The authors describe a cammunicatian model for measuring the
flectkoness of advertising, Th;s
.?, :;t-~etcribes the process in
terms of a series of steps.
12
A model for predictive
measurements of
advertising effectiveness
Roberf J. Laviclge
and Gary A. Steiner
,,..
What arc the functions of advertising? Obviously the ultimate function is
to help produce sales. But all advertising is not, should not, and cannot be
designed to produce immediate purchases on the part of all who are cxposccl to it. lmmedi~te sales results (even if measurable) are, at best, an in. . . .,
complete criterion of advertising effectiveness.
In other words, the effects of much advertising are “long-term.” This
is sometimes taken to imply that all one can really do is wait and see—ultimately the campaign will or will not produce.
,l-lowever, if something is to happen in the long run, something must
,,
bc happening in the short rim, .something thrst will ultimately Icad to ?
5
cvcntua] sales results. And this process must be measured in order to pro<
vidc anything approaching a comprehensiv~ ~hti~~f the effectiveness J
( of the advertising.
<P Ultimate consumers normally do not switch from disinterested indi-’>
;) Vlduals to Convinced Pllrchascrs in one instantaneous stcP” Rather? theY
approach the ultimate purchase through a process or series of steps in
‘, which the actual purchase is but the final threshold.
/
...4....,,,,., >, ~
k? “:
Cn
‘1
“
+
Reprinted from the )ournal of M~rketing, national quarterly publication” of the American Marketing Ass(~ciotimr (October 1961), pp. 59-62.
(
137
I
138
Applications in marketing reseorch
seven steps
.“ >..’$s~?k+
Advertising may be thought of as a force which must move people
up a series of steps:
Near the bottom of the steps stand potential purchasers who are
completely unuwure of the existence of the product or service in
question.
Closer to purchasing, but still a long way from the cash register,
are those who are merely ~wure of its existence.
Up a step are prospects who kmw what the product ha to ofler.
Still closer to purchasing are those who have favorable attitudes
toward the product—those who like the product.
Those whose favorable attitudes have developed to the point of
prejererrce over all other possibilities are up still another step.
Even closer to purchasing are consumers who couple preference
with a desire to buy and the conviction that the purchase would bc
wise.
Finally, of course, is the step which translates this attitude into
actual Purchu.se.
Research to evaluate the effectiveness of advertisements can be designed to provide measures of movement on such a flight of steps.
The various steps are not necessarily equidistant. In some instances
the “distance” from awareness to preference may be very slight, while the
distance from preference to purchase is extremely large. In other cases, the
reverse may be true. Furthermore, a potential purchaser sometimes may
move up several steps simultaneously.
Consider the following hypotheses. The greater the psychological
and/or economic commitment involved in the purchase of a particular
product, the longer it will take to bring consumers up these steps, and tbc
more important the individual steps will be. Contrariwise, the less serious
the commitment, the more likely it is that some consumers will go almost
“immediately” to the top of the steps.
An impulse purchase might be consummated with no previous awareness, knowledge,, l~kin~ or conviction with respect to the product. On the
other hand, an ihdustrbl good or an important consumer product ordinarily will not be purchased in such a manner,
(“.
A model for predictive measurements of advertising et?eetiveness
139
to move people up the final steps toward purchase, At an extreme is the
“Buy Now” ad, designed to stimulate. ~ overt action. Contrast
this with industrial advertising, most of which is not intended to stimulate
immediate purchase in and of itself, Instead, it is designed to help pave the
way for the salesman by making the prospects aware of his company and
products, thus giving them knowledge and favorable attitudes about the
ways in which those products or services might be of value. This, of course,
involves rnovcmcnt up the lower and intermediate steps.
Even within a particular product category, or with a specific product,
different advcrtiscmcnts or campaigns may be aimed primarily at different
steps in the purchase process—and rightly so. For example, advertising for
ncw automobiles is likely to place considerable emphasis on the lower steps
when ncw models are first brought out. The advertiser recognizes that his
first lob is to make the potential customer aware of the new product, and
to give Ilim knowledge and favorable attitudes about the product. As the
year progresses, advertising emphasis tends to move up the steps. Finally,
at the cnd of the “model year” much emphasis is placed on the final step
—the attempt to stimulate immediate purchase among prospects who are
ass(mlcrl, by tbcn, to have information about the car.
‘1’IIc simple morfcl assumes that potential purchasers all “start from
scratch.” EIowever, some may have developed negative attitudes about the
product, which place them even further from purchasing the product than
those completely unaware of it. TIIC first lob, then, is to get them off the
negative steps—before they can move up the additional steps which lead
to purchase.
three functions of advertising
‘IIIc six steps outlined, beginning with “aware,” indicate three major
v
f(lllctions of advertising, ( 1 ) The first two, awareness and knowledge, , ~z
relate to inlorrnution or ideus. ( 2 ) The second two steps, liking and prcf- i
ercnce, have to do with favorable uttitudes or feelings toward the product.
(3) ‘l’he final two steps, conviction and purchase, are to produce uction–
the acquisition of the product.
I
“1’hcsc three advertising functions are directly,, related to a classic
psychological model which divides behavior into three components or dimensions:
1, “l’he cognitive component-the intellectual, mental, or “rational”
different objectives
Products differ markedly in terms of the role of advertising as related
to the various positions on the steps. A great deal of advertising is designed
sta tcs.
2. ‘1 ‘hc affcctivc component-the “emotional” or “feeling” states. ;
3. ‘1’bc conativc or motivational component–the “striving” states,
relating to the tendency to treat obiccts as positive or negative
\ goals.
.-c
S ~OAJfl S
ssaueJDMo puDJg
sanb!uqso+ OA!$>e!OJd
slo!\ueJejj!p
a!luowes puo 5+5 II
● SOIj>Jnd q uo!+ua~ul
senb!u~>et ● A!/3e!OJd
4!/!qoa!/ddv ~se~oe,ii
jo sde$s w pe+nlei
seqavoiddo qaioaseJ
jO sa\dunx3
su6!oduoa Jesoal
5U!J!JM Aqs
se[6u!r
SU0601S
SPO P!}!SSO12
s$ueue>unouuv
Ados eA!+d!J3sea
sloeddo JOIIIO[6 ‘snbo~g
SPO
eA!+!+ed~oD
Ado> eA!+O+UaUn6Jv
Slo!uou!+sel
sloeddo a>!Jd
sJejjo ,,a2uo~3-4sol,,
sloes
spo a40~s l!o+e~
eSOq>Jnd-jO-+U!Od
sde~s sno!JoA w $uoAe/el
6U!S!JJ&ApD JO uo!/owoJd
jo sed~ jo seldwnxa
1-Z1
379V1
P
NOI 1A ~
k.
{’4“
=SVH3Nnd
\e-o~~d ..–. . . . . .
\
pJOMO~
$uewaAow
laPOUJa~fO~ PS+O[aJ ~2JOaSaJ 6U!S!JJaAp0 PU06U!S!+leApv
..
—
“seJ!Sap JaaJ!p
Joa+olnw!~s spv .saA!+
7
-ow jo wlaaJ a~-eA!Jouo~ ,z,O’
...
Suo!suaw!p
/OJO!AOyaq
pa+ola~
.,
97
(
142
Applications in marke~ing research
A model for predictive measurements of advertising effectiveness
testing the model
Many common measurement of advertising effectiveness have been
concerned with movement up either the first steps or the fiml step on the
primary purchase flight. Examples include surveys to determine the extent
of brand awareness and information and measures of purchase and repeat
purchase among “exposed” versus “unexposed” groups.
Self-administered instruments, such as adaptations of the “semantic
differential” and adjective check lists, are particularly helpful in providing
the desired measurements of movement up or down the middle steps. The
semantic differential provides a means of scaling attitudes with regard to
a number of different issues in a manner which facilitates gathering the
information on an efficient quantitative basis. Adlectivc lists, used in
various ways, serve the same general purpose.
Such devices can provide relatively spontaneous, rather than “considered,” responses. They are also quickly administered and can contain
enough clcmcnts to make recall of specific responses by the test participant
clifficult, especially if the order of items is changed. This helps in minimizing “consistency” biases in various comparative uses of such mcasurcmcnt
tools.
