Brand Loyalty Measures

Tijdschrift voor Econoniie e n Management Vol. XLI, 4, 1996

A Review of Brand-Loyalty Measures in Marketing

by M. MELLENS", M. G. DEI<IMPE'%nd J.-B. E.M. STEENIUMP""

I. INTRODUCTION

The success of a firm depends largely on its capability to attract consumers towards its brands. In particular, it is critical for the survival of a company to retain its current customers, and to make them loyal to the brand. Former Ford vice president Basil Coughlan estimates that every percentage point of loyalty is worth $100 million in profits to his firm (Serafin and Horton (1994)), and major enterprises like Del Monte, Harley Davidson and General Motors are spending large sums ot money to induce brand loyaity (Monzo (1994); Lefeon (1993)). Firms selling brands w ~ t h a high rate of ioyai consumers nave a competitive advantage over other firms. Brand loyal consumers reduce the marketing costs of the firm as the costs of attracting a new customer have been found to be about six times higher than the costs of retain- ing an old one (Rosenberg and Czepiel(1983)). Moreover, brand loyal consumers are willing to pay higher prices and are less price sensitive (see e.g. IO-ishnamurthi and Raj (1991); Reichheld and Sasser (1990)). Brand loyalty also provides the firm with trade leverage and valuable time to respond to competitive moves (Aaker (1991)). In sum, loyalty to the firm's brands represents a strategic asset which has been identified as a major source of the brands' equity.

"Faculteit Econoniische en Toegepaste Economische Wetenschappen, IC.U.Lcwen, Leuven. l"' Faculteit Economische en Toegepaste Econon~ische 'flete~ischappen, K.U.Leuveil. L e ~ ~ v c i l

en Wageningen University, Wageningen, The Netherlands. The authors are indebted to the Belgium National Science Foundation (N.F.W.O.) for providing financial support under grant no.2.0062.94.

Given the importance of brand loyalty, it is not surprising that it has received considerable attention in the marketing literature since Copeland's seminal work which was published over 70 years ago (Co- pelavid (1923)). Studying and managing brand loyalty, however. should start with a clear definition of the construct involved, and with the de- velopment of valid measures.

Unfortunately, while there seems to have emerged considerable agreement on the conceptual definition of brand loyalty since the work of Jacoby and Chestnut (1978), no unified perspective to measure it has yet emerged. Still, a valid measure is essential for a better understanding of the concept by marketing researchers and marketing managers alike. Moreover, knowing the limitations of a measurement method is crucial for a correct interpretation of the results of a study. The purposes of the paper are therefore: (1) to present a structured review of the major categories of brand-loyalty measures, with an emphasis on the developments since Sacoby and Chestnut's (1978) monograph; and (2) to provide directions to marketing managers with respect to the use of brand-loyalty measures in applied marketing settings. The review starts with a detailed discussion of the brand-loyalty concept. Next, we evaluate four main types of brand-loyalty measures. Finally, conclusions are drawn and recommendations for the managerial use of brand-loyalty measures are provided.

11. THE CONCEPT O F BRAND LOYALTY

It is convenient to distinguish conceptual definitions, which are ab- stract descriptions of the phenomenon being studied, and operational definitions, which are measurement methods (see e.g. Jacoby and Chestnut (1978); Peter (1981)). Conceptual definitions are necessary to assess the construct validity' of the adopted measurement methods. Without them, the correctness of specific brand-loyalty measures cannot be evaluated, and meaningful and meaningless results cannot be distinguished.

Perhaps the most elaborate conceptual definition of brand loyalty was presented by Jacoby and Chestnut (1978). We will use and discuss this definition, because it covers the most important aspects of brand loyalty, and siiice it enjoys widespread support in the marketing literature, either in its original form or in slightly modified ver- sions (e.g. Assae1 (1992); Mowen (1993); Wilkie (1990)).

According to this definition, bland loyalty is: "The ( a ) biased, (b) behavzor~l r esposzse, (c) e~pressed over tznze, ( d ) by some decror d lon- i i~a- l~zng unzt, (c) wzth respect to orze or Inore alferl?atzi/e b~a'clnds 001 of li set of sriclz brarzds, and (f) zs afi~izcaon ofpsychologzcal (decision-n~akzng. evab~ciflr~e) processes (Jacoby and Chestnut (1978, p.80))". This deh- nition identifies six requirements for brand loyalty. Below, each of thern is d~sctnssed in scmex,.ihat more d e t d .

Biased behavioral response (a-b) First, brand loyalty is a biased response. This implies that there has to be a systematic tendency to buy a certain brand or group of brands. which means that brand choice should not foilow a zero-order process. A nrocess Y is zero-order if each brand is chose11 by the ccns::mer with a certain probability which is independent of the consumer's past purchase decisions. Nothing that the consumer did or is exposed to alters the probability to purchase a specific brand (Massy, Montgo- mery, and Morrison (1970))'. Zero-order behavior is not part of the brand-loyalty construct, because this would imply that brand loyalty is beyond control by any marketing action, and hence a meaningless concept for marketing managers. Brand loyalty also entails actual purchases of a brand. Verbal statements of preference towards a brand are therefore not sufficient to ensure brand loyalty.

Expressed over time (C) An incidental bias towards a brand does not guarantee brand loyalty. As the process is dynamic, some consistency is needed during a cer - tain time span. This suggests that one should not only consider the number of times a specific brand is purchased during that period, but also the purchase pattern over successive purchase occasions. As such, one can distinguish partially loyal behavior from completely (non)loyal behavior. Considering a purchase sequence for brands A and B, Brown (1952) distinguished consistent loyalty towards brand A (indicated by a purchase sequence AAAAAA), divided loyalty (ABABAB), and unstable loyalty (AAABBB). For brand A the situation is much dimmer under unstable loyalty than under divided loyalty. These simple examples show that the purchase pattern over a given time span contains valuable information about brand loyalty.

Decision-making unit (d) Brand loyalty is defined by the purchase pattern of a decision-making

unit which may be an individual, a household or a firm. Important to notice is that the decision unit does not have to be the actual purcha- ser. For example, the purchases of a household are often made by one of the parents, but other members of the household may also be involved in the decision process (see e.g. Agnew (1987); Davis (1976)). This issue becomes important when the members of a household have different product needs and use goods for different purposes. In that case, we might observe switching behavior on the household level which represents different needs or usage purposes by different fa- mily members rather than an absence of brand loyalty.

Selectiorz of bmnds (e) The fifth condition Is that one or more brands are se!ectec! nut of a set of bmnds. This condition implies that consumers may actually be loyal to more than one brand, a phenomenon observed by many researchers (e.g. Ehrenberg (1972); Jacoby (1971); 0' Leary (1993)). Espe- cially for low involvement goods, the consumer often does not evaluate brands on a continuous scale, but classifies them discretely as acceptable or unacceptable. If more than one brand is acceptable, an individual might be indifferent between them, and exhibit loyalty to a group of brands rather than to a single brand. A problem with multi- brand loyalty is that it is hard to distinguish this kind of behavior from brand switching, especially if there are only a few brands available. An individual who buys brand A and B with the purchase sequence ABBABAAB may be defined as a multi-brand loyal consumer if more than two brands are available. But if enly brands A and B ran be selected, the behavior can be interpreted as brand switching, since every brand available is used regularly.

The fifth condition also implies that in order to have brand loyalty, there must be an opportunity to choose among alternatives. Jacoby and Chestnut (1978, p.82) expressed it as follows: "Before one could speak of brand loyal, one must have the opportunity of being dislo- yal". As such, brand loyalty cannot exist when a brand has a mono- poly position. The determination of the product category therefore becomes of major importance. We will return to this issue in section IV where recommendations to marketing managers are formulated.

