Vroomian expectancy theory v as used to shov/that the motivational perceptions

attributed to a set of sales "incentives" by a sample of life insurance salesmen

were related to two performance criteria.

Expectancy Theory Predictions of Salesmen'sPerformance

INTRODUCTIONThe determinants of salesmen's performance have

long intrigued researchers and practitioners alike.Unfortunately, little information has been amassedin this area. Rather, the bulk of current knowledgehas been accumulated in the domain of sales aptitudewhere the evidence is substantial. Based on resultsreported in numerous studies, Ghiselli found that theaverage validity coefficient between sales proficiencycriteria and tests of intellectual abilities was .33 whilethat for personality (including interest) tests was .30[15, p. 470]. Although these average coefficients arehigh relative to those obtained on other occupationalgroups, the two aptitude measures taken together asindependent predictors would not explain more than20% of the variance in performance. Thus, over 80%of the variance remains to be explained by otherconstructs.

The subject of motivation has also received exten-sive discussion in the sales management literature butlittle substantive knowledge has emerged. This may,in part, reflect the fact that authors usually subsumemotivation under the broad category of compensationor under other financial remuneration schemes some-times classified as "stimulators" (see, for example,[10, 35, 43, 44]). Writings on the subject frequentlycontain an implicit assumption that the basiccompensation package is the primary regulator of

*Richard L. Oliver is Assistant Professor in the Department ofBusiness Administration, University of Kentucky. He wishes tothank Professors Shelby D. Hunt, Donald P. Schwab, and HerbertH. Heneman, III of the University of Wisconsin for their counseland assistance.

motivation. It is further assumed that other incentivesor stimulators (e.g., contests, bonuses, conventions)operate only to induce performance over and abovethat which can be engendered from the basic plan.

While many of the managerial tools used by the salesadministrator provide strong motivating forces, in in-stances they are insufficient and additional incentivesare required. Most sales executives agree that a soundcompensation plan can be the strongest force to moti-vate salesmen [43, p. 430].

Conclusions such as the one just cited are oftensupported by reference to an early study by Haringand Myers [21]. Questionnaires, designed to probethe use of special incentives for salesmen, were mailedto members of the National Sales Executives repre-senting approximately 8,000 companies. Of 542 re-spondents, 396 answered a question pertaining to theeffectiveness of various incentives in stimulating theaverage salesman to better his usual performance. Theresearchers found that the basic compensation planwas cited as the first ranked incentive by 243 respond-ents and concluded that "basic compensation is theprimary motivator of salesmen" [21, p. 158].

This conclusion is unwarranted for three reasons.First, the authors have made inferences about therelative effectiveness of various "motivators" solelyon the basis of superiors' reports. However, it hasbeen noted elsewhere [28, 34] that, when asked torank the importance subordinates attach to variousjob factors, managers typically overemphasize theimportance of pay. The results of the Haring andMyers [21] study may have been similarly biased.Second, the study provided no evidence of validity.While a number of possible incentives were identifiedin the study, Haring and Myers fell short of providing

243

Journal of Marketing ResearchVol. XI (August 1974), 243-53

244 JOURNAL OF MARKETING RESEARCH, AUGUST 1974

empirical support for their assertion. A predictivestudy testing the validity of the various "motivators"would be required to substantiate their conclusion.

Third, the results of one mail survey do not providesufficient evidence for the authors' sweeping conclu-sion, especially in view of the paucity of empiricalresearch on financial compensation. Opsahl and Dun-nette observed that:

(W)e know amazingly little about how money eitherinteracts with other factors or how it acts individuallyto affect job behavior. Although the relative literatureis voluminous, much more has been written about thesubject than is actually known. Speculation, accompa-nied by compensation fads and fashions, abounds . . .[34, p. 94].

Opsahl and Dunnette further noted that compensationas a primary motivator of organizational behavior isonly one of many possible theoretical roles positedin the literature. In short, no evidence exists to suggestthat money is either a primary motivator or is primaryon a hierarchy of motivators.

To a lesser extent, the motivating qualities of non-monetary or psychological incentives have also beenprominently argued in the literature, although theyare usually suggested as adjuncts to the basiccompensation package. Lists of psychological needsrequiring satisfaction have been proffered by manywriters in the sales management field (see, for example,[4, 14, 31]) but have not been subjected to empiricalinvestigation. As a result of the lack of evidence inboth the monetary and nonmonetary areas of motiva-tion, little is known about how compensation and otherpossible rewards motivate salesmen to produce.

Preoccupation with the content of motivationalschemes (e.g., pay, psychological needs) has led writ-ers to ignore the cognitive processes by which behavioris initiated, directed, and continued. That is, theoristshave been content merely to suggest specific thingsthat motivate behavior rather than to delineate theprocesses by which major classes of variables interactto produce behavior. Campbell, Dunnette, Lawler,and Weick summarized this point succinctly:

A motivational theory is useful for making predictionsonly to the extent that it specifies both content andprocess, that is, to the extent that it specifies the identityof the important variables and the processes by whichthey influence behavior [5, pp. 341-2].

As a result of the lack of a comprehensive motiva-tional model, the determinants of salesmen's motiva-tion remain essentially unknown. Without this knowl-edge, one is unable to assess empirically the propor-tion of variance in performance attributable to motiva-tion. The implications for the sales manager aresubstantial. In order to increase production throughan optimal allocation of resources, the relative impactof motivation as opposed to ability and other factorson performance must be known.

In contrast to the paucity of theory and evidenceregarding the performance of sales personnel, consid-erable research has been undertaken in the area ofindustrial psychology and significant progress has beenmade in increasing the state of current knowledgeon employee productivity as a function of motivationat both the theoretical and empirical levels. Muchof industrial motivation theory has direct applicabilityto the motivation of sales personnel.