Elliciency of these self-administered devices makes it practical to obtain responses to large numbers of items. This facilitates measurement of
elements or components differing only slightly, though importantly, from
each other.
Carefully constructed adicctive check lists, for example, have shown
remarkable discrimination between terms differing only in subtle shades of
meaning. One product may be seen as “rich,” “plush,” and “expensive,”
while another one is “plush,” “gaudy,” and “cheap.”
Such instruments make it possible to secure simultaneous measurements of both global attitudes and spen”fic image components. These can
be correlated with each other and directly related to the content of the
advertising messages tated.
Does the advertising change the thinking of the respondents with
regard to specific product attributes, characteristics or features, including
not only physidl c%;~kteristics but also various image elements such as
“status”? Are these changes commercially significant?
The measuring instruments mentioned are helpful in answering these
questions, They provide a means for correlating changes in specific attitudes concerning image components with cltanges in global attitudes or
position on the primary purchase steps.
143
:,. . , .;~,m
‘,
When groups of consumers are studied over time, do those who show
more movement on the measured steps eventually purchase the product
in greater proportions or quantities? Accumuktion of data utilizing the
stair-step model provides an opportunity to test the assumptions underlying the model by answering this question.
Three concepts
I’his approach to the measurement of advertising has evolved from
three concepts:
1. Realistic measurements of advertising effectiveness must be related
to an understanding of the functions of advertising. It is helpful
to think in terms of a model where advertising is likened to a force
which, if successful, moves people up a series of steps toward purchase.
2. Measurements of the effectiveness of the advertising should provide
measurements of changes at all levels on these steps—not lust at
the Icvcls of the development of product or feature awareness and
the stimulation of actual purchase.
3. Changes in attitudes as to specific image components can be evaluated together with changes in over-all images, to determine the
extent to which changes in the image components are related to
movement on the primary purchase steps.
I
(
The hypothesis of a hierarchy of effects: a pasiial evaluation
The a u t h o r exam;nes the widespread hypothesis in advertising
that a “’hierarchy of off ects” follows upon an individual’s perception of an advertising message and
before he buys.
I
Lavidge and Steiner claim that this sequence is based on what they term a
classic psychological model, which divid~.= into mgnitive, affective
and conative (or motivational) states.* “ ‘ ‘‘”
Ille Lavidge-Steiner hypothesis of a hierarchy of effeets offers in a
concise and clear manner viewpoints widely held in advertising circles for
many years. Attention, interest, desire and action [30]; awareness, acceptance, preference, intention to buy and provocation of sale [38];’ awareness,
comprehension, conviction and action [4] are but a few of similar but more
sketchy views of the internal psychological process a typical consumer is
supposed to experience from the perception of an ad to purchase; ‘“ ‘
Some recent refinements of the hypothe-sis are symptoms of its growing popularity. Copland, for instance, states that it cannot be expected
that a purchase will take place only if the individual has passed through
each of the stages of awareness, comprehension and conviction. Rather,
it is to be expected that some people who are merely aware of the existenee
of the brand will be buyers, rather more of those who comprehend the
message will be buyers, and even more of those who are convinced of the
truth of the claim will be found to be taking the final behavioral jump [5].
The broadly held agreement on this subject and the practical consequences which flow from this agreement provide an incentive to subject
the hypothesis to critical scrutiny. For if it is true that a one-way flow of
progression from message reception to overt behavior exists, then sales as a
.
13
The hypothesis of a
hierarchy of effects:
a partial evaluation
Krisfian S. PaIda
.
In 1961 Lavidge and Steiner presented a model for the predictive measurement of advertising effectiveness [12]. They postulated a hierarchical sequence of effects, resulting from the perception of an advertisement, which
moves the consumer ever closer to purchase. In diagram form the model is
as follows:
Movement
toword
purchose
Purchase
T
Canvictian
?’
Preference
?
Liking
‘r
Knowledge
‘r
Aworeness
criterion of effectiveness can be dispensed with and “substitute” variables
used instead.z It is notorious that sales measures of advertising effectiveness
are employed scantily, and a good case can be made for the claim that the
general acceptance of the idea of a hierarchy of advertising effects is to a
large extent responsible for this?
Voices of skepticism and dissent have not been entirely absent in
advertising literature [16]. Criticism seems, however, to have been directed
at each individual step in the hierarchy rather than at the hypothesis as a
whole.’ Furthermore: the criticism appears to have been predominantly
concerned with the methodological soundness of the research methods
employed to ascertain the effectiveness of an advertisement to bring about
.
Behavioral
Relotesf research
dimension
Cagnative-the realm
af matives
Split-run tests
Intention to buy
Projective techniques
Affective-the realm
Brand preference
mea sures
Image measures
Praiective techniques
Awareness surveys
Aided recall
‘“-”’”’~’’:~ematians
Canotive-the realm
of thoughts
Reprinted from the Journdl of M~rketing Resedrch, American Marketing Association
(February 1966), pp. 13-24.
144
145
I It is interesting to note that the branch of social psychology known as mass commuilications has not yet, with a single esception, offered s ~~thesis about the
decision prrrecss the individual goes through after the ~rception of an “actionoriented”
message. Mendelssohn [15] provides both the exeeption and a critique of mass comnrunications theory on this point.
z I have dealt elsewhere with the assertion that a firm ears find the optimum size of its
advertising appropriation, even though avoiding the use of sales as the ultimate yardstick of effectiveness [25]; it cannot.
~ A good example is the privately available [27].
4 In their recent book, Lllcas and Britt present a systematic evaluation of the research
methods dealing with each step in the hierarchy [14].
&
u
(
146
The hypothesis of a hierorchy of effects: o poriial evaluation
Applications in rnorketing reseorch
are obtained to the contrary, this is an assumption that advertising management and research personnel accept aud.y, mrrcludes Heller.
Such data to the contrary, in fairly per.suas}ve form and quantity, have
been assembled by Haskins [9]. He is interested ;rr the relationship between
factual learning and attitudm, or factual learning and sales. (This article
centers on the latter relationship. )
Haskins first surveyed those few private advertising research studies
which he could find. Two used sales as the criterion. In the first, awareness
was, but gain in knowledge was not related to increased sales. From tbe
second, which was a massive experiment, it appeared that knowledge was
not a prerequisite to sales, but attitude and belief changes preceded sales,
and there was a relationship between the two measures.
Searching through Psychological Abstracts for 1954-1963, he found 17
studies dealing with the correlation of knowledge changes with attitude or
behavioral changes. Two showed a positive relationship between changm in
knowledge/recall and the criterion variable (mostly attitudes ), two a negative relationship and the remaining 13 little or no relationship.
Against this evidence might be set the ancient popularity of Starch’s
service and his series of articles about Netapps [32]. Starch asserts that when
users arc divided into those who recall and those who do not recall tbe
arlvertisemcnt, the difference between them can be attributed to advertising. And that, typically, those who recall buy more than those who do not.
Rotzoll, in a recent summary of the usual objections to Starch, points
out why bigb purchase and high recall could both be present without one
necessarily being the cause of the other:
awareness, recall, attitude chang% etc. It has not tended to the question,
until very recently, the plausibility of assumption that each of these steps
contributes to an in~~ “@bability of purchase.5
In the following seetions, the assumption that movement up each step
of the hierarchy increases the probability of purchase on the part of a consumer will be critically evaluated, (The methodological soundness of the
various methods measuring the impact of advertising on the “intermediate”
variables will only & touched upon.) First, a survey of some of the literature and empirical evidence concerning the various hierarchical steps will
be made; second, an analysis of some data which have been gathered; third,
and finally, objections in the form of a table will be presented, and questions on the economic soundness of not using sales as the criterion of advertising effectiveness will be posed.
Published sources have been checked that would bc of relevance to the
testing of the hypothesis of hierarchical effects with regard to the link between each step and buying behavior. Confidence in this bibliographical
survey was strengthened when the bibliographical survey of the. Advertising
Research Foundation appeared and dovetailed with it [2].