Function of U ysychologicul process V) Brand loyalty is a function of psychological (decision-making, evalua- tive) processes. Brands are chosen according to internal criteria re-

sulting in a conzrrzztrnerzt towards the brand, which. according to Ja- coby and Chestnut (19781, is an esserztzal elernerlt ofbrand loyalty. This point of view is in line with the information-processillg paradigm, which is the dominant point of view in consumer behavior (Bettman (1979)). Although consumers do not always seek information acti- vely, they do receive some information, e.g. due to advertising cam- paigns. which may be used to form certain beliefs about brands. Based on these prior beliefs, brands are evaluated and some are preferred over others. In time, the consumer may develop a commitment towards a brand and become brand loyal. Hence, brand loyalty implies consistent repurchase of a brand, resulting from a positive affection of the consumer towards that brand.

We should point out, however. that the importance of commitment is not supported by some researchers who argue that buying behavior is caused by instrumental conditioning (see Foxall (1987) for a review). They posit that observed behavior alone is capable of explaining brand loyalty. According to this view, the purchase will lead to a "reward" (the brand is adequate) or a "punishment" (the brand is in- adequate). The former induces the repurchase of a brand while the latter induces brand switching. In this approach, brand loyalty is re- garded as a consequence of behavior, rather than as an explanation.

We do not subscribe to the point of view that observed behavior alone is capable of fully explaining "orancl loyalty. "fle support Szcoby and Cizestnut 'S (i978j argumen~a~ion that ~ur rnr r l i t~ l i e~~t is dn esseii- tial element of brand loyalty, as it allows to separate brand loyalty from repeat buying. Repeat buying may be due to inertia. which means that consumers stay with the same brand because they are not prepared to spend effort and time to search for other brands. A study of Hoyer (1984) concluded that inert consumers have different motives, different decision rules and require other marketing actions than brand loyal consumers. In particular, they do not evaluate a large set of alternatives but use simple decision heuristics like "Always buy the cheapest brand" or "Always buy the same brand". Repeat buying may be influenced by variables such as e.g. the amount of shelf space or distribution intensity, which are supervised by the retail manager. In contrast, brand commitment is more likely to be influenced by a brand's distinguishing characteristics, design features or images (Riezebos (i994)).

111. MEASURES OF BRAND LOYALTY

Thc six criteria identified Ir; o i ~ disciissio!: of the conceptual definition can in a next step be used to evaluate specific operation3.1 measures. Rather than discussing ail indiviciual operationaiizatio~ls in de- i-ail, and since measures which common characteristics have similar strengths and ~~eaknesses , we classify them into four groups. based on the foliowing two dimensions: (I) attitudinal versus behavioral measures, and (2) brand-oriented versus individual-oriented mea-

m. s~l-es. B hese dimensions are usecl since ehey appear frequent!^ in the marketiilg literature (e.g. Bicemer- (1993); Jacoby and Chestnut (1978)), are related to specific recpiire;nents of the cmceptuai definition (which makes it easier to point our the advantages and dra:wbacks of a group), and provide a wa~kable distinction for n~arketiilg managers.

Behavioizrl versus aititudirznl r-nenszires The majority of the operational measures can be categorized as either behavioral or attltrndinal depending on their relative emphasis on; respectively, the purchasing or the cognitive component. Their popula- 1-$7 has .raried over time and anlong researchers", as both categories have their specific strengths and weaknesses (see 'Fable I).

-7 l. be.traviorai measure? define brand ?oya.ity in terms of the actnal pUrc l''""S erVo.,' "-- 2." " a.." ."

~ u d i - L J a CLL b d i b i ii~lie p e r % ~ J , t j l i j ~ focusing on COD- , dii;iorj-s $jlzse-j Leha.410rar A G,-.- -ponse. expresseci over time) of the

conceptual definition. Their advantages are that t h q are: (1) based on actual purchases. which az-e directly related to the performance and exisrcnce of the firm; (2) ncE likely to be incidental as they are usually based 011 bellavior over a period of time; and ( 3 ) relatively easier to collect than axtitudinal data.

The ~n?iost important Iinritation of behavioral measures is "cat they make no distinciion bePween brand 1~yaPt-y and repeat buying, and therefore may contain spuriosas Isyalty (Day (1969)). Furthermore, although behaviorai data are the most accurate representatio~ls of past bel.zvior, they are act neccssa;-ily a gor;d r i - , p ~ ~ ~ e ~ t ~ t i o ~ 0:' fut:lre bek~avior: especially under changed circumstances (Day, Shocker, and Sriva-s~am (1979)). In particular, behavioral measures are sensitive to sllort-run ili_lcluations, caused ibr example by the fact tinat the customer's ;3lefzi:zcl Ir;rancl is i:t-.Lnporarj~:j ,:,hj[ s~;;lt.::~. F;iiiti?:i. is ilai,/J tcl select iEle right decision uz~i! as no ,nfnrms.to:z is csllecled 0 3 1 ti;e 1111- trierking 1-casor2 for a particular behaviol-.

(1) Rnscd or> acfiial behavior:

~ ~ h ~ ~ i ~ ~ d (2) "011-incidental, Meaauizr (3) easy to coliect.

I

frorn brand loyaity; (2) :nore ssnsitive to sliort-run

iluituaTsons; (3 ) difiicuit LO pick fighi decision-

Attimdinal from bnu~C ioydt)-.

, (3) easier to pick iigli! dccislo;~

I (i: Valid representation of I reali?~ not guaranteed;

(2) incidentai: (3j harder to colicst.

T k r i ., c~nt l ;as l~ attitrrdinal ixeasures Lire able to disiil~guish bi-aad 10-

yalty from repeat buying. They are based on stated p-references, commitment or purchase intentions of the cons-~lrness? thus emphasizing tile cognitive element of brand loyalty (condl'iiens e and f of the con.- c e p ~ n a ~ ~ e f ~ l i i t i o c ) ~ 3~:itudir;sl m-ax>h;ircs. 2 iriight ;-',IS<> be easier to zhoose the right decision unit (condition d). They are usualiy based on surveys, and it may be possible to get data from the decision maker rather than the puxchaser by asking questions 410 the righe individual, Finakiy, they give insight into the rnoiivations for the corrsr~mer's cinoice

q ~ v behaviur, and these mativations are less ~ri:eiy to be influenced by random short-run fluctuations:

Hoy:ex~er> attitu.dfl:ii measgrcs mqi net be scc-rafe rep;esell- taticn of ~eaii ly as they are nstbbasea on actual prrrrcmses. A consu-- mer ma.y r~~tionalize his choice when questioned by ths researcher, and make up a~ zvaluaticjn of brands even when no explici! evaluation is made in r e d shopping siiuarions. Moreoverj other variables t h a ~ at- titudes are I~aown to inRuence actxai purchases. For example. an In- dividual n a y have a favo.=sble attit~ade tchwards Porsclle, but still not buy it due to hudget .;onsrr:iii~rs. Xenze, the validiijl of attitiid'l~~al

measures depends on the strength of the attitude-behavior relationship. Furthermore, attituclinal measures are often based on data observed at a sing12 point in time. Their incideiltal n a b r e might be atte- nuated by collecting attitudillal data on a longitudinal basis, but the costs in doing so may quickly become prohibitive.

I~:cii1)r'dzrn!-cirit.?7fei! ! * P ~ . Y L ! s h ra~~! -~r ie l? fe ( ! rzcaslires Brand loyalty is the result of a mental processing of the brand's features by the consumer, which is implied by condition f. Hence, brand loyalty may be seen mostly as a property of the brand ('S features) (Aaker (1991); Rossiter and Percy (1987)) or nlay be considered more as a characteristic of the respective consumers who process that information (Hafstro~n, khae and Choung (1992); Sproles 2nd Kenda?! (1986)) so. Along those lines, we can classify brand-loyalty measures as, respectively, brand-oriented or individual-oriented. This distinction is not always as clear-cut as between attitudinal and behavioral measures, and some operationalizations may even be labeled as brand- oriented in one study and as individual-oriented in another" In our subsequent discussion, we classify these measures according to their most common use in the marketing literature.

