Research Paper No. 1421 Relative Explanatory Power of Agency Theory and Transaction Cost Analysis in German Salesforces Manfred Krafft Rajiv Lal Sonke Albers December 1996 Graduate School of Business Stanford University Manfred Krafft is Assistant Professor of Marketing, University of Kiel, Germany. Rajiv Lal is Professor of Marketing and Management Science, Graduate School of Business, Stanford University. Sonke Albers is Professor of Marketing and Management Science, University of Kiel, Germany.
43
Embed
Relative explanatory power of agency theory and transaction cost analysis in German salesforces
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
ResearchPaperNo. 1421
RelativeExplanatory Powerof AgencyTheoryand Transaction CostAnalysis in German Salesforces
ManfredKrafftRajiv Lal
SonkeAlbers
December1996
GraduateSchool of BusinessStanfordUniversity
Manfred Krafft is Assistant Professorof Marketing, University of Kiel,Germany.
Rajiv Lal is Professorof Marketing and ManagementScience,GraduateSchoolof Business,StanfordUniversity.
SonkeAlbers is Professorof Marketing andManagementScience,Universityof Kiel, Germany.
RelativeExplanatory Powerof Agency Theory and
TransactionCost Analysis in German Salesforces
December1996
Manfred Krafft*, Rajiv Lal** and Sonke Albers***
* AssistantProfessorof Marketing, University of Kid, Germany.
** Professorof Marketing and ManagementScience,GraduateSchool of Business,
Stanford University.
~ Professorof Marketingand ManagementScience,University of Kiel, Germany.
Germany again gives rise to the possibility of finding a significant impact of this
variablein our data.
4. RESEARCH DESIGN
‘To test the hypothesesdescrihedin the previoussection,we maileda questionmiaire
to 1099 chiefsalesexecutivesof Germansalesforces,noting the fact that firms mostly
useonly one compensationschemefor the whole salesforce..Ini this way, we seek to
capturetime perceptionsof chief salesexecutiveswho are involved in the designof
salesforcesand their incentiveplans. The questionnairewas developedarid pretested
throught,he direct help of 10 top salesexecutives.Severalefforts were madeto solicit
time responseof the chief salesexecutives. The initial mailing wa.s followed imp by a
secondmailingafter 4-6 weeks. Furtherefforts to improvetheresponserateweremade
througha contactby fax and promiseof sharingtheresultsof ourstudy. The survey
wascompleteabout twelve weeksafter the first mailing and resultedin a response
rate of about 25 %. Of the 270 respondingfirms, 61 firms usedrepresentativesonly,
173 useda direct salesforceand 36 useda hybrid organization. ‘~ A comparisonof
our samplewith other Germanstudiesis reportedin Table 2 and a comparisonof
the different characteristicsof the empirical studiesinvestigatingthe direct vs. rep
decisionand the designof compensationplans is presentedin Table 3.
4Thereasonsfor the existenceof a hybrid organizationseemto be rather arbitraryand includesalesforcesin transition, vacantterritoriesbeing temporarilyassignedto repsand reps being usedas a threat to companysalespeopleto improve their performance.
12
4. 1. 11’!easures
The two dependentvariablesin ourstudy are, thekind of salesforce(direct salespeo-
ple or reps)and the proportion of salaryto total pay in caseof the direct salesforce.
(1992) and Lal, Outlandand Staehin (1994). The impact of factorssuch as timnme de-
voted to selling, easeof measuringinputs and inadequacyof output as measureof
performanceis also consistentwith the effects reportedin John and Weitz (1989).’~
Similarly, theinsignificanteffect of uncertaintyobservedin ourdatais consistentwith
the findings in othercross-sectionalstudiessuchasCoughlanand Narasinmhaim(1992)
and Johnand Weitz (1989). In contrast,the effect of easein replacingsalespeople
in our datadoes not support the findings in John and Weitz (1989). In thisway
we areableto documentthat mostof the findings reportedfor Americansalesforces
generalizeto Germany.