Cogntiive dimensions
oworeness
Studies of change of awareness accompanying changed amounts of
advertising effort are so numerous that it is impossible to survey them. With
one exception, there is no good evidence that such changes in awareness
precede rather than follow purchase? Several studies, in which the relationship between change in awareness and change in sales was investigated,
in addition to other hierarchical eff ccts, are reviewed below.
knowledge,
147
recall ond recogntiion
The posit%#’@’6&~rtisers and their agencies with regard to measurement of awareness, recognition or recall was stated by Hcller as follows
[10]. Advertisers hypothesize that if advertising is to sell, it must communi,
cate, and that the ad that communicates best is tbe one that will produce
the greatest memory impression. Tbe assumption is that the memory production of an advertisement is related to its sales effectiveness. Until data
5 The important reeent exception is Haskins [9].
6 Tile exmption is [21], discussed later.
z“
=
a ) Product interest could affect willingness to be exposed to advertising
for a product;
b) Post-purchase doubts about the wisdom of choosing a product could
Icad to exposure to that prod~lct’s advertising in order to allay such
doubts;
c) “Yea-saying” tcndcncics may in ftatc readership-purchase correlation;
d ) Starch relics on the assumption, possibly unsound, that perceivcrsb~lycrs closely rcscmblc norrperccivcrs-nonbuycrs in all significant aspects except for exposure to the advertising message and purchase
of the brand [29].
1,1 /4. *,$.*.S*
To those objcctioos the following, from among many others, might
be added:
e) The possibility that many factors, other than the printed ad in question, were not eliminated from the test situation: frcqrrerrcy of exposure to previous ads of the brand, similarity to previous or other ads,
nurnbcr of claims in ad, etc.
f) l’urcbascrs of the brand may be better “rememberers,” because it may
bc easier to associate rcccnt cxpcricncc and recent perception.
(
148
The hypothesis of a hierarchy of effects: a partial evaluation
Apphccrt;ons hr marketing research
The methodological pitfalls surrounding recognition and recall tests
are discussed in Lucas and Britt [14].7
A search uneov~ th$”lol]owing studies in which recall is measured
against purchasing and which appear relevant to testing the hierarchical
hypothesis.
A study undertaken at Scott Paper deserves brief mention because it
claims to have used experimental methods to determine the relationship
between changes in recall, attitudes and market shares [28]. From the published report it is, however, impossible to determine with any degree of confidence what went on.”
NBC undertook an expensive study in the Davenport, Iowa, metropolitan area in 1952 to determine the effects of television viewing on purchasa [21]. A s~ial effort was made to find out whether viewing leads to
purchase, rather than purchase to the remembering of viewing.
Interviewed were 2,452 women, first in February and then in May, on
a particular T’V program viewed and the brand advertised. Table 13-1 was
TABLE 13-1
Respondents viewing a porticulor TV progrom end/or buying the brand
advertised, in February and May
BUYING
+-
-+
- -
Total
460
76
86
175
173
59
27
104
191
44
53
113
351
113
80
347
1175
292
246
739
797
363
401
891
2452
Viewing
++
++
+ -+
- Total
.
&urce: [21, Tssblo 5 in Appendix A].
constructed, where the first algebraic sign (+ or — ) refers to buying or
viewing in February, and the second to ‘buying or viewing in May.
Achi-square s~tistic of 89 calculated from this table points to a strong
relationship ~&h’ti&ing and buying.
7 Partierrlarly pp. 60 and 101. Note also their conclusion that @rt/olio tests of advertisements are so insensitive to changes in variables of interest as to be of little practical
value (pp. 77-78).
6 Thus, [or instanm, on page 66 it is stated that there was no advertising or promotion
in control markets (while test markets remived special direct mail advertising). But a
chart on page 68 gives an index of changes in recall of such advertising in control mar.
kets as well. A section heading is General Industry Awareness and Attitude Study, but
the word attitude cannot subsequently be found in the entire section.
149
If it were true that buyers are better rememberers of advertising exposure than nonbuyers, then it would,,.~
ble to expect that: (a)
c aim they started viewing
relatively more buyers than nonbuyers shoti~
T
between surveys; (b) relatively more buyers should claim they mntinued
to view the program. But the Davenport dab did not bear this out. Though
not conclusive, the evidence presented in this NBC study appears to indi- <
cate tha=ll
A SI- 1 ar study,
on magazine r~dership as well
- gathering
- Information
‘—
as television viewing, was reported by the project director for the NBC
Davenport study [3]. It appeared that TV viewing did, but magazine reading did not, contribute to an increase in sales. A subsequent critique by
Semen of this article points to some of the weak spots in this and other
NBC studies [31 ]. In particular, it attempts to show that the “start and
stop” viewing or reading analysis is of dubious value in this eontext.
Bridging the cognitive and affective behavioral dimensions is survey
evidence gathered by the NBC Hofstra television study [19]. Matched
panels of Ncw York metropolitan area television owners and nonowners
( 3,270) were interviewed about buying behavior on ho occasions, in January and May, 1949. The surveys covered purchases of brands advertised on
television and competing brands not on television, among such products
as gasoline, cigarettes, coffee, soap, watches, refrigerators. etc.
Most striking are the results of the semnd survey, which asked about
17 brands bought lately and about recent (last month) and remote (before
last month ) TV exposure. Table 13-2 summarizes the results. The base for
TABLE 13-2
Relative sales increases at various levels af expasure to TV advertising
(base o unexpected non-TV-owners)
Nonowner, seen TV
Total owners
Seen program recently
Seen program regulorly
Seen commercial recently
Liked commercial recently
1 2 . 8 %
41,7
54.9
59.6
64.3
I
,,,,,,L , *WV%:O:2
,.
Source: [19, p. 4 9].
the figures is the percentage of nonowners unexposed to TV who have
bought the brand in the past month: 23.5 percent. There appears to be a
steady progression in sales increases for ever higher levels of awareness, re- 1
call and liking of commercials.
(
150
The hypothesis of a hierarchy of effects: a partial evaluation
Applications in morket;ng research
Nevertheless, more detailed analysis shows that the memory ( cognitive) effect may be mo~e irn rtant than the liking (affective effect) of
E ousands of “likers” and “dislikers” intercommercials: of the ~i~
viewed; 34.5 percent who have seen the commercial recently and disliked
it, purchased the brand; only 33.8 percent of those who liked the commercial and saw it “remotely” did buy a brand [19, p. 42].
‘llle key issue in this study is how well the set owners were matched
with the nonowners. For every TV owner, a nonowner was obtained from
the same block, resembling the TV owner as closely as possible in family
composition and standard of living. Eleven variables were used in pairing
responclcnts. However, the fundamental question was not resolved: was
~not remembrance of advertising increased as a result of buying rather than
I!vice versa?
I
Af?ecfive
In any case, a very large number of attitude studies undertaken by
social psychologists are not concerned ,~~ntual prediction of behavior on the basis of ascertained attitudes. A~most 30 years ago it was
pointed o,ut that while critics deem the “merely verbal” aspect of attitude
mcasurcmcnt to be its Achilles’ heel, actions are no more “valid” inherently than words. Actions are frequently designed to conceal or distort
“true” attitude quite as fully as verbal behavior [18]. This caution about
tllc meaning and purpose of attitude measurement is not, however, observed universally either among psychologists or advertising men. As Festinger puts it:
What I want to stress is that we have been quietly and placidly ignoring a
very vital problcm. We have essentially persuaded ourselves that we can simply
assume that tberc is, of course, a relationship between attitude change and
(
\ subsequent behavior and, since this relationship is obvious, why should wc labor
to overcome the considerable technical dificrrlties of investigating it? But the
fcw relevant stodics certainly show that this ‘obvious’ relationship probably
dots not exist and that, inclccd, some non-obvious relationships may exist [8].
dimensions
Without question the strongest conviction held by the advertising
community with regard to the hi~rarchy relates to the ~nk bctwccn attitude (or change in attitude) and sale of the advertised product (or change
in sales).
What criterion, then, can be measured wllicll provides a predictive
151
COPY
test; what criterion is related to sales or brand usage? My answer is attitude
shift. . . . And why do I believe this? Because there is considerable and growing evidence that attitude shift is related to brand image; and because there is
an abiding logic backing up this evidence [1].
Is there such a logic; is there such evidence? There are really two
aspects to the problem of a link between attitude (or attitudinal change)
i~nd behavior (or behavioral change): ( 1 ) Is attitude a mechanism which
tends to direct behavior? (2) Must a change in attitude precede, rather
than follow, a change in behavior?