VROOMIAN EXPECTANCY THEORYExpectancy theory as applied to employee motiva-

tion is generally attributed to Vroom [46], althoughits historical roots are in the works of Lewin [29]and Tolman [45] who postulated that organisms de-velop cognitive expectancies regarding the outcomesof behavior and consequently behave in a mannerwhich is likely to result in preferred outcome states.

Generally, expectancy theory posits that the mo-tivational force experienced by an individual to selectone behavior from a larger set is some function ofthe perceived likelihood that that behavior will resultin the attainment of various outcomes weighted bythe desirability (valence) of these outcomes to theperson [25]. Thus, it is essentially a process theoryin that its focus is on the major classes of motivationalconstructs and the manner in which they interact asopposed to detailing the specific outcomes or needsthat presumably motivate behavior [5]. As such itmay help to integrate previous writings on the contentof sales motivation.

Vroom's [46] theory hypothesizes that employeejob performance (P) is a function of the multiplicativeinteraction between motivation (M) and ability (A).'Thus:

(1) P=f(MxA).

The rationale for the multiplicative relationship is thatif an individual is low on either performance compo-nent, then his performance must be necessarily lowas well.

Motivation, in turn, is hypothesized to be a functionof the multiplicative interaction of the valence of one'sperformance goal / (Vp and the subjective probabilityor expectancy that one's efforts will result in theattainment of that performance goal (Ej). Thus:

(2) M=f(V,. x £ p .

A performance level is seeh as acquiring valenceonly if it is perceived as leading to the attainmentof desired job-related outcomes such as pay or recog-

' In the strict theoretical sense, Vroom, [46] posits adiscrete level of motivation for each effort and performancelevel. However, when effort is construed generically andone performance level is singled out as the criterion ofinterest, Vroom's conceptualization reduces to (1) [33].

EXPECTANCY THEORY PREDICTIONS OF SALESMEN'S PERFORMANCE 245

nition, or to the blocking of undesirable outcomessuch as being terminated. The desirability of a job-related outcome, k, is specific to the individual andconstitutes his valence for that outcome (V^). Valenceis positive if the outcome is desirable, negative ifthe outcome is undesirable, and zero if one is indiffer-ent toward the outcome. One's perception of thedegree to which performance at level / will result inor block the attainment of outcome k is termed theinstrumentality^ of performance level j for outcomek (7j ). Instrumentality is positive if performance re-sults in attainment of outcome fc, negative if it blocksattainment of outcome k, and zero if it has no effecton attainment of outcome k.

The theory posits that the valence of a performancelevel (Vj) is a function of the multiplicative interactionof the valence of the fcth outcome and the instru-mentality that performance level / will result in out-come k, summed over all (n) outcomes. Thus:

(5) V, =

Evidence as to the predictive validity of the theoryin industrial settings is mounting [23, 32], althoughthe performance variance explained, while significant,is low. Generally, performance measures are regressedon the valences and instrumentalities of a prioridetermined outcomes, expectancy, and ability in addi-tive and/or interactive form. Tlie motivation andperformance valence constructs are usually not mea-sured directly because of operational difficulties.While self-reported effort is sometimes used as a proxyfor motivation, this variable can be seen as completelydetermined by the antecedent constructs.

It should be noted here that the theory does notspecify the content of the motivationally relevantoutcome set. It has been shown elsewhere, particularlywith regard to the formation of attitudes and alsoin simulated choice behavior, that only a small subsetof all possible affective and belief cognitions aboutan object are "salient" [6, 19, 37]. This issue takeson added significance given the large number of"motivators" suggested in the sales management liter-ature. The motivating qualities of various compensa-tion plans, incentive type "stimulators" (e.g., salescontests, conventions, production clubs) [20], andhigher order psychological needs [47] have, at onetime or another, been afforded prominent status inthe motivational literature. While at least one writerhas argued that little evidence exists to support thealleged potency of any one set [8], investigations of

^ Similarities between the instrumentality construct as usedhere and as used in attitude models attributed to Rosenberg[37] and Fishbein [12] and also the possible isomorphismbetween expectancy and instrumentality-value theories havebeen elaborated by Jacoby [25].

this nature have not been forthcoming. Thus, a testof the separate effects of each set should, in part,resolve some long-standing issues regarding the effi-cacy of monetary, organizational, and psychologicaloutcomes.

Of special import to the present investigation isthe fact that no previous study has attempted tovalidate expectancy theory on a sample of salesmendespite the great emphasis placed on motivation asa determinant of performance in the sales managementliterature. The purpose of the study reported hereis to provide such evidence.

METHOD

Subjects

Subjects were all full-time life insurance agents ofa medium-sized Midwestern life insurance company.A questionnaire tapping motivational perceptions wassent to 99 full-time agents in late March, 1972. Inall, 95 (96%) of the full-time field responded to therequest. Of the 95 respondents, three were eliminatedbecause of incomplete and unusable questionnairesand three were terminated before the performanceperiod expired.

In addition, nine "agency building" agent-managerswere eliminated from the study because they werenot sufficiently involved in production activities tohave meaningful performance goals. This resulted ina final sample size of 80. The average subject was38 years old, had been with the company 8 years,and had 2 years of education at the college level.

Design

A cross-sectional nonexperimental design was em-ployed. Because of time and resource constraints,performance data were collected over the six-monthperiod of January through June, 1972. The question-naire was administered midway through this time span.While an annual performance period may have beenmore meaningful in terms of reliability, analysis ofthe previous years production figures showed thatproduction for the first six months correlated .936with the annual data.

MeasuresDependent variables. Two criteria were investigated.

The first, six-month production volume, was usedbecause of the pervasive nature of absolute volumeas (he criterion in the insurance industry. The produc-tivity of agents, agencies, companies, and the industryitself is measured in terms of this factor. Productiondata were supplied by the company.

