Australasian Marketing Journal - Volume 13 - Number 2 - 2005

Editorial
Editorial
amj Progress and Prospects

Paul Patterson & Mark D. Uncles
The Australasian Marketing Journal (amj) is a quality academic journal for scholars, educators and practitioners covering all core areas of marketing, from branding and advertising to customer relationship management and services management, and from marketing metrics and knowledge management to consumer behaviour and cross-cultural studies. It is the official journal of the Australian & New Zealand Marketing Academy (ANZMAC) and is published two to three times a year. In this editorial we comment on the health of the journal today and offer some thoughts about future opportunities and challenges. A Story of Steady Progress Building on a history that extends back to 1993, the amj has progressed to become the leading Australasian marketing journal for researchers and educators in the region. It prides itself on being a generalist journal in which key marketing issues are examined. A sign of this is the breadth of topics that have been featured. The most heavily represented topic over the 5 year period 20012005 was relationship/loyalty marketing (8 papers), followed by exporting/international marketing, social marketing, education, and methods/techniques (6 papers each), advertising, consumer behaviour and e-commerce (5 papers each), branding (4 papers), NPD/diffusion (3 papers), customer service, networks, pricing and strategy (2 papers each), and country-of-origin, culture, complaints/service recovery, and product strategy (1 paper each). Considerable impetus has been given to a number of key topics through the publication of special issues, notably: Services Marketing edited by James Barnes, Greg Elliott and Paul Patterson [2002, 10 (1)]; Social Marketing edited by Janet Hoek [2003, 11 (1)]; Emerging Issues in Services Marketing: emotions, emarketing and encounters edited by Janet McCollKennedy and Paul Patterson [2003, 11 (2)]; and
Empirical Generalisations in Marketing edited by Mark Uncles and Malcolm Wright [2004, 12 (3)]. The emphasis throughout has been to publish quality research. In 2001, as a visible sign of this commitment, the amj best paper prize was instigated. The editors select a shortlist of papers from the previous year, which is put to a distinguished panel to select a winner. The prize is awarded annually at the ANZMAC conference. Previous winners have been: Ian Wilkinson, A History of Network and Channels Thinking in the 20th Century [2001, 9 (2), 23-52]; Liliana Bove and Lester Johnson, Predicting Personal Loyalty to a Service Worker [2002, 10 (1), 24-35]; Josee Bloemer and Gary Odekerken, Antecedents and Consequences of Affective Commitment [2003, 11 (3), 33-42]; John Scriven and Andrew Ehrenberg, Consistent Responses to Price Changes [2004, 12 (3), 21-39; Sara Dolnicar, Roman Freitag and Melanie Randle, To Segment or Not to Segment? An Investigation of Segmentation Strategy Success Under Varying Market Conditions [2005, 13 (1), 20-35]. The publication process is supported by a group of expert referees and a double-blind peer review process. In this the editors are greatly assisted by members of the Editorial Review Board and, in addition, they have been fortunate in the advice received from ad hoc reviewers. Not only has this resulted in constructive feedback for authors, but it also has been instrumental in keeping review times to an average of only 8.2 weeks with 76% of manuscript reviews turned around in 6 weeks (based on figures for 1st and 2nd reviews, November 2004November 2005). In 2002, to recognise the contribution of reviewers, the amj best reviewer prize was instigated. The editors select the recipient on the basis of the comprehensiveness of the reviews, constructiveness of feedback and response time. The award is granted at the annual ANZMAC conference. Past winners have been
Australasian Marketing Journal 13 (2), 2005
Editorial
Tony Pecotich (2002), Louise Young (2004), Rob Lawson (2004) and Janet Hoek (2005). Internationally, the profile and reputation of amj is growing, assisted by on-line access and systems such as Proquest and ELMAR. In addition, the Editorial Review Board not only includes experts with Australian and New Zealand affiliations, but also lists leading researchers from China, Korea, Netherlands, Singapore, Thailand, UK and USA. It remains the case that the majority of authors have an Australian or New Zealand affiliation 59% and 18% respectively (based on figures for 20012005, whereby if two authors affiliate in different countries then 0.5 is allotted to each and so on for greater numbers of authors). Nevertheless, 14% of papers are written by authors with a European affiliation, 6% with a North American affiliation, and 3% with a SE Asian affiliation. The amj primarily exists as a vehicle for publishing peerreviewed scholarly research, but as the journal of a professional society, ANZMAC, it also has an important role disseminating ideas on educational philosophy and practice. To this end, amj has carried articles on student learning, the barriers to deep learning in student teams, the pros and cons of journal rankings, advice for early career researchers, the writing of research theses, and related themes. More broadly, the journal is seen as a resource for researchers and educators in marketing, and carries research notes, case studies, book reviews, executive summaries, invited commentaries and editorials. Noteworthy too is the investment made in the development of a journal web-site [accessible through ANZMAC (www.anzmac.org) or through the School of Marketing at UNSW (www.marketing.unsw.edu.au)]. All back issues from the first issue in 1993 to the present day are freely available, providing access for marketers wherever they are located in the world. Future Prospects Over the coming years the journal will continue to evolve and, we trust, gain in stature. Crucial to this process is the continued flow of manuscripts these are the lifeblood of the journal. We have every confidence that quality manuscripts will continue to be submitted. Nevertheless, it would be good to see the number of desk rejections fall since August 2000 roughly 16% of submissions have been desk-rejected because papers have not met even minimal submission requirements or because they have not fallen within the coverage of a
general marketing journal. Prospective authors are advised to give their work a final polish before submission, thereby reducing the chances of desk rejection, minimising the number of revision cycles and increasing the chances of eventual publication. A challenge will be to continue to raise the profile and reputation of the journal. On-line access has been of immense importance, and this has seen a growth in citations to articles published in the journal. But more can and should be done in this regard. One avenue for achieving this is to build on the evident success of our professional society, ANZMAC. The journal provides a very tangible benefit to members it is a means of helping members keep up to date with the latest scholarship in the region and beyond, as well as giving access to resources on educational practice, cases, commentaries and so forth. It is to be expected that these links will be strengthened. Also we would expect to see the journal reaching out beyond the confines of the academic and educational communities of Australia and New Zealand. Commercially, the journal already benefits from links with marketing and marketing research organisations in Australia and New Zealand, notably its long-term association with the Australian Marketing Institute (AMI). In recent years Executive Summaries have been distributed to members of various professional associations in the region, with very positive responses. In future it may be possible to exchange and cross-link more resources, with a view to stimulating greater dialogue between the research and business communities. Geographically, the amj has an important role to play in the Asia-Pacific region, as is already reflected in published articles focusing on Asia, the international composition of the Editorial Review Board and plans for a special issue on Marketing in China. In saying this we do no more than echo the remarks of Peter Chandler who, in the very first issue, wrote: For a marketer, there is surely no more exciting place in the world to be than in the Asia Pacific region. For most of the countries which comprise the region, gross national product is outpacing the rest of the world and associated with this, there is rapid economic development which in turn is generating many new business opportunities. Both marketing practitioners and marketing academics at tertiary institutions such as mine need to stay abreast of the innovative approaches which are working in the region. They also need to be better acquainted with the social, political and economic fabric of the countries
Editorial
within which they operate. I hope that the amj will help them in this quest and that as a result, they will share in the excitement and rewards available during this golden age [1993, 1 (1)]. If anything, these words are of greater significance today than they were when first uttered over a decade ago. Parting Thanks For any enterprise of this kind to succeed there must be a team of committed helpers and assistants. We have already acknowledged the importance of our ad hoc reviewers and Editorial Review Board. We also acknowledge the help and advice of the ANZMAC Executive and particularly the publications subcommittee. Production of the journal would not be possible without the financial support provided by ANZMAC members and personal and institutional subscribers, however we are indebted to the School of
Marketing, University of New South Wales (UNSW), for generously providing the greatest proportion of funds. As editors, we have had the benefit of invaluable support, dedication and commitment from our Editorial Assistant, Nadia Withers, and Production Editor, Irine Kay at Blubeetle Studio. Additional assistance in managing the web-site and distributing hardcopies has been provided by Margot Decelis and Paula Aldwell. Our editorial predecessors, Frank Alpert and Lester Johnson, gave us charge of a thriving amj in 1999. Over the intervening five years the journal has gone from strength to strength. Now we have pleasure in wishing our successor, Richard Varey, every success for the future. Paul Patterson & Mark Uncles School of Marketing University of New South Wales
Executive Summaries
Executive Summaries
Consumer Loyalty: Singular, Additive or Interactive? Robert East, Philip Gendall, Kathy Hammond & Wendy Lomax Do profit-making actions such as customer retention and recommendation have a common basis in a single measurable form of loyalty? The dominant theory is that combination measures of past behaviour and attitude will serve best in predicting future loyal behaviours, though evidence on this matter is scant. We test the dominant theory using attitudinal and behavioural measures to predict the retention and recommendation of supermarkets, cars and a variety of services. We find that attitudinal measures predict recommendation well but behavioural measures do not. We find that behaviour measures predict later retention but attitudinal measures give weak or no prediction of such retention. Finally, we find no evidence that attitudinal and behavioural measures work synergistically. We conclude that there is no benefit in using a combination measure of attitude and behaviour to predict recommendation and retention. This evidence implies that practitioners should be cautious about the inferences that they make about retention and recommendation. Customers with a strong patronage history cannot be expected to recommend more, though they may be retained better than others. Conversely, more satisfied customers may recommend more (and thereby raise customer acquisition) but their satisfaction will have little effect on their retention. Repeated Binary Logit: Analysing Variation in Behavioural Loyalty Cam Rungie & Gilles Laurent Large data sets recording detailed repeat purchase behaviour are now available to brand and store managers.
