Sie sind auf Seite 1von 17

GILBERT A. CHURCHILL, JR., NEIL M. FORD, STEVEN W. HARTLEY, and ORVILLE C.

WALKER, JR.*

The authors use meta-analysis techniques to investigate the evidence that has been
gathered on the determinants of salespeople's performance. A search of the pub-
lished and unpublished literature uncovered 116 articles (the list of which is available
upon request) that yielded 1653 reported ossociations betv/een performance and
determinants of that performance. The results indicate the determinants can be or-
dered in the following way in terms of the average size of their association with
performance: (1) role variables, (2) skill, (3) motivation, (4) personal factors, (5)
aptitude, and (6) organizational/environmental factors. When ordered according
to the amount of the observed variation in correlations across studies that is real
variation (i.e., not attributable to sampling error), the determinants ronk as follows:
(1) personal factors, (2) skill, (3) role variables, (4) aptitude, (5) motivation, and
(6) organizationol/environmentol factors. To investigate whether the associations
between each of the categories of predictors and performance could be partially
accounted for by the presence of moderator variables, the results were broken out
by customer type, product type, and type of dependent measure used. The results
indicate that the strength of the relationship between the major determinants and
salespeople's performance is offected by the type of products salespeople sell. The
authors discuss the impiicotions of these findings for soles managers and researchers.

The Determinants of Salesperson Performance:


A Meta-Analysis

Sales managers have always tried to understand the classification of phenomena and limitations of the stud-
determinants of good sales performance. In response to ies, it is not the most effective way of generating a con-
this interest, researchers have examined many possible sistent summary of results.
determinants of sales performance over the past 75 years. Meta-analysis is an attractive alternative for integrat-
The studies have produced very inconsistent results with ing research findings (Glass 1976. p. 4; Glass. McGaw,
respect to what factors affect sales performance and the and Smith 1981). It treats the fmdings of individual stud-
strength of the relationships. ies as dependent variables and examines those findings
Past reviews of the salesperson performance literature as a function of one or more independent variables in an
have not provided an adequate explanation for the ob- attempt to account for the variation in results across stud-
served inconsistencies. Part of the problem is that all of ies. The independent variables investigated typically in-
the reviews have been narrative. Though a narrative re- clude both the substantive and methodological factors that
view affords the opportunity for special insight about the can influence results. In essence, meta-analysis is the
application of the principles traditionally employed in
primary research studies to the review and integration of
'Gilbert A. Churchill. Jr is Dunald C. Slichter Professor in Busi-
ness Research and Neil M. Ford is Professor of Business, University the findings in a body of studies.
of Wisconsin-Madison. Steven W. Hartley is Assistant Professor of We use meta-analysis to analyze the evidence about
Marketing, University of Denver. Orville C. Walker, Jr.. is Professor the factors that affect salesperson performance. In the
of Business, University of Minnesota. first section we describe the process followed in locating
The many helpful suggestions of the three anonymous JMH review-
ers and the Editor are gratefully acknowledged. the studies that served as input to the investigation. We
then look at some structural characteristics of these stud-

103

Journal of Marketing Research


Vol. XXII (May 1985). 103-18
104 JOURNAL Of= MARKETING RESEARCH, MAY 1985

ies, including when they appeared, where, the number correlation. No differences were found. Consequently,
of companies, and the number of salespeople involved. the measure of association used was ignored in subse-
In the next two sections we overview the analysis pro- quent analyses.
cedure and examine the evidence on how much of the
variation in salesperson performance is associated with STRUCTURAL CHARACTERISTICS OE STUDIES
the major determinants—aptitude, skill levels, motiva-
tion, role perceptions, other personal, organizational, and Temporal Patterns
environmental factors. Finally, we discuss the implica- An investigation of the publication dates of the studies
tions of the findings for sales managers and researchers. revealed that 57% were conducted in the 20-year period
beginning in 1951—probably because of the lack of
PROCEDURE computer capability before then. The temporal pattern
The primary goal in selecting data sources for a meta- for the average number of companies and average sam-
analysis is to secure a representative sample and avoid ple size per study was very irregular, as Table 1 indi-
potential bias. The lack of a sampling frame, however, cates.
ultimately dictates that a meta-analyst attempt a census The process of knowledge development suggests that
of all studies pertaining to a research question. as evidence accumulates, theories improve, fewer as-
Figure 1 illustrates the procedure used to identify in- sociations are hypothesized, and the evidence for those
vestigations of salesperson performance for our study. associations is stronger. These expectations were not
Though a typical literary review focuses on published supported. Rather, later studies contained more reported
literature, a meta-analysis acknowledges the fact some associations between predictors and performance mea-
research may not be published. Thus, in the first "stage" sures, and the average size of the correlation coefficients
of the search we utilized published reference sources, was no larger. In fact, the strongest associations between
indices, and journals, whereas in the second—and more predictors and performance measures were found in the
difficult—search effort we attempted to identify unpub- earliest studies when the theories about the determinants
lished literature by contacting businesses, consultants, of salesperson performance were least developed.
government offices, and trade associations. Not surpris-
ingly, many sources were reluctant to release what they Size of the Correlations
considered proprietary information. Finally, references The generally small size of the correlations between
from the studies identified in the first stage were ex- predictors and performance criteria is somewhat dis-
amined for new citations. couraging. As Table 1 indicates, the average r across the
The literature search identified 409 citations from 63 1653 correlations is .188; on average less than 4% of the
joumals covering the period from 1918 to 1982. We were variation in salesperson performance is associated with
able to obtain copies of 393 of the studies. The search a single predictor of that performance. In most studies,
of unpublished literature identified 36 dissertations, 27 of course, the proportion of variation explained is higher
of which were released by university libraries, and lim- than 4% because the variables were combined in devel-
ited industry selection material. oping the prediction equation. Nonetheless, the fact that
Of all the studies obtained. 116 contained enough in- so little of the variation in performance is associated with
formation to be included in the meta-analysis in that they any single predictor supports the notion that models of
reported some empirical evidence about the relationship the determinants of salesperson performance must in-
between salesperson performance and one or more log- corporate multiple causes.
ical predictors of that performance.' Most studies re-
ported associations between performance and several ANALYSIS PROCEDURE
variables. For purposes of many of the subsequent anal- Given the large number of factors that have Deen in-
yses, we treated each individual association as a separate vestigated as possible determinants of a salesperson's
observation, producing 1653 observations in all. performance, we need to categorize the reported corre-
Most studies reported measures of association be- lations in order to make sense of them. One useful tax-
tween performance and a predictor variable in the form onomy is provided by the Churchill, Ford, and Walker
of a correlation coefficient. Though most studies used (1985) model of salesperson performance. Their model
product-moment correlation coefficients, some used other holds that a salesperson's performance is a function of
measures of association. The results were analyzed to five basic factors: (1) aptitude, (2) skill level, (3) mo-
determine whether the measure of association used in the tivation, (4) role perceptions, and (5) personal, organi-
study was in any way related to the size of the reported zational, and environmental variables.
Using their definitions, we classified the 1653 obser-
vations into one of six categories—aptitude, skill level,
motivation, role, personal factors, or organizational and
'A list of the studies as well as a more detailed version of the article environmental factors. In coding the studies, we record-
is available by request. Please write Gilbert A. Churchill, Jr., Uni-
versity of Wisconsin-Madison, School of Business, 1155 Observa- ed each variable at the greatest level of detail possible.
tory Drive, Madison, WI 53706. For example, the use of either the Gordon Personal Pro-
DETERMINANTS OF SALESPERSON PERFORMANCE 105