If brand-oriented measures are ~ased, a value of brand loyalty is derived for each brand. Differences in loyalty between illdividuals are of less importance, and data are often aggregated across individuals. With these measures, it is possible to compare brands, and to study the influerice of their respective marketing strategies on the resulting brand loyalty. Hoiwever, they are less suited to study the influence of individual characteristics on brand loyalty. Moreover, aggregation problems may arise if the consumer population is heterogeneous with respect to brand preferences. If this is not taken into account, the resulting estimates will be biased (Massy et al. (1970)).

On the other hand, if an individual-oriented measure is used, the loyalty of specific customers is estimated, and it is of less importance to what specific brand that individual is loyal. We may further distinguish individual-oriented measures which quantify brand ioyalty within a specific product category (e.g. cars, soft drinks), and individual- oriented measures which treat brand loyalty as a general characteristic of the consumer (i.e. as a character trait). This information is useful to segment the consumer population, or to study the influence of certain consumer characteristics such as risk avoidance, innovati- veness, or shopping-proneness on brand loyalty. Because little atten-

tion is paid to specific brands, these nleasures are less suited to make co~nparisons between brands.

Sumr?zaly Based on the aforementioned dimensions, four main categories of measurement categories can be distinguished: 1. brand-oriented attitudinal measures (e.g. the proportion of con-

sumers who intend to buy Stella Artois beer the next purchase occasion);

2, individual-oriented attitudinal measures (e.g. the score on an agreement-disagreement scale with the statement: "I like to stick to well known brands");

3. brand-oriented behavioral measures (e.g. the fraction of repeat buyers of Stella Artois beer);

4. individual-oriented behavioral measures (e.g. an individual is brand loyal in his beer consumption if he buys his favorite brand of beer in more than fifty percent of the purchase occasions).

These four categories form the overall framework for our subsequent discussion. A detailed outline is presented in Table 2.

2 measures based on r~~div~dual-level data.

2 general measures D2, sequence-of-porchas measures.

Indiv~di~al-oriented

In this category. we examine (l) measures that use stated purchase intentions or stated preferences, and (2) measures that utilize commitment as an indicator of brand loyalty. The difference between these two subgroups is that the former is measuring intended befiavior, while rile latter is measuring direcrly an essential element or co~npol~ent of brand loyalty.

7. Stated purci~are-intentionlpreference measures

A brand-loyal consumer is iikdy to prefer a certain brand, and has the intention to bluy that brand on h t u r e pwchase occasiol~s. This has lead some researchers to use measures based on stated preferences or on purchase intentions, after which one could derive the propor- tiol; of people preferring the brand.

The earliest effort to quaarhify brand loyalty in this way was made by Guest (1942) who asked individ~~als "Which brand do you prefer?", Since then similar measures have been employed quite frequently in marketing practice (see e.g. Brown (1993); TFest-Aanlisop Fdagazine (Januaay 1992)). Their main disadvantage is that they only ii~dicatt: the :tyndenc9 to buy a upecifir brand: azrd may therefore he only a weak indicato~ O F hnlh ash~; r I behnvioa- and the und~rlylng br2l.d loyalty. On the positive side, "Li-iey are we,?; interpretable, easy to C O ~ - lect within a short period of time, and m a y tlaerefore be an appropriate alternative when actual purchase data are i i ~ d to get (e.g. Srz the case of durable goods with long interpurcilase times). Hence, although the theoretical base sf these measures is weak, they may be quite for practical purposes,

As indicated in cur discussion ol' r he conceptual defirriti~n~ commitment tswarcls a brand is an essential cor~dition for brand loyalty. Hence, it seems logical that brand loyalty can be estimated ic terms of t l i ~ level of cemmitrnent towards a brand (see e.g. Bloerner (1993); Mar- tin and Goodeil (1991 ). Tk~iy j~ i - (1 981)), To &a-sin .. brand..o:iienrec1 measure, the number of customers coi~~mitled to the brand, rr the mean level of ccr~nrnitn~ent is computed,

In the literature, several operatiorializations of cornlnitment have been proposed including direct ratings (see e.g. Traylor (1981)) and i~riirezt appic:aches such as the e:;rant 1 3 v,rhich crle recorni~~cncis ";he

prodr:ct io other people (Baker (1991)). Compared to other attitudinal meastares; commilrnent r-neasolres of brand loyalty are superior as: (Ij an additional element of the ccinceptual definition (condition f) is expiicit!y incorporated, and (2) the link i7etween commitment and behavior is liiteiy to he stronger.

Twcv categories of' individual-oriented a1 tltudinal measures can be identified: (l) measures which define brand loyalty within a specific product category, and (2) those which specify brand Ioyaiiy as a general characteristic of the individual. When general measures are used, ilnique brands are not specified. In contrast, measures on the product- ,-.. ,~ctegory p level explicitly consider the evaluation of a number of brands.

l. Measures on the product-caregoq kvei

individual is likely ?= be brand li;.jraI if he has a E.iigh?y favorable arritude L01~ards cerlain braiids. Therefold, braild atiitlades may - - ". . " used ro r,ons~r~!cr iilal'li~auai-o~ien~ed branc't-ioyaily measures. For expository purposes, we d l ~ u s s "is category using tEic measure deve- Ic>ped by Jacai'by (1 97 1); xvlaich ha:: received considerable atteaation in lnarketlng literature (see e.g. Bennett and Kassarjian (1942); Jacoby and Blssn (19'80); Jacoby, Chestnut, and Fisher (1978); Jamis and Wil- cox (1976)).

The basic idea is that an lndi~id;~~ai 's attitude towards any brandva- ries ffom abssl~~telji acceptable to absolutely unacceptable. If the number of acceptable brands increases, brand switching is more likely to occsjr, an3 !hi: inde'virlrixl .j;rilE become less brand loyal. This notion is I-cilecizd in tile ~ccepta11ce-rejeci1011 scale (scc Figure P 1'. Individual i is expected to be more loyal than individual 2 as only one brand is acceptable for him and he is likely to buy rhat brand on every purchase occasion. In this approach, brand loyaltji can be estimated by:

i-\ '1 (1) the number of brands -11 tile acceptance region; 01- rrie disian- ce between the acceptance and the rejection region, which becomes larger when brand loyalty is stronger.

FIGURE 1

Accepm11ce/l.ejectio7z scales of hvo ilzclividllals

Individual 2

A B C E

C A B E P G

Acceptable Neutral

The Jacoby method incorporates the evaluation process of the individual and allows for multi-brand loyalty. It requires, however, an expensive and time consuming data-collection method, especially if the number of biands unclei study is ~eiativtly large. The measure seeilia rdtiie~ scl~siiive LU 111~ s p c ~ i f i ~ brands chosen. For example, the evaluation of brands becomes blurred if unknown brands are taken into account, since they are unlikely to be rated higher than acceptable brands (Sabonmatsu, Kardes, and Gibson (1989)). Therefore, obscure brands are hardly ever included in the acceptance region, but that may be due to a lack of awareness rather than to an explicit evaluation.

Unacceptable

Acceptable

2. General individual-oriented attitudinal measures

General individual-orientcd attitudinal measures regard brand loyalty essentially as a personality characteristic. Brand loyalty is not pri- marily the result of an evaluation of a specific set of brands but is caused by the consumer's personality or decision-making style. In this t~adit ion, brailcl loyally is estirnaiccl by a battery of statements concerning general individual behavior rather than statements about specific brands. Examples of this approach are the measures of Raju

Neutral Unacceptable

(1980), Sproles and Kendall (1986), and Hafstrom et al. (1992). For instance, Raju's measurement instrument included statements like: "If I like the brand, I rarely switch to another brand" or "I get bored buying the same brands, even if they are good". A score is obtained for each individual, depending on the level of agreement or disagreement with such statements, and this score is interpreted as a general hrand-loyalty measure.