In an effort to overcome the problems introducedwhen comparing the performance of salesmen indifferent territorial environments (for discussion, see[18]), a second relative criterion was used. Asa matterof company policy, agents were encouraged to estab-

246 JOURNAL OF MARKETING RESEARCH, AUGUST 1974

Table 1JOB OUTCOMES

Making the (Lives) clubReceiving more responsibility in my positionGetting improvements in home office practices and proce-

duresHaving better working relations with my supervisorFurthering my professional growth and developmentReceiving more recognition and appreciation for my produc-

tion effortsMaking the (Premium) clubGetting better agency office facilitiesFeeling more secure in my jobFeeling a greater sense of accomplishment from the work

I'm doingAdvancing within the company to a (higher) field management

positionMaking greater use of my skills and abilities on my jobHaving my family and friends view my job as having greater

prestigeGetting better supervisionFeeling greater self-esteem from my jobGoing to the company conventionHaving better working relations with the agents in my agencyReceiving more income from my jobMaking the ($1 Million) clubFeeling a greater sense of self-fulfillment from my job

lish a production objective or goal for the calendaryear. As agent goals presumably incorporated anumber of situational factors (e.g., market potential,intensity of competition) and individual factors (e.g.,tenure), they were used to provide a common basisfor making interagent production comparisons. Thus,a goal attainment criterion was established in this studyby dividing an agent's six-month volume by his pro-duction objective. Goals were obtained from thequestionnaire.

Outcomes. Twenty job outcomes were selected soas to include compensation, incentive-type awardspeculiar to the research setting, and a number of otherjob dimensions identified by Herzberg et al. [24] ashaving implications for motivation. These outcomeswere aggregated into sets so that competing interpreta-tions of a "motivationally salient" set could be tested.The various outcome sets are listed in Table 1.

Compensation. In addition to the strong nonempiri-cal emphasis on pay in the sales management literature,some evidence exists to suggest that pay is a significantpredictor of productivity when tested alone in anexpectancy theory format [26, 36,39,40]. The potencyof pay as a predictor may arise because it can beseen as instrumental in the attainment of many otherneeds, job related or otherwise [28]. The outcome,receiving more income from my job, was used forthis variable.

Incentives. Four production incentives specific tothe research setting were investigated. These includedmembership in three production clubs and invitationto the company's annual convention. Club membership

was based on three annual performance criteria.Agents selling one million dollars in paid annual volumewere admitted to the "$1 Million" club, while thoseplacing life insurance on 100 people in a calendaryear were awarded with membership in the "Lives"club. The top 16 agents yielding the greatest paidannualized premium (a profitability measure) wereadmitted to the "Premium" club. Finally, the conven-tion qualifications were based on a combination ofvolume and commission. As has been noted, a testof the separate effects of this set should, in part,resolve some long-standing issues concerning the be-havioral implications of sales incentives over andabove the effect of the basic compensation plan.

Intrinsic outcomes. Lawler [27] has drawn a dis-tinction between intrinsic and extrinsic outcomes.Intrinsic outcomes are self-bestowed and are thus moreavailable to the individual while extrinsic outcomesmust be mediated through another entity. In addition,intrinsic outcomes are of higher order in terms ofMaslow's [30] paradigm and conform to what salesmanagement writers refer to as "psychological"needs. The five intrinsic outcomes used in this studyincluded furthering one's professional growth, receiv-ing a sense of accomplishment from one's job, thesatisfaction of using one's skills and abilities on thejob, furthering one's self-esteem, and gaining a senseof self-fulfillment.

Most desirable outcomes. Researchers in organiza-tional behavior [17], social psychology [37], andmarketing [6, 19] have suggested that only a smallset of the most salient or important outcomes areused by individuals in behavioral decisions and attitudeformation. This position assumes that a motivationallysalient outcome set is peculiar to the individual. Theoutcomes used here included the three most desirableoutcomes as selected by the subject in a special sectionof the questionnaire.

All job outcomes. In early writings on expectancytheory, no distinction was made between the salienceof various sets of outcomes. Rather it was assumedthat all outcomes had some behavioral implicationsbecause if any did not, they would receive an assignedvalence or instrumentality weight of zero. A separatetest of all 20 outcomes chosen for the study wasincluded to provide evidence for this notion.

Valence. Outcome valences were measured on ascale from extremely undesirable (-3) through neitherdesirable nor undesirable (0) to extremely desirable(+3). This scoring method was used for all analysesexcept those in which the three most desirable out-comes were used. Because considerable restrictionof range was expected, a procedure used by Goodman,Rose, and Furcon [17] was applied. The most desirableoutcome was assigned a valence weight of 3, the secondranked outcome was given a weight of 2, and thethird outcome was given a valence score of unity.

Instrumentality. Instrumentality items were phrased

EXPECTANCY THEORY PREDICTIONS OF SALESMEN'S PERFORMANCE 247

in terms of subjective probabilities. Subjects wereasked to scale their beliefs of the effect of attainingtheir production goal on the occurrence of each ofthe outcomes. The scale ranged from no effect (0)to certain to occur (1). Intermediate values includedlow probability (.25), 50-50 chance (.5), and highprobability (.75).

Expectancy. For the expectancy measure, eachsubject was asked to indicate the chances in ten thathis efforts would lead to the attainment of his produc-tion goal.

Ability. While the literature on the ability of sales-men [15, 18] strongly suggests that both intelligenceand personality inventories are predictive of salessuccess, the company unfortunately did not administerstandard versions of these tests to its agents. Onlya recently introduced "in-house" intelligence test wasavailable. To confound matters, test scores wereobtained under both employee and applicant conditionsand were not available for roughly 10% of the sample.As the company was not willing to have additionaltests administered to its field, the scores on the"in-house" test were used as a sole measure of ability.The mean score was substituted for missing scores.

ANALYSISWhile Vroom's [46] model is stated in terms of

multiplicative interactions between variables, re-searchers [13, 23] have raised the issue of whetherthe multiplicative model contributes any variancebeyond that explained by the main effects in additivecombination. In order to provide evidence relatingto this issue, the contribution of a multiplicativeinteraction above that of the main effects was assessedin terms of the incremental variance explained in thecriteria [7].