Figure 1
SEARCH PROCESS

Unpublished Published

1
Government Industry Other Reference Indices Other

Equal Life Disser- AMA Biblio Work Books


Employment Insurance tation #36: Selling Related
Opportunity Market Abstracts and Sales Abstracts
Research Management Search
Association Biblio-
Air Force Business graphies
Human Mental Periodicals of
Resources Measurements Index Previous
Laboratory Yearbook Reviews

Personnel
National Tests in Management
Technical Print Abstracts
Information
Service
ASAP Hand- ERIC
book of
Smithsonian Personnel
Science Psych
Information Abstracts
Exchange Dartnell
Sales
Managers Journal
Handbook Indices

Handbook of
Industrial &
Organiza-
tional Psych

Collect Articles

Check References

No
106 JOURNAL OF MARKETING RESEARCH, MAY 1985

Table 1
USE OF INDEPENDENT VARIABLES AS A FUNCTION OF PUBLICATION PERIOD

Organi-
zational/
Average Role environ-
no Average Moti- percep- Personal mental
Number com- sample Total Aptitude Skill level vation tion factors factors
Publication of panies size per no. Mean No./ No./ No./ No./ No./ No./
period studies per study study of r's r No.' study*" No. study No. study No. study Nn. study No. study
Before 1920 1 1.0 18.0 11 .266 22 22.0
1921-1925 7 17.3 626.1 56 .218 2 2.0 54 9.0
1926-1930 3 1.0 63.7 38 .354 8 4.0 26 8,7 2 2.0 2 2.0
I93I-I935 2 13.5 245.0 13 .324 4 2,0 1 1,0 8 8.0
1936-1940 3 1.0 291,7 9 .190 3 1.5 5 5.0 1 1.0
1941-1945 3 24.0 41,7 35 .142 34 17,0 1 1.0
1946-1950 4 11.5 306,5 21 .223 16 5.3 1 1,0 4 4.0
1951-1955 13 1.1 121.7 101 .175 51 7.3 14 3,5 1 1.0 32 5.3 3 3.0
1956-1%0 29 1.1 215,8 367 .234 222 12.3 88 8,8 13 1,9 44 5.5
1961-1965 12 9.2 195,5 321 .131 275 27.5 13 4,3 6 6,0 17 8.5 10 10.0
1966-1970 12 2.9 298,8 131 .269 34 4.3 8 2,7 47 9.4 42 8.4
1971-1975 7 33.4 144.1 109 .113 34 11,3 20 10,0 27 13.5 10 5.0 18 18.0
1976 and after 20 21.0 491.7 430 .166 127 14.1 1 1,0 30 5.0 59 14.8 193 12.1 20 10.0
Total
(percentage) 116 1653 .188 832 (50) 178 (10) 126 (8) 59 (4) 407 (25) 51 (3)
'Refers to total number of correlations between predictors and performance reported in the period. This number is greater than lhe number of
studies because most studies reported multiple pairwise correlations between predictors and performance criteria.
""Refers to average number of correlations per study among those studies in which the predictor was used. By dividing the number of correlations
of a given type by the average number per study, one can determine the number of studies in a period in which the particular predictor was used.

file or the Bernreuter Personality Inventory to measure concentrate on the weighted average correlation, but also
the same personality characteristic (e.g.. sociability) was report the unweighted or simple average correlation in
recorded by using separate four-digit codes. One person the analysis that follows. More specifically, we report;
coded all of the studies. Coding quality was checked by
having an independent investigator code a random sam- 1. The mean of the observed correlations across studies us-
ing sample-size-weighted and unweighted r ' s . The
ple of 5% of the study observations. No errors were weighted mean provides the best estimate of the popu-
identified with any of the four-digit codes that were as- lation correlation between the predictor and the criterion
signed. The four-digit codes subsequently were col- assuming there is no range variation in the independent
lapsed into the six major categories of predictors by the variable and no error of measurement.
authors in a group meeting. 2. The attenuation-corrected mean which is the best esti-
mate of the population correlation between the latent
Statistics Used to Test the Model constructs after allowing for the errors associated with
To assess the strength of the relationship between each their measurement,
of the hypothesized major determinants and actual per- 3. The total variance in the observed correlations and an
estimate of how much of it is due to sampling error. The
formance, we followed the analytic procedures sug- corrected variance which represents the difference be-
gested by Hunter, Schmidt, and Jackson (1982). They tween the total variance and the sampling error variance
suggest that the best estimate of the size of the corre- is also reported. The square root of the corrected vari-
lation between predictor and criterion is not the simple ance is used to form a 95% credibihty interval around
mean r across studies, ". . . but a weighted average in the weighted mean. The credibility interval for each de-
which each correlation is weighted by the number of per- terminant provides an indication of the size of the popu-
sons in that study" (p. 41). Similarly, the appropriate lation correlation likely to be observed in a given study.
variance in the r's across studies is the frequency-weighted Large credibility intervals suggest the observed correla-
average squared error, rather than the usual sample vari- tion between the predictor and performance might vary
dramatically from study to study because of the influence
ance. Their rationale is that a correlation based on 500 of other factors present in one situation but not in others.
persons contains more information than one based on 50 When the "real variation' in results is large, one searches
persons because the estimate based on the larger number for moderator variables that may account for it.
of observations has smaller sampling error. There ap-
4. The 95% credibility interval for the attenuation-cor-
pears to be some risk, though, in relying exclusively on rected mean. Though the attenuation-corrected results were
the weighted average correlation in that a few large-sam- used to calculate these credibility intervals, the attenua-
ple studies could dominate the analysis. We therefore tion-corrected variances are not displayed because gen-
DETERMINANTS OF SALESPERSON PERFORAAANCE 107