Consider the first aspect. Many psychologists have for a long time
been very careful about the definition of attitude.
Helen Peak dcscrvcs to
.
be quoted on ~$,@jcct:
Attitude is (defined as) a hypothetical construct which involves organization around a conceptual or perceptual nucleus and which has affective properties. . . . It is often said that an attitude is a “readiness for action” which
seems to imply that behavior is directly determined by attitudes. Wc regard
this at best as a greatly oversimpliticd statement of the relationship between
attitude and irction . . . an attitude should not bc expected to serve as an
adequate basis for predicting all behavior, since it is rarely more than one of
several components of motive structure [26].
It is symptomatic that “applied” social psychologists, working in fields
in which it is important to predict behavior, have been aware for some
tirnc that it is not easy to infer behavior from attitudes and vice versa.
‘1 ‘hc best summary of their thinking on this subject was recently presented
by Vroom [35]. 1 le points out that there appears to be no tendeney for
persons with prejudicial attitudes toward Negroes and Jews to express their
prcjurficcs when their interaction is within the eontext of a formal role
rclationsbip demanding a lack of discrimination. He also stresses that in
no sense should employee attitudes be regarded as causes of effective job
pcrforrnancc. 1 Ie argues that “the conditions which produce positive attitudes on the part of the employees toward their jobs are not necessarily
those that motivate thcm to perform effectively on these jobs.” Substitute
“consurncr, product, purchase” for “employee, job, perform” and the similirrity of this problem between industrial relations and advertising becomes
apparent.
Consicler now the second aspect. It is easy to imagine a purchasing
situation in which advertising, effective as a reminder of a particular brand
name, caused the consumer to select this rather, tiaa.~tier brand. Satisfaction with the consequences of the purchase evoked a favorable attitude
where none existed before, or strengthened a weak preference. That attitude change ean follow behavioral change is now widely accepted–the,
literature stretches from racial prejudice research to empirical studies of,
cognitive dissonance [6, 33]. Yet, there seems to be only one published
cmpirica] study of advertising effectiveness in which attitudes are measured nftcr exposure to advertising, but before any buying takes place, and
rnatcllcd against subsequent purcbascs [7].
I
152
Applications 1ss market;ng research
This should not imply that there are many published reports of
studi=delving into ~e.~~ent relationship &tieen attitude (change)
and buying (changer. fie~r ~ative scarcity is indicated by the fact that
of the 98 cases presented in the National Industrial Conference Board’s
monograph Mauring Advertising Results [22], only one makes some attempt in the direction of attitude—sales measurement (Case 35, Tea
Guncil of the USA).
te, cone and Belding designed to
DuBois reported on a study b
[71. Data were obtained On use
find out whether attitudes in fluen
and attitude for 40 assorted grocery store brands by 228 respondents in a
panel in one city. Most of the analysis was based on a composite of all 40
brands. which, when multiplied by 228 respondents, gave a total of 9,120
cases for the &mposite b r a n d ( X L ) .
Among users who called the brand “one of the best” at the outset, 68
percent continued using the brand. Among those who were only mildly
favorable, 50 percent continued using it. Among the handful less than
favorable, only 28 percent continued use.
Among nonusers who called the brand one of the best at the outset,
25 percent became users in the next few months, in contrast to 17 ~rcent
of those nonusers who had called the brand “good,” and nine percent of
those who rated it at anything less than good.
Both types of attitude effect, holding users and generating new users,
were observed to continue for a giderable -d of Wfter lwo
after a full year,
six months, and even [in a seDa
md of analysis of the same data, the 40 individual
brands were ranked according to- the percentage of users who called them
one of the best. ~wfound-that.the .higber thi~pro rtion, the higher
-the
alw.the-pr~p~tion dms-who+~
brands were reranked aarding to the percentage of nonusers who called
them one of the best. Again, the larger this percentage, the larger also was
the percentage of nonusers who became users within the next few months.
The study also provided evidence that changes in attitudes are conducive to changa in action. If the attitudes became more favorable, the
users were more likely to stay with the brand than if attitudes remained
mildly favorable. And, on the other hand, a deterioration of attitude
among users ~,.ta,lead to a falling off in the percentage of those remaining with the brarid. Similar effects of attitude change were observed
among nonusers. An increase in favorability went with larger percentages
of conversion. A decline in favorability of attitude led to lower perccntages of conversion.
Here the evidence, as reported, appears to bear out the idea that attiI
~ tudes do precede and causally influence buying. Unfortunately, the paucity
I1> of technical information (how were the observations pooled, what com-
The hypothesis of a hierarchy of Weets: a partiol evolution
153
posite measure constituted a “favorable atti~de,” how was the pretest
effect COp& with, etc. ) given in this ~-~ published in the proceedings of a conference precluda the thorough critique such a study
-.
would merit.
!(, :
A very interesting study of recall, association and, especially, attitude
change (as measured by the semantic differential), offering a wealth of
technical detail, was reported by
w 06] It stopped short, however,
of looking at purchasing behavior.
cvertheless, a brief passage from it
merits quotation:
From a research point of view, a further-reaching consequence was the introduction of more evidence s[lpporting the hypothesis that high levels of asso.
ciation or recall did not necessarily mean a favorable attitude or disposition to
buy the product. In addition, it seemed that high levels of recall did not even
necessarily mean that the consmncrs nndcrstood the core idea of intensive and
cxpcnsivc advertising campaigns [16, p. 371].
%“”. (;L 9;$-$@,qi,
Conative dimensions
intention to buy
There appear to be no published studies of the predictive power (with
r&gard to sales) of udverfking induced intentions to buy.o The classic
warnings about difficulties of using intentions to forecast sala, mncisely
uttered 15 years ago by brie and Roberb, are still valid [13]. The pro!
lcms faced in this area, but restricted to the relatively easier subject matter
of consumer durables, are exhaustively discussed by Juster [1 1].
More fhan one dimension
The following two studies which are briefly reviewed cover more than
onc behavioral dimension of the hierarchy.
At first sight, NBC’S Fort Wayne study [20] IOOh like a massive test
in which the hierarchy of effects is well ddtiheri~%~ “How television
works to condition customers all along the road to purchase is a research
area explored for the first time in this study. : . . It advances viewers
along every step in the creation of customers for a brand. It turns strangers
into acquaintances . . . acquaintances into friends , . . friends into cuss Wells and Dames drew attention to the effeet “cxaggeraton” might have on survey
resldts when bnth ex~~osure to media and intentions to buy are measured [37]. But this
problcm has no relation with the }licrarchy of effects.
(
154
Applications in marketing reseorch
tomers [20, p. 9].” After wading through the quagmire of results and technical appendim, it ~,rn~t,apparent that no such thing was documented.
Over 5,000 ho&%iv~ were interviewed in Fort Wayne in the Fall
of 1953 before the first local television station went on the air. They were
reinterviewed six months later. The emphasis was on respondents who
acquired a TV set between the two surveys. Some or all of the TV buyers
were measured on brand awareness, brand-product association, slogan identification, trademark recognition, brand reputation, or brand preference
with regard to abut 35 advertised brands and then compared to the “unexposed” respondents. There is little question about the sturdy increase
in the level of the “communications” variables among the set buyers as
opposed to those “unexposed.” However, information about percentage
changes in the various effects, Qrs.d the percentage change in purchases, is
given out on Scotties face tissues only; no mention is made of Halo in
this contex~ although all of the measurements taken on Scotties were
also taken on Halo; and the percentage change in purchases of other
brands is presented in aggregate form only. Scotties registered a net absolute increase (difference between set buyers and the unexposed ) in slogan
identification of 46 percent, in brand reputation of 6 percent, in brand
preference of 20 permnt. Characteristically, only the absolute increase (percent of set buyers who bought Scot ties during the last four weeks before
the second interview m“nus the percentage of the same interviewees who
bought Scotties in the last four weeks before the first interview) in
Scotties’ purchases is given— 20 percentage points. The increase in buying
in the control (unexposed) group is not presented.
Anyone who is seriously interested in the technical quality of advertising effectiveness research is advised to go carefully over this study, which
is ambitious, costly and typical.