The measures in this subgroup are relatively easy to apply and quantify brand loyalty directly as a propel-ty of the individual. They are useful in studying the influence of consumer characteristics on brand loyalty, and when dealing with new products for which it is uncertain which individuals are most likely to become brand loyal. However, one may question whether it is actually justified to treat brand loyalty as a general characteristic (Assael (1992)). The problem is that although some consumers may, in general, tend to be more brand loyal than others, many other variables (e.g. consumer knowledge of a product- category) also influence their behavior. As their knowledge is not equally high for every product-categoiy, the consumer's loyalty may differ among product-categories, and the predictive validity of these general measures may be limited. A final drawback is that the evaluation and selection of specific brands (condition e of the definition) is not incorporated.

C. Brand-oviented behavioral measures

After discussing brand-oriented (section A) and individual-oriented (section B) attitudinal measures, we now focus on brand-oriented behavioral measures. A number of subgroups are distinguished within this cell on the basis of the measures' data requirements. We distinguish: (1) measures based on aggregated data, and (2) measures based on individual-level data.

1. Measures based on aggregated data

Aggregation across individuals is a common way of obtaining brand- oriented loyalty measures. We discuss measures based on two kinds of aggregated data, namely switching matrices and market shares.

a . M e a s u r e s b a s e d o n a g g r e g a t e d swi tching m a t r i c e s

Brand loyalty may be quantified by distinguishing consec~~tive purchase occasions, and observilig which braids are purchased. If an individual sticks to tile s a n e brand, his behavior cari be characterized as brand loyal. This notion forins the basis for brand-loyalty mea- sl_ares derived from aggregate switching matrices.

k'or a simple two brand scenario, a switching lnatrix indicates how many coilsumers stick to the same brand or switch to another brand on five consecutive purchase occasions. As illustrated in Figure 2, these aggregate switching matrices can easiiy be transformed into a Markov matrix of conditional switching probabilitiesf'. Component (1,2) ef this ma"rix indicates the conditional probability of choosi~~g brand B given that brand A was chosen on the previous purchase occasion. The diagonal elemelmts then represent the probability of staying with the same brand, and can be interpreted as a measure of brand loyalty.

A first-order Idarksv process implies that consecutive purchases are statistically dependent (i.e. the probability of h y i n g brand B in period 1- depends on wha; brand was purchased in ; - I ) and therefore satisfies condition (a) of the concegt~lal definition. Markov matrices have been used quite frequently to study brand loyalty (see e.g. Massy et aB, (1970)), They are easily interpretable, and their analysis is straight-

~. ~

fniward. However, same researchers have c r~ i :~c i~ed rhe use 9f Mzxkov matrirea ii. s ~ ~ ~ r j y i n g brer,d L~'iia-I?~v , ., (see E:lge! 2nd slacI~~!r?I (Ig82t I for a review), One of' the disadvantages is that the consumer pspralationl is assumed to he homogeneous, i.e. that all consumers have the same coladitlonai probabilities. '%'his ass~~irflption Is rather restrictive as consumers are likely to have different preferences towards brands. If the consumer population is indeed heterogeneous in their preferences to-wards "wands, the Markov-based estimate of brand loyalty will be biased (Massy, et al. (1970)).

/ Clirrcnt brand 1 I l

I n 1 70 G O

Mnricov matrix

A parsimonious way to incorporare heterogeneity was developed by Colombo and Morrison (1989)? who distinguished two groups of buyers7: (1) l~ard-core ioyals who buy the same brand with absolute certainty at every single purchase occasion: and (2) pote~itial switchers who cl~oose at every purchase occasioll one of the brands according to a certain probability distribution (e.g. they choose brand X with probability 0.4- and brand B with probability 0.6). The proportion of hard- core foyals can be interpreted as reflecting the magnitude of a brand's Pkjyaltgi base, These estimares are more realistic than those obtained using Markov matrices as they account for the fact that- a repeat purchase does not always inrply brand logialty. Hence, Coiombo and Morrison's measure will contain less spurious loyalty than estimates of brand loyalty based on Markov matrices.

'The measure of Colombo and Morrison is a. special case of latent- class nrodels. For other (more complex) applications of late~t-class madels in the context of brand-loyalty measures, the reader is refer- red to Grover and SI-inivasan ((4383), (1989)), Jain, Bass, and Chen (1990) and Jaii-4 and Rao (2994). The underlying idea of these studies is that the entire consumer population can be divided into different segments. The probability to choose a brand is the same for all consumers of the same subgroup but differs between subgroups? .As with the Colambo and Moi-rison model, the size of the group choosi~zg a brand v~ i th probability one (and ayhicl1 tilerefore is curmpleie:y loyal) is used as tile brena-ioya,liy measure". These complex Sareni-class mu- dels are theoretically superior to the Colornbs and W'orrisofi r~iodel as populatiora of potential svitchers ic divicied f~rrti~ei- into different segments? whicl: is more reaiiskic. Therefore, extended latent class models will provide better estimates of brand loyalty. However, the xiociels are s~ather~~atically comple::, which may inhibit their widespread implementation among n~arketing managers.

b . M e a s u r e s b a s e d o r m a r k e t s h a r e s

Brand loyalty can be yuanej:'ied usillg [he brand-siiccific in1er- cepts in market-share attraction models (Cooper and Nakanish? (1988)). 11-1 lhose iriodels a brand's markct sharc is determined by its relative attractiveness vis-A-,is the other bran~ds. This at"iactiveness itself is 6-t tr;.-ined . # 7 by (1 j ?he .,,raise and effeciiveness or" irs markelii~g-mixva- riables, arnd ( 2 ) a constanh which is assumed to reflect the brand's Io- yaity.

The advantage of brand-loyalty measures based on market shares is that the data are often available at low cost. Moreover, brand 10- yalty is directly related to a manegerially important perfor~nance variable. However, condition (1) of the conceptual definition (biased response) is not incorporated as a high market share might be the result of a zero-order process.

2. Measures based on individual-level data

So far, individual behavior has been aggregated (either in switching matrices or market shares) before deriving brand-loyalty estimates. In the last decade, several measures based on individual-level data have been developed. We discuss brand-loyalty measures that are included as an independent construct in discrete-choice models to predict brand choice1'.

Discrete-choice models are used increasingly to model the selection of brands out of a finite set of alternatives. A particularly promi- nent model is the one developed by Guadagni and Little (1983). They used an individual's sequence of purchases to derive a brand-loyalty estimate for that individual for each brand on every purchase occasion] l. Brand 1's loyalty measure for individual h on purchase occa- s i o ~ n. WL", (n), is defined as a weighted average of this value at the previioils purcliast occasicl; ( )L-I) ancl the picvious p ~ i c h ~ ~ d ~ ~ i b i ~ i i . fulllldl1y.

where HISTORY is a dummy variable which equals one if alternative j is chosen by individual h at purchase occasion n - l and zero other- wise. The implication of equation (1) is that at a given purchase occasion the purchase history of an individual is exponentially weighted1'. Brand loyalty is high for a particular brand, if that brand is bought frequently on recent purchase occasions. The relative influence of the most recent purchase is given by the parameter a. If a is zero, the first factor in equation (1) vanishes, and only the last purchase decision determines the value of the brand-loyalty measure: i.e. the most re- cently bought brand has a brand loyalty of one, and all others have a brand loyalty of zero. In contrast, it is only determined by the very first

purchase if a is one. Hence, the value of a is of utmost importance but unltnown to the researcher. The estimation of this parameter is rather cumbersome (Hadel, Eattin, and Little (1992)), and often the value of a is detel-mined by a grid search on a hold-out sample. In dis- clete-choice models, the variable BL", ( 1 1 ) is incorporated along with marketing mix variables to predict the individual's brand choice, and has 0fte11 bee11 fo~.uld to have significant explanatoly power.