erally they were very close to the sampling-error-cor- mean with the attenuation-corrected mean. The atten-
rected variances. The attenuation-corrected credibility uation mean is calculated by dividing the weighted mean
interval describes the size of the real variation in results by the prcxluct of the average reliabilities of the inde-
across studies when proper allowance is made for the pendent and dependent variables, where the average re-
errors with which the independent and dependent con- liabilities are calculated from those studies in which they
structs were measured. are reported (Hunter. Schmidt, and Jackson 1982, p. 7 4 -
IMPACT OF MAJOR DETERMINANTS 83; Schmidt and Hunter 1977). These average reliabil-
ities are .860 for aptitude and .831 for pedbrmance. The
Aptitude errors of measurement cause the mean to be depressed
The very first study on salespersoti performance, which from .193 to .138. Even with no errors in the measure-
appeared in 1918 (Oschrin), examined the relationship ments, then, salespeople's aptitude on average has a low
between aptitude and performance. More than 50% of correlation with their performance.
the pairwise correlations reported iti the literature sinee
then also use an aptitude measure as the predictor of per- Skill Levels
formance (see Table i). Though the emphasis given to Selling skills have received less attention as predictors
aptitude seems logical, the empirical evidence on the of pertbrmance than aptitude. Only 178 of the total 1653
strength of the relationship between aptitude and per- correlations use a salesperson skill as a predictor of per-
formance is weak. As Table 2 and Figure 2,A indicate, formance. Further, the attention paid to skills has been
820 correlations were found in the literature. The simple sporadic (see Table I) and the number of skills inves-
correlations range from 0 to .85. The higher correlations tigated per study has been low. The heaviest emphasis
suggest that in at least some studies salespeople's apti- on skills as a predictor of performance was in the 1956-
tude was a good predictor of their performance. How- 1960 period when 10 studies employed at least one skill
ever, the weighted mean correlation is only .138, sug- variable and there were 88 reported associations in all.
gesting that, on average, slightly less than 2% of the The average size of the correlations with performance
variation in salespeople's performance can be accounted is greater when skill is used as a predictor than when
for by variations in their aptitude. Further, a large por- aptitude is used. As Figure 2,B indicates, the distribu-
tion—more than 49%—of the total variation in observed tion of simple correlations is also much less skewed. The
correlations across studies is attributable to sampling er- sample-size-weighted mean correlation of .268 indicates
ror. The credibility interval for the weighted mean after that on average somewhat more than 7% of the variation
adjusting for sampling error is - . 0 5 4 :^ p :^ .331. The in performance is associated with variations in skill. The
size of the adjusted interval suggests there may be one attenuation-corrected mean of .320 suggests that more
or more variables that moderate the relationship between than 10% of the variation in the latent constructs is shared
aptitude and performance, a point to which we return variance.
subsequently. Not only is the total variation in observed correlations
The strength of association between aptitude and per- larger with skills than with aptitude, but a much smaller
formance is biased downward only slightly by measure- proportion of the total variance is sampling error vari-
ment errors as can be seen by comparing the weighted ance—25% for skills versus 49% for aptitude. The 95%

Table 2
VALUES OF r FOR TYPES OF INDEPENDENT VARIABLES

Variance Credibility interval for


Mean Sampling Attenuation-
Number Attenuation- Total error Corrected Ratio Weighted corrected
Independent variable of r's' Simple Weighted corrected T (SE) tC) SE/T mean mean
Aptiiiidc 820 .170 .138 .193 .018 .009 .009 .494 -.054 -.076
.331 .461
Skill 178 .276 .268 .320 .037 .009 .028 .253 -.067 -.077
.602 .716
Molivation 126 .228 .184 .258 .017 .009 .008 .533 .005 .012
.363 .505
Role 59 .302 .294 .379 .020 .008 .on .422 .079 .102
.508 .655
Personal factors 407 .166 .161 .292 .043 .002 .041 .058 -.240 -.438
.560 1.000
Organizational/ 51 .142 .104 .104 .010 .004 .006 .405 -.051 - .051
environmental factors .259 .259
There were 12 correlations, all of which pertained to aptitude, for which the sample size was not reported and which had to be excluded from
the analysis bccau.se the emphasis is on sample-size-weighted mean r's.
lOB . •• J O U R N A L O F M A R K E T I N G RESEARCH, M A Y T985

Figure 2
DISTRIBUTION OF CORRELATION COEFFICIENTS FOR EACH MAIN PREDICTOR OF PERFORMANCE

A. APTITUDE B . SKILL LEVEX

(S 140

^ KB

t .13 . » .m .m .44 . » . » . • .71 .B4 • • um .04 . u .JO . a .m .44 . » .B . « . n .04 .mum


Size of Correlation Coefficient Size of Correlation Coefficient

C, HCmVATION D. ROLE

.04 .12 .30 . » .« .44 . » . n .ai .m .M . « i . a o


Size of Correlation Coefficient Size of Correlation Coefficient

E. PERSONAL FACTORS F . OHCANIZATIONAL/ENVIROKMENTAL FACTORS

.04 .12 .ao .20 .IB .44 .» .m .m .» .** .SB 1.00 .04 . 1 2 .ao .ao .BB .44 . » .BD .BB .7B . M .OB 1.00
Size of Correlation Coefficient Size of Correlation Coafficiant
DETERMINANTS OF SALESPERSON PERFORMANCE 109