The criteria were first regressed on the componentsof the additive model under investigation and thevariance explained in each criterion was observed.Next, the multiplicative interaction of components wasintroduced in the model as a third variable. Thecoefficient of determination yielded by this secondmodel was then compared to that obtained in the first

'Negative instrumentality responses (corresponding to ablocking effect) were not included in the study for tworeasons. First, it was felt that the inclusion of four corre-sponding negative items would make the instrumentalitysection more burdensome and difficult to comprehend.Instrumentality is a rather abstract concept requiring condi-tional probabilistic judgments. In order to minimize thepossibility of subject frustration, efforts were made tofacilitate completion of this section. Second, it was difficultto imagine goal attainment as having a blocking effect onany of the outcomes. Because of the autonomous natureof the agents' job and high importance placed on production,even peer relations were not expected to be adverselyaffected. Thus, it was hoped that the omission of negativevalues would not truncate the range of probable responses.

model and the incremental variance explained wastested for significance. If the increment was signifi-cant, the interaction was concluded to have made acontribution to the explained variance above that ofthe main effects.

HypothesesThe research hypotheses investigated were as fol-

lows:Hypothesis 1: Aggregate valence and instrumentali-ty

(a) P = a + p , 2 V.+ P^S/ j ,(b) P = a + p, 2 V, + p , 2 I,, + p3 2 VJ^,

Hypothesis 2: Performance valence and expectancy(a) P = a + p . i : V,7,.,+ P2E,.(b) P = a+^,lV,I^,+ ^^Ei

Hypothesis 3: Motivation and ability(a) P = a + p . E j S VJ^, +(b) P = a + p , E , . S V , r

p, A

RESULTS

Hypothesis 1Simple correlations between aggregate valence,

instrumentality, their interaction, and the performancecriteria are presented in Table 2. Aggregate valencefigures for the most desirable outcome set are notpresented because of the artificial weighting schemeused. Two-tailed tests were used for the aggregatevalence and instrumentality figures while one-tailedtests were used for the interaction term. Directionallytesting the interaction term but not its componentsderives from the theory. While both positive andnegative valence scores may be related to a motivatedstate if the corresponding instrumentality term is ofthe same sign, the interaction term should be relatedto high motivation only if the valence-instrumentalityproduct is positive. A negative product would beassociated with reduced motivation.

It is apparent from inspection of Table 2 that onlythe incentive outcomes predicted the performancecriteria. Aggregate valence, instrumentality, and theinteraction for this outcome set were highly correlatedwith the volume criterion although only aggregatevalence and the interaction were correlated with thegoal attainment criterion. While the instrumentalityterm for the compensation outcome predicted thevolume criterion, the correlations between the interac-tion term and both criteria were effectively zero. Thus,it appears that only perceptions regarding the incentiveoutcomes were effectively related to agent motivationin this study.

Results obtained from regressing the criteria on an

248 JOURNAL OF MARKETING RESEARCH, AUGUST 1974

Table 2CORRELATIONS BETWEEN THE PERFORA(\ANCE CRITERIA AND AGGREGATE VALENCE, INSTRUMENTALITY, AND THEIR

INTERACTION FOR FIVE JOB OUTCOME SETS

Aggregatevalence

Aggregateinstrumentality

Aggregatevalence times

instrumentality

Outcome setCompensationIncentiveIntrinsicMost desirableAll outcomes

Volume-.036

.433"-.150

—.043

Goalattainment

.068

.238"-.070

—-.018

Volume

.225"

.429"

.093

.153

.168

Goalattainment

.112

.148

.070

.090

.070

Volume

.022

.480''-.105

.172

.112

Goalattainment

055201 °

- 032102

.048

"ps .01.

Table 3RESULTS OBTAINED WHEN THE PERFORAAANCE CRITERIA WERE REGRESSED ON TWO AGGREGATE VALENCE AND

INSTRUMENTALITY MODELS (HYPOTHESIS 1)

Outcomeset

Compensation

Incentive

Intrinsic

Most desirable"

All outcomes

"2 V,, term omitted.

Model

ab

ab

ab

ab

ab

R

.271

.283

.470

.484

.225

.266

.153

.185

.194

.201

R^

.073

.080

.220

.234

.051

.071

.023

.034

.038

.041

Volume

p:R^

n.s.n.s.

.01

.01

n.s.n.s.

n.s.n.s.

n.s.n.s.

pAR^

n.s.

n.s.

n.s.

n.s.

n.s.

R

.113

.136

.239

.240

.129

.136

.090

.113

.114

.144

Goat attainment

R^

.013

.019

.057

.058

.016

.018

.008

.013

.013

.021

pR^

n.s.n.s.

n.s.n.s.

n.s.n.s.

n.s.n.s.

n.s.n.s.

pAR^

n.s.

n.s.

n.s.

n.s.

n.s.

additive combination of (a) aggregate valence andinstrumentality and (b) aggregate valence, instru-mentality, and their multiplicative interaction areshown in Table 3. The results show that, when usingthe incentive set to predict the volume criterion, asignificant multiple correlation coefficient was ob-tained although it was not significantly higher thanthe simple correlation obtained between the criteriaand the components taken separately. However, noother outcome set predicted this criterion nor wasany outcome set related to the goal attainment criteri-on. In addition, the contribution of the interactionterm over and above that of its components wasnonsignificant in every test for both criteria.

The pattern of results above can best be explainedby the correlations between the aggregate terms pre-sented in Table 4. Valence and instrumentality werehighly correlated as were the component terms withtheir interactions. It appears that the high multicollin-earity of the predictor variables obscured any possible

unique contribution of the individual terms, a phenom-enon also observed in a study of attitude formation[42].