The Guadagni and Little measure in equation (1) does not filter out the effects of marketing-mix variables which may have affected the consumer's purchase history. As shown in Srinivasan and Miba- rian (19901, this may both mask the effect of marketing-mix variables and overstate the loyalty estimate. Moreover, several researchers have argued that the expr ession captures the heterogeneity among consumers rathei than their brand loyalty (Kanetkar, Weinberg, and Weiss (1990)).

Individual measures related to discrete-choice models offer vast op- portunities for brand-loyalty research as (1) they are behavioral measures at the individual level, (2) choice dynamics are incorporated, (3) explanatory variables describing brands and consumers can be added so that both the individual-related component and the brand-related component of brand loyalty are implementable, and (4) the relative irifluence of brand loyalty on brand choice compared to other variables can be studled. However, the data requirements are high as individual purchase data over long time per~ods are needed. Tills issue is becoming less burdensome with the growing availability of scannlel clat-. Qn !he ether hsnd, it is u n r l ~ a r w h ~ t l ~ ~ r the?? mea~14t.e~ give an accurate and unbiased estimate of brand loyalty, which may inhibit their usefulness for marketing managers.

D. Belznviornl individ~~al-orientede measures

Hn the last cell, we consider behavior based approaches that regard brand loyally as a property af the hdividual. Tn this respect. we discuss two main categories: (l) proportion-of-purchase and (2) sequencc- of-purchase measures.

1. Proportion-of-purchase measures

As brand loyalty is a behavioral tendency towards a brand, one could argue that an individual is brand loyal if one particular brand accounts

for a high proportion of his total purchases in the product categoq. This insight is used $1 proportion-of-purchase measures. An example is Cunni~~gham's (1956a,b) llnlirket share criterion, which compn- tes the market share of brands witlziil a iio~~.sehold. This method is a common way to separate loyal from non-loyal consumers (see also Helsen and Schmittlein (1994); Sohnsor? (1984)). An individual is considered brand loyal for a given product czitegory, if the brand purchased most frequently has a market share higher than same cut-off value (often fifty percent)13.

Pn-oportion-of-purcl~ase measures are easy to use and easy to im plement. Their main disadvantage is that they oversiniplify the issue. For instance, more recei-st purchases are not weighted more heavily. Mon-eoiie-d, a high proportion of 1)urch;7se can be the result of a zero- order process which means that condition (1) of the conceptual definition can be violated. Like purchase intention measures, the theore- eicaf value of proportion-of-purchase measures is Limited. However, the researcher may prefer s ~ c h a measure because of practical consl- derations.

2. Sequence-of-purchase measures

The sscond may to obtain an iildivid-sal-oricnted brand-lcyaity ~ncasu- re based or1 che individual consurrmers' pwchase behavior, uses purcllase sequences. A consistent bias an a purchase sequence towards a brand is an indication of brand loyalty. A simple procedure using purc!~ase ,eqlien..:cc i~ the "1-hree i1.i a r ~ ~ . v " criferjr.~! ('rui.kp_r (1454); h!drC<:nl nell (1968)). According to this measure, an individual is considered brand Loyal if he buys a pzrtlcular brand on ~ h r e e cunseci~tive purchase occasions. Thzse rulcs of thumb have similar (dis)advantages as the proportion-of-purcllase lneasures discusses1 in DP. A ~.,dvanced Ivieasrnre is related to the number of brand runs.

A brand run is any sequence rsf consecutive purchases of the same brand. For exr.rq>le, i niil-chase sequence AABWBBABB of braad -C. and B consists of three brand runs. If brand loyalty exists, the fiunlber of brand runs will be szr~ail. AA added benefit of the more advanced ineasures is that they can be used to study Ihr, order of the choice process (Bass etaal. (1984); Massji nr al. (1970)). For exar-srple, the bina- mlal runs rest ~ s e s the izct that, 11 the cllorce process is zea-Q-order, the nur-fiber of brand runs containing a particular brand is distributed 1:ypergeanaetric. Using this result. the expected number of brand runs

is calculated and is compared to obser;>ed ~ ~ n m b e r of runs. If the former is significantly higher than the Iatrer, the process is not zero- oldel znd conditicl~~ (a) of the conceptual defi~iitislr is sztisficd. Hencc, sequence-of-puschaseras measures aie of specirl sheoretical interest as they txahle trs to test an essential c o r i d i t ~ o ~ of brand loyalty.

The four main categories of measurement methods cover different elements of the conceptliai brand-loyalty corpstrrrct. Par example, beha- . . \!iorai measures stress the importance of actual purchase behavior ro detecr brand loyalty, but neglect underlying cognitive processes. i n contrast, attitudinai measures emphasize the imporlance of the cognitive processes, but ignore actual behavior.

Given these measilres' one-sidedness, it seems desirable to construct mixed measures. A number of measures have been developed tkat sirn~altai1eousij7 incorporate attitudinal and behavioral elements (see e.g. Gay (1969); Mehrotra (1984); Newman aa~d Vv'erbei (1973)). An example af such an approach is the dollar-metric procedure des- cribed in Pessemler (19591, which formed the basis for the brand 10- yalty operatisnalization in Raju, Srinivasan and La1 (1 990), and was used by Park and Srinlvasan (1904) in their study oai the measure- p--en",f tralld .-;T--;?.ST X ~ L L I I ~ The idea of t!;i; 1 ~ ~ ~ a s ~ ; r e is ddeteri-fliae ",c

. "

pric,-pi"*Ttinr;r a :,orse;;rlo;- ;; -&.;ijjng .sxj fGsr 1;;; f;;.ori:c brand. I!' . .

that price is high, the consumer is likely to be brand loyal. Since it as impossible to measure this premium by loolilng at actual price and purchase data, a iaboratoq experiment is needed. -As scch, data requirements may be expensive and inhibit the widespread use of the method. Moreover, it is questionable whether laboratory strldies present valid representations of actual "uehvior. Because of these disadvantages: the dollar-metric procedure has only b e n ased sporadical- ip in the marketing Piteraeiia-e. However, the principle of a price-pre- ~nirmm paid by the customer to obtain his preferred brand is increasingly used in recent brand-equity studies (see in this respect Fraliqais and MacLacl~Tan (1995) and Swait et al. (1993)).

Brand loyalty has been studied extensive!?. for academic zts &ell as pactica! reasons. As-{xis emphas;zed in ths In~trocluction, a loyal cuslo-

mer base is a competitive advantage for a brand. However, in order to manage brand loyalty effectively, good measurement methods are necessary. For that reason, this paper has focused on alternative operationalizations of the construct. So far, we have concentrated on the methodological characteristics of brand-loyalty measures, and have only given limited attention to their managerial usefulness. In this section we first point out some academic issues which have not yet been solved in satisfactory way. Next, we consider key issues for proper brand-loyalty research in applied marketing settings.

A. Recomi~zendations for rnarlceling academics

Despite extensive research, many problems still have to be addressed before brand loyalty is fully understood. We concentrate on two of them, namely: (1) improvement of the brand-loyalty measures adopted, and (2) construction of brand-loyalty measures for marketing practice.

Improvement of brand-loyalty measures There is substantial agreement among researchers about the conceptual definition of brand loyalty. Most adhere to the definition of Ja- coby and Chestnut (1978). However, the publication of the Jacoby and Chestnut (1978) monograph has not led to measures that incorporate more conditions mentioned by the conceptual definition than previous operatio~lalizations. This is probably due to the fact that: (1) the background of the re~earcherq qtadying brand loyalty. p~ychologists. economists, and statisticians, differs considerably; (2) a significant number of researchers claim that individual behavior is too complicated to explain, and therefore advocate the use of stochastic models to fit aggregated observed behavior rather then explaining individual differences in behavior.