credibility interval ranges from a low of - . 0 6 7 to a high not part of the aptitude, skill level, motivation, and role
of .602. The credibility interval for the latent constructs perceptions components. Past studies have included such
is larger still, ranging from a low of —.077 to a high of factors as the salesperson's age, height, sex, weight, race,
.716. The large amount of variation in the correlations appearance, education, marital status, number of depen-
that remains after removal of the effects of sampling er- dents, club memberships, and other similar character-
ror provides strong evidence that one or more variables istics. As Table 1 indicates, the interest in personal char-
moderate the relationship between skill level and per- acteristics as a predictor of salesperson performance has
formance. been pervasive and continuing. It is the second largest
category of associations, accounting for approximately
Motivation 25% of all reported correlations.
Researchers' interest in motivation as a predictor of Table 2 indicates that the simple mean correlation for
salespeople's performance is relatively recent. Except for personal factors is .166 and the weighted mean corre-
one early study, it was not until the 1950s that research- lation is .161. On average, the variation in performance
ers empirically explored motivation as a predictor of sales accounted for by variations in personal factors is rela-
perfomiance. tively equal to that accounted for by aptitude.
The weighted mean eorrelation of .184 and attenua- The major difference between personal factors and ap-
tion-corrected mean correlation of .258 in Table 2 in- titude as predictors of performance is related not to what
dicate that the shared variance between motivation and is true on average but to what happens across studies.
pertbrmance averages more than 3% with the variables First, the variance in observed correlations is much larger
as measured and less than 7% with the latent constructs. (more than two times larger) when personal factors are
The simple mean correlation is larger than the weighted used to predict performance than when aptitude factors
mean correlation and as Figure 2,C indicates, most of are used. Moreover, a much smaller portion of the vari-
the simple correlations range from .02 to .48. The over- ation is attributable to sampling error, less than 6% with
all pattern of results suggests that, on average, motiva- personal factors and more than 49% with aptitude fac-
tion is a better predictor of performance than aptitude, tors. The upshot is that analysts can expect a wider range
but not as good a predictor as skill level. in the strength of the observed associations when per-
Table 2 also indicates that when motivation is used to sonal factors rather than aptitude are used to predict per-
predict pertbrmance more than half of the observed vari- formance. The credibility interval representing the ob-
ation in results is attributable to sampling error. Alter- served correlations that might be realized after the effects
nately, the credibility interval—which indicates the size of sampling error are taken into account when personal
of observed correlations likely to be realized in future factors are used to predict performance is - . 2 4 0 ^ p ^
studies that use motivation to predict performance—ranges .560.
from .005 < p < .363. which is approximately the same
as for aptitude but smaller than for skill. Organizational and Environmental Factors
The investigation of how organizational and environ-
Role Perceptions
mental factors affect salespeople's pertbrmance has been
Researchers' interest in role as a predictor of sales- concentrated in a few studies, namely one in each of the
people's performance is even more recent than their in- 1951-1955. 1961-1965, and 1971-1975 periods and two
terest in motivation. The first empirical study using role after 1976, which yielded 51 empirical associations in
did not appear until 1978, and only four studies have all.
used role as a predictor of performance. As Table 2 indicates, the average correlation between
Table 2 indicates that the 59 correlations between role predictor and performance criterion is lowest for orga-
and performance average .294 with the weighted mean nizational and environmental factors among the six cat-
and .379 with the attenuation-corrected mean. These egories of predictors studied. The average simple mean
findings suggest that less than 9% of the variance is shared correlation is .142 and the weighted mean correlation is
when the constructs as measured are used whereas more .104. The attenuation-corrected mean correlation also is
than 14% of the variance is shared when the latent con- shown as .104 because not enough evidence is available
structs are used. Not only is the average correlation with for calculation of the attenuation-corrected results. The
performance higher for role variables than for other pre- results suggest that, on average, only 1% of the variation
dictors, but as Figure 2.D indicates, the spread of cor- in performance is associated with variations in organi-
relations is more symmetrical around the average. Also, zational/environmental factors. The variation in ob-
more than 42% of the variation in correlations is sam- served correlations when organizational/environmental
pling-error-induced. The weighted mean credibility in- factors are used to predict performance is also small, and
terval after allowing for sampling error ranges from .079 more than 40% of that variation is attributable to sam-
to .508. pling error. The resulting credibility interval is conse-
quently narrow. The combination of a low average value
Personal Variables and narrow credibility interval suggests that analysts
Personal variables are intra-individuai factors that might should expect to observe relatively small pairwise cor-
be related to salespeople's performance but which are relations when using organizational/environmental fac-
no JOURNAL OF AAARKETING RESEARCH, MAY 1985

tors to predict performance; the best estimate is that these variable. The meta-analysis within subsets also tells us
correlations will be in the range from - . 0 5 1 to .259. how much of the residual variance within subsets is due
One factor that might partially account for the poor to sampling erTor and how much is real (Hunter, Schmidt,
predictive power of the organizational/environmental and Jackson 1982, p. 36).
factors vis-a-vis other predictors is the general lack of
variability and the restriction of range of the organiza- To investigate whether the associations between each
tional/environmental factors. As only five studies in- of the categories of predictors and performance could be
vestigated the impact of organizational/environmental partially accounted for by the presence of moderator
factors on performance, the observations came from only variables, the results were broken out by customer type,
a few companies. Consequently, they reflect only a lim- product type, and type of dej>endent variable. Moderator
ited set of external conditions, and the limited variability variables tend to show themselves in two ways: (1) the
of the predictor naturally constrains the amount of vari- average correlation varies across subsets and (2) the
ance in the criterion for which it might account. sampling-error-corrected variance averages lower in the
Figure 3 summarizes the fmdings for the six hypoth- subsets than for the data as a whole (Hunter, Schmidt,
esized determinants of salespeople's performance. The and Jackson 1982, p. 48).
determinants are ordered in the figure according to their
degree of association with performance. The credibility Customer Type
intervals for the observed variables are shown in Figure Table 3 reports the results when the values of the cor-
3,A and those for the latent constructs are shown in Fig- relation coefficient are broken out by customer type and
ure 3,B. the six main categories of predictors. Customers were
categorized broadly as either "institutional" or "individ-
IMPACT OF MODERATOR VARIABLES ual" because there were too few observations for fmer
The ability of the individual determinants to predict distinctions.
performance seems rather unimpressive. The highest av- Several facts stand out in Table 3. First, though most
erage correlations are produced by the role perception, studies identified the type of customers being served, some
skill level, and motivation measures, but these measures did not. Second, the historical emphasis has been on firms
account for only 3 to 9% of the variation in performance selling to institutional customers rather than individuals.
with the results unadjusted for measurement error. How- Finally, some variation is present in the size of the av-
ever, the unimpressive results are somewhat counterin- erage correlations across customer categories and some
tuitive. It seems, a priori, that salespeople's perfor- variability is present in the sampIing-erTor-corrected
mance would be related to their ability to perform or to variances across subsets as well.
the skills they bring to the job, or to their motivation One could focus on the weighted mean within each
levels, and so on. Perhaps the disappointing results are subset to determine for what types of customers the pre-
due to a failure to model adequately the full set of de- dictor most affects performance, because the weighted
terminants of sales performance. Walker. Churchill, and mean provides the best estimate of the mean population
Ford (1977), for example, suggest that " . . . the ap- correlation. One also could use the credibility interval
propriate definition of aptitude, and the appropriate mea- around each weighted mean to estimate how much of the
sures of the construct, may vary greatly from industry variation within each subset is real. Though we mention
to industry, firm to firm, and product line to product several of the main differences hereafter, we do not con-
line, depending on what specific tasks must be per- centrate on these comparisons because Tables 3, 4, and
formed and what dimensions of performance are consid- 5 can be inteipreted in exactly the same way as Table
ered important" (p. 167). The argument implies that the 2, and readers can make such judgments themselves. In-
type of customers to which the firm sells or the types of stead, we concentrate our discussion on comparisons of
products it sells or the types of measures it uses to eval- the "corrected variances (CV) ratios." The CV ratios,
uate performance all might affect the observed relation- which are listed in the last line under each of the major
ships between sales performance and various predictors categories of predictors, show the ratio of average cor-
of that performance. rected variance within subsets to the total corrected vari-
The empirical evidence in Table 2 and Figure 3 sup- ance for that predictor. A large ratio indicates the av-
ports such an argument. For all six categories of pre- erage corrected variance is only marginally lower within
dictors, the variation in population correlations across subsets than for the predictor as a whole; for example,
studies after allowing for sampling error is still large. a ratio of .90 means that the average corrected variance
When such a condition arises, it is appropriate to look within subsets is 90% of the overall corrected variance
for moderator variables to explain that variance. for the predictor. Conversely, small ratios indicate the
average corrected variance is substantially lower within
To test a hypothesized moderalor variable, we break the subsets than for the predictor as a whole. In sum, the
set of studies into subsets using the moderator variable. smaller the ratio the more the cross-classification vari-
. . . We then do separate meta-analyses within each sub-
set of studies. If we find large differences between sub- able in question moderates the relationship between the
sets, then the hypothesized variable is indeed a moderator predictor and performance.
DETERMINANTS OF SALESPERSON PERFORMANCE 111