While one canoften provide ex-postrationalizationsfor discorifirming results, it
is importantto note that oneof the more importanthypothesesof AT with regards
5While Johnand Weitz (1989) reporta direct andsignificant impactof this variable,the effectof this variableis significant only throughits interactionwith time devotedto selling in our data.
18
to uncertaintYhuts riot found significantsupport in any empirical study using cross—
sectionaldata. Tins is particularly interestimmgsince other stuidlies usimig imidividual
level (lata (Lal, OutlandandStaelin 1994 and .Josephand Kalwarmi 1995) have found
support for this hypothesis. To investigatethis issue, we interviewed some of the
executivesin oursample.Theseinterviewsleadus to believethat executivesreact to
uncertaintysomewhatdifferently thanassumedin AT basedanalyses.1mm particular,
since uncertainty is beyond the control of the firm, the managementtreats it as
sonmmethirmgthat the salesforceshould beable to dealwith, amid if not, salesexecutives
prefer to find salespeoplewho can. In order to he ableto do so, a sufficient increase
imi uncertainty is accompaniedby higher commissions.Salespeoplewho are unable
to dealwith uncertaintywill find suchcontractsto be unattractivearid thereforenot
want to stay with time company; while those who can, get rewardedthrough a plami
that consistsof a higherfractionofincentivebased!cOmnj)ensation.Hence,we conmcli.mde
that while the AT literatureon salesforcecompensationplans(leals with umicertainty
only through the design of compensationplans, managersin our sampleseem to
dealwith uncertaintyby offering the right incentivesto signal the higheruncertainty
in the selling environmentand attract an appropriatelyskihledi salesforce. This is
especiallytrue of Germanywhere the legal restrictionsmake it extremelydifficult
to fire salespeople,as a consequencethe hiring decisionis of much more importance
than in the U.S.A.
Among themoreuniqueaspectsof ourstudy is the inclusionof theconstructtime
devotedto selling and its interactionwith easeof measuringinputs. We find that
while the hypothesiswith respectto time devotedto selling is supportedin ourdata,
the hypothesiswith respectto easeof measuringinputs is also supported,although
indirectly through the interactionwith time devotedto selling. In other words, we
find that for any level of time devotedto selling, the proportionof salaryincreases
19
with um(:m’ease in easeof nmeasurimmginputs, as preilicteni. On time other hanni, the
relationshipbetweenthe fraction of salary ann! time devoted to selling (iCpCuIds on
the easeof mneasurringinputs. III particular,we find that if it, is sufficiently difficult
to measureinputs, theeffect on conmpensationplansis dominatedby the direct effect
of time time devotedto selling; and includesa higher percentageof incentiveswith
immcreasimmgtime devotedto selhimmg. However, if inputs are rather easy to measure,
therm the net effect on the cornipemmsationmplan is (line to easeof measuringinputs andh
resultsin a higherpercentageof salary with increasingeaseof measuringinpnmts for
any level of time (levoted to selling. The inclusion of thesevariablesseemsto he
important given that. the ~3-coefficientsfor thesevariablesare relatively high. )
Finally, ourstudy providesa strict testof the competinghypotheseswith respect
to input and output mnmeasuresin AT and TCA frameworks. Recall that AT P1~Ol)O5e5
that when it is easy to measureinputs, compensationplans should consist mainly
of salary;however if immputs are difficult, to measure,the conipenisatiomiilanm should
increasinglyemphasizeincentiveswith increasein adequ.iacyof outputsas a measure
of performance.We thereforetest thesepredictiomis by specifyingeaseof measuring
inputs as a main effect and include an interactionterm that picks up the effect of
inadequacyof output measureswhen inputs are difficult to measure(see AT model
in Table Sc). In contrast,TCA suggeststhat the designof compensationplan should
focus more on the outputs. However, if output is an inadequatemeasureof per-
formance,TCA hypothesizesthat compensationplans should include an increasing
fraction of salarywith increasingeasein measuringinputs. We testthesepredictions
by specifyinga model where adequacyof output measuresis a main effect and the
interactiontermcapturesthe effect of useof measuringinputswhenoptputmeasures
areinadequate.Finally, thesetwo constrainedmodelscan be comparedwith a model
wherewe include easeof measuringinputs and adequacyof outputsas a measureof
20
J)erformnanceas ummain effects in addition to time interaction ternim (hetweemi thesetwo
main effects).6 The resultsare presentedin Table Sc. It is observedthat while the
interactionbetweendifficulty in measuringinputs andoutputasa measureof perfor-
manceis highly significantand in the hypothesizeddirection in the AT model, the
interactionbetweenoutput beinga bad measureof performanceand easeof measur-
ing inputsis insignificant in theTCA model. Thusour resultslendsupport to the AT
framework. Furthermore,the resultsof an F-testindicatethat thereis no loss in the
explanatorypowerof the AT and TCA models(constrainedmodels)whencomparedl
with that of the unconstrainedmodel.