A recent massive and analytically conscientious study, John B. Stewart’s Re~titive Arfvetiising in Newspu@rs [34], turned some of its attention to the hierarchical steps. Chapters 5, 6 and 7 are devoted to consumer
awareness of brands, Chapter 8 to product knowledge imparted by advertising, Chapter 9 to product images and 10 to purchase intent.
The book reports on a massive experiment in Fort Wayne, Indiana,
with advertising campaigns for two products. As usual, a higher level of
awareness and .b~ product knowledge was exhibited by purchasers than
nonbuyers; as usual there is no information on precedence in time, except
that “those who planned to try” scored about one-third higher on product
knowledge than those who did not. With regard to the attitude towards
the advertised products, Stewart says “Apparently product usage was a
more powerful influence on the image than was exposure to advertising
[34, p. 192].”
Intent to purchase was not matched with actual purchase. Five re-
The hypothesis of a hierarchy of effects: a partial evaluation
155
spondents out of the 1,314 subjecb who were not exposed to advertising
declared that they intended to purch~t~ product; of the 1,903 subjects who were exposed, 17 declared their readiness to buy one product,
and 27 declared intent to purchase the other product [p, 211]. After the
campaign ran for several weeks, however, the differences between the exposed and nonexposed groups tended to disappear,lo
In his concluding chapter, Stewart writes:
What means can management use to evaluate a specific advertisement? Judging
from this campaign, the only safe measure would appear to be trial purchase.
As a more thorough understanding of persuasion through advertising is obtained, it may become possible to evaluate advertisements accurately before
they are actually run. But it would not have been possible to do a good job in
evaluation with the ‘before purchase’ measures used in this study [p. 300].
Two private studies of advertising effect which throw some light on
the hierarchical hypothesis, are now analyzed; first, because they provide
an opportunity of looking at some length at original data, and, second,
bccallsc they are typical of many other unpublished research studies. The
first study attempts only to link awareness of product with purchasing, but
it gives a substantial amount of data. The second deals with many of the
hierarchical steps; the data, however, are not abundant.
‘Illc first study concerned a newly launched brand of a household
utility product, not a regularly purchased itcm, priced between $6 and $8,
which was to some extent functionally differentiated from other brands of.
the same product class. Three tc]cphone surveys were undertaken in
March, May and July of the same year. In March and July the same 39
mcclium-to-srnall sized cities from coast to coast were the locale of the
survey; in May, however, only 30 of them. The universe from which the
sample was taken were households listed in telephone directories; about
400 names were selected at random in each city’s directory on each occasion. I’bus, the sample had different subjects for each time.
First a question was asked designed to yield information about the
respondent awareness of the brand. Then a question was asked whether
the respondent purchased the product within the last three months and,
if so, which brand. The two variables in the ~tu,d~,~re awareness and market share, and arc expressed in percentages of total respondents. Information was also gathered about level of advertising activity (Table 13-3).
I
‘l’able 13-4 cross-classifies period-to-period changes in awareness with
changes in market share. Since x 2 based on the cross-classification of
10 The study did find that the advertising campaign for one product was successful and
that it probzbly was not for the otbcr; this, bowever, has no bearing on the subject of
this paper.
I
●
Applicot;ons in marketing research
156
TABLE 13-3
Market shares, awareness and advertising intensity in 39 cities disclosed
during March, Moy a-tiy *eys
City
Y,
Y,
1
6.3
1.4
11.8
2.7
1.9
2.2
15.1
12.0
6.6
6.2
11.8
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
1.5
9.4
4.8
9.9
7.5
2.8
11.5
6.1
10.0
14.8
6.6
2.1
9.0 .“
5.9
-.
7.3
8.7
8.0
11.1
9.9
21
22
23
24
25
26
27
28
29
30
12.8
20.5
11.1
15.0
12.9
1.6
5.7
17.4
21.1
9.6
5.0
15.3
7.4
18.3
13.8
6.9
3.2
23.3
11.9
8.7
31
32
33
34
35
36
37
38
39
19.0
25.6
9.3
21.2
23.7
25.0
21.9
9.4 ;,y a *.J?.? .
15.2
21.7
10.4
13.7
29.1
12.5
30.0
13.3
24.2
25.3
Y,
x,
x,
x,
D
3.9
4.1
5.3
6.1
6.2
6.3
6.7
6.9
7.1
7.4
14.4
5.6
11.3
8.3
7.4
11.4
10.8
11.0
12.8
12.4
13.8
..
9.4
13.1
8.1
11.7
9.1
12.8
12.3
14.7
14.8
13.0
1.5
0
1
1
1
0
1
1
1.5
1
7.5
7.8
8.4”
8.4
9.9
10.1
10.4
11.6
13.5
13.6
I
14.6
22.9
15.7
15.7
15.8
16.0
16.7
18.1
18.8
18.9
9.8
16.3
5.4
13.3
14.3
10.9
18.0
8.0
12.8
21.8
9.1
14.6’
9.9
10.6
12.4
14.7
9.1
13.9
18.3
13.3
18.4
13.8
19.6
18.8
10.6
18.9
12.6
14.4
27.4
1
1
1
2
1
1.5
2
1
1.5
3
15.4
24.8
7.7
20.7
25.0
9.8
18.9
35.2
16.2
22.5
21.5
19.6
11.8
23.8
23.7
14.4
15.9
34.4
14.3
23.6
22.2
18.7
21.8
23.6
30.8
17.9
25.2
31.8
23.0
29.3
4
1
3
1.5
4
1
3
1.5
3
3
19.0
19.4
19.8
20.0
22.2
24.7
25.0
28.6
29.2
24.9
15.1
27.3
31.4
15.4
29.5
29.9
35.1
19.5
32.5
30.2
25.7
30.6
35.6
29.6
33.8
40.0
43.5
31.5
4
0
I
1
2
1
1
1
3
13.4
22.2
35.7
34.5
22.9
37.0
27.1
36.1
28.3
Note: The Y’s r-present merke? shores in the various ci)ies In percentages. The
X’S afo percentages
(
J
of respondents in corresponding cities who are owore of the
b r a n d . T h e s u b s c r i p t s stand for the March (l), Moy (2) and July (3) surveys. D is
o variable w h i c h roprosonts six different levels 0 1 c o m b i n e d T V a n d newspoper
a c t i v i t y s u p p o r t i n g the brand in question in the various cities. The data ore ar ronged in ascending order of magnitude of Y,.
The hypothesis of a h;erarchy
of ef?ectst a part;al evaluotian
157
AX8 with AYE is equal to 5.9 (Xal,.W = 3.8), it appears that a relationship
between the two vanablm exists which U, ~j~butable to chance alone.
Further estimation of this relationship led to regrmion analysis.
TABLE 13-4
Number of changes in market shares and awareness from March-May to
May-June
Change in awareness
Change ;n
mark et share
-AX,
AX,
Total
AX,
-AX,
AX,
-AX,
AY,
5
2
2
0
9
0
2
7
AY,
4
1
---
-AY,,
12
Subtotal
4
2
AY,
-AY,
16
4
0
0
7
0
1
0
12
2
-AY,
Subtotol
Total
Note: W h e r e
A% e X;-l
6
8
14
18
12
30
,A~i = ~1 - Y1-l,
The following linear multiple regressions were run:
group i
a) Yi = f(Xl)
Y,= I(X, , D,)
b ) Y,= f(log X,)
Y,= f(log X, , D,)
c) log Y: = f(logxt)
log Y\ = f(log X,, D,),
,, ,; L>,?, *%,,.+
where t = 1, or 2, or 3 representing, respectively, data from the first,
second and third survey, and D/s are dummies representing four levels of
combined TV and newspaper advertising activity: D1 = 1 when D = 1,
otllcrwisc its value is zero; and, similarly, D2 = 1 when D = 1.5 or 2;
D, = 1 when D = 3 or 4. The values of these dummy variables do not
change from period to period to period.