We show how these data can be used to identify segments on the basis of the propensity to repeat purchase or to switch. A new method is introduced which analyses multiple characteristics of shoppers (this is the Repeated Binary Logit model). The method determines for individual brands the characteristics of shoppers that reflect typically higher (or lower) levels of repeat purchase. The implications for the marketing mix are substantial. Segments with greater repeat rates will have reduced elasticity and small changes in the marketing mix will have less impact. Furthermore, the method indicates that the differences between brands can be reflected less in their typical customer profiles and more in their loyalty structures. With this method there is now a completely new and relatively simple approach to segmentation on the basis of loyalty. Environmental Attitudes and their Relation to the Dominant Social Paradigm Among University Students In New Zealand and Australia William E. Kilbourne & Michael J. Polonsky Traditionally, there has been a view in Western societies that in order to get consumers and businesses to be more responsible, there needs to be concerted effort to get these groups to better understand their impact on the natural environment. As such, many governmental bodies and social marketers have tried to focus on increased awareness of how consumption and production impact the natural environment. Raising consumer and industry awareness is indeed important. However, it is suggested that there is an underlying set of attitudes regarding the dominant social paradigm- i.e. how humans see their role in the biosphere - that may be more important in regards to changing consumption behaviour. The paper suggests that it is this underlying philosophical view that in fact dictates, or at least moderates, broader environmental attitudes and behaviour.
Executive Summaries
As such, those seeking to minimise environmental degradation resulting from mass consumption, need to focus on re-orienting societal views towards consumption and humans role in the environmental system. Such changes will open the way for more pragmatic shifts in behaviour making it more environmentally responsible. This means that much of present socially oriented environmental marketing may fail to be successful as there are underlying impediments to changes in attitudes and behaviour. For any governmental action to be effective in changing behaviour, it needs to reposition how individuals see their role in the world- i.e. changing the dominant social paradigm. The research suggests that changes in the dominant social paradigm will bring about changes in both attitudes and behaviours within the Western societies of Australia and New Zealand. As such, this approach may be more appropriate in all western societies. Evaluating a Model of Industrial Relationship Performance: A Comparison of European and Chinese results using the IMP Data Base James Wiley, Ian Wilkinson & Louise Young Business relations are the means by which key resources are accessed and co-developed, non-value added costs are reduced, and value is co-created and delivered to customers and other stakeholders. The development and management of relations pose particular problems for firms engaged in international business activities because they involve dealing with counterparts from different social and business cultures with the attendant problems of communication and understanding. As the world globalises, such relations and networks become more frequent and necessary. This paper examines the extent to which a model of international business relationship performance developed in a Western business context applies in China. The role and importance of relations and networks in Chinese business is emphasised in concepts such as guanxi, obligation driven contracting and face. Business relations might function in different ways as a consequence. Our results show that the same model generalises across European and Chinese contexts, not withstanding the fact that Chinese international business relations generally were initiated more recently and are
less developed because they have only recently been allowed to trade internationally as independent enterprises. Though some measurement issues need additional research, our results provide some comfort that business relations tend to function in similar ways in such different contexts. Perhaps the nature of international business relations forces Chinese firms to adopt more Western styles of relationship development. Designing and implementing Internet questionnaires using Microsoft Excel Iain R. Black, Alejandra Efron, Christina Ioannou & John M. Rose This research note details the use of Microsoft Excel to design and implement internet based questionnaires. Data for two different projects was collected using specially configured workbooks as a platform to ask questions and collect answers. These were then saved and distributed as Excel files. It was possible to design a broad range of question types and answer formats using Excel. In a study on customer satisfaction typologies, open and closed questions were used and answers were provided on multiple choice, Likert, semantic differential, binary and ratio scales. In the second study, a choice experiment, the choice profiles and answers were collected along with a range of demographic information. The design of the questionnaires, collection of data, coding and preparation were found to be straightforward, involved minimal training (for intermediate level Excel users), allowed flexibility in format and structure and saved considerable time compared to paper and pencil methods. We believe that Microsoft Excel provides a legitimate option for designing internet questionnaires when compared to paper and pencil methods. In addition, where the cost of specific software such as Websurveyor is too high (and added to by training requirements), Excel provides a less expensive alternative with costs restricted to additional coding time and becoming familiar with a range of functions. It may be of particular interest therefore to academic researchers, those completing research degrees and internal research in commercial organisations. Its usability in these situations is helped by pre-existing relationships that may exist. This is vital so as to minimise the difficulty of reaching appropriate samples because of internet security programs.
Consumer Loyalty: Singular, Additive or Interactive?

Robert East, Philip Gendall, Kathy Hammond & Wendy Lomax
Abstract Consumer loyalty may be defined as a singular concept, usually as an attitude toward the loyalty object or as repeatpatronage behaviour; alternatively, the definition may combine attitude and behaviour in either an additive or an interactive expression. We argue that definitions of loyalty are useful if they predict phenomena such as recommendation, search and retention (loyalty outcomes). In three consumer fields, we find that combination measures of customer loyalty often perform poorly as predictors of loyalty outcomes compared with singular measures since recommendation is predicted by attitude but not by repeat patronage, whereas retention and search behaviour are predicted better by repeat patronage than by attitude. We also find that the prediction of loyalty outcomes is not improved by the inclusion of an interaction term in the model. On this evidence, we argue that combination concepts of loyalty are of limited value. Further, we find that there is no form of loyalty that consistently predicts all the different loyalty outcomes and, therefore, we should abandon the idea of a general concept of loyalty. Keywords: Consumer loyalty, Recommendation, Search, Retention
1. Introduction 1.1 Aims Concepts of loyalty and their associated measures are of value if they can be used to predict behaviour such as customer recommendation, reduced search for alternatives and high customer retention (hereafter termed loyalty outcomes). The issue addressed here is how well different measures of loyalty predict these loyalty outcomes. We compare singular definitions, based on either behaviour or attitude, with definitions that combine behaviour and attitude. We test which of these definitions of loyalty help us to predict different loyalty outcomes, and we ask whether one definition alone will usually predict all three outcomes, thus justifying a general concept of loyalty. We conduct these tests using three sets of studies on supermarket, car and service categories respectively (described and analysed in sections 5-7).