This result could be explained by cognitive consis-tency principles. As predicted by balance theory [22],the agents may have oriented their instrumentality

Table 4CORRELATIONS BETWEEN AGGREGATE VALENCE,

INSTRUMENTALITY, AND THEIR INTERACTION

Outcome setCompensationIncentiveIntrinsicMost desirableAll outcomes

Aggregatevalence with

instrumentality.441.685.402

.677

Aggregatevalence with

theinteraction

.847

.902

.878

.885

Aggregateinstrumentality

with theinteraction

.665

.822

.714

.978

.811

EXPECTANCY THEORY PREDICTIONS OF SALESMEN'S PERFORMANCE 249

perceptions so as to coincide with valence perceptionsor vice versa. On the one hand, behavior may beseen as instrumental only for valent outcomes. Thatis, no thought may be given to unimportant outcomeswhether performance is instrumental for their attain-ment or not. As an alternative hypothesis, outcomesmay be desirable only to the extent that they areattainable. That is, only a dreamer would desiresomething wholly outside of his grasp. This argumentis consistent with recent thinking in attitude theorywhere it is believed that people change their beliefs

Table 5CORRELATIONS BETWEEN EXPECTANCY, ABIUTY, AND

THE PERFORMANCE CRITERIA

1

1.2.3.4.

VolumeGoal attainmentExpectancyAbility

.262° .065

.309» .029— .087

°p < .01, one-tailed test.

Table 6CORRELATIONS BETWEEN THE MOTIVATION VARIABLE

AND THE PERFORAAANCE CRITERIA

Outcome set

CompensationIncentiveIntrinsicMost desirableAll outcomes

Volume

.131.502"

- . 0 2 5.269".174

Goalattainment

.171.259=".060.245".135

.05..01.

and attitudes to bring them in accord with theirbehavior [3].

As a corollary to the relationships suggested above,a third hypothesis may be tenable. Perhaps the valence-^ instrumentality relation reflects the state of naturefor some outcomes, while the instrumentality -* va-lence relation best models reality for other outcomes.Based on the results presented here, these hypothesesare worthy of investigation.

Hypothesis 2Simple correlations between the expectancy variable

and the performance criteria are presented in Table5. Significant correlations were obtained for bothcriteria although the magnitude of the goal attainmentrelationship was higher. Clearly performance was afunction of expectancy perceptions.

The correlations between the motivation term andthe performance criteria are presented in Table 6.One-tailed tests were assumed. Both the incentive andmost desirable outcome sets predicted the criteriasignificantly. However, with the exception of therelation between the volume criterion and the motiva-tion term attributable to the incentive set, it appearsthat the significant results derived largely from theexpectancy measure. Note further that the multiplica-tive combination of expectancy and performance va-lence for the incentive set did not appreciably increasethe correlation obtained with the performance valenceterm alone.

The potency of the expectancy measure is reflectedin the pattern of results obtained when the criteriawere regressed on an additive function of performancelevel valence, expectancy, and their interactionreported in Table 7. Contrary to the earlier findings,the goal attainment criterion was significantly predict-ed for all outcome sets. However, the multiple cor-

Table 7RESULTS OBTAINED WHEN THE PERFORMANCE CRITERIA WERE REGRESSED ON TWO EXPECTANCY AND

PERFORAAANCE VALENCE MODELS (HYPOTHESIS 2)

Volume Goal attainment

Outcomeset Model

Compensation

Incentive

Intrinsic

Most desirable

All outcomes

ab

ab

ab

ab

ab

.262

.313

.498

.503

.290

.294

.290

.291

.269

.276

.069

.098

.248

.253

.084

.086

.084

.085

.072

.076

n.s..05

.01

.01

.05n.s.

.05n.s.

n.s.n.s.

n.s.311344

331341

314322

312314

309330

.096

.119

.109

.117

.099

.104

.097

.099

.096

.109

.05

.05

.05

.05

.05

.05

.05

.05

.05

.05

250 JOURNAL OF MARKETING RESEARCH, AUGUST 1974

relation coefficients were not significantly larger thanthe simple correlation between expectancy and thegoal criterioij. Because the correlation between thevolume criterion and expectancy was of a lowermagnitude, the results using the volume criterion wereattenuated for all outcome sets except the incentivegroup where the coefficient of determination washighly significant. Almost 25% of the volume criterionvariance was explained when expectancy and theS V /,fc term were regressed in additive combination.

When the interaction between expectancy and per-formance valence was introduced in the regression,its contribution to the explained variance was nonsig-nificant for all outcome sets using both criteria. Infact, addition of this term dropped the multiple cor-relation coefficient below the .05 significance levelfor two of the outcome sets when regressed on thevolume criterion. On the other hand, it increased thepredictive power of the compensation constructs al-though this increase was not significant.

The reason for the nonsignificant increase in predict-ability when the interaction term was introduced canagain be attributed to the inherent multicollinearitybetween interaction terms and their components. Boththe expectancy measure and the performance valenceterm were significantly correlated with their interac-tion. Thus the additive (a) models appear to be atleast as predictive as higher order models incorporatinginteraction terms.

Hypothesis 3Simple correlations between the ability measure and

the performance criteria are shown in Table 5. Thestatistics show that ability as measured in this studywas not correlated with the criteria. In fact, the abilitymeasure made no further predictive contribution whencombined either additively or multiplicatively with theother variables. For this reason, further discussionof the findings is not presented here.

The lack of correlation between ability and perfor-mance should be viewed only in the context of thisstudy, however. As the reader may recall, threeproblems were evident with the available scores forthe "in-house" intelligence test used for the abilitymeasure. First, the test was of uncertain origin; henceits reliability and validity were essentially unknown.Second, the technique of substituting mean predictorscores for missing scores was used. Generally, thiswill attenuate the magnitude of a correlation unlessthe corresponding criterion values are also at the mean.