Figure 3
SUMAAARY OF FINDINGS ON IMPACT OF MAJOR DETERMINANTS

A. USING OBSERVED VARIABLES

ROLE SKILL MOTIVATION PERSffliAL APTITUDK ORGANIZATIONAL/


FACTORS FNVIKONMENTAL
— FACTOBS

B. USING LATENT CONSTHUCTS

ROLB SKILL MOTIVATION PERSONAL APTITUDE ORGANIZATIONAL/


FACTORS ENVIRONMENTAL
FACTORS
112 JOURNAL OF MARKETING RESEARCH, MAY 1985

Table 3
VALUES OF r FOR TYPES OF INDEPENDENT VARIABLES AND CUSTOMER TYPES

Customer type
Individual
Independent variable consumers Institutions Unknown
Aptitude
Number of r's 260 392 168
Simple mean .180 .186 .119
Weighted mean .156 .171 .085
Attenuation-corrected mean .218 .240 .118
Corrected variance (C) .012 .011 .001
Credibility interval for
Weighted mean -.062 £ p s; .373 -.035 S p s .378 .018 s p s .151
Attenuation-corrected mean -.084 £ p <; .579 -.046 < p := .525 .028 s p s .209
CV ratio .972
Skill level
Number of r's 64 101 13
Simple mean .254 .287 .294
Weighted mean .265 .272 .279
Attenuation-corrected mean .317 .324 .333
Corrected variance (C) .033 .017 .013
Credibility interval for
Weighted mean -.100 £ p s: .631 -.010 s p := .534 .048 s p s .510
Attenuation-corrected mean -.118 s p <; .751 -.014 s p '.= .634 .061 < p < .605
CV ratio .812
Motivation
Number of r's 55 43 28
Simple mean .289 .226 .110
Weighted mean .211 .231 .073
Attenuation-corrected mean .296 .325 .103
Corrected variance (C) .005 .005 .001
Credibility interval for
Weighted mean -.071 s p < .350 -.088 < p -= .374 .014 £ p S .132
Attenuation-corrected mean -.109 < p :s .484 -.134 s p == .515 .022 ^p^ .183
CV ratio .509
Role
Number of r's 52 7
Simple mean .316 .194
Weighted mean .323 .194
Attenuation-corrected mean .418 .251
Corrected variance (C) .006 .019
Credibility interval for
Weighted mean -.172 s p =s .475 - .078 < p :s .467
Attenuation-corrected mean -.223 s p < .613 -.101 s p< .603
CV ratio .629
Personal factors
Number of r's 161 197 49
Simple mean .220 .126 .154
Weighted mean .238 .129 .087
Attenuation-corrected mean .434 .235 .159
Corrected variance (C) .072 .002 .008
Credibility interval for
Weighted mean -.297 ^ p < .773 .040 £ p < .217 -.093 s p s .268
Attenuation-corrected mean -.542 s p < 1.00 .073 :£ p S .396 -.170 :< p ^ .489
CV ratio .756
Organizational/environmental factors
Number of r's 7 16 28
Simple mean .147 .236 .087
Weighted mean .145 .251 .051
Attenuation-corrected mean .145 .251 .051
Corrected variance (C) .002 .010 0
Credibility interval for
Weighted mean .059 £ p s .232 .051 s p == .451 -.001 ^ p s .102
Anenuation-corrected mean .059 < p S .232 .051 £ p -= .451 -.001 S p :< .102
CV ratio .558
DETERMINANTS OF SALESPERSON PERFORMANCE 113

The CV ratios suggest that type of customer has very The table distinguishes between two types of objective
little effect on the relationship with performance when data—those which control for externalities and those
either aptitude, skill level, or personal factors are used which do not. This distinction is consistent with the con-
as predictors. With motivation, as well as organiza- ceptual separation of "performance" and "effective-
tional/environmental and role factors, customer type does ness," as discussed by Walker, Churchill, and Ford (1979.
moderate the relationships because the average corrected p. 33-5). Objective company data that control for ex-
variance within subsets is approximately 60% or less of ternalities include .such items as sales as a percentage of
what it is for each predictor as a whole. quota or market potential, sales corrected for economic
conditions, or the salesperson's route difficulty. Objec-
Product Type tive company data that are not corrected for externalities
Product type and customer type clearly overlap in that include such items as total sales volume, expenses, num-
certain types of goods and services are sold to either in- ber of total calls or new account calls, or the number of
stitutional customers or individual customers but not both. demonstrations. Some of these measures reflect sales-
However, certain other goods and/or services might be person inputs (e.g., number of calls) and some refiect
sold to either type even though the product is primarily outputs (e.g., sales). The use of both inputs and outputs
a consumer got^d or an industrial good. Table 4 conse- as performance criteria was observed for all four types
quently shows the results by type of product, using the of dependent measures.
broad classes consumer goods, industrial goods, and ser- The evidence in Table 5 supports authors who argue
vices. that the size of the correlations between predictors and
Several items in Table 4 are notable. First, the re- criteria is not inflated if self-report measures of perfor-
search effort is more balanced for type of product than mance are used as criteria. The corrected variance within
it is for type of customer, though historically research subsets averages only slightly less than the total cor-
emphasis has been greater for consumer goods than either rected variance for all of the main categories of predic-
services or industrial goods. More important, tyj>e of tors, except for organizational/environmental factors, and
product seems to affect the relationship between each of the differences in means across subsets are small. Also,
the main determinants of performance, with the excep- the direction of the differences is mixed. Indeed, for some
tion of aptitude, and salespeople's actual performance. predictors, self-report measures of performance produce
The type of product sold has the greatest moderating lower average associations than do more "objective"
effect on the relationship between motivation and per- measures. The general conclusion that emerges from Ta-
formance where the average corrected variance within ble 5 is that there is no basis for generalizations that
subsets is only 38% of what it is for motivation overall. higher correlations can be expected when particular types
The weighted means suggest that motivation bears the of performance measures are used as criteria.
strongest relationship with performance when salespeo-
ple are selling industrial goods. Product type also mod- DISCUSSION
erates somewhat the impact of skill, role, personal fac- The results of our meta-analysis of 116 studies indi-
tors, and organizational/environmental factors, in that cate that the determinants can be ordered in the follow-
the average corrected variance within subsets ranges from ing way in terms of the average size of their association
50 to 60% of the total corrected variance for each of with sales performance: (1) role variables, (2) skill, (3)
these variables. Most of the difference in the size of the motivation, (4) personal factors, (5) aptitude, and (6) or-
associations across subsets seems to be concentrated in ganizational/environmental factors. However, when or-
one product subset for each of these basic determinants. dered according to the amount of the observed variation
Personal factors make the most difference, for example, in correlations across studies that is real variation (i.e.,
when salespeople are selling services, whereas organi- not attributable to sampling error), the determinants rank
zational/environmental factors make the most difference as follows: (1) f>ersonal factors, (2) skill. (3) role vari-
when salespetiple are selling industrial goods and the least ables, (4) aptitude. (5) motivation, and (6) organiza-
difference when they are selling consumer goods. tional/environmental factors. The latter result suggests
that those studies which employ personal factors to pre-
Dependent Measures dict performance have the greatest deviation from what
The "appropriate" way to measure performance is a is true on average in the size of the correlations they
dispute that lingers in the literature. Some authors argue obtain because of situational or other environmental fac-
that much evidence about job performance is flawed be- tors. Another conclusion is that none of the predictors
cause it relies on self-report measures which may have by themselves account for a great amount of the vari-
an upward bias. Others believe that self-report measures ation in performance—less than 10% on average—^though
are fme because even if they are biased there is no reason it could be much higher in any single study. Finally, the
to suspect the amount of bias varies systematically across results indicate that the strength of the relationship be-
salespeople. To investigate whether the results depend tween the major determinants and salespeople's perfor-
on the measure of performance used, the results are re- mance is affected by the type of products salespeople
ported in Table 5 by type of dependent measure. sell. A salesperson's aptitude, for example, has a stronger
114 JOURNAL OF MARKETING RESEARCH, MAY 1985