From a theoretical point of view, one could argue that the ideal measure should include attitudinal and behavioral components. and should be able to reflect both individual-level and brand-related differences. Indeed, the attitudinal and behavioral aspects are two key components of the widely accepted conceptuaI definition of Jacoby and Chestnut (1978). Furthermore, we have argued that both indivi- duai- (e.g. degree of risk aversion) and brand- (e.g. their quality) related characteristics may induce differences in brand loyalty. An aggregation across either dimension could obscure important manage-

rial insights and lead to sub-optimal decision making. Even though we agree that, from a practical point of view, it may not be straight- fo~ward to incorporate in a single operationalization all four dimensions, we feel this remains an important goal for applied researchers.

A promising opportunity in this respect is offered by the emergen- ce of discrete-choice models. Thus far, however, these models have most!y beex estimated on actual purchase data, and have thereby igno- red the cognitive aspect. These models could also be applied to preference data (see e.g. Kamalura and Srivastava (1986)), in which case they would capturelreflect the cognitive dimension. The data collection efforts would quickly become excessive, however, when doing this on a regular basis to incorporate dynamic effects (see in this respect our earlier discussion or, the disadvantages of attitl~dina! measurec\ -1.

Construction for rnarketirzgpvactice Another major avenue for future brand-loyalty research is to bridge the gap between measures used in the academic marketing literature and measures used in marketing practice. Since the start of brand-loyalty research, the technical complexity of the methods to analyze brand loyalty has increased dramatically. However, this may hamper their widespread use in practical applications (Little (1970)). More- over, due to budget- or time constrailrts marketing managers may actually prefer simpler measures over theoretically better ones. More research is needed on the consequences (e.g. in terms of predictix~e validity) of using simple rather than advanced measures.

B. Recommendations to marketing managers

Finally, we provide some recommendations for applied researchers, based on our theoretical discussion of brand-loyalty measures.

Carefully define the product category Great care should be exercised in adequately defining the product category. Indeed, this will determine which brands are included in the analysis, and will therefore influence the resulting brand-loyalty estimates.

Keep it sivzpl2 In applied marketing settings, it may be advisable to use simple measures, as they are often cheaper, easier and faster to obtain. More-

over. more complicated techniques often require data of higher quality. If ~ h e s e data are not available (or are too expensive to collect), increased measurenlent erior inay offset the theoretical advantages of the advanced methods. Also, theoretical research has not yet adequately shown the severity of the (potentially negative) consequences of using simple measures, as was indicated in section 1V.a.

Re care,fill with uni-dimensional nyeasures The vast majority of brand-loyalty measures is uni-dimensional in the sense that they either emphasize the cognitive, behaviorai, brand-related, or individual-related component of brand loyalty. Because of this, the validity of the measures used today is limited. The manager should always consider the specific limitations of the selected me2su- rernent method. If two different measures yield the same conclusions, confidence in the validity of rhe findings increases substantially.

Select a brand-loyalty measure corresponding to the intended purpose As every category of brand-loyalty measures emphasizes different

elements of brand loyalty, no method is suitable for every intended purpose. Therefore, the method chosen should correspond to the purpose of the brand-loyalty study. If the manager wants to use brand loyalty for segmentation purposes, an individual-oriented measure should be used. In this case attitudinal measures (general or at the product- category level) may be most appropriate. The stability of segments based on these measures is greater as they are based on preferences of C O I I ~ U L ~ ~ ~ T S ~vhich are less influenced by short-term fluctuations. When the marketing manager wants to investigate whether repeat buying is either due to inertia or brand loyalty he might use commitment measures. Finally, behavioral measures are inore appropriate when the influence of marketing-mix variables on brand loyalty is important to the marketing manager. In this respect, measures related to discrete- choice models are particularly useful as they estimate brand loyalty at the individual level and offer the possibility to study the interac- tion between brand loyally and marketing-mix variables.

1. Corirtruct validity i~uplies that the measure is measuring thc concept it is iupposed to measurc. A detailed cliscussioil abo~i t the cliii'ereni aspect?, ofvalitliiy i, bcyoilii the scope of this paper. The reader is refevred to Petcr (1981) for a delriilcd trcatmcnt of the issue.

2. A process is first-order if the probability to choose a certain brand depends only on the previous purchase of the consumcr. If more past purchases influence the current choi- cc probability the process is said to he of highcr order (e.g. of second-order if the last t ~ v o p ~ ~ ~ i h i i s e s iillli~ence the ~ u i l ~ l l t blill~d ci~uicc).

3. Until Day (1969), brand loyalty was mcasured almost exci~~sivcly as a behavioral construct. In the beginning of the seventies, more atlentioi~ was paid to the cognitive component of consumer behavior, and attitndinal measures became quite popular. Nowa- days, the use of attitudinal or behavioral il~easurcs depends on tlie purpose of the study. In the marketing-science literature there is a tendency to use behavioral measures. Part of this might be due to the increasing availability of scanner data.

4. For example, we may consider an individual as bl-and loyal if the brand purchased most frequently is bought in Inore than fifty percent of the purchase occasions (Cunning- ham (1956a,b)). This operationalizalionresultsin aninclividual-oricntedmeasul-c. How- cvcr. we Inay also look at the proportion of consumers buying brand A in more than fifty percent of the purchase occasions. Then a similar operationalization (i.e. the 50% rule) results in a brand-oriented measure. Because it is usually einployed in the first way, we will categorize this measure as individual-oriented.

5. We refer the interested reader to Jacoby (1971) for a discussion on the mechanics involved the construction of such a scale.

6. When it is assumed that only the last brand purchased affects the current purchases, we call the Markov process first-order. We refer to Lilien; Kotler, and Moorthy (1992) for a discussion on higher order Markov models.

7. See Bayus (1992), Bultez (1990a,b), and Kannail and Sanshez (1994) for other applications and extensions of Colombo and Morrison's model.

8. It is possible to relax this assumptio~i and account for lieterogeileity within switching segments. This is worked out in more detail by Jain et al. (1990). Although it slightly cliailgca tile iiiicipieiaiiun v1 sulrre of i i ~ c para~ileicrs. ii doea not airer thc basic ideas of the method.

9. Similar measures may also be derived from ~uarke t shares using the stochastic preference model of Bass (19741.

10. One could also interpret the brand-specific intercepts in discrete-choice inodels as a measure of the brand's loyalty (see Cooper and Nakanishi (1988) for a similar interpretation in the context of market-share attraction models). However, these intercepts do not reflect individual differences in brand loyalty. and observed differences in the value of these intercepts Inay reflect differences in other (omitted) variables such as their quality level or their intensity of distribution. rather than true differences in the loyalty of the custoiner base (sec also Iiamakura and Russell (1993)).

I l. For expository purposes. we focus on the nieasure proposed by Guadagni and Liitlc since their measure was tlie first effort to incorporate hraiid loyalty in a discrete choice model. Since then, the ineasurc is used 2nd discussed extensively in the marketing literature (2.9. Gupta (1988); Tcllis (?')SS)), and several rci'i~~crncntz and exlcnsions have been offered (Fader and Lattin (1993); Ortmeycr. Lattin. and Montgomery (1991)).

12. Consider for example an individual i \vith thc purchase sequence ABBAA. For this ill- dividual

RL' (S) = aRI,' (4) + ( ] - U ) . I irnd RL' (4) = nRI.' (7) + jl-u).0. 4 4 4 1

After making the appropriate substitutions. we are able to express

BLf,,(5/ OS: U ' BL' , (-7) + (l-a).

In a similar Lvay B L ' , (5) can bc computed. 13. For slightly different opcratio~lalizatio~ls ofthis criterion. sec e.g. Cliarlton and Ehre11-

berg (1976).