Before this, we examine the different definitions of loyalty (section 1.2), report previously established evidence on combination measures of loyalty and the predictors of retention (section 2), set out a theory predicting retention, search behaviour, and recommendation (section 3), and draw hypotheses based on this theory (section 4). 1.2 Definitions of Loyalty Loyalty to an object (e.g. a brand, store, service or company) is shown by favourable propensities towards that object. These propensities may be behavioural or attitudinal. In industrial and service marketing, behavioural loyalty is viewed as retention of the brand (e.g. Reichheld 1996; Reinartz and Kumar 2000). For services, particularly those in semi-continuous use such as mobile-phone airtime, such retention can be measured by the duration of time that the customer has used the service and, for durables, by the customers repeat
10
purchase of the brand. In markets such as groceries, where customers may use several brands in a category, researchers have used the share-of-category expenditure to measure customer loyalty (e.g. Baldinger and Rubinson 1996; Bhattacharya 1997; Bhattacharya, Fader, Lodish and DeSarbo 1996; Deighton, Henderson and Neslin 1994). Another behavioural measure of customer loyalty, which was used by Hauser and Wernerfelt (1990), is portfolio size; this is the number of brands used in a period (the larger the number, the lower the loyalty). Turning to attitudinal measures, liking the brand has been used as a predictor of retention (e.g. Baldinger and Rubinson 1996) while other researchers have explained loyalty in relation to satisfaction (e.g. Oliver 1999; Shankar, Smith and Rangaswamy 2003), commitment (e.g. Bloemer and de Ruyter 1998; Pritchard, Havitz and Howard 1999), and trust (e.g. Ennew and Binks 1996; Morgan and Hunt 1994). Definitions may be singular, in the sense that they focus on single concepts, either attitudinal or behavioural. However, much of the discussion of loyalty has centred on more complex definitions that may include both antecedents and consequences of loyalty. These more elaborate treatments have often represented consumer loyalty as a combination of concepts. For example, Jacoby and Chestnut (1978) proposed a six-component definition of loyalty that included both attitude and behaviour. This approach to the conceptualisation of loyalty seems to be partly related to a need to represent the causes of loyalty, and partly semantic, that this is what the term loyalty means. Semantic considerations provide a start in scientific definition but the idea that attitude-behaviour congruence is required for loyalty seems ill founded. In interpersonal contexts, loyalty is shown when persons do not undermine others by what they say or do. A person who has undisclosed misgivings, yet still behaves supportively, is seen as loyal. Thus, everyday usage appears to link loyalty more with behaviour than attitude. The inclusion of potential causes in the definition of loyalty is typified by Oliver (1999); he defines loyalty as a deeply held commitment to re-buy or re-patronise a preferred product/service consistently in the future, thereby causing repetitive same-brand or same brand-set purchasing, despite situational influences and marketing efforts having the potential to cause switching behavior. This definition implies a positive correlation between attitude and behaviour in loyal groups but Oliver reports that this association is usually quite weak and concludes
that true loyalty, with emotional commitment to the brand, is rare. A problem raised by complex definitions such as that of Oliver (1999) is that, by incorporating the causal explanation of loyalty into its definition, it becomes impossible to test this causality without circularity. It may be better to focus on the essence of a concept and to exclude potential causes and effects from the definition. Here, we can learn from an earlier and related controversy in the attitude-behaviour field, which was described by Fishbein and Ajzen (1975). These researchers were faced with a wide variety of definitions of attitude and commented (1975, p10): Clearly, there exists a great diversity of viewpoints concerning the attitude concept, and this state of affairs is reflected in a multitude of definitions of attitude. Many of the disagreements among investigators are questions of theory rather than definition. For example, we saw that many definitions of attitude make explicit reference to the nature of the disposition or to factors that influence it. Theorists usually have not made clear which aspects of an elaborate theoretical description of attitude are essential in defining aspects of the concept and which are speculative arguments that require empirical verification. Fishbein and Ajzen (1975) stripped attitude back to its essential evaluative meaning and treated cognitive and behavioural factors as separate concepts that were linked to attitude in a testable causal model. The theories that used this simplified definition of attitude (reasoned action and, later, planned behaviour) have provided major explanations in social and consumer research. We propose a similar approach to the definition of loyalty and thus prefer simple definitions that express the essence of the concept rather than definitions that incorporate possible antecedents and consequences. Another complex definition of loyalty treats it as an interaction between attitude and behaviour. A definition of this form is justified when each component facilitates the other and one component on its own is insufficient, for example when motivation and ability are both required for high performance. Day (1969) measured loyalty in this way and found that this gave stronger associations with customer characteristics than a measure of behaviour on its own. Mellens, Dekimpe and Steenkamp (1996), favoured a combination conceptualisation of loyalty but pointed out that the operational definitions of researchers did not, in
11
Figure 1: Dick and Basus (1994) Framework for Customer Loyalty Relationship
Cognitive antecedents
Social norm
Affective antecedents
Relative attitude
Repeat patronage
Word-of-mouth Search motivation Resistance to persuasion CONSEQUENCES
LOYALTY RELATIONSHIP
Conative antecedents
Situational influence
Figure 2. Dick and Basus (1994) Typology of Loyalty
Repeat patronage High Relative attitude High Low True loyalty Spurious loyalty Low Latent loyalty No loyalty
practice, correspond with this conceptualisation. One of their conclusions was that more research is needed on the consequences (e.g. in terms of predictive validity) of using simple rather than advanced measures. In this paper, we report on this needed research. We test whether combination measures of loyalty predict specific loyalty outcomes better than singular measures and whether one measure predicts a number of different outcomes, thus justifying a generalised concept of loyalty. From here on we use the term repeat patronage to cover the behavioural predictor of loyalty outcomes. 2. Previous Evidence Below we review evidence on the issues raised above.
We examine the support for conceptualising loyalty either as an interaction or as an addition of repeat patronage and attitude, and, because of the affinity between satisfaction and attitude, we also examine the evidence that customer retention can be predicted from satisfaction measures. 2.1 Loyalty as a Combination of Attitude and Behaviour Day (1969) found only weak evidence that repeat patronage, measured as share-of-category purchase, was associated with customer characteristics and suggested that this was because many behaviourally loyal customers were influenced mainly by opportunity and
12
routine rather than by preference. Day proposed that attitude to the product would distinguish the intentionally based truly loyal customers from the spuriously loyal customers whose retention arose from convenience, environmental pressure, or habit. When Day used a product of attitude and repeat patronage as a measure of loyalty he found that both consumption and demographic variables were related to this measure better than they were to a measure of repeat patronage on its own. Day did not compare his interaction model with an additive model based on repeat patronage plus attitude, so it was not clear that the product term raised explanation beyond that of the main effects in an additive model. Thus, Days evidence gave support for a two-dimensional measure but it remained open whether this should have an additive or an interactive form. We have only identified one study in which an interaction term enhanced the prediction of a loyalty outcome; Bolton (1998) found that the retention of mobile phone airtime was predicted better when the model included a term for the product of customer satisfaction and tenure duration. Dick and Basu (1994) presented the framework model of loyalty that is shown as Figure 1. They suggest that relative attitude drives repeat patronage, subject to antecedent and situational constraints, and that an attitude-behaviour association leads to further loyalty behaviours. They state customer loyalty is viewed as the strength of the relationship between an individuals relative attitude and their repeat patronage. (By relative they mean compared with available alternatives because it is the contrast between alternatives that is likely to motivate behaviour such as recommendation). An association between attitude and repeat patronage will not necessarily raise the prediction of loyalty outcomes. To our knowledge, this aspect of the theory has not been tested. It might be tested using a series of studies, but this is not attempted here. However, the model does raise the question of how much attitude and repeat patronage are correlated and we do report on this matter. Dick and Basu also present the typology shown in Figure 2; this divides consumers into four segments using two levels of behavioural loyalty and two levels of relative attitude to the brand. Figure 2 suggests that we should find most of the expected consequences of loyalty (e.g. word of mouth recommendation, reduced search and retention) in the high-high top left quadrant of Figure 2. This Figure seems to indicate a combination rather than a correlational basis for loyalty and we test this, using hierarchical regression to separate main and interaction
effects. Dick and Basu make it clear that their model of customer loyalty should apply to retail, service, frequently purchased goods and industrial contexts, but they do not report systematic predictive tests of the models shown in Figures 1 and 2. 2.2 Related Evidence Baldinger and Rubinson (1996) found that both repeat patronage (share-of-category requirement) and attitude predicted the retention of grocery brands over one year though the effect of attitude appeared to be quite weak. In this work, there was no test for an interaction effect. Baloglu (2002) found that high attitudinal loyalty raised positive word of mouth and reduced search for alternatives but that a repeat-patronage measure did not affect word of mouth and had little effect on search. Another study by Pritchard and Howard (1997) used the Dick and Basu typology to segment travellers; this work showed a number of differences between the four segments but the authors did not investigate loyalty outcomes such as recommendation and retention. Macintosh and Lockshin (1997) also found significant differences between customers divided according to the Dick and Basu typology, but they predicted intention to repurchase, rather than actual retention. Mittal and Kamakura (2001) compared satisfaction-intention and satisfaction-retention links and found limited correspondence between the two. They also argue that when intention is measured at the same time as its predictors, as is usually done, common method effect may inflate the association. Furthermore, Chandon, Morwitz and Reinartz (2005) have found that correlations in surveys between intention and subsequent behaviour are artificially increased by the process of asking about the respondents intentions. This evidence implies that an intention measure may be a poor proxy for actual retention. 2.3 Satisfaction and Retention Studies predicting retention from attitudinal measures have mainly employed satisfaction as the independent variable. If we exclude studies in which intention to rebuy has been used as a proxy for retention, we find only three cases showing a strong association between satisfaction and behaviour. Andreasen (1985) studied ten patients who reported serious dissatisfaction with their medical care and found that six of them switched physicians; Bolton (1998) found that dissatisfaction had a strong predictive power in interaction with customer tenure; and Bolton and Lemon (1999) found that
13
increases in satisfaction were associated with later increase in usage of mobile phones. There are more studies showing weaker effects. Crosby and Stephens (1987) found that life assurance renewal had a low positive correlation with satisfaction with the provider. Kordupleski, Rust and Zahoric (1993) found limited evidence for a lagged effect of satisfaction on retention in company research by AT&T. Reichheld (1993) reported that between 65 percent and 85 percent of customers who defect say they were satisfied with their former supplier. Ennew and Binks (1996) did not find clear evidence of a positive association between service quality (usually closely related to satisfaction) and retention. Additionally, Hennig-Thurau and Klee (1997) have reviewed this field and generally find moderate association between satisfaction and retention. This review shows that satisfaction generally has a positive association with retention but that the effect is often small. One reason for this is that the measure of satisfaction usually employed is not relative. People are likely to be motivated to recommend or retain a supplier because of the superiority of that supplier over others and a relative satisfaction measure should be more sensitive to this difference between alternatives. In our work, we use relative measures of attitude. A second reason for a weak association between satisfaction and retention can be found in the pattern of defection. At least with regard to services, defection often occurs as a consequence of specific failures at the point of delivery (Keaveney 1995); such episodes are unlikely to be anticipated well by an attitudinal measure. Prior satisfaction may sometimes allow the customer to discount the failure, as Bolton found (1998), but there may also be other situations in which a high prior satisfaction makes service failure all the more upsetting. A third reason why defection is often unrelated to satisfaction is that it may be involuntary; this is often found with regard to retail services where location may constrain usage (East, Lomax and Narain 2000). We conclude from this review that attitudinal measures are unlikely to give a strong prediction of brand retention. 2.4 Research on Customer Tenure and Recommendation As noted, one measure of repeat patronage is customer tenure duration and, later, we test whether this is related to customer recommendation. Here, we review evidence on the association between recommendation and tenure.