Table 4
VALUES OF r FOR TYPES OF INDEPENDENT VARIABLES AND PRODUCT TYPES

Product type
Independent variable Consumer goods Industrial goods Services Unknown
Aptitude
Number of r's 238 237 172 173
Simple mean .194 .186 .164 .121
Weighted mean .176 .170 .146 .088
Attenuation-corrected mean .246 .238 .205 .123
Corrected variance (C) .019 .010 .007 .001
Credibility interval for
Weighted mean -.102 < p £ .455 -.032 £ p £ .372 -.017 £ p £ .309 .032 £ p £ .144
Attenuation-corrected mean -.141 £ p S .634 -.041 £ p £ .518 -.020 £ p £ .430 .048 £ p £ .198
CV ratio 1.000
Skill level
Number of r's 48 22 46 62
Simple mean .238 .250 .228 .351
Weighted mean .239 .210 .264 .346
Attenuation-corrected mean .285 .251 .315 .413
Corrected variance (C) .026 .015 .038 0
Credibility interval for
Weighted mean -.081 £ p £ .558 -.033 £ p :S .454 -.128 £ p £ .656 .152 £ p £ .540
Anenuation-corrected mean -.095 £ p £ .664 - .037 £ p £ .540 -.151 £ p £ .781 .175 £ p £ .651
CV ratio .672
Motivation
Number of r's 56 32 18 20
Simple mean .268 .255 .236 .067
Weighted mean .209 .264 .174 .051
Attenuation-corrected mean .294 .372 .245 .072
Corrected variance (C) .002 .006 .004 0
Credibility interval for
Weighted mean .118 £ p£ .300 .112 £ p £ .417 .042 < p £ .306 .012 £ p £ .090
Attenuation-corrected mean .179 £ p £ .408 .170 < p < .574 .066 £ p£ .425 .016 £ p £ .129
CV ratio .379
Role
Number of r's 16 36 7
Simple mean .306 .321 .194
Weighted mean .306 .326 .194
Attenuation-corrected mean .395 .422 .251
Corrected variance (C) .006 .006 .019
Credibility interval for
Weighted mean .146£ p£: .466 .178 £ p <; .495 -.078 £ p £ .467
Attenuation-corrected mean .189 £ p s .601 .231 £ p £; .613 -.101 £ p £ .603
CV ratio .635
Personal factors
Number of r's 166 92 102 47
Simple mean .163 .112 .216 .177
Weighted mean .158 .115 .238 .052
Attenuation-conected mean .288 .209 .434 .095
Corrected variance (C) .011 0 .067 .006
Credibility interval for
Weighted mean -.051 £ p £: .367 .013 £ p <; .217 -.280 £ p £ .755 - .097 £ p £ .201
Attenuation-corrected mean -.094 £ p <; .669 .023 £ p s: .396 -.510 S p £ I.OO -.177 S p £ .366
CV ratio .546
Organizational/environmental factors
Number of r's 4 16 13 18
Simple mean .098 .236 .187 .036
Weighted mean .098 .251 .203 .036
Attenuation-corrected mean .098 .251 .203 .036
Corrected variance (C) 0 .010 0 0
Credibility interval for
Weighted mean .029 £ p <; .166 .051 £ p ^^ .451 .091 s p £ .314 .052 £ p £ .124
Attenuation-corrected mean .029 £ p <• .166 .051 £ p s; .451 .091 - P - .314 .052 £ p £ .124
CV ratio .531
DETERMINANTS OF SALESPERSON PERFORMANCE 115