The third problem concerns using scores from testsadministered under both employee and applicantconditions. Rothe [38] has shown that the mean ofa test distribution of applicant scores tended to behigher than the mean of scores for employees whowere asked to take the test for validation purposesand attributed this to a higher level of "test-takingincentivation" on the part of the applicants. Thus,

pooling test scores obtained under both conditionsas was done in this study may also have contributed tothe low correlation between ability and performance.In view of these problems, the reader is cautionednot to interpret the findings reported here as discon-firming evidence for the role played by ability as adeterminant of performance. Rather, these resultsshould further underscore the necessity of using reli-able and valid instruments in a conscientious selectionprogram.

In summary, a "motivationally salient" outcomeset was identified using expectancy theory constructs.Valence and instrumentality perceptions of the fourincentive-type outcomes used by the company understudy were strongly related to productivity and, toa lesser extent, goal attainment. No other outcomeset was consistently related to the criteria. While thegeneral expectancy measure was also correlated withboth performance criteria, the ability measure wasnot. When the volume performance criterion wasregressed on an additive combination of expectancyand the S V^I^^ term for the incentive set, almost25% of the variance in performance was explained.Given the paucity of empirical evidence on the roleof motivation in salesmen's performance, these resultswill hopefully encourage further research in this area.

DISCUSSIONThe general predictive superiority of motivational

perceptions attributed to the incentive outcome setover those of other sets has significant implicationsfor sales management. First, this result shows thatexpectancy theory may be useful in identifying whichof many possible outcomes in a sales situation areeffective motivators and which are unrelated to per-formance. This information will permit the sales man-ager to shift the resources at his disposal to the mostproductive incentive programs and to improve or phaseout those with little effect.

Second, the finding that various sales "stimulators"were, in fact, related to productivity supports in partwhat had been heretofore accepted as "lore" in thesales management literature. While this result mightbe attributed to spurious association in that a numberof outcome sets were tested against the criteria,incentive outcomes are generally recognized as havingmotivational implications. The great frequency withwhich these devices are used [20] attests to thepopularity of this belief.

The poor results obtained using the other outcomesets are noteworthy. In particular, the valence timesinstrumentality product for the compensation outcomedid not yield significant results, although instrumental-ity alone was significantly correlated with the volumecriterion. While this finding would appear to disputethe central postulate of many theories of sales motiva-tion, it could be attributed solely to a restriction ofrange in the valence of pay measure. However, the

EXPECTANCY THEORY PREDICTIONS OF SALESMEN'S PERFORMANCE 251

Standard deviation of this variable (LIO) was notsignificantly smaller than the mean standard deviationof all 20 outcomes (L32) or of the 4 incentive outcomes(1.25). Thus it appears that the desirability of furtheringone's income was, in fact, randomly related to bothperformance criteria in this study.

The low validities attributed to the most desirableoutcome set may have arisen from the possibility thatoutcome desirability does not necessarily connoteoutcome instrumentality although, as Table 4 shows,there is a marked tendency for these constructs tobe correlated. Conceivably, some outcomes could behighly desired but be relatively unattainable. Expec-tancy theory would predict that outcomes of this naturewould have little motivational impact.

The poor results obtained with the intrinsic outcomeset suggest that outcomes of this nature may bemotivationally dysfunctional. In particular, agentsmore desirous of intrinsic rewards showed a slighttendency to be poorer performers. Thus, until furtherevidence is collected, it is suggested here that themotivational implications of higher order psychologicalneeds have been overemphasized in the literature.

Finally, the notion that all job-related outcomes aremotivationally salient was not supported. Rather, asa number of writers [3, 12, 37] have suggested inother contexts, only a small subset of salient cognitionswere related to the sales behavior investigated in thisstudy. That is, while an individual may harbor manybeliefs about the effect of his performance on variousoutcomes and be able to elicit them when asked, hemay only act on a limited number.

Conceptual RefinementsA number of conceptual modifications may help

to improve upon the performance variance explainedin this study. Campbell et al. [5] and Jacoby [25]have suggested promising theoretical refinements tothe basic expectancy model. Campbell et al. posita hierarchy of outcome levels and suggest that per-formance is only instrumental for some (lower level)outcomes (e.g., pay). Higher level outcomes are seenas contingent not on performance, but on the attain-ment of lower level outcomes. Thus the motivationalforce attributed to a higher level outcome is a functionof its desirability, the probability that lower leveloutcomes will result in attainment of the higher need,and the desirabilities and instrumentalities of the lowerlevel outcomes. While this approach may be a moreaccurate representation of the relationship betweenoutcomes residing at various levels and performance,problems may be encountered in specifying the partic-ular need hierarchy for each individual.

Jacoby [25] has distinguished between the motiva-tional force attributed to outcomes of behavioral actsand that attributed to antecendent (input) values. Inputvalues are those which "push" an individual to per-form at a certain level apart from the outcomes of

performance and can be loosely classified as thosefactors which influence behavioral norms (i.e., whatone should do in a situation). Jacoby hypothesizesthat individuals can respond to questions concerningthe "significance" of a particular input and also the"likelihood that that input will influence selection"of a behavioral alternative. Jacoby's revised modelcan be stated as:

where:

M = motivation to perform (at level j),S( = significance of the Jth input,

Ly = likelihood that input i will lead to selectingperformance level j ,

V^ = valence of outcome k, andIj^ = instrumentality of performance level / for

attaining outcome k.

Hopefully future tests of expectancy theory predic-tions will profit from incorporation of the two theoret-ical refinements suggested here.

Apart from basic extensions of the theoreticalmodel, the relationships reported in this study mayhave been attenuated by improper specification ofthe combinatorial mode between variables. Expec-tancy theory assumes that an individual mentallymultiplies valence and instrumentality to arrive at amotivational force deriving from his desire for oraversion to each outcome. Yet no evidence existson how individuals actually process valence andinstrumentality perceptions. Conceivably this processmay be linear as opposed to multiplicative or maybe of some other interactive form. A growing bodyof research is emerging in the area of clinical judgment[16] that may provide insights to this issue.