(
The hypothesis of a hierarchy of effeets: a partial evaluation
Applicsst;ons in mrrrkotirtg research
158
The first three equations listed below show the three ‘&t” estimated regressions for the three time peri
“$.s for Y with subscripts t,
t + 1 and t + 2, respectively). This stit~ti
‘ includes not only, for inY
group ii
a) Yt+l = f(xt}.q-.~c:*i
Y,+, = f(x,+l, xt)
Y,+, =f(x,)
Y,+, = f(xt+l)
Y,+2 = f(xt+2 , Xt+l)
YI+2 = f(xt+z, Xt+l , Xt)
stance, the form Yt+2 = f(Xl+2), but also Y~+~ = f(Xt+~ , X1+l , Xt ,
D,) etc. The criterion for inclusion into this group was the size of the
standard deviation of regression residuals (also called standard error of
estimate ). This statistic is the one most closely associated with forecasting
performance–the smaller it is, the smaller the forecast -error is likely to
bc [23].”
Y,+l = f(Xf ,Dt) etc.
Subgroups (b) i.e., the semilogarithmic, and (c) i.e., the logarithmic,
versions, were also run.
(1)
...
group III
a ) Yt+l=f
Y,+l=f
Y,+, =f
(
(
2X,+1 + Xt
2
)
’
Subgroups (b) and (c) were also calculated, where Fisher’s term was
3 log X,+ z + 2 log Xt + 1 etc., but the denominator was not transformed.
log Y,+2 = —0.321 + 1.090 IOgxt+z
(:.099!
7.03?
ND39
. Not expressed in logs
R2 : 0 . 7 6 5 .
The following two regressions are also listed for further discussion:
group iv
Koyck’s distributed lag [24]:
a) Yt+I = f(Xc+l, Y:)
Y,+l=f(Xt+],Y~,D,)
Y,+2= f(xt+z, Yt+l)
Y~+z = ~(xt+2 , Y/+1”, D~).
Subgroups (b) and (c) were also calculated.
(4 *.J .“+*J,... . ;
group v
AY,+l = f[A(xt+l)]
,
(3)
)
6
)
3x,+2+2x~+l +xt, D, .
Y,+2 = f
6
)
(
\
>
Y,+l = -0.722 + 0.667 Xi+l
(0.078)
7.051
3.775
N=30
R’= 0.723.
3X,+2 + 2X,+1 +x,
(
Y,= –1.176 + 0.741 X~
(0.087)
7.444
4.391
N=39
R2 = 0.661.
(2)
2xt+1 + ‘t D,
2
159
where AYI+l = Yl+l — Yt, etc.
AY,+2= f[A(Xt+, )]
AY, +,= f[A(X,+l)]
AY, +Z= f[A(X,+z,Xt+l)].
All in all, 70 regressions were run.
(4)
(5)
Y,+Z = —26.793 + 27.031 log X ( + 2 + 6.013 log Yt+l
(2.834)
(4.354)
6.687
3.619
N=30
R2 = 0 . 7 2 7
Ay, + I = –1.470 + 0.896 AXI+l
(;.;;:)
6.289
Nx29
.
R2 : O.joo.
i,,
,J
.!S.*I*,;,,(<
How should these results be interpreted? These three consecutive surveys are in many ways typical of much commercial advertising research,
in that quite a few data are generated without a rigorous attempt at getII Y is tfre market share and X awareness in periods indieatcd by the subtipts; N is
sample size; Rz is the coefficient of multiple determination; the figures in parentheses are
standard errors of regression coefficients and tile figures on the last line below each regression are, respectively, the standard deviation of the dependent variable and the
standard deviatioil of regression residuals.
(
160
Applications in marketing research
ting unambiguous results. This was the reason why it was considered important to subject the data @a? figorous an analysis as Possible.
Just as the pre~ti~~i-square analysis indicated, the regression
results confirm the pr~nm of a strong concurrent relationship bctieen
I awareness and market share. One could have, however, more confidence
in the existence of a causal relationship flowing from awareness to purchasing, if two phenomena potentially obtainable from the data had been
detected. First, the presence of higher levels of advertising activity should
(
have strengthened the awarenas-purchase relationship. But regression
equations incorporating dummies, which represented varying levels of
advertising activity, were of a lower quality than those which ‘did not.
Second, lag correlations between awareness and market shar)would
give a better indication of the direction of the causal flow between variables. But regressions using lagged awareness, the more refined Irving
Fisher lag, or the sophisticated Koyck lag distribution simply did not fit as
well as those of the concurrent form. (See, as evidence, the “best” lag
regression shown above, Equation 4,) Even first differences performed
poorly (Equation 5 shows the best first-difference equation fitted).
Thus, it cannot be said that the data from these surveys confirm the
hypothesis that awareness tends to precede or even to contribute to rate of
purchase. They only show that higher awareness coexists with higher purchasing rata.
A few years ago a Canadian producer of a brand of a frequently purchased, very widely used, low cost consumer product started cosponsoring
a highly popular television program in a mrtain provinm. (The other
sponsor was not new to the program. ) While his sponsorship was new, his
brand was well established in that province, holding a market share of almost 20 percent. To assess the effativeness of his advertising venture, the
manufacturer’s marketing research agency conducted three telephone surveys.
The purpose was, roughly, to get information on the awareness of the
sponsor’s identity, on the viewers’ reaction to the sponsorship of such a
program, on the extent of recall of the commercials, on the attitude toward
the company and its products, and on buying.
The telephopc<in~ews took place in the four principal cities of the
province. The first one was staged just before the start of the telecasts, the
second four weeks later, the third lust after the TV series finished several
months later. Five-hundred different subiects were selected in each of the
three surveys by a random procedure from the telephone directory. Two
quotas were imposed: the subiects had to be users of the product (not the
brand ) and half of them had to be male. Thus, the universe was one of
product users. The sex restriction (achieved by alternately asking to speak
The hypothesis of a hierarchy of effects: a partial evaluation
161
to the housewife or the husband) probably had little distorting effect,
users and viewers not likely being-oq :i
“ori grounds-differentially
distributed between the sexes. The time ow
“, e survey was after 5 P. M.,
and three callbacks were made.
awareness of sponsorship
“Could you tell me what company, or companies, sponsor each of
the following TV programs (six programs, rotating in order) ?“
During the second interview, 23 percent of all respondents correctly
mentioned the sponsor’s name. At the time of the third interview 48 percent did. Among the users of tbc brand the percentage went from 33 to
57, among nonusers it increased from 21 to 46. The results do not appmr
1
to need further statistical analysis.
opinion of sponsorship
“As it happens, the telecasts are sponsored by Company A and B
(rotating company name from respondent to respondent). Are there any
comments you would like to make about either or both companies and
their sponsorship of the TV program?” The results arc set forth in the
following two tabulations. for brand users and those who are not ~ur.
cbascrs if the brand:
BRAND USERS
Fovoroble
Second survey
Third survey
Totol
(o) 44
Other opinion
opinion
persons
(C) 46
—
90
(b)
60
(d)
71
Total
104
(n,)
(n,)
117
%
i
z = - . S9
USERS OF OTHER, ERA,NDS
survey
Third survey
Second
Totol
Favoroble opinion
Other opinion
Totol
186 persons
184
210
lW
396
383
G
G
z = .23
—
779
(\
162
The hypothesis of a hierarchy of effeds: a partial evaluation
Applications in madeting research
The analysis of the differences between proportions follows Wallis
and Roberts [36]. The follo ,~g statistic, z, is computed and reference is
made to the normal disbiiron tablg:
It is apparent that analysis of the total sample (500) will not yield a
higher value of z.
163
For the change recorded from the first to second sumey, it is clear that
since the two subgroups’ z statistic exceed, ~.the z statistic for the whole
,,
group will also.
For the change recorded from the second to third survey, the change
is clearly of little significance among brand users-5 t3-57 percent. This
small amount of change is enough to swamp the change among nonusers.
TABLE 13-5
Attitudes toward sponsor as indicated by respondents’ opinions obout
faur campanies, including the sponsor
recall of commercial cantent
PERCENT OF RESPONDENTS MENTIONING SPONSOR
“Let’s discuss . . . telecasts. Think back to the last few TV programs
and describe the [first/semnd] commercial you can remember. . . .“
ATTITUDE TOWARD
COMPANY
Sponsor’s
All re-
brond users
123
RESULTS FOR ALL RESPONDENTS
Second survey
Third survey
11%
20
52
Sponsor’s commercial
Recoiled first
second
20
Most progressive
Best reputation
Most community minded
Most interested in his
customers
Most popular
High quality products
Fostest growing
~ese figures do not need further statistical analysis.