In 1990, Reichheld and Kenny suggested that long-term customers might refer more new customers than recently acquired customers. In 1996, Reichheld repeated this claim. It seems likely that long-term customers will come to like their suppliers more and that this could then raise the rate of recommendation of these suppliers. Those who are dissatisfied with a supplier are more likely to switch so that survival bias should leave longterm customers who are more positive toward their supplier. But other factors could oppose any positive association between tenure and recommendation. Less interesting services do not bear repeated recommendation to the same person so long-tenure customers may use up their opportunities to recommend. Also, novelty wears off over time so that the suppliers product may become less salient to a user with longer-term use and, as a result, less talked about. This loss of novelty is likely to be greater in frequently used services, such as credit cards, compared with infrequently used services, such as car servicing. Loss of novelty will be offset if the offering changes frequently, e.g. fashion stores. These considerations suggest that recommendation rates could differ between categories and we would expect recommendation rates to fall with tenure more for services that are less interesting, less changing and more frequently used. Previous evidence suggests that the rate of recommendation usually declines with the duration of customer tenure. East, Lomax and Narain (2000) showed that recommendation rates fell with tenure for supermarkets and womens hairdressing. Naylor and Kleiser (2000) found that first-time users of a health resort used more word of mouth compared with repeat buyers. Gremler and Brown (1999) found only a modest growth in the total number of claimed recommendations made by customers of banks and dental practices over time, which suggested that the rate of recommendation fell off with tenure duration. Wangenheim and Bayn (2004) found that new utility customers recommended more than established users. These studies indicate that recommendation rates are negatively related to tenure. Two studies show no relationship between tenure and recommendation rate (Kumar, Scheer and Steenkamp 1995; Verhoef, Franses and Hoekstra 2002). Also, a comment by Reinartz and Kumar (2000) indicated that recommendation was associated with attitude than customer tenure.
14
3. The Effect of Similarity between Measures We argue that recommendation will be predicted mainly by relative attitude, and retention will be predicted mainly by repeat patronage. This is because of measurement similarities within each pairing. Our argument starts with the observation of Ajzen and Fishbein (1977) that two measures are more likely to be related if they are similarly specified by reference to object, action and context (and time, where applicable). For example, if we compare the attitude to Woolworth and the attitude to shopping at the nearest Woolworth store, the latter will provide a better prediction of actually shopping at the nearest Woolworth store because of the common specification. This issue of measurement compatibility may be seen as an example of common method effect: when two measures have features in common they are more likely to be associated. More specific explanations will also apply and we introduce these in the context of the studies that are reported. Below, we explain the effect of measurement similarity in relation to recommendation, search and retention. 3.1 Recommendation We suggest that the relative attitude to a brand will predict recommendation of that brand because the reasons for liking one brand more than another are often the same as the reasons that are given when making a recommendation in favour of one brand over another. In this way, relative attitude and recommendation have features in common. By contrast, repeat patronage has no strong features in common with recommendation. This means that we would expect a lower association between repeat patronage and recommendation than between relative attitude and recommendation. 3.2 Search When spending is high at the main store (high repeat patronage), there is less to spend at other stores and this will have the effect of reducing the number of other stores used (search measure). Thus, the repeat-patronage and search measures are, to a degree, competing for the same effect and, because of this commonality, we expect them to show a negative correlation. By contrast, the search measure does not have features in common with the attitude measure and this leads us to expect a weaker correlation between these variables. But when we measure search by the number of car makes considered before purchase, there are no obvious features in
common with either relative attitude or repeat patronage and, in this case, we cannot predict which of these variables will relate most to search. 3.3 Retention Retention has much in common with repeat patronage since these two behaviours may be seen as the same behaviour at different times. This suggests that retention will be predicted better by repeat patronage than by relative attitude, which has no obvious features in common with retention. 4. Hypotheses Based on our review, we frame three hypotheses. We test these Hypotheses in three sets of studies. The categories investigated and some of the measures used are changed for each Set. In particular, we change the measure of repeat patronage. This is because the conventional measure of repeat patronage varies with the category and we wished to test the measures normally used. Our Hypotheses provide predictive tests as suggested by Mellens, Dekimpe and Steenkamp (1996) and relate to ideas found in the work of Day (1969), Dick and Basu (1994), Jacoby and Chestnut (1978) and Oliver (1999). H1: Recommendation, search and retention will be significantly associated with either relative attitude or repeat patronage, but not both. Specifically, we expect repeat patronage to be positively associated with retention, negatively associated with search (in Set 1) and not associated with recommendation; relative attitude will be positively associated with recommendation but not with retention or search in Set 1. H2: The addition of an interaction term (relative attitude by repeat patronage) to any of the main effects models predicting recommendation, search, and retention will not increase R2 substantially. It is possible for an interactive term to add substantially to the explanation provided by a model, even when one or both of the components of the interaction term do not do so on their own. Thus, support for H1 does not preclude support for H2. However, our review shows little previous evidence for the superiority of an interaction model. H3: No single definition of loyalty will predict all three loyalty outcomes (recommendation, search and retention). If we find that recommendation, search and retention are predicted by different variables, H3 will be supported.
15
Table 1: Supermarkets - Predictions of Loyalty Outcomes Britain Estimate (a) Recommendation of main store (UK, N=869, NZ, N=1157) Main Effects Model Relative Attitude Repeat patronage Interaction Model Relative Attitude (1) Repeat patronage (2) 1 x 2 term (b) Number of stores used (UK, N=871, NZ, N=1140) Main Effects Model Relative Attitude Repeat patronage Interaction Model Relative Attitude (1) Repeat patronage (2) 1 x 2 term (c) Retention of main store over 12 months (UK, N=577, NZ, N=976) Main Effects Model Relative Attitude Repeat patronage Interaction Model Relative Attitude (1) Repeat patronage (2) 1 x 2 term 0.07 0.25 -0.03 0.13 0.04 .64 .02 .93 .78 .79 0.17 0.47 0.003 0.29 0.06 .11 .000 .99 .42 .60 0.08 -0.68 0.08 -0.68 -0.19 .42 .000 .77 .02 .99 -0.10 -0.90 0.44 -0.36 -0.18 .23 .000 .12 .21 .05 1.05 -0.20 0.43 -0.88 0.24 .000 .02 .23 .02 .07 1.33 -0.10 0.77 0.67 0.20 .000 .21 .03 .06 .11 Sig. Cox & Snell R2 % Estimate New Zealand Sig. Cox & Snell R2 %
9.2
14.4
9.5
14.6
9.5
14.3
9.5
14.6
1.0
3.6
1.0
3.6
5. Set 1: Supermarket Loyalty We surveyed supermarket customers, using as a measure of repeat patronage the proportion of spending given to their primary store-group. Data were collected in two countries, Britain and New Zealand, to provide a broader test of the Hypotheses. These countries have some differences in retail structure and population density, which could affect results. If the same results are obtained in both countries, the findings are more generalisable.