REi:ERENCES

Aakcr, D.A., 1991, Managing BI-and Equity, (The Frcc Press. New York). Agnew. J.. 1987, Children Coine of Age as Consumers. iV1(irketi~ig ~Ve\i,s 21. 25. 8-9. Alonzo, V.. 1994. Til Death Do Us Part, ince~ltil,e 168. 4, 37-42, Assael, I-I., 1992. Consumer Behavior and Markcling Action. 4th ed.. (PWS-Kent Publis-

hing Company; Boston). Bass. EM.. 1974, Thc Theory of Stochahtic Preference and Brand Switching, Jolii7inloj'n/lili1-

ketilig Research 11, 1. 1-20. Bass; EM., Givon. M.M., Icalwani, M.U., Reibstcin, D. and Wriglit; G.P., 1984, An Invea-

tigatiorl into the Order of the Brand Choice Process. 11fni.kelii1g Science 3, 4, 267-287. Bayus. B.L., 1992, Brand Loyalty and Marketing Strategy: an Application to Home Ap-

pliances. 2v;Ciiijici;iig Scieiice 11. 1. 21-38. Bennett, P.D. and Kassarjian. H.1-I., 1972; Consumer Behavior (Prentice Hall, New-Jer-

S"). Bettlnan, J.R., 1979, An Information Processing Theory of Consurner Choice, (Addison-

Weslep Reading, Massachusetts). Bloemcr, J.; 1993, Loyaliteit en tevredenheid (Universitaire Pers Maastricht, Maastricht). Brown, G.H., 1952. Brand Loyalty: Fact or Fiction?,Acil~erlisi?zgAge 23, June 19,53-55; June

30, 45-47; July 14. 54-56; July 28. 46-48; August 11. 56-58; September 1. 76-79. Brown, R., 1993, Keeping the Cable Customer Satisfied, Brondcastir~g nr~rl Cable 123, 10,

10. Bultcz, A., 1990a, Competitivite: Modkles et rkalitts; Partie 1: Modkles descriptifs. (Euro-

pean Institute for Advaiiced Studies in Management), Working Papcr 90-04. Bultez, A., 1990b. Competitivite: Modkles et rkalitks; Partie 2: Modeles explicatifs. (Euro-

pe:ln Institute for Aclvanccd Stirdies in Management). Working Paper 90-05. Charlton, P. and Ehrenberg, A.S.C.; 1976, An Experiment in Brand Choice. Jourrzal ofMnr-

IcetingResearclz 13, 2, 152-160. Colombo, R.A. and Morrison, D.G.. 1989. A Brand Switching Model with Implications for

Marketing Strategies. >ilirrketing Science S, 1, 89-99. Couper, L.G. and ?YTakanidii, AI., 1988. M;irkct-Sl;;irc A;:a!j.sis: Evalaaticg Co~::pcti:ivc

Marketing El'fectiveness, (Kluwer Academic Publishers, Boston). Copeland; M.T., 1923. Relation of Consumer's Buying Habits to Marketing Methods, H N ~ -

vcird Bzlsi~~ess Review 1, 2, 282-289. Cunningham, R.M., 195ha, Brand Loyalty - What. Where. I-Iow Much?, Hrrn~ard Busiizess

Re1:iew 34. 1. 116-128. Cunningham, R.M., 1956b. Brand Loyalty - What, Where. How Much?. Jourrlnl of~\.Ial.ke-

ti~ig 21, 2, 206. Davis, H.L., 1976, Decision Making within the I-Iousehold. Jour~lnl of C o ~ ~ s r ~ m e r Resenrcll

2, 4; 241-258. Day. G.S., 1969, A Two-Dimensional Concept of Brand Loyalty, Jozrr?~nl oj'Adver/isiilg Xe-

searclz 9, 3, 29-35. Day, G.S., Shocker, A.D. and Srivastava, R.K., 1979, Customer-Oriented Approaches to

Identifying Product-Markets, .lou~iznl of ~Marlcelirig 43, 4 , 8-19. Ehrcnberg, A.S.C., 1972, Repeat Buying: Theory and Applications, (North-Holland Pu-

blishing Company, Amsterdani ). Engcl, J.F. and Blackwell, K.D., 1982, Consumer Bchavior, 4th ed., (The Drydcn Press, New

York). Fader, P.S., Lattin, J.M. and Little, J.D.C., 1992, Estimating Nonlinear Parameters in the

Multinomial Logit Model; hlarketillg Science 11, 4, 372-385.

Fader. P.S. and Lattin. J.M., 1993. Accounting for Neterogeneity and No~istationarity in a Cross-Sectional Model of Consumer Purchase Beilavior.!bIalketiilg Science 12.3. i0J-317.

Foxall, G.R.. 1987. Radical Behaviorism and Consu~ner Research: Theoretical Promise and Empirical Proble~ns. I~1i~i7raiioi1ciI Jo~iu~( i l oj'Rescrirc11 iil Ahiiketiilg 4. 2, 111-129.

Frailpis. P. and MacLachlan. D.L.. 1995. Ecological Validation of Alternative Customer- Basccl Brand Strength blcasures, 1niei.rlcitionnl Joilrnal of liesei~rcll 111 ~Vl~rrkriiilg Inirrria- iioilal 12, 1. 321-332.

Grover. R. and Srinivasan. V., 1987. A Simultaneous Approach to Market Segmentation and Market Structuring, Joloiii~ritrl oJ~VIarl~eting Reseiirch 24, 2, 139.153.

G~ove i , R. ;uld Sri~iivil,ar~. V., 1989; An Approach for Tracking Within-Segrnr~it Shifts in Market Sharcs. Jolorirnol oJ'~\.lnrlceiii~gResearc/~ 26, 2; 230-236.

Guadagni, P.M. and Little, J.D.C.. 1983, A Logit Model of Brancl Choice Calibrated on Scan- ner Data, ~\darketing Science 2, 3, 203-238.

Guest, L.P.. 1942. Last vs. U s ~ ~ a l Purchase Questions. Journal ofApplieil P.~!'cliolog~ 26. 2, 180-186

Gupta, S.. 1988. Impact of Sales Promotions on When, What. and How M~lch to Buy. Jo~ii.- iinl of ~Mcrriceting Researcl1 25; 4. 342-355.

Haistrom, J.L., Chae. Jung Sook and Cholung. Young Sook, 1992, Co~~sumer Decision-$La- king Styles: Comparison Between United States and Korean Young Consumers. The Jolcr- ilnl of Coirsr~n~erAairs 26, 1. 146-158.

Helsen, I<. and Schmittlein, D., 1994, Understanding Price Effects for New Nondurables: How Price Responsiveness Varies Across Depth-of-Repeat Classes and Types of Con- sumers; Ez~ropenn Jounznl ofOyerotiona1 Resenl~l~ 76; ?. 359-374.

Hoyer, W.D., 1984, An Examination of Cons~~mer Decision Making for a Conimon Repeat Purchase Product, .loz~rnal of Corzsuiner. Kesearcli. 11; 3, 822-829.

Jacoby, J., 1971, A Model of Multi-Brand Loyalty, Jolrrnnl ofAdvertising Reseorcli. 11, 3. 25-31.

Jacoby, J . and Chestnut, R.W., 1978, Brand Loyalty: Meas~~rement and Management (John Wiley & Sons, New York).

Jacoby. J., Chestnut, R.W. and Fisher. W.A.. 1978. A Behavioral Process Approach to Information Acquisition in Nondurable Purchasing; Jololirnitl oflMnrketi~igiZeseal.clz 15.4. 532-544.

Jacoby, j. and Olson, J.C., lY'7U, An At t~tud~nal Model of Brand Loyalty: Conceptual Un- derpinnings and Instrumentation Research, Paper Presented at the University of Illi- nois Confere~lce on Attitude Research and Consumer Behavior, Urbana, Illinois, (Avai- lable 3s Purdue Papers in C n ~ i i ~ m c r Psychology. 159).

Jain; D.Y.M., Bass, F.M. and Cllen, Y.M., 1990, Estin~ation of Latent Class Models wit11 Heterogeneous Choice Probabilities: an Application to Market Structuring. Jo~irnal of Marketing Researciz 27, 1, 94-101.

Jain, D.C. and Rao, R.C.. 1994, Latent Class Models to Infer Market Structure: a Compa- rative Analysis, European Jolirnal of Operrrtior~al Research 76, 2, 331-343.