Table 5
VALUES OF r FOR TYPES OF INDEPENDENT VARIABLES A N D DEPENDENT MEASURES

Dependent variable
Objective company
Self-report Manager and peer Objective company data with control
Independent variable performance ratings data for externalities
Aptitude
Number of r's 61 444 271 44
Simple mean .153 .174 .153 .268
Weighted mean .150 .156 .118 .209
Attenuation-corrected mean .209 .219 .165 .293
Corrected variance (C) .010 .006 .010 .019
Credibility interval fw
Weighted mean -.053 s p s .352 -.004 < p < .317 - .080 ^ p s .316 -.065 s p S .483
Attenuation-corrected mean -.071 < p :e .490 -.002 :5 p S .440 -.110s p< .441 -.087 < p S .673
CV ratio .909
Skill level
Number of r's 1 43 106 28
Simple mean .350 .234 .302 .239
Weighted mean .350 .219 .283 .225
Attenuation-corrected mean .418 .262 .338 .269
Corrected variance (C) 0 .032 .032 .001
Credibility interval for
Weighted mean .250 < p <; .450 -.138 < p :£ .576 -.074 S p S .639 . 162 < p £ .288
Attenuation-corrected mean .287 s p s: .549 -.163 < p!5 .687 -.086 :£ p s .761 .202 :£ p S .336
CV ratio .968
Motivation
Number of r's 18 32 65 11
Simple mean .236 .253 .195 .338
Weighted mean .208 .213 .146 .287
Attenuation-corrected mean .293 .299 .205 .404
Corrected variance (C) .007 0 .009 .006
Credibility interval for
Weighted mean .046£ps .370 .148 s p ^ .278 -.042 s p < .374 .132 < p :£ .442
Attenuation-corrected mean .073 :£ p <: .513 .!91 S p £ .408 -.057 s p s .467 .201 s p s .607
CV ratio .756
Role
Number of r's 44 15
Simple mean .315 .264
Weighted mean .321 .247
Attenuation-corrected mean .415 .319
Corrected variance (C) .006 .018
Credibility interval for
Weighted mean . 168 :< p s: .474 -.021 < p s .515
Attenuation-correeted mean .217 :£ p s: .612 - . 0 2 7 £ p £ .665
CV ratio .798
Personal factors
Number of r's 83 132 170 22
Simple mean .129 .157 .199 .112
Weighted mean .152 .069 .225 .085
Attenuation-corrected mean .277 .125 .410 .155
Corrected variance (C) .012 .004 .060 0
Credibility interval far
Weighted mean -.069 < p s .373 -.064 s p < .201 -.265 s p s .714 .024 :S p S .146
Attenuation-corrected mean -.126 < p < .681 -.117 :S p :S .368 -.483 s p s I.OO .044 < p < .266
CV ratio .725
Organizational/environmental factors
Number of r's 14 13 18
Simple mean .160 .229 .036
Weighted mean .136 .199 .036
Attenuation-corTected mean .136 .199 .036
Corrected variance (C) .003 .004 0
Credibility interval for
Weighted mean .028 < p <: .243 .071 s p s .326 - . 0 5 2 S p £ .124
Attenuation-corrected mean .028 :£ p :£ .243 .071 < p s .326 - .052 s p < .124
CV ratio .393
116 JOURNAL OF AURKETING RESEARCH, MAY 1985

impact on that individual's performance when he or she research, then, is that studies should include salespeople
is selling products rather than services. who leave the organization, particularly those who leave
after only a short time on the job.
Limitations Another important issue related to performance is
The preceding analysis has several limitations. First, whether the individual studies measured performance
the analysis is based on "static" data in the sense that correctly. Industrial psychologists have begun to distin-
the correlations do not reflect salesperson-customer in- guish among behavior, performance, and effectiveness
teractions in specific sales situations. Some authors ar- when discussing the job accomplishments of individuals.
gue that salespeople's behaviors must be tailored to in- Behavior refers to what the job occupant does—that is,
dividual customers and that the nature of the salesperson's the tasks on which the individual expends effort while
actions in specific selling situations ultimately deter- working. Performance is behavior evaluated in terms of
mines how successful the salesperson will be (Weitz its contributions to the goals of the organization. Effec-
1979). None of the studies examined focused on the dy- tiveness refers to some summary index of organizational
namics that make up individual sales transactions. outcomes for which an individual is at least partly re-
A second limitation is related to the conceptualization sponsible, such as sales volume or market share. The
and measurement of the variables. A variety of variables crucial distinction between performance and effective-
are captured under each major grouping. The aptitude ness is that the latter does not refer to behavior directly,
category, for example, reflects a number of personality but rather is a function of additional factors not under
characteristics as well as some other general ability char- the individual salesperson's control—such as the poli-
acteristics. Further, the quality of the measures capturing cies of top management, the sales potential in a territory,
the constructs was not investigated in many studies. Even and competitive actions.
when psychometric properties were investigated, only a The distinction between outcomes and desirable be-
few of the indicants of quality were typically reported, haviors is an important one, particularly in the selling
most generally a reliability coefficient. TTiough we made context. A salesperson's efforts or behaviors are much
use of the reliability coefficient to adjust the results for more controllable than the results produced by those ef-
attenuation correction, we recognize that reliability is only forts. If a representative's sales fall short of quota, the
one of several indicants of the quality of a measure. It problem may lie with the person, the quota, or perhaps
addresses random enors in a measure but does not get a change in the environment. If the number of calls a
at method variance, false reporting, and other systematic salesperson makes falls short of the target, however, the
errors. Though the grouping of variables into broad cat- problem lies much more directly with the individual. In-
egories was necessary and though the analysis did at- dustrial psychologists are increasingly arguing that job
tempt to incorporate the measure quality information that occupants should be judged solely on those phases of
was available, readers must keep these measurement their jobs over which they exercise control, and should
limitations in mind when interpreting the results. not be held responsible for results beyond their control.
A third limitation is related to the selling situations Unfortunately, most performance appraisal systems^—-and
reflected in the studies. The selling situations and sample most of the performance measures used in the studies
of firms were broader for some variables than for others analyzed here—emphasize outcomes rather than behav-
simply because of the number of studies that used a par- iors, a condition one of the leading writers in personnel
ticular determinant. has labeled the "Achilles heel of our profession" (Hene-
A fourth, and probably the most serious, limitation man 1975, p. 91).
pertains to performance and its measurement. Salespeo- In sum, determining the appropriate measure of per-
ple who were very poor performers are probably under- formance in a given situation is a complex issue. We
represented because most of the studies were cross-sec- took the measures as reported at face value. We did not
tional. Salespeople who had quit or had been fired because quesiton whether the measure used was the most appro-
of poor performance were not included. Though most of priate one for the situation; to do so would have required
the samples of salespeople probably did include poor insight into the various selling situations that simply was
performers who may have been fired or quit after the not available from the articles. Though we did explore
study was completed, the number of such people in the the issue of whether type of performance measure used
sample depends greatly on the companies' hiring and had any impact on the results, the fact remains that the
turnover rates. It also seems likely that new recruits, who correlations might be low simply because a number of
may include a disproportionate number of poor perform- inappropriate "performance measures" were used in the
ers, were underrepresented because of the lack of infor- first place.
mation about their sales performance. These conjectures
suggest that the range of performance observed in the Implications
studies may be artificially narrow and skewed upward. Our fmdings suggest several implications for sales
This, in turn, would help to account for the low corre- managers and researchers. The fact that no single deter-
lations between the various determinants and variations minant can explain a very large proportion of the vari-
in sales performance. One obvious implication for future ation in sales performance has two obvious implications.
DETERMINANTS OF SALESPERSON PERFORMANCE 117