5.1 Data We conducted initial surveys in 1998. Twelve months later, the respondents were re-surveyed so that the retention of the main store group could be assessed. In Britain, a sample of 2000 names and addresses was drawn from the electoral registers of England and Wales. In New Zealand, 2200 names were drawn from the electoral rolls. In both cases, at each address, the first womans name was selected since supermarket shopping is predominantly a female role; if the household did not
16
contain a woman, the first mans name was used. An accompanying letter indicated that a response was sought from the person who usually did the shopping in the household. In both cases, a two-wave postal survey was used but, additionally, an intervening reminder card was used in New Zealand. The usable response rates were 46% (Britain) and 61% (New Zealand) in 1998. In 1999, when respondents were re-contacted, 77% of the British respondents and 82% of the New Zealand respondents returned questionnaires. Single-item measures were used, which are shown in Appendix A. The number of supermarkets used in the last four weeks served as the measure of search behaviour. 5.2 Results Retention rates were 76% in the UK and 72% in New Zealand. We used ordinal regression to predict recommendation, number of stores used and retention because the predicted variable was ordinal. For twovalued outcomes such as retention/defection, ordinal regression provides the same coefficients as logistic regression. Table 1 shows how well relative attitude and repeat patronage predict the three loyalty outcomes. In this table we describe the analyses dealing with main effects as the Main Effects Model and the analyses that include the interaction term as the Interaction Model. H1: Hypothesis 1 was that each loyalty outcome would be predicted by one variable and therefore that an additive model would not be supported. The three forms of the Main Effects Model in table 1 relate to this. First, we found that recommendation was strongly predicted by attitude in both countries (p < 0.001) but was weakly predicted negatively by repeat patronage in Britain (p = 0.02) and showed no relationship to repeat patronage in New Zealand. Second, the number of stores used (search) was negatively predicted by repeat patronage (p < 0.001) in both samples, as hypothesised. There was no significant association between attitude and number of stores used. Third, primary-store patronage was significantly associated with retention in both countries while attitude was not, though the models were very weak. Thus, the three behaviours were each predicted in the expected direction by one variable only, H1 stands and an additive model is not supported. These findings do not occur because attitude and repeat patronage are so highly correlated that one is eclipsed in
the regression analysis. The Spearman rank correlations between the measures of relative attitude and repeat patronage were fairly low (rs= 0.13 for Britain and 0.15 for New Zealand). H2: Hypothesis 2 was that the inclusion of an interaction term would not add substantially to R2. This is tested by a comparison between the Main Effects and Interaction Models (see Endnote). In examining the gain in R2, we are concerned with effect size rather than with significance and we set a relatively low hurdle of an improvement in R2 of two percentage points. Using this criterion, there are no substantial gains in R2 when the interaction term is added in respect of any of the three outcomes and H2 is therefore supported. H3: Hypothesis 3 was that there would be no measure of loyalty that predicted all three behavioural outcomes (recommendation, search for alternatives and retention). We have shown that recommendation is predicted by attitude, while the number of stores used and retention are predicted by repeat patronage. Thus, there is no consistent loyalty predictor across the three outcome behaviours and H3 is supported. 5.3 Discussion of Set 1 The research findings presented above confirm all three Hypotheses. An additional finding was that retention was poorly predicted by our measures. One explanation for this is that, in retail, brand switching is controlled by environmental contingencies that are not well anticipated by measures of either attitude or repeat patronage. Environmental control of everyday habits might also explain the lack of any effect of attitude on the number of stores used. Also, the fairly low correlation found between the measures of relative attitude and repeat patronage suggests that there is little direct causal connection between these variables and that each is affected by different contingencies. The lack of any effect of repeat patronage on recommendation may have arisen because those people who use a wide variety of stores (low repeat patronage) are the ones who can best compare those stores and this could offset any tendency to recommend derived from more exclusive use. If this is so, a different measure of repeat patronage might show a positive relationship with recommendation. 6. Set 2: Car Repeat-Purchase Brand Loyalty In order to widen the test of our Hypotheses and to examine further the issues raised above, we conducted
17
research on car purchase (new and second-hand combined). 6.1 Data An Internet study on UK car purchase was conducted in 2003; there were 495 respondents. Internet studies have low response rates; in this case it was 2%. For cars, we could use a direct measure of repeat patronage. We investigated the main car make of those who had made at least three successive car purchases, designating the last three purchases as 1, 2 and 3 with 3 the most recent. Our measure of repeat patronage is greater if purchases 1 and 2 are the same make and our measure of retention is greater if 2 and 3 are the same make. We also asked about recommendation and the number of alternative cars considered (search) so that we could predict these
outcomes. The key questions used are shown in Appendix A. We used the repeat-patronage measure and the relative attitude to the second car to predict both the number of alternative car makes considered and the retention of the car make when choosing the third car. For recommendation, we focused on recommending the third car and therefore used the rating of this car for the relative attitude measure and the makes of cars 2 and 3 to establish the measure of repeat patronage. 6.2 Results There was a fair degree of behavioural loyalty when this was measured as buying the same car make as last time. From purchase 1 to 2, and 2 to 3, the car make comparisons showed that 30% and 31% of respondents
Table 2: Cars: Prediction of Loyalty Outcomes Estimate (a) Recommendation (N=495) Main Effects Model Relative Attitude Repeat patronage Interaction Model Relative Attitude (1) Repeat patronage (2) 1 x 2 term (b) Number of other makes considered Main Effects Model Relative Attitude Repeat patronage Interaction Model Relative Attitude (1) Repeat patronage (2) 1 x 2 term (c) Retention of car make at third purchase Main Effects Model Relative Attitude Repeat patronage Interaction Model Relative Attitude (1) Repeat patronage (2) 1 x 2 term 0.91 0.30 0.92 0.64 0.16 .000 .10 .000 .004 .009 Sig. Cox & Snell R2 %
17.5
18.6
-0.03 -1.05 0.34 -1.75 -0.30
.74 .000 .19 .001 .135
7.2
7.7
0.23 0.58 0.55 2.02 0.62
.03 .005 .08 .001 .01
3.4
4.8
18
bought the same make again on successive purchases. When 1 and 3 were examined, the rate dropped to 19%. Some part of these repeat rates may be ascribed to chance; using the market shares disclosed in the survey, we estimated that approximately 10% should be deducted from the rates observed to correct for chance. H1: Are loyalty outcomes predicted by relative attitude, repeat patronage or both? From H1 we expect only one significant predictor for each loyalty outcome. In table 2, the Main Effects Model shows that, for car purchase, attitude is a significant predictor of recommendation, but repeat patronage is not, as in Set 1. On the number of
other makes considered (search), we find that the repeatpatronage measure is significantly related in the direction expected (negative) whereas relative attitude is not significant, as in Set 1. On retention, the model was weak (R2 less than 0.05) but here both attitude and repeat patronage are significant with the latter dominating. So, in Set 2, H1 is rejected with regard to retention but supported for the other two loyalty outcomes. The correlation between relative attitude and repeat purchase was significant but not large (rs= 0.22, p < 0.001). H2: Does the product of relative attitude and repeat patronage add significantly to the prediction of loyalty
Table 3: Correlations (R2) Between Variables in 23 Service Studies (1) Service (country, response rate %) (2) N (3) Method (4) (5) (6) Repeat Relative Repeat patronage attitude patronage (tenure) & recom- (tenure) & recommendation & relative mendation attitude -0.44* -0.39* -0.36* -0.28* -0.09 -0.04 -0.03 -0.03 -0.02 0.02 0.04 0.04 0.06 0.07 0.07 0.08 0.10 0.12 0.13 0.18* 0.20* 0.20* 0.25* -0.01 0.32* 0.10 0.19* 0.12 0.32* 0.32* 0.21* 0.36* 0.41* -0.05 0.24* 0.32* 0.42* 0.38* 0.01 0.21* 0.14* 0.45* 0.34* 0.35* 0.43* 0.03 0.42* 0.26 0.05 0.22* 0.23* 0.04 0.10 0.03 0.14 0.06 0.15 0.23* 0.05 0.20* 0.11 -0.06 0.17* 0.04 0.27* 0.24* 0.30* 0.12 0.33* 0.19* 0.21* 0.15
Cheque book service (UK, 81) Credit card (UK, 81) Car insurance (UK, 81) Credit card (UK, 43) Main supermarket (UK, 81) Mobile airtime (UK, 86) Motor insurance (UK, 58) Dentist (UK, 57) Dry cleaning (UK, 52) Internet provider (UK, na) Leisure centre (UK, 40) House contents insurance (UK, 58) Main supermarket (Mexico, 40) Main fashion store (UK, 81) Car insurance (Mauritius, 45) Favourite restaurant (UK, 86) Email (UK, na) Hairdresser (Mexico, 40) Search engine (UK, na) Main fashion store (Mexico, 40) Car servicing (UK, 52) Car servicing (Mauritius, 45) Car servicing (UK, 81) Means * significance < 0.05
187 165 156 140 164 266 109 208 111 118 95 121 166 152 201 187 223 163 206 158 131 167 110
Deliver to household and collect Deliver to household and collect Deliver to household and collect Deliver to household and post return Deliver to household and collect Class distribution to students Deliver to household and post return Deliver to household and post return Deliver to household and post return Email via friends Mail survey of members Deliver to household and post return Mail delivery and return Deliver to household and collect Mall intercept Class distribution to students Email via friends Mail delivery and return Email via friends Mail delivery and return Deliver to household and post return Mall intercept Deliver to household and collect
19
outcomes using the additive model and thus support an interaction model? There are no substantial changes in R2 from the Main Effects Model to the Interaction Model, as in Set 1, and so H2 is supported. H3: Does any one measure of loyalty predict all three behavioural outcomes (recommendation, search for alternatives and retention)? No single formulation consistently predicts the different outcomes; repeat patronage fails to predict recommendation at a significant level and attitude fails to predict the number of alternative makes considered. In Set 2, we also measured the intention to re-buy the current car make. When we conducted a regression analysis on this measure of intention, the R2 was 0.28 and both relative attitude and repeat patronage were significant. The addition of an interaction term did not raise the R2. 6.3 Discussion of Set 2 With the exception of the prediction of retention, the findings of Set 2 support our Hypotheses. In particular, the results suggest that there is no connection between the degree of past patronage and the degree of recommendation. As in Set 1, this study showed weak prediction of actual retention. If the intention to re-buy is predicted instead of true retention, a much stronger explanation is obtained. We review this matter in the General Discussion. 7. Set 3: Customer Tenure Duration and the Recommendation of Services In Set 3, we focus on the recommendation of services. We chose recommendation because the evidence from Sets 1 and 2 had shown that this outcome was better explained than retention or search and we chose services to widen the scope of our work and because recommendation appears to be particularly influential in the adoption of new service suppliers. For example, Keaveney (1995) showed that 50% of service customers found a new supplier through recommendation. In this work, we used the duration of customer tenure as a measure of repeat patronage. 7.1 Data From 2000 to 2003, we conducted surveys to investigate factors associated with recommendation in 23 service categories used by respondents. Most of the surveys were conducted in the UK but one (covering three categories)
was carried out in Mexico and one (covering two categories) was carried out in Mauritius. Response rates were in the range 40%-86%. Table 3 shows the categories, the response rate, the numbers of users responding, and the method used to apply the questionnaire. We present the rank correlations between repeat patronage (measured as customer tenure) and recommendation, relative attitude and recommendation, and relative attitude and repeat patronage. Examples of the items covering repeat patronage, relative attitude and recommendation are shown in the Appendix. 7.2 Results The findings reported in Table 3 are ordered by the correlation between repeat patronage and recommendation (column 4). Overall, the mean correlation of column 4 is close to zero but this average obscures the fact that, in some services, recommendation is negatively related to repeat patronage (e.g. cheque service and credit cards) while in other categories it is positively related (e.g. car servicing, fashion stores in Mexico). Column 5 shows that 22 of the 23 correlations between relative attitude and recommendation are positive and 18 are significant. Column 6 of Table 3 shows that relative attitude is often positively associated with customer tenure (average rs= 0.15). These data indicate that some process, such as loss of salience, is counteracting the tendency for recommendation to rise as attitude rises with tenure. Using regression, we probed the four cases in column 4 showing a significant positive association between repeat patronage and recommendation; only in one case did repeat patronage remain significant when relative attitude was included in the equation. These data therefore give general support for H1 since, in most cases, recommendation is strongly and positively associated with relative attitude and only one case was found when repeat patronage was significantly associated with recommendation, after controlling for the effect of attitude. In order to test H2, we selected the three cases in Table 3 where both attitude and repeat patronage were significantly and positively associated with recommendation, since these conditions made an interaction effect more likely. Table 4 shows the results. In no case was the R2 raised by as much as two percentage points by the addition of the interaction term to the regression analysis.