Most reliable compony
Independent compony
Average
atiitude toward company
“I am going to read a list of descriptions and for each one, would you
please name the one of the four manufacturers that it best fits. If you
think it describes more than one, name them.” (Rotate name of manufacturer; keep reminding respondents who the four manufacturers are. )
The z statistics appropriate to evaluate changes in attitudes toward the
company are given for the averages:
..%’.
.l
.{:
L
First
to second survey chonge:
Sponsor’s brond users
Other brond users
1.76
2.36
Second to third survey chonge:
Other brond users
All respondents
2.38
1.28
Nonusers
spondents
123
!23
40
46
18
54
46
56
‘ 41
72
48
62
30
64
11
16
10
29
13
22
13
39
15
27
13
39
17
22
12
34
20
29
19
46
40
63
38
57
38
52
78
40
72
61
50
78
37
71
69
11
30
10
27
13
12
34
15
38
27
15
45
14
46
40
17
36
16
33
17
20
43
23
53
21 19
45 52
34 47
44
58
57
17
24
28
23
31
23
35
17
45
35
I Iowcvcr, the amount of change recorded between the first and the
third surveys is considerable for all groups.
brand
usage
, , ,‘$ -$~-: “.
“Now, speaking about (’the product’ ) only, which brand of it do you
use mainly?”
‘Illc proportion of all respondents claiming to use sponsor’s brand
mainly iocrcascd from 19.0 percent on the first survey to 20.8 percent on
the second and 23.4 on the last survey. ‘rhe first change being a small one,
the change between the first and third surveys will be assessed.
‘rhc z statistic oscd to evaluate this change yields the value of 1.62.
It SI1OUICI be pointed out that the dcpcndcnt variable here is not
.
.
.<
.
sasoyaJnd
,,oslndus!,, jo walqoJd ayl
-DZ!U*40 ~UD
Uo!+
~’q IID 4D P0406
-!4SISAU! UOaq ~le2JO>S SDq
sclqolnpuou popu DJq OJ pJ05
-oJ +!* suo!+ua$u! pe>npu!
-6 U! S!\ J0ApD jO Jo#od OA!\J!p
4Jd ‘U06!q2!W jO &! SJOA!Un
‘iquo~ qsloasa~ 4aAJnS
a~ Aq patusiwn>op klaA!
-SSDW uaaq scJq spoo6 slqo
-anp Jawnsuo>
jo sassola oh
pJo6aJ qJ!A sAaAJns ~nq.o~
-uo!~ua~u!
jo
Jamod aA!~s!paJd
sAaA)ns
eldwos
uo!JD+aJdJa&ul
.[~l] S6U!
-UOaU1 S$! JO! AOljaq UO
sJajuo3 q3!qA A+!lopoul
,e..’
o
so Jo ‘JO! ADqSq 6u!l>aJ!p
ws!u Dq>aw D so Ja~!.s
paua!. a q sn+ AOUI apn~
-!$JV ‘JO! AOqSq paAa!qao
LpmJ10 a~ 0+ 6u!uDaw
aA!6 0+ saA4as puo PaAl OA
-u!
s!
•p~!uo” u! e6u0q2
O ‘tsJ!j sa6uoq> JO! ADqaq
j ! ‘puoq Ja~o a~ UO .JO!
-Aoqaq
(PIJOM
U!
u! a6u0qa asno> OJ
s ! ILI!D a~ ‘JsJ!j pa6uoq2
s ! apn+!$~o uaqM ‘paqs!n6
-U!JS!p aq UD> JO!ADq
loa4
UOq+ AJOJOJOq Ol U! ‘pUOIAOH
0! 6u!pJ02>0 ‘aAa!q>b OJ Ja!
-sDa uo!~o>!unwwo> q6n0JqJ
a6uc.qo apn~!t~o ‘.6.a) 6u!4+as
ssauloznlou ‘aldwDs jo ~![on~
u!4a*q>S
spu D1q 6u01uD
-aq pu. ap~!uo uaa~aq
d!qsuo!+olaJ
,6u!uoaw, o
p u o d!~suo!~ola, Iosnaa v,,
J>a>!pu! jo AJ!p!l DA
‘6u!uoaw jo ~![ouo!suaw!p!ufi
sasuaJajaJd puo
suo!hsanb
sLaAins
apn~!b~v
pJD*oJ
sapn~!+~v
~JajjO ,,pu DJquOu,, o
s ! snlnw!~s
~uoAalaJ uaqm
qanow jo pJOm Aq
spo 6UOUID a>ua~ajald
~noqo p!os aq 6u!wAu0 uo~
uo!4Du!uo4u02
‘aA!k>oIJ$D
Jajjo
j ! !stlnsai so!q s,add!l> uodno~
~!l!q W!joJd alq!ssod
j!a~ jo uo!~o>!pu!
a A ! 6 $Ou
s a o p sal!joJd jo uos!Jodwoa
!suo!u!do ou JO} a>uo*ol
-10 ou !saA!+>ajpo jo 4!4010d
IO!J
-uajajj!p 5!JuDwaS
6u!44as jo
ssaul DJn$ou ‘aldwos jo A+!lonb
JazAlouD
uIn860ad
puo~ Anq Q UO!4
.!sods!p
puo pD jo ~uau
.Aotua (>!+aqJsa)
- q
Uo!J>auuoa
“
taajja
uaa*
oloq !paq!l s! po
auo Isoal tD +DW uo!~dwnssv
1DS!601 ON
SPO Jo ‘0s
-! JoduIoa paJ!od
pus. 4! Jalu-jO-JapIo
spo uaehuq sa>ua
.JajaJd eqt jo 6u!q!l
S4SaJ
&! A!\!suas 0144!1
● A!i2ajjV
O!lOfiJOd
lo!aJ8wwos Jo ~uautas!~
,,sJaJaqwawai,,
-4aApD ● u o ay~ uDqh Jaq40 SJOJ
]m~aq
q Aow pUO~ jO SJDSDqJJnd
.% .Jqs
●
->oj jo uo!/ou!uI!la
.
•~aldwo>ul
llD2aJ
1uo!t!u60>a~
a! A,as-jlcs u! AIJDl
-flJ!~JOd ‘aw!~ jo uo!\>Dq
.<
.,
4uo>!j!u6!s AuD Aq ● soqs
. .. .
L -Jnd 5$! apa>ad OJ puoq
~ jo ssaua,oMo Joj ~!ssaa
-au lB3!601 ou ‘loJaua6 ul
+noIu jo pJOM jo
●
auanbasuos
:.
SO SS@UaJOmD puD 6U!S!$JaA
-po jo
●
>uanbasuo> so ssau
-OJD*O uaa~q uo!40u!u!Jas!a
●
AaAJns
-i
t
ssauaJo& puDJg
ldwq
●A
!$ !u603
;,
SSSU~OaM
q2J0esaJ
aA!JUOJS9ns
qsns u!
lj>Joasal
swlqoid Pahlosaiun ● uos
pwola. /o>!dAL
!Ij aq~ jo sessau~oam lo>!6010po~aw
s!saqtod.4q stsajjo-jo-Ay>JoJa
J~&OS UO!JOUIJ~U!
jO so/dwox3
p u o aii!~uo~sqns jo
Uo!suelu,lp
/o! JO! AD1/a8
AJOIUUInS v
9-C1 319V1
(
166
The hypothesis of a hierarchy of effects: a partial evaluation
Applications in marketing research
~,y’~k~ “;”’
First survey
Third survey
95
117
,’
Nonusers
405
383
Total
Soo
500
sales, but rather the number of respondents who say they use the sponsor’s
brand rnuirrly. This is far from a reliable revenue yardstick. Thus, it is not
possible to say whether, as a result of the advertising, revenue increased, or
users bought at a higher rate than previously while some new customers
were acquired, etc. It is only possible to say that among the total respondents, the number of users grew from 95 to 117 out of 500.
What overall mnclusion can be reached from this study with regard,
to the hypothmis of hierarchical effects? It is clear that many of the communication objectives of the sponsor were reached: awareness of his sponsorship increased strongly from the time the TV series started; recall of
commercial content was very good and increased considerably. Slightly
more ambiguous results were obtained when opinions were sampled. Some,
but only a small amount, of increase of favorable opinion towards the
sponsorship of the TV series (really, toward this type of advertising) was
registered. However, on the highest rung of the hierarchical ladder a considerable increase in favorable attitude toward the company was recorded,
especially between the first and third surveys.