Methodological RefinementBass and Wilkie [2] and Wilkie and Pessemier [48]

have recently drawn attention to a potential problemthat may arise when cross-sectional regression isperformed on multi-attribute attitude models. Unlessthe very strong assumption of homogeneity acrossrespondents is met, the results obtained in tests ofsuch models may be attenuated because various seg-ments of the sample perceive different rankings ofattribute importance [41] or display divergent re-sponse sets [2]. These problems can be solved byperforming an intra-individual analysis if independentvariable data are collected on more than one criterion(e.g., preferences for competing brands). Criterionrankings can then be compared to rankings predictedby the attitude model for each individual takenseparately. Bass and Wilkie have shown that the useof intra-individual analysis improved upon the number

252 JOURNAL OF MARKETING RESEARCH, AUGUST 1974

of correct preference predictions from a cross-sec-tional analysis of the same data by 15% [2, p. 267].

Although the data collected in this study did notpermit intra-individual analysis in that the motivationalperceptions of various performance levels were notinvestigated, small samples of sales personnel fromthe same company selling to a regional market suchas that used in this study may be "homogeneousenough" for cross-sectional analysis [48] whencompared to national consumer samples used in manystudies of brand preference. However, the possibilityremains that the results obtained here may have beenattenuated by regressing across heterogeneous re-spondents. To overcome this problem infuture studies,researchers are advised to investigate the motivationalperceptions attributable to various performance levelsrather than focusing on one (e.g., goal attainment)as was done here. For example, subjects could beasked to scale the instrumentalities of falling shortof, making, and exceeding their goal for each outcome.The predicted performance level having the highest2 Vf^Ij^ score could then be compared to the actualperformance level for each salesman taken separately.

Sources of Unexplained VarianceWhile the multiple correlation obtained when the

volume criterion was regressed on expectancy theoryconstructs was highly significant, the theory as testedhere explained only 25% of the performance variance.Three sources of unexplained variance were not testedin this study that may have increased the coefficientof determination measurably. The first encompassesa number of situational variables including marketpotential, territory workload, and the company's effortand experience [10, 11,35] that may appreciably affecta salesman's productivity. While an attempt has beenmade to include these variables in a multivariate model[9], the initial results have not been productive.Second, the potential contribution to the perfor-mance variance explained by pure measures of abilitywas not adequately tested. Had scores from a validselection battery been available, intelligence and per-sonality constructs may well have been predictive.As suggested in the introduction to this article, stan-dard tests of this nature have demonstrated validityon sales occupations.

Third, much of a salesman's decision to producemay be stochastic. Bass [1] has demonstrated thatbrand switching may be a stochastic process. It isalso conceivable that a salesman's performance maynot be entirely deterministic because of a "stochasticelement in the brain" which influences his productiondecisions. This may be especially true for sales oc-cupations where a high degree of autonomy is present.Future researchers may wish to provide an estimateof the performance variance remaining after stochasticprocesses have been accounted for so that potentially

promising theories can be evaluated against the re-maining variance attributed to deterministic processes.

REFERENCES1. Bass, Frank M. "The Theory of Stochastic Preference

and Brand Switching," Institute Paper No. 415, Institutefor Research in the Behavioral, Economic, and Manage-ment Sciences, Krannert Graduate School of IndustrialAdministration, Purdue University, 1973.

2. and William L. Wilkie. "AComparative Analysisof Attitudinal Predictions of Brand Preference," Journalof Marketing Research, 10 (August 1973), 262-9.

3. Bem, Daryl J. Beliefs, Attitudes, and Human Affairs.Belmont, Calif.: Brooks/Cole, 1970.

4. Burton, William L. "There's More to Motivating Sales-men than Money," in Taylor W. Meloan and John M.Rathmeli, eds.. Selling: Its Broader Dimensions. NewYork: Macmillan, 1960, 62-5.

5. Campbell, John P., Marvin D. Dunnette, Edward E.Lawler, III, and Karl E. Weick, Jr. Managerial Behavior,Performance, and Effectiveness. New York: McGraw-Hill, 1970.

6. Churchill, Gilbert A., Jr. "Linear Attitude Models: AStudy of Predictive Ability," Journal of MarketingResearch, 9 (November 1972), 423-6.

7. Cohen, Jacob. "Multiple Regression as a General Data-Analytic System," Psychological Bulletin, 70 (December1968), 426-43.

8. Cotham, James C , III. "Comments on: 'Sales Contestsas a Motivating Device,' " in Thomas R. Wotruba andRobert M. Olsen, eds.. Readings in Sales Management:Concepts and Viewpoints. New York: Holt, Rinehartand Winston, 1971, 297-300.

9. Cravens, David W., Robert B. Woodruff, and Joe C.Stamper. "An Analytical Approach for Evaluating SalesTerritory Performance," Journal of Marketing, 36 (Jan-uary 1972), 31-7.

10. Davis, Kenneth R. and Frederick E. Webster, Jr. SalesForce Management. New York: The Ronald Press, 1968.

11. Dodge, H. Robert. Field Sales Management: Text andCases. Dallas: Business Publications, 1973.

12. Fishbein, Martin. "Attitude and the Prediction of Beha-vior," in Martin Fishbein, ed.. Readings in AttitudeTheory and Measurement. New York: John Wiley andSons, 1967, 447-92.

13. Gavin, James F. "Ability, Effort, and Role Perceptionas Antecedents of Job Performance," The ExperimentalPublication System, Issue 5, Manuscript 190A. Wash-ington, D.C.: American Psychological Association, 1970,1-25.

14. Gaythwaite, Myles S. "Motivating Salesmen: SomeEnds and Means," in Albert Newgarden, ed.. The FieldSales Manager. New York: American ManagementAssociation, 1960, 161-85.