In summary,we have provided the first empirical investigationof both AT and
TCA basedconstructsthrougha single diata source. We providea testof the com-
peting hypotheseswith respectto input arid outputfactors amid find support for the
predictionsbasedon Agency Theory. Furthermore,while manyfindings in the com-
pensationliteratureareconfirmedand thereforegeneralizeto Germany,we also offer
somenew findings in theform of rejectingsomehypothesesandsupportingnew ones.
The main conclusionwith respectto rejecting thesehypothesesis that someof time
operationahizationsmaynot bevalid; but moreimportantly, our findings suggestthat
an importantconsiderationhasbeenleft out in modelingcompensationplans. While
the extant literature investigatesthe impact of changesin the environmenton the
designof compensationplans holding everythingelse constant(with the exception
of Lal and Staelin1986), salesexecutivesseemto dealwith different selling environ-
mentsthroughnot only the designof compensationplansbut alsoby attractingthe
right type of salespeople.In otherwords, the practiceof salesmanagementsuggests
an interactionbetweendesignof compensationplansandrecruitment/selectionwhile
the theoreticalliterature treatstheseissuesindependently.
6We aregrateful to ProfessorRichardStaelin for his help in developingthe appropriatetestforthesehypotheses.
21
5. EXPLANATORY POWER OF TCA AND A~ENCYTHEORY
Time more recent theoreticaland the empirical salesforcemaumagemnentliterature
has benmefitted inmmensely from two different theoreticaldevelopmnemitsimm the ceo—
imomnics literature, namely TransactionCost Analysis andi Agency Theory. While
sonicauthorshaveusedAgencyTheory as time fommdat.ionifor their investigation,oth-
Weinberg,CharlesB. (1975), “An Optimal CommissionPlan for Salesmen’sControl
over Price”, ManagementScience,21, 937- 943.
Williamson, Oliver E. (1975), Markets and Hierarchies: Analysis and Anti-Trust
Implications, New York: The FreePress.
Williamson, Oliver E. (1981), ~‘TheEconomicsof Organization:The Transaction
Cost Approach”,AmericanJournal of Sociology,87, 548 - 577.
Williamson, Oliver E. (1985), The EconomicInstitutionsof Capitalism, New York:
The Free Press.