20
Since we only considered one outcome, recommendation, we cannot test H3 in Set 3. 8. General Discussion 8.1 The Findings Reviewed Our research has used a variety of measures and different categories to examine three Hypotheses about loyalty. We show that recommendation is significantly predicted by relative attitude in the supermarket and car Sets, and in a substantial majority of the 23 service cases. By contrast, recommendation is generally not predicted significantly by repeat-patronage measures. With regard to search, the model outcomes are weaker but the repeatpatronage measure is significant and the relative attitude measure is not significant. On retention, we find that neither attitude nor repeat patronage give a strong prediction but that repeat patronage is significant in both the supermarket and car cases and relative attitude is significant only in the case of cars. The inclusion of an interaction term never had more than a marginal effect on the R2, even in Set 3 where we selected cases that would give the best chance of finding this effect. This means that, on a principle of parsimony, there is no case in our studies for invoking interaction effects in order to predict loyalty outcomes. Our evidence suggests that attitudinal and behavioural measures of loyalty do not have much in common. Empirically, we can see this in the relatively low correlations found between the repeat patronage and relative attitude measures in our studies. These low associations suggest that attitude is conditioned in ways that may be unrelated to repeat-patronage behaviour.
Although a combination of attitude and repeat patronage may sometimes predict further loyalty behaviour, the assumption that this is normally the case does not appear to be justified. We believe that our findings should cause marketers to question whether loyalty should be seen as some combination of relative attitude and repeat patronage. This approach to loyalty is often found in textbooks, which may cite Day (1969), Dick and Basu (1994, particularly the typology, figure 2), Jacoby and Chestnut (1978), Mellens Dekimpe and Steenkamp (1996), and Oliver (1999) in support. However, we point out that none of these sources provided evidence showing that loyalty outcomes were predicted better by a combination measure. 8.2 What is the Alternative? Instead of asking what loyalty is and what its consequences are, it might be more productive to focus on loyalty outcomes such as recommendation and retention and to ask about the conditions that produce these outcomes, including loyalty measures. When this is done, different measures of loyalty can be tested alongside other variables such as expertise, opportunity and weight of purchase. The interest in loyalty stems from its potential outcomes, so let us examine these outcomes directly. 8.3 Explaining Findings The general explanation for our findings is in terms of common method effect and is very simple. In particular, we explain that relative attitude predicts recommendation because attitudes are based on particular attributes of the object and that it is these
Table 4: Predictors of Recommendation Car Services Study 1 (N=131) Estimate Main Effects Model Relative Attitude (1) Repeat patronage (2) Interaction Model Relative Attitude (1) Repeat patronage (2) 1x2 1.44 0.19 -0.01 .12 .87 .97 2.89 1.95 -0.37 .05 .13 .21 0.61 -0.58 0.26 .44 .36 .20 1.40 0.14 .00 .38 Sig Cox & Estimate Snell R2 % 18.3 1.16 0.37 Sig .00 .05 Study 2 (N=110) Fashion Stores (N=158) Sig .00 .06 Cox & Snell R2 % 19.6
Cox & Estimate Snell R2 % 17.6 1.50 0.22
18.3
18.6
20.5
21
attributes that are also likely to be the content of recommendations. Similarly, we argue that past patronage and retention are essentially the same behaviours, though separated in time. This may prompt the response that the explanation is obvious and near to a tautology. Once stated, the explanation is obvious but no one to our knowledge has previously stated it in this form. We argue that obvious explanations that no one has yet recognised are useful. For example, this particular explanation leads to a focus on those evaluative attributes that may be used in word of mouth. 8.4 The Weak Prediction of Retention Brand retention was poorly predicted in our studies. We accept that there may be circumstances in which retention is better predicted, for example when the predicted behaviour follows soon after measurement, but, generally, our findings indicate that brand retention is difficult to predict in the longer term. This finding suggests that attempts to raise retention by enhancing satisfaction may have limited effect on this variable though they could impact on recommendation. Against this argument, Anderson and Mittal (2000) suggest that some product attributes have more effect on retention than others and that some customers are more responsive than are others. When these attributes and customers are targeted, retention may show more relationship with satisfaction, but the scope for applying such a focused approach may be limited. One explanation for the weak prediction of retention is that consumer volition is heavily constrained by opportunities and social rules. Supermarket use is restricted by the proximity of stores, and car purchase by company policy. Attitude may be crowded out by such constraints so that it has little effect on retention. For cars, we show that the intention to repeat purchase is much more strongly determined than retention and here attitude has a predictive role. One interpretation of this is that intention is much less constrained by external contingencies (which may not yet have occurred) and therefore relative attitude can have more predictive effect. However, whatever the explanation, the practice of using an intention measure as a proxy for retention seems unjustified. Although there is published evidence that intentions do predict behaviour quite well, this evidence deals more with the prediction of category purchase than with the prediction of brand purchase, as studied here.
8.5 Implications This work has practical value. Customer tenure and satisfaction are often available in databases. Our data suggest that practitioners might use satisfaction scores to pick segments for incentivised recommendation but they should be careful about using tenure for this purpose. In general, the categories that are frequently used and unchanging are recommended more by recent customers. Long-term customers may recommend a category more only if it is complex and used infrequently. 9. Conclusions In a series of studies, we tested whether additive or interactive combinations of relative attitude and repeat patronage predict loyalty outcomes better than singular measures. A number of measures were used for repeat patronage and recommendation and the work was conducted on retail, durable and service products. We found that recommendation was predicted by relative attitude and not by repeat patronage, while search behaviour and retention were predicted more by repeat patronage than by relative attitude. These findings cast doubt on the value of combination concepts of loyalty and also on the idea that there is a generic form of loyalty that will predict a variety of different loyalty outcomes. Endnote
The testing and interpretation of interaction effects is discussed by Jaccard and Turrisi (2003). The hierarchical regression method for testing for interaction may be subject to error when the product term is formed from ordinal measures since non-linearity may distort the product term. However our expectation is negative, that there will be no effect, and it seems unlikely that distortions could lead to the elimination of a real effect.
References Ajzen, I., Fishbein, M., 1977. Attitude-behavior relations: a theoretical analysis and review of empirical research. Psychological Bulletin 84, 888-919. Anderson, E.W., Mittal, V., 2000. Strengthening the satisfaction-profit chain. Journal of Service Research 3 (2), 107-120. Andreasen, A.R., 1985. Consumer responses to dissatisfaction in loose monopolies. Journal of Consumer Research 12 (Sept), 135-141. Baldinger, A.L., Rubinson, J., 1996. Brand loyalty: the link between attitude and behavior. Journal of Advertising Research 36 (6), 22-34.