First, managers should be wary of any personnel ad- dom that the determinants of sales performance are job-
ministrator or consultant who contends that measurement specific. Indeed, this point provides another possible ex-
of a single factor—or even several factors within a sin- planation for the observation made before that "influ-
gle category of predictors (e.g., personality tests mea- enceable" characteristics do a better job of explaining
suring several traits)—can accurately predict salespeo- variations in performance than "enduring" personal
ple's future sales performance. Second, theoretical models characteristics; infiuenceable characteristics are more likely
which hypothesize multiple determinants—and cate- to be job-specific and also are more likely to be mea-
gories of determinants—of sales performance are prob- sured with instruments that reflect the unique require-
ably on the right track. ments of the job. Researchers in sales management must
The multiple-determinant perspective may also ac- pay more attention to how sales jobs differ from one
count for the facts that (1) the number of correlations another before they will be able to explain a more sub-
per study seems to have increased rather than decreased stantial proportion of the variance in sales performance
and (2) the size of the individual correlations did not across individuals occupying different .sales jobs.
change as much as one might have supposed they would We suspect the last point will not be a revelation for
as theories about the determinants of salesperson per- most sales managers; many of them already have a good
formance improved over time. As researchers discov- subjective feel for the specific requirements of the job
ered that performance is influenced by multiple factors, they supervise, and for the kinds of abilities, skills, and
their theories and subsequent research began to reflect motivation a recruit needs to do the job well. Many of
this fact. Even though (W/v/dutt/correlation coefficients these perceptions seem to be subjective in that managers
did not improve over time, the amount of total variance often rely more heavily on the impressions they gain in
in sales performance explained by the predictors is greater personal interviews with job candidates than on objec-
in the more recent studies that have a multiple-deter- tive tests of personal traits or skills. It is also possible,
minant framework. though, that many companies are ahead of academic re-
The findings about the relative ability of the various searchers in developing objective measures that accu-
categories of determinants to explain variations in sales rately predict their salespeople's performance, because
performance suggest that "enduring" personal charac- they can tailor such measures to the limited range of jobs
teristics such as aptitude variables and personal/physical and job requirements germane to their own flnn. Indeed,
traits do have some relationship to performance, but not if there is a "file drawer" problem inherent in academic
as much as those characteristics which are "influenca- researchers* inability to gain access to proprietary stud-
bie" through increased training and experience or more ies, it may be exactly the opposite of the problem usually
effective company policies and procedures (e.g., skill encountered by social science researchers. Instead of
levels, role perceptions, and motivation). The implica- published studies having more positive results than those
tion is that from a manager's point of view, whom one hidden away in people's files, in this case the hidden
recruits is important, but probably not as important as company studies may be more positive. The reason is
what one does with the recruits—and to them—after they that in the published studies the researchers often have
have been hired. The fmding also supports the wisdom tried to predict performance across salespeople in dif-
of the preference many sales managers have for hiring ferent kinds of sales jobs and in different firms and in-
people with experience in the same or similar industries, dustries using the same set of performance criteria and
because such people are more likely to have developed the same set of predictor variables and measures.
the necessary skills and may be less likely to suffer from
inaccurate or ambiguous role perceptions. REFERENCES
The finding that self-report measures of performance
do not lead to significantly higher correlations than other Churchill, Gilbert A , Jr.. Neil M. Ford, and Orville C. Walker,
"more objective" performance measures has an obvious Jr. (1985), Sales Force Management, rev. ed. Homewood,
IL.: Richard D. Irwin, Inc.
and important implication for future research. Though Glass, Gene U. (1976), "Primary. Secondary, and Meta-Anal-
other measures may be preferred on conceptual grounds ysis of Research," Educational Researchers. 5. 3-8.
(e.g., to reduce method variance when a number of pa- . Barry McGaw, and Mary Lee Smith (1981), Meta-
per-and-pencil instruments are used), self-report mea- Analysis in Social Research. Beverly Hills, CA.: Sage Pub-
sures do not create an upward bias in the findings, and lications.
the tendency that is often observed to discount flndings Hcneman. Herbert. Jr. (1975), "Research Roundup." The Per-
simply because of the "natural" upward bias produced sonnel Administrator. 91.
by self-report measures of performance is unfounded. Hunter. John E., Frank L. Schmidt, and Gregg B. Jackson
However, the use of multiple indicators of different types (1982), Meta-Analysis: Cumulating Research Findings Across
to assess the performance construct is desirable. Studies. Beverly Hills. CA.: Sage Publications.
Oschrin. Elsie (1918), "Vocational Tests for Retail Sales-
Perhaps the most important finding in our analysis is women," 7ourna/o/App/i>t^/*iTcfto/o^_v. 2 (June). 148-55.
the observation that the type of product sold influences Schmidt, F. L. and J. E. Hunter (1977),' "Development of a
the correlations between the various predictors and per- General Solution to the Problem of Validity Generaliza-
formance. This finding supports the conventional wis- tion," Journal of Applied Psychology. 62, 529-40,
118 JOURNAL OF MARKETING RESEARCH, MAY 1985

Walker. Orville C., Jr., Gilbert A. Churchill, Jr., and Neil M. A. Churehill, Jr., eds. Eugene: College of Business Admin-
Ford (1977), "Motivation and Performanee in Industrial istration, University of Oregon, 10-75.
Selling: Present Knowledge and Needed Researeh," Journal Weitz, Barton A. (1979), "A Critical Review of Personal Sell-
of Marketing Research, 14 (May), 156-68. ing Research: The Need for Contingency Approaches," in
, , and (1979), "Where Do We Go From Critical Issues in Sales Management: State-of-the-Art and
Here? Selected Conceptual and Empirical Issues Concerning Future Research Needs, Gerald Albaum and Gilbert A.
the Motivation and Performance of the Industrial Sales- Churehill, Jr.. eds. Eugene: College of Business Adminis-
force," in Critical Issues in Sales Management: State-of-the- tration, University of Oregon, 76-126.
Art and Future Research Needs, Gerald Albaum and Gilbert

JMR REPRINT POLICY AND PROCEDURE


All reprints are sold according to the following price schedule, in minimum orders of
50 copies.

50 copies with covers $ 75.00


100 copies with covers 125.00
(multiples of 100 may be ordered)

There is NO RETURN and NO EXCHANGE on reprints.


Under the "fair use" provision of the new copyright law taking effect January 1978,
anyone may make a photocopy of a copyrighted article for his or her own use without
seeking permission. Also, a single copy reprint or an order of less than 50 copies may
be obtained from University Microfilms International, 300 N. Zeeb Road, Ann Arbor,
MI 48106. Articles are priced prepaid at $6.00 plus $1.00 for each additional copy of
the same article. Complete issues are obtainable at IOC per page, minimum order $10.00.
To obtain permission to reproduce one's own reprints in quantity, please contact the
Permissions Department, American Marketing Association. 250 S. Wacker Drive, Chicago.
IL 60606.

Das könnte Ihnen auch gefallen