29
APPENDIX I: Description of DependentVariables
DependentVariable, Operationatization, Re,nark.s
Kind of salesforce(direct salespeopleor manufacturer representatives)How many salespeople(employed and independentrepresentatives)are primarily working for your salesorganization (without satesmanagement)?
salespeopleNumber of independentsalesrepresentativesworking in your salesOrganization(A comparisonof thefirst and secondquestionreveals, whethera company uses representatives only, adirect salesJt~rce01(1 hybrid)
Proportion of salary to total pay *)How niany of your salespeopleare compensatedby the following plans ?• Straight Salary %• Straight Commission %• A Combination Plan, such as
- Salary plus Comniission(SALCOM) %- Commissionand Drawing Account %- Salary plus Bonus (SALBON) %- Commissionplus Bonus %- Salary plus Commissionplus Bonus(SACOBO) %
If you are offering combination plans, what level doesthe proportion of variable inconie to total pay typically reach? (VARPERC)
(The variablepropornon wascomputedvia the followingformula: (VARPERC*(( SALCOM÷SALBON+SACOBO)/100)).The dependemitvariable is th,e,i thecomplementto 100%).
*) There is nosinglecompany in the ‘direct salesforce’subsamplethat uses~commissiouanddrawingaccount’or ‘commissionplus bonus’ plans.Thereforetheformula hasbeen simplified and only containscombination plans with somesort of salary.
Salespeoplehardly replaceable(a = .64)When a salespersonquits, wecan easily hire a good replacementOur products can be explainedeasily toward our accounts,such that salespeoplecould sell all of our products already after a short trainingi rile
What are the total costsfor training new salespersons(costsof training, compensationof the salesperson)?about DM per salespersonHow much timeof the training is neededfor
% customerspecific know-how,% product specific know-how,% company specific know-how,% know-how of general selling techniques,% other’ ?
Ouput is bad performancemeasure(a = .72)Using outcome measures(i.e. overall sales)how preciselydo they represent the actual effort?How preciselycan you infer the actual individual selling effort from the outcomemeasures?How many factors beyond the control of your salespersonsdo influence the selling outcome?
Easeof measuring inputs (a = .69)It is just not possibleto superviseour salespeoplecloselyIt is difficult to evaluatehow much effort any individual in this group really puts into his jobThesesalespeopletravel somuch that closesupervision is impossibleIt is easy for thesesalespeopleto turn in falsified salescall reports if they want to
Source
John/Weitz(1989)new
new
new
new
similar to Anderson (1985)similar to Anderson (1985)
Company(frequency of transaction)Sizeof the salesforce
How many salespeople(employedand independentrepresentatives)are primarily working for your salesorganization (without salesmanagement)? salespeople .
John/Weitz (1989)
Time devotedto sellingOf their total working time your salespeopleuse % for salescalls new
Travel requirementsHow many nights do your salespersonsspend in a hotel in atypical month? about nights Anderson (1985)
Environment
Environmental uncertainty (a = .64)How often do you or one of your competitors introduce competitive new products’? (seldom - often)How fast doesthe environment of yourcompanychange (i.e. technology,intensity of competition’?(slowly - fast)How strong do you perceivethe intensity of competition in your market (segment)‘?(low - high)
Salesvolatility (a = .73)How much did the market volume of your industry vary on averageover the last five years?How much did overall salesof your entire salesforcevary on averageover the last five years’!How much did your actual overall salesdiffer from your expectedoverall sales’?
Diversification of riskOne salespersonis on averageresponsiblefor about accounts
similar to Anderson (1985)similar to Anderson (1985)similar to Anderson(1985)
How many of your salespeopledid prior to the job in your company have• less than three years %• threeto sevenyears %• more than sevenyears % experiencein selling?
How many of salespeoplehave worked in your salesorganization for% lessthan one year,% one to less than three years,% three to lessthanfive years,% five to less than, ten years, resp.% ten yearsor longer
How many of your salespeoplehave receivedtheir highest degree from% Hauptschule(9 yea,-shig/ischiool),
% Mittlere Reife (10yearshighschool),% Abitur (13yearsspecialtrack of highschool),% college-degree,% university-degree’?