22
Baloglu, S., 2002. Dimensions of customer loyalty: separating friends from well wishers. Cornell Hotel and Restaurant Administration Quarterly 43 (1), 47-59. Berry, L.L., 1983. Relationship marketing. In L.L. Berry, G.L. Shostack, G.D. Upah, (Eds), Emerging Perspectives on Service Marketing, Chicago: American Marketing Association. Bhattacharya, C.B., 1997. Is your brands loyalty too much, too little or just right? Explaining deviations from the Dirichlet norm. International Journal of Research in Marketing 14 (5), 421-435. Bhattacharya, C.B., Fader, P.S., Lodish, L. M., DeSarbo, W.S., 1996. The relationship between the marketing mix and share of category requirements. Marketing Letters 7 (1), 5-18. Bloemer, J., de Ruyter, K., 1998. On the relationship between store image, store satisfaction and store loyalty. European Journal of Marketing 32 (5/6), 499-513. Bolton, R.N., 1998. A dynamic model of the duration of the customers relationship with a continuous service provider: the role of satisfaction. Marketing Science 17 (1), 45-65. Bolton, R.N., Lemon, K.N., 1999. A dynamic model of customers use of services: usage as an antecedent and consequence of satisfaction. Journal of Marketing Research 36 (2), 171-187. Chandon, P., Morvitz, V.G. and Reinartz, W.J., 2005. Do intentions really predict behavior? Self-generated validity effects in survey research. Journal of Marketing, 69 (2), 1-14. Crosby, L.A., Stephens, N., 1987. Effects of relationship marketing on satisfaction, retention, and prices in the life insurance industry. Journal of Marketing Research 24 (4), 404-11. Day, G.S., 1969. A two-dimensional concept of brand loyalty. Journal of Advertising Research 9, 29-35. Deighton, J., Henderson, C.M., Neslin, S.A., 1994. The effects of advertising on brand switching and repeat purchasing. Journal of Marketing Research 31, (1), 28-43. Dick, A.S., Basu, K., 1994. Customer loyalty: towards an integrated framework. Journal of The Academy of Marketing Science 22 (2), 99-113. Dwyer, R.F., Schurr, P.H., Oh, S., 1987. Developing buyer-seller relationships. Journal of Marketing 51 (2), 11-27.
East, R., Lomax, W., Narain, R., 2001. Customer tenure, recommendation and switching. Journal of Satisfaction, Dissatisfaction, and Complaining Behavior 14, 46-54. Ennew, C.T., Binks, M.R., 1996. The impact of service quality and service characteristics on customer retention: small businesses and their banks in the UK. British Journal of Management 7, 219-230. Fishbein, M.F., Ajzen, I., 1975. Belief, Attitude, Intention, and Behavior. Reading, Mass.: AddisonWesley. Gremler, D.D., Brown, S.W., 1999. The loyalty ripple effect: appreciating the full value of customers. International Journal of Service Industry Management 10 (3), 271-291. Hauser, J.R., Wernerfelt, B., 1990. An evaluation cost model of consideration sets. Journal of Consumer Research 16 (March), 393-408. Hennig-Thurau, T. Klee, A., 1997. The impact of customer satisfaction and relationship quality on customer retention: a critical reassessment and model development. Psychology and Marketing 14 (8), 737764. Jaccard, J., Turrisi, R., 2003. Interaction Effects in Multiple Regression, 2nd Edition. Thousand Oaks: Sage University Paper. Jacoby, J., Chestnut, R.W., 1978. Brand Loyalty Measurement and Management. New York: Wiley. Keaveney, S.M., 1995. Customer switching behavior in service industries: an exploratory study. Journal of Marketing 59, 71-82. Kordupleski, R.E., Rust, R.T., Zahoric, A.J., 1993. Why improving quality does not improve retention (or whatever happened to marketing?). California Management Review, Spring, 82-95. Kumar, N., Scheer, L.K., Steenkamp, J-B.E.M., 1995. The effects of supplier fairness on vulnerable resellers. Journal of Marketing Research 32 (1), 54-65. Macintosh, G., Lockshin, L.S., 1997. Retail relationships and Store Loyalty: A Multi-Level Perspective. International Journal of Research in Marketing 14 (5), 487-497. Mellens, M., Dekimpe, M., Steenkamp, J-B E.M., 1996. A review of brand loyalty measures in marketing. Tijdschrift voor Economie and Management 41 (4), 507533.
23
Mittal, V., Kamakura, W.A., 2001. Satisfaction, repurchase intention, and repurchase behavior: investigating the moderating effect of consumer characteristics. Journal of Market Research 38 (1), 131142. Morgan, R.M., Hunt, S.D., 1994. The commitment-trust theory of relationship marketing. Journal of Marketing 58 (July), 20-38. Naylor, G., Kleiser, S.B., 2000. Negative versus positive word-of-mouth: an exception to the rule. Journal of Satisfaction, Dissatisfaction, and Complaining Behavior 13, 26-36. Oliver, R.L., 1999. Whence customer loyalty? Journal of Marketing 63 (Special Issue), 33-44. Pritchard, M.P., Howard, D.R., 1997. The loyal traveler: a typology of service patronage. Journal of Travel Research 35 (4), 2-10. Pritchard, M.P., Havitz, M.E., Howard, D.R., 1999. Analyzing the commitment-loyalty link in service contexts. Journal of the Academy of Marketing Science 27 (3), 333-348. Reichheld, F.F., 1996. The Loyalty Effect. Boston: Harvard Business School Publications. Reichheld, F.F., 1993. Loyalty-based management. Harvard Business Review 71 (2), 64-73. Reichheld, F.F., Kenny, D.W., 1990. The hidden advantages of customer retention, Journal of Retail Banking 12 (4), 19-23. Reinartz, W., Kumar, V., 2000. On the profitability of long-life customers in a non-contractual setting: an empirical phase and implications for marketing. Journal of Marketing 64 (4), 17-36. Shankar, V., Smith, A., Rangaswamy, A., 2003. Customer satisfaction and loyalty in online and offline environments. International Journal of Research in Marketing 20 (2), 153-175.
Verhoef, P.C., Franses, P.H., Hoekstra, J.C., 2002. The effect of relational constructs on customer referrals and number of services purchased from a multiservice provider: does age of relationship matter? Journal of the Academy of Marketing Science 30 (3), 202-216. Wangenheim, F.V., Bayn, T., 2004. Satisfaction, loyalty and word of mouth within the customer base of a utility provider: differences between stayers, switchers and referral switchers. Journal of Consumer Behaviour 3 (1), 211-220. Biographies Robert East is Professor of Consumer Behaviour at Kingston Business School, London. Philip Gendall is Professor and Head of the Department of Marketing at Massey University, New Zealand. Kathy Hammond is Managing Director at Duke Corporate Education and previously Assistant Professor at London Business School. Wendy Lomax is Professor and Head of the School of Marketing, Kingston Business School, London. Correspondence Addresses
Robert East, Kingston Business School, Kingston Hill, Kingston KT2 7LB, Surrey UK. Telephone: +44 (0) 20 8547 2000 ext 5563. Facsimile: +44 (0) 20 8547 7026. Email: R.East@kingston.ac.uk; Philip Gendall, Department of Marketing, Massey University, Private Bag 11-222, Palmerston North, New Zealand. Telephone: +64 (6) 350 5590. Facsimile: +64 (6) 350 2260. Email: P.Gendall@massey.ac.nz; Kathy Hammond, Duke Corporate Education, Houghton Street, London WC2A 2AE. Telephone: +44 (0) 20 7107 5250. Facsimile: +44 (0) 20 7107 5280. Email: kathy.hammond@dukece.com; Wendy Lomax, Kingston Business School, Kingston Hill, Kingston KT2 7LB, Surrey, UK. Telephone: +44 (0)20 8547 2000. Facsimile: +44 (0) 20 8547 7026. Email: W.Lomax@kingston.ac.uk
24
Appendix A: Measures with Response Frequencies Frequencies % % Set 1: Supermarkets Relative attitude Compared with other stores, would you rate your main store as ? Repeat patronage What proportion of your total grocery spending is made in your main store Recommendation Have you ever recommended your main store to others? Search In the last four weeks, how many different supermarkets or other grocery stores have you used? Poor Adequate Good Excellent 50% or less 51 80% 81 95% More than 95% Rarely or never Occasionally Quite often One Two Three or four More than four [1] [2] [3] [4] [1] [2] [3] [4] [1] [2] [3] [1] [2] [3] [4] UK 2 22 60 16 11 36 34 19 35 52 13 18 45 31 6 NZ 1 22 55 22 5 28 42 25 38 50 12 22 40 34 4
Retention of the main store group. This was measured over a 12-month duration by comparing the main store chain reported in 1999 with that reported in 1998.