Jarvis, L.P. and Wilcox J.B., 1976, Repeat Purchasing Behavior and Attitudinal Brand Lo- yalty: Additional Evidence, in Bernhardt, 1C.L.. ed., Marketing 1776-1976 and Beyond, (American Marketing Association), 151-152.

Jol~nson, T., 1984, The Myth of Declining Brand Loyalty, .lournnl ofAd~'ertising Researclz 24, l , 9-17.

ICamakura, W.A. and Russell, G.J., 1993, Measuring Brand Value with Scanner Data, Inter- notional Joz~rnnl of Research in Marketing 10, 1, 9-22.

Kamakura. W.A. and Srivastava? R.K., 1986, An Ideal-Point Probabilistic Choice Model for Heterogeneous Preferences, Marketing Science 5, 3, 199-218.

Kanetkar, V., Weinberg, C.B. and Weiss, D.L., 1990, Alternative Brand Loyalty and Ad Ex- posure Definitions in Brand Choice Models, Paper Presented at the Marketing Science Conference, March 22-25, (University of Illinois, Urbana-Champaign).

Kannan, P.K. and Sancl-rez, S.M., 1994, Competitive Market Structures: a Subset Selection Analysis, Mnnogement Scieiice 40, 11, 1484-1499.

Krisiinamurtlii. L. and Raj. S.P., 1991. An Empirical Ai~alysis of tlic Reiationship hciwecii Brand Loyalty aiid Consumer Price Elasticity, ~Vl(iil%-etiiig Scieiice 10, 2. 172-183.

Lefton. T.. 1993. Del Monte Puts House on the Market to Cross-Sell Brands, Umnrliveek 34.48. 34.

Lilien. G.L., Kotler; P. and Moorthy, K.S., 1992, Marketing Models, 3rd <:d., (Prenticc- I-Iall. New Jersey).

Little. J.D.C., 1'170. Models and Managers: thc Coilccpi of a Dccisioil Calci l l~~s, ~\Icii~nge- ~ I I C I I T Science 16. S, B466-B485

Martin. C.L. ancl Goodell. P.W. 1991, Historical Descriptive ancl Strategic Pcl-bpectives on tlic Cci~~striict of Fiocii~z; Coiiiriiitii~iiit, Ll~i.up;~iii .T~'oiiiii~il ~J'il'?li~i-Xcii;;g 35. 1. 5.3-60.

Massy, W.F.; Moiltgomery, D.B. and Morrison, D.G., 1970. Stochastic Models of B ~ ~ y i n g Bchavior, (1VI.I.T. Press, Cambridge Massacliusctts).

McConncll, J.D., 1968. The Dcvelopmcnt of Brand Loyalty: an Experimental Study, Jo~ir- niil of ~l.lnrlcetiilg Resenrcl~ 5,1, 13-19.

Mehrotra. S., 1984, Mow to Mcasure Marketing Productivity, Jolrmnl ofAclveriisirrg Rr- serrrclr 24; 3 , 9-15,

Mowcn. J.C.. 1993, Colisurner Behavior, 3rd ed.. ( MacMillan Publishing Company, New York).

Ncwrnan, J.W. and Werbel, R.A., 1973, M~lltivariate Analysis of Brand Loyalty I'or Major Household Appliances; Joziv~al o f ~V/lnrlieti~rg Research 10, 4. 404-409.

O'Leary. N., 1993, Brands on Trial,Aclweek 34, 21, 24-31. Ortnieyer, G., Lattin, J.M. and Montgomery, D.B.. 1991, Individual Differences in Res-

ponse to Consumer Promotions, I~irer71ntioirrrl.lo1i~~nl ofResearcl7 in ~VIarketing 8.3, 169- 185.

Park. C.S. and Srinivasan, V., 1994. A Survey-Based Method for Measuring and Under- standing Brand Eq~ii ty and its Extendibility, .lo~{rnnl of Mrrrketing Resenrcl~ 31, 2; 271- 288.

Pessemier, E.A., 1959, A New Way to Determine Buying Decisions, Joirnral of~llrr~.keting 24, 2, 41-46.

Peter, J.P., 1981, Construct Validity: a Reviea of Basic Issues and Marketing Practices, Joiir- 1101 of 1I4rri-ketiilg Research 18, 2. 133.145.

Raju, J.S., Srinivasan, V. and Lal, R., 1990, The Effects of Brand Loyalty on Competitive Priuc Prurliutio~lal Strdtcgics; ? ~ I ~ ~ ~ i c i g c ~ ~ ~ e i z i S L ~ L I I I L C 36, 3, 276-30;.

Raju, P.S.. 1980, Optimum Stinlulation Level: Its Relationship to Personality, Demogra- p h i c ~ , and Exploratory Behavior, Jorirnnl oJ' Coilsriiner Researcl1 7, 3, 272-282.

Rrirclihrl( F.F. iinil Siltrer. WE.. 1990. Zero Defections: Quality Comes to Services. IIrir- i,aicl B~lsiness Review 68, 5, 105-111.

Riezebos. H.J., 1994, Brand-Added V a l ~ ~ e , (Eburon Publishers, Delft). Rosenbcrg. L.J. and Czepicl, J.A., 1983, A Marketing Approach to Custoiiier Retention;

Joririz(r1 OS Coi~sroner M~rrlreiing 2, 45-51. Rossiter. J.R. aiid Percy, L., 1987, Advertising and Promotion Management. (McGraw-

Hill, New York). Saboninatsu, D.M., ICardcs. E R . and Gibson, R.D., 1989, Tlie Impact of I~iitial Proccssiilg

Goals on Memory-Based Brand Comparisons, in S~LIII, T.K.. ed., Advaiices in Coilsumer Research 16. (Provo, Utah). -129-432.

Sel-afin, R. and Horton. C., 1'194. ALItoma~crs Cio a Loiig Way to Keep Fiock Loyai,/ii/~,er- iisingAge 65, 13, S2+.

Sproles, G.B. ancl Keiidall, E.L., lC186, A Methodology for Profiling Consumers' Decision- Making Styles, Tlze .loirlnnl q/'Co17surnerA,ffrrirs 20, 2, 267-279.

Srinivasaii. V. and Kibnrian, T.. 1990, Purchase Event Feedback: Factor or Fiction?. Papel- Presented at t11e Marketing Science Conference, March 22-25, (University of Illinois, Ur- bana-Chaiiipaig~i).

Swait, J.. Erdem, T., Louviere. J., and Dubelaar. C., 1993, Tlie Equalization Price: a Measu- re of Consumer-Perceived Brailcl Equity, In/erizntioi~al Jorirrial of Resenrch in ~VInrketii~g 10. 1, 23-45.

Tellis. G.J.. 19SS. Advei-tisiiig Exposure. Loy:illy. and Uraiid Purchase: a X\(>-Stage Mode: of Cl~oice, Jolii.i1~1 o,/'l\lii%('tii~g RCSCI I~CI I 25, 2. 1.34-l4-J.

Test-Aankoop Magazine. 19$)2. !Costprijs en hctro~~w-i?ani-lleiiivlarliei van wagens. P ~ / - / l i ~ i ~ k o ~ j ) 1110-

,~iizi/ ie 340. l. 0-13. Tvayloi-. M.B.. i9:;l. Product Irivolvciiient and Brariii Corni?iitiiicnt. .lorir.il/il c!fAtil~i.i . i i~i~ig

Resi.cii.cl~ 21, (1. 51-50. X~clier. WT., 1904, The Dzvelopmenr of l3ra1i.d Loyalty. Jouri!iii ofJliir.krtir~g Rc,si.~ii.c ii 1. 3.

32-35. Wilkie, W.L.. 1'190. Consuil~er Be11avio1-, 2nd ed.. (Johii W i l y Kr Sons. New Y~>i.k)

Brand Loyalty Measures

Documents

Transcript of Brand Loyalty Measures