No such clear<ut results were obtained with regard to brand usage.
Keeping in mind that brand usage is not a perfect substitute for sales, it
cannot even be said, at least on non-Bayesian grounds, that it show~d a
significant amount of change. Thus, while “significantly” large numbers of
respondents moved “up the hierarchical ladder” of awareness, liking of
the advertising and of attitudes, performance on the last “rung” is difficult
to assess. Add to this the uncertainty of the causal direction (awareness + opinion + recall + attitude ~ usage) and we end with an unsatisfactory feeling. At that, this study appears to have been a rather good
one, unwittingly perhaps, but nevertheless persistently testing the hierarchical hypothesis.
As a cone-.~rk, it is suggested that the only satisfactory and
lasting answer to the doubts or unwarranted assertions concerning the
hierarchical hypothesis would be a well-designed experiment. Only an experiment can approach the assessment of the direction in the causal flow
unambiguously. Many problems can be foreseen, especially with pretest
effects in such experimentation, but it is not the task here to advise on its
feasibility.
A table was constructed (Table 13-6), leaning on the Lavidge-Steiner
‘ representation, to point out in condcnscd form most of the substantive
167
weaknesses of the hypothesis of hierarchical effects. Singled out for inclusion are also certain methodological w
,, most of them not menm~i ‘ ed with ~ch step in the
tioned above, which appear to behierarchy. The considerable number of these methodologi~l problems,
quite apart from the substantive objections brings up forcefully this question: Is it, on balance, really more difficult and expensive to investigate the
direct link between advertising expenditure and sales, than it is to undertake research into each step of the hierarchy-even if the existence of u
hierarchy of effects were actually established?
REFERENCES
1. ACHEN~AVM, Alvin A., “Is Copy Testing a Predictive Tool?” Proceedings,
10fh Anrruul Conference ( Ncw York: Advertising Research Foundation,
1964) , p. 66,
2. Sales Measures of Advertising: An Annotated Bibliography, compiled by
L. Krueger and C. Raymond (New York: Advertising Research Founda/
tion, 1964 ).
3. C OFFIN , Thomas E., “A Pioneering Experiment in Assessing Advertising
Effectiveness,” Jourrrul of Marketing, 27 (July 1963), pp. 1-10.
4. COLEY, Russell H., &d., Defining Advertising Gods for Measured Advertising Results (New York: Association of National Advertisers, 1961), p.
55.
5. C OPLAND , Brian, “An Evaluation of Conceptual Frameworks for Measuring Advertising Results,” Proceedings, 9th Annual Conference (New York:
Advertising Research Foundation, 1963 ).
6. DEUTSCH, M., and COLLINS , M, M., Interracial Housing—A Psychological
Evaluation of ~ Social Experiment (Minneapolis: University of Minnesota
Press, 1951),
7. DUBOIS, Cornelius, “The Story of Brand XL: How Consumer Attitudes
Affcctcd its Market Position,” Proceedings, i Sth Annual Conference,
American Association for Public opinion Research, Public Opinion Quarterly, 24 (Fall 1960), pp. 479-480.
8. FESTINCER, Leon, “Behavioral Support for Opinion Change,” Public
Opinion Quarterly, 27 (Fall 1965), pp. 4~17.
9. HASKINS, Jack B., “Fact ual Recall as a Measure of Advertising Effectiveness,” )ourn~l of Advertising Research, 4 (March 1964), pp. 2-8.
10. [IELLER, Norman, “An Applicati&n of Psychological Learning Theory to
Advertising,” Journal of Marketing, 20 (January 1956), pp. 248-254.
11. JUSTER , F. Thomas, Anticip~tions-und Purchases ( Prince};n, N.J.: Princeton University Press, for the National Bureau of Economic Research,
1964).
12. LAVIDGE, Robert C., and STEJNER, Gary A., “A Model for Predictive
(
168
App/icat;ans in market;ng research
Measurements of Advertising Effectivenessfl )ournaf of Marketing, 25
(October 1961
13.
14.
15.
16.
17.
’18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
), pp. 59-62.
James ~; ~ ~OSIERTS, Harry V., Basr”c Methods of Marketing
Research (New York: MeCraw-Hill, 1963).
L UCAS , Darrell B., and BRSTT, Steuart H., Measuring Advertising Electiveness (New York: MeCraw-Hill,, 1963).
MENDELSOIIN, Harold, “Measuring the Process of Communications Effect,” Public Opinion Quarterly, 26 (Fall 1962), pp. 411-416.
MINOAK, William A., “A New Tcchniquc for Mcas{lring Advertising Effectiveness,” Journal of Marketing, 20 (April 1956), pp. 367-378. -MOSCOVICI, Serge, “Attitudes and Opinions,? Annual Review of Psychology, 14 ( 1963), pp. 249-250.
M U R P H Y , C., MU R P H Y , L. B., and N E W C O M B, T. M., Ex~”merrtul Soci~l Psychology (New York: Harper & Row, 1937), pp. 909-912.
National Broadcasting Company, The Hofstra Study: A Measure of Sules
E~ectiveness of TV Advertising (New York: National Broadcasting Company, 1950).
—, Strangers into Customers (New York: National Broadcasting
Company, 1953 ).
—, Why SaZes Comes in Curves (New York: National Broadcasting
Company, 1953).
Measuring Advertising Results (New York: National Industrial Conference
Board, 1962).
P ALDA , Kristian S., ‘The Evaluation of Regression Results,” in Stephen
A. Greyser, cd., Toward Scientific Marketing, Proceedings, Winter Conference of the Amm”can Marketing Association, 1963, pp. 279-290.
—, The Meuauremerrt of Cumulative Advertising Eflects ( Englewood
Cliffs, N.J.: Prentice-Hall, 1964), Ch. 2.
—, “Sales Effects of Advertising,” /ourti of Advertising Research,
4 (September 1964), pp. 12-16.
P EAK , Helen, in Marshall R. Jones, cd., Nebrasku Symposium on Motivation-l 955 (Lincoln, Neb.: University of Nebraska Press, 1955), pp.
151-152.
Program for Measuring the EfectiveneS of General Motors’ Advertising
(Probable date: 1963, 100 pages).
ROENS, Burt B., “’New Findings from Scott’s Special Advertising Research
Study,” Proceedings, 7th Annud[ Conference (New York: Advertising Research Foundation, 1961 ), pp. 65-70.
ROTZOLL & B4 “The Starch and Ted Bates Correlative Measures of
Advertising Effecti~eness,” Journal of Advertising Research (March 1964),
pp. 22-24.
S ANDACE , C. H., and F RYBURGER , Vernon, Advertising Theory and Pructice, 6th ed. ( Homewood, Ill.: Irwin, 1964), p. 240.
S EMON , Thomas T., “Assumptions in Measuring Advertising Effectiveness,” Journu[ of Marketing, 28 (July 1964), pp. 43-14.
STARCH , Daniel, Measuring Product SaZes M~de by Advertising ( Mamaroncck, N.Y.: Daniel Starch and Staff, 1961 ).
bRIE,
The hypotheses of a hierarchy of effects: o patiial evaluation
169
33. S TRAITS , Bruce C., “The Pursuit of the Dissonant Consumer,” /OUfnd/
of Murketing, 28 (July ]964), pp. 6
g ~; IVews@@s (Boston: Har34. S T E W A R T , John B., Repetitive Advefi’ti
?
vard Business School, 1964 ).
35. V R O O M , Victor A., “Employee Attitudes,” in George Fisk, cd., The Frontiers of Management Psychology (New York: Harper & Row, 1964),
pp. 127-143.
36. W ALLIS , Alan, and ROBERTS , Harry V., Statistics-A New Approach (New
York: Free Press, 1956), p. 430.
37. W E L L S, William D., and DAMES, Joel, “Hidden Errors in Survey Data,”
]ournaf Of Marketing, 26 (octobcr 1962), pp. 50-53.
38. W OLFE , Harry D., B ROWN , James D., and T H O M P S O N , C. Clark, Measuring Advertising Results ( Ncw York: National Industrial Conference Board,
1962), p. 7.