Education:(weightedindex)
new
CoughlanfNarasimhan (1992)
Coughlan/Narasimhan (1992)
Coughlan/Narasimhan(1992)Averageindustry income: What is the annual total pay (salary + variable income), that your industry offers an averagesalesperson’?________ DM
Table La.’ Hypothesesandfindingson the vertical integrationissue
- Minimum utility• Tenure• Education• Averageindustry income
—
+
+
+
+~—]: The higher the variable thehigher[lower] is theprobabilityof choosingemployedsalespeople*: significant at the 10%level **: significant at the5% level
***: significant at the 1% level n.s.: not significant
Table lb.’ Hypotheses(1/1(1Findings Regam’dingSaIa,’y asa Percemitageof’ Total Compensation
- Transaction specific assets- In vestment in training- Output is a had performancemeasure- Easeof measuring inputs * time devoted to selling- Difficultiness of measuring inputs
+
+
+
—
+
+
—
— * a)
n.s.n.s.
+ * V
+ * f.
V
f.
Company (frequencyoftransaction)- Sizeof the salesforce- Time devoted to selling- Travel requirements
+
—
—
÷
— ** f.— *** V
— * f fV
Sellingenvironment- Uncertainty
• Environmentaluncertainty• Salesvolatility• Quota achievement• Diversification of risk• Uncertainty* Replaceability
0000
÷
+
÷+
—
n.s.
+ * a)
+ l”I”l’ V
+ ~,“ V
— * f.
+ *** V mixed
~./
Salesperson- Risk aversion- Effectiveness
• Sellingexperience• Age
• Salesper call- Minimum utility requirement
• Tenure• Education• Average industry income
+
—
—
—
+
+÷
.
— *** V
,
— ~ b)
, n.s.
+ ** f.
+ *** V
+ *** V— ~ b)
+ ‘~~‘~‘ V V
f
V
V
R2 (adjusted) I ‘ .22 .68 LR: .15 not reported
+~—j: The higher the influence of the variable the higher (smaller) should be the proportion of salary of the employedsalespeople***:significantat the 1% level**: significant at the 5% level *: significant at the IO% level n.s.: not significant V hypothesis supported f: hypothesis getsno supporta)~Variables are multicolhinear. If included separatelyeach variable is significant(p < .10). When both are included their combined effects are significant.b): Variables are multicollinear. Each construct is significant whenit alone appearsin the model (p < .01) with ‘selling experience’ in the right and ‘average
industiy income’ in the wrong direction. Both are insignificant whe1i introduced into the model at the sametime.
Table2.’ Comparisonofthesample(direct salespeopleonly) with commercialcompensationstudies
+~—]: The higher the variable the higher [lower] is the probability of choosingemployedsalespeople*: significant at the 10% level **: significant at the 5% level ***: significant at the 1% leveln.s.: not significant V: hypothesissupported f: hypothesisnot supported
Table 41,: Comparisonof’Anderson‘.v andourfindings on the vertical integration issue
+~—]: The higherthe variable the higher [lower] is the probability of choosingemployedsalespeople
*: significant at the 10% level**: significant at the 5% level ***: significant at the 1% leveln.s.: not significant V: hypothesissupported f: hypothesisnot supported
)
)
)
)
Variable
Hypotheses 13 coefficients
TCA AT (Logit regression)
one-tailed
p-values
Empirical
findings
Market (market failure)
- Salespeopleeasily replaceable
- Transactionspecificassets
- Difficultiness in evaluating performance
• Output is bad performance measure• Input easy* time devotedto selling• Easeof measuring inputs
—
+
+
+
+
+
+
0.452
0.022
, 0.1590.4080.003
0.094 *
0.406 n.s.
0.027 ‘~
0.003 ‘~‘~‘~‘
0.485 n.s.
f
•
VV
Company (frequency oftransaction)- Sizeof the salesforce
Note: To investigate if uncertainty and the interaction term are not significant due to multicollinearity,we ran regressionswith only one of thesethree variables (uncertainty, replaceability, interaction term) andfind that only replaceability is significant, albeit at a higher significance level (lower p-value).