15. Ghiselli, Edwin E. "The Validity of Aptitude Tests inPersonnel Selection," Personnel Psychology, 26 (Winter1973), 461-77.

16. Goldberg, Lewis R. "Simple Models or Simple Pro-cesses? Some Research on Clinical Judgments," Ameri-can Psychologist, 23 (July 1968), 483-96.

17. Goodman, Paul S., Jerry H. Rose, and John E. Furcon."Comparison of Motivational Antecedents of the Work

EXPECTANCY THEORY PREDICTIONS OF SALESMEN'S PERFORMANCE 253

Performance of Scientists and Engineers," Journal ofApplied Psychology, 54 (December 1970), 491-5.

18. Guion, Robert M. Personnel Testing. New York: Mc-Graw-Hill, 1965.

19. Hansen, Fleming. Consumer Choice Behavior: A Cogni-tive Theory. New York: The Free Press, 1972.

20. Haring, Albert and Malcolm L. Morris. Prizes, Awardsfor Motivation. New York: Sales and Marketing Execu-tives-International, 1968.

21. Haring, Albert and Robert H. Myers. "Special Incen-tives for Salesmen," Journal of Marketing, 18 (October1953), 155-9.

22. Heider, Fritz. The Psychology of Interpersonal Rela-tions. New York: John Wiley and Sons, 1958.

23. Heneman, Herbert H., Ill and Donald P. Schwab."Evaluation of Research on Expectancy Theory Predic-tions of Employee Performance," Psychological Bulle-tin, 78 (July 1972), 1-9.

24. Herzberg, Frederick, Bernard Mausner, and BarbaraBlock Snyderman. The Motivation to Work, secondedition. New York: John Wiley and Sons, 1959.

25. Jacoby, Jacob. "Consumer and Industrial Psychology:Prospects for Construct Validation and Mutual Contri-bution," in Marvin D. Dunnette, ed.. The Handbookof Industrial and Organizational Psychology. Chicago:Rand McNally, forthcoming.

26. Lawler, Edward E., III. "Ability as a Moderator ofthe Relationship Between Job Attitudes and Job Per-formance," Personnel Psychology, 19 (Summer 1966),153-64.

27. . "Job Attitudes and Employee Motivation:Theory, Research, and Practice," Personnel Psychology,23 (Summer 1970), 223-37.

28. Pay and Organizational Effectiveness: A Psy-chological View. New York: McGraw-Hill, 1971.

29. Lewin, Kurt. The Conceptual Representation and theMeasurement of Psychological Forces. Durham, N.C.:Duke University Press, 1938.

30. Maslow, A. H. "A Theory of Human Motivation,"Psychological Review, 50 (July 1943), 370-96.

31. McDermid, Charles D. "How Money Motivates Men"in Robert F. Gwinner and Edward M. Smith, eds.. SalesStrategy: Cases and Readings. New York: Appleton-Century-Crofts, 1969, 321-32.

32. Mitchell, Terence R. and Anthony Biglan. "Instru-mentality Theories: Current Uses in Psychology," Psy-chological Bulletin, 76 (December 1971), 432-54.

33. Oliver, Richard L. "Expectancy Theory Predictions ofSalesmen's Performance and Determinants of the Mo-

tivational Antecedents," unpublished doctoral disserta-tion. University of Wisconsin—Madison, 1973.

34. Opsahl, Robert L. and Marvin D. Dunnette. "The Roleof Financial Compensation in Industrial Motivation,"Psychological Bulletin, 66 (August 1966), 94-118.

35. Phelps, D. Maynard and J. Howard Westing. MarketingManagement, third edition. Homewood, 111.: RichardD. Irwin, 1968.

36. Porter, Lyman W. and Edward E. Lawler, III. Manage-rial Attitudes and Performance. Homewood, 111.: RichardD. Irwin, 1968.

37. Rosenberg, Milton J. "Cognitive Structure and Attitudi-nal Effect," Journal of Abnormal and Social Psycholo-gy, 53 (November 1956), 367-72.

38. Rothe, Harold F. "Distributions of Test Scores ofIndustrial Employees and Applicants," Journal of Ap-plied Psychology, 31 (October 1947), 480-3.

39. Schuster, Jay R., Barbara Clark, and Miles Rogers."Testing Portions of the Porter and Lawler ModelRegarding the Motivational Role of Pay," Journal ofApplied Psychology, 55 (June 1971), 187-95.

40. Schwab, Donald P. and Lee D. Dyer. "The MotivationalImpact of a Compensation System on Employee Per-formance," Organizational Behavior and Human Per-formance, 9 (April 1973), 215-25.

41. Scott, Jerome E. and Peter D. Bennett. "CognitiveModels of Attitude Structure: 'Value Importance' isImportant," Proceedings. American Marketing Associa-tion, 1971, 346-50.

42. Sheth, Jagdish N. andW. Wayne Talarzyk. "PerceivedInstrumentality and Value Importance as Determinantsof Attitudes," Journal of Marketing Research, 9 (Febru-ary 1972), 6-9.

43. Stanton, William J. and Richard H. Buskirk. Manage-ment of the Sales Force, third edition. Homewood, 111.:Richard D. Irwin, 1969.

44. Still, Richard R. and Edward W. Cundiff. Sales Man-agement, second edition. Englewood Cliffs, N.J.: Pren-tice-Hall, 1969.

45. Tolman, Edward C. Purposive Behavior in Animals andMen. New York: Century, 1932.

46. Vroom, Victor H. Work and Motivation. New York:John Wiley and Sons, 1964.

47. Whitmore, E. "What Really Motivates Salesmen," SalesManagement, (October 7, 1960), 115-8.

48. Wilkie, William L. and Edgar A. Pessemier. "Issuesin Marketing's Use of Multi-Attribute Attitude Models,"Journal of Marketing Research, 10 (November 1973),428-41.