Table5a: Findingson .ralary as’ a percentageof total compensation
n.s.:
The higher the variable the higher [lower] the percentageof salary should besignificant at the 10% level **: significant at the 5% level ***: significant at the 1% levelnot significant v’: hypothesissupported f: hypothesisnot supported
- Transaction specificassets- Investment in training- Output is a bad performance measure- Ease of measuring inputs * time devotedto selling- Difficultinessof measuringinputs
—
+
++
—
+
+
—
— * a)n.s.n.s.
+ * V
+ * f.
+ *
n.s.
+ **
+ ~
n.s.Company (frequencyof transaction)-Sizeofthesalesforce- Time devoted to selling- Travel requirements
- Minimum utility requirement• Tenure• Education• Average industry income
+
—
—
—
++±
.
— *** V
.
— b)
n.s.+ ** f.
+ *** V
+ ~— ~ b)
+ *** V
— *
.
+ ~
+ ~— *
R2 (adjusted) .22 .68 LR: .15 not reported .28
+~—]: The higher the influence of the variable the higher (smaller) should be the proportion of salary of the employedsalespeople***:significantat the 1% level**: significant at the 5%level *: significant at the 10%level n.s.: not significant V hypothesissupported f: hypothesisgetsno supporta): Variables are multicollinear. If included separately eachvariable is significant (p < .10). When both are included their combined effects are significant.b). Variables are multicollinear. Each construct is significant when it alone appeal’s in the model (p < .01) with ‘selling experience’in the right and ‘average
industry income’ in the wrong direction. Both areinsignificant whenintroducedinto themodel at the sametime.
- Minimum utility• Tenure• Education• Averageindustry income
—0.124 *
0.261 ~
0.272 ~
— 0.087
—0.118 *
0.254 ***
0.269 ‘~‘~
— 0.080
—0.123 *
0.261 ~0.272 ~‘K
— 0.086
Model Statistics:R2
adjustedR2
F
0.3720.2814.072
0.3680.2834.314
0.3720.2874.383
Table6a: Explanatorypower of TCA andAT spec~ficvariablesfor the decision to employdirectsale~forcevs. independentrepresentatives
Total modelTotal model withoutAT-specific variables
Total model withoutTCA-~pecificvariables
Numberof variables 14, 9 * 11 ~
—2Log Likelihood(d.f./significance)
ModelX2
(d.f. / significance)
85.040(151/1.000)
68.762(14 / 0.000)
93.265(156/ 1.000)
60.537(9 / 0.000)
107.514(154/0.998)
46.288(11 / 0.000)
Improvementx2-testofthe —2LL-changeagainstthetotal model if theory-specificvariablesareremoved
Improvementx2: 8.225Significance 0.1443
Improvement x2: 22.474Significance 0.0001
*: The five variables not included are ‘Time devoted to selling’, ‘Selling experience’, ‘Tenure’,‘Education’, and ‘Average industry income’
t: The three variables not included are ‘Salespeopleeasily replaceable’, ‘Transaction specificassets’,and ‘Size of the salesforce’
Totalmodel
Total model withoutAT-specific variables
Total model withoutTCA-specific variables
Number of variables 15 9 * 11
R2
adjusted R20.3630.277
0.1470.082
0.3410.278
Partial F-test of the R2 changeagainst the total model if theory-specific variables are removed
F-to-change: 6.268Significanceof F-to-change
<0.000
F-to-change: 0.951Significanceof F-to-change
0.438
*: The six variables not included are‘Input easy* time devotedto selling’, ‘Time devotedtoselling’, ‘Selling experience’,‘Tenure’, ‘Education’,and ‘Average industry income’
t: The four variables not included are ‘Salespeopleeasily replaceable’, ‘Transaction specificassets’,‘Size of the salesforce’,and ‘Uncertainty * Replaceability’
)
)
Table6b: Explanatorypower of TCA and AT spec~flcvariablesfor salary as a percentageof totalcompensation