25
Appendix A: Measures with Response Frequencies (Cont.) Frequencies % Set 2: Car purchase Relative attitude How do you rate your current car, compared with the best alternative make of car that you could have realistically chosen? Much worse [1] 1 Worse [2] 8 The same [3] 27 Better [4] 38 Much better [5] 26 Recommendation In the last six months, how many times have you actually recommended <car make> to others? Never Once 2-3 times 4-6 times More than 6 times Search How many other makes did you consider before you chose your current car, the <car make>? No other car brands considered 1 other brand considered 2 other brands considered 3-5 other brands considered More than 5 other brands considered Set 3: Service Use, Sample Questions Relative attitude Compared to other car servicing firms, how do you rate the firm you use? Poor Below average Average Good Very good Repeat patronage How long have you been using this firm? Less than 1 year 1-2 years 2-4 years 4-8 years More than 8 years Recommendation How many times have you recommended this firm to others in the past six months? 0 1 2 3-4 5-8 More than eight times [1] [2] [3] [4] [5] [6] 62 15 15 5 1 2 [1] [2] [3] [4] [5] 11 21 26 25 17 [1] [2] [3] [4] [5] 1 1 17 59 22 [1] [2] [3] [4] [5] 27 17 27 22 7 [1] [2] [3] [4] [5] 31 18 41 6 7
26
Repeated Binary Logit: Analysing Variation in Behavioural Loyalty

Cam Rungie & Gilles Laurent
Abstract Brand and store managers are anxious to obtain high loyalty, as operationalized by repeat purchase behaviour. In this paper, we introduce the Repeated Binary Logit (RBL) model, which analyses directly the impact of covariates on repeat purchase, and which can be described as an extension of traditional logistic regression. We present empirical applications of RBL, and we discuss its relationships to several classical models. Keywords: Brand choice, Buyer behaviour, Choice models, Data mining, Marketing research, Segmentation
1. Introduction Brand and store managers are anxious to obtain high loyalty from their consumers. And, very often, they equate loyalty and repeat purchase behaviour. Assessing the level of repeat purchase, in a category at large or for a specific brand, is a key step in diagnosing a market, and in defining a marketing strategy. We argue that market analysts must go beyond the sheer measurement of loyalty through repeat purchase. Rather, they should identify the causes of loyalty, and measure their impact, through a statistical analysis. In this paper, we introduce the Repeated Binary Logit (RBL) model that analyses directly the impact of covariates on loyalty. RBL can be described as an extension of traditional logistic regression. We present several empirical applications of RBL. Finally, we discuss its relationships to several classical models.
2. The Repeated Binary Logit (RBL) Model 2.1 Functional Form We first describe the simple case where there are no covariates impacting loyalty. RBL is constructed from the beta binomial distribution. Let a randomly selected decision maker make k selections from a binary choice set containing alternatives A and B (A could be a specific brand A, and B all the other brands; or A could be a specific store A, and B all the other stores). Let the number of times alternatives A and B are selected be rA and rB where rA + rB= k . Over the population of decision makers, these selection rates are random variables, RA and RB. They have a beta binomial distribution (BBD) conditional on k and on the parameters A and B of the beta distribution. The probability density function for the BBD is:
Equation 1
27
2.2 Link to Marketing Statistics For the BBD the mean, , for the proportion of times alternative A is selected is (Johnson, Kotz, and Balakrishnan 1994; Johnson, Kotz, and Balakrishnan 1997; Kotz, Balakrishnan, and Johnson 2000): Equation 2
Equation 6
Thus, RBL models the market share and repeat rate simultaneously. Finally it is interesting to consider the special case where k=1, i.e. when each consumer makes a single choice. Obviously, in such a case, the study of loyalty loses its interest. However, we can observe that, in this condition, RBL reduces to: Equation 7
In many marketing applications this is equivalent to, or a close estimate of, the market share for alternative A. A simple indicator of loyalty is the repeat rate, . It is defined to be the probability of selecting alternative A at the next choice occasion, given it was selected at the last; i.e. Pr(RA =2|k=2)/Pr(RA =1|k=1). (A choice occasion is an event where a selection from the choice set is made, e.g. a purchase in the category for a brand choice, or a shopping trip for a store choice.) Each decision maker makes two independent choices from the same choice set with replacement and under stationary conditions. From Equation 1: Equation 3
which is the basis of MNL and logistic regression, a tool widely used in marketing for analysing binary choices (Ben-Akiva and Lerman 1985; Guadagni and Little 1983; Louviere, Hensher, and Swait 1999; McFadden 1974; McFadden 1984). Thus, from Equation 5 and Equation 7, RBL is an extension of logistic regression. 2.4 Estimation The Likelihood Function The likelihood function can be specified as follows. Let there be a sample of n decision makers. For decision maker i, where i=1,2,n: Equation 8
2.3 Introducing Covariates Covariates are introduced by building a generalised linear model based on the BBD and the exponential linear link functions (McCullagh and Nelder 1989). These link functions are: Equation 4
From Equation 1 and Equation 8 and using the property of the gamma function where k!= (k+1) then the likelihood function is (Edwards 1976; Eliason 1993): Equation 9
The vector of covariates is x and there are two vectors of regression coefficients, A and B. The repeated binary logit (RBL) model is specified by inserting the link functions from Equation 4 into Equation 1. The resultant likelihood function is given below. The RBL model has the following properties: Equation 5
Regarding implementation, the vectors of parameters A and B are identified provided that k>1 (and that, as in general linear regression models, the vector of covariates x does not include excessive colinearity). Log gamma is widely available in computer packages. It is computationally more feasible to directly calculate the
28
log of Equation 9 rather than to calculate L and then take the log. In optimizing log(L) local maxima can be encountered so it is wise to consider multiple start points. The statistical significance of each of the covariates can be established using the likelihood ratio test. A macro, written in MATLAB, for likelihood estimation and likelihood ratio testing is available for beta evaluation from the first author. 3. First Example: Store Loyalty and Brand Loyalty We present an empirical example of RBL analysis, in which a covariate, while it has no impact on market share, has a significant impact on loyalty. Marketing Scan provided purchase data for one year in Angers, France. This is a closed zone panel, which follows all the supermarkets in the city and its suburbs. This means that the panel records all purchases by the panelists from supermarkets in the city, whatever the store in which the purchase takes place. For each purchase occasion by a panelist, we therefore record both the brand choice, and the supermarket where the purchase took place. There are 3,459 households followed over one year, who made a total of 21,982 purchases in the category while 313 households made no purchases. We analyse the binary choice between the leading brand, with 15.7% market share, and all other brands. The covariate we consider in this example is the proportion of each households purchases of all supermarket items, over a year, which is in their favourite supermarket, i.e. the individual share of category requirements for the favourite supermarket. This is a measure of their global loyalty to their favourite supermarket. The dependent variable, the market share for the leading brand on the market, does not change as the covariate, the loyalty to each consumers favourite supermarket, changes. A traditional logistic regression in the 21,982 purchases shows that the coefficient of the covariate is neither statistically significant (LR=1.950, p=0.163 with 1 df), nor managerially relevant. Given the large sample size, this is strong evidence that the leading brands market share does not vary according to shoppers loyalty to their favourite supermarket. However, the application of the RBL model leads to a more complex diagnostic: a significant value of the parameters for the covariate leads to the prediction of no significant impact on the market share, but a significant impact on brand loyalty. The estimation of the model
according to the procedure described earlier indicates that the covariate, the loyalty to favourite supermarket, is significant (LR=19.1176, p= .00007 with 2 df). For each distribution parameter A or B there are two estimated coefficients, a constant, and a coefficient for the covariate. The empirical estimates are: A=[-1.0139,0.9702] and B=[0.6481,-1.0705]. These estimates predict no change in the market share of the leading brand when the covariate, the loyalty to the favourite supermarket, increases. This is illustrated in Figure 1, where the plot of the estimated change is practically a horizontal line (we also plot in Figure 1, for comparison purposes, the observed raw values for subsamples based on the value of the covariate). However, there is a considerable change in the repeat rate, , as shown in Figure 2 (which plots the repeat rate predicted by the model, as well as the observed raw values for the sub-samples). This example demonstrates the potential of RBL to produce differentiated diagnoses. The covariate of interest, loyalty to supermarket, has no impact on the market share of the leading brand, but it does relate to the loyalty for that leading brand. The market share of the leading brand does not increase with consumers loyalty to their preferred supermarket. However, consumers who are more loyal to their preferred supermarket are also more loyal to the leading brand. This outcome results directly from the numerical values of the estimates A =[-1.0139,-0.9702] and B=[0.6481, -1.0705]. The coefficients of the covariate are similar (-0.9702 and 1.0705). As a consequence, the predicted market share, given by A/(A+B), does not change much when the covariate increases. However, the predicted repeat rate, given by (1+A)/(1+A+B), increases markedly. These results are very important from a managerial point of view. From a diagnosis approach, one should not interpret in the same manner the same observed repeat rate for the leading brand, depending on whether it is observed for a consumer who is highly loyal to one supermarket, or for a consumer who allocates her purchases across many supermarkets. From a decisionmaking point of view, the promotional program directed towards consumers of the leading brand should be very different, depending on whether or not they are loyal to their favourite supermarket.
29
Figure 1: Market Share for the Leading Brand Does Not Change with Loyalty to Favourite Store
50% Market Share for Leading Brand 40% 30% 20% 10% 0% 30% 40% 50% 60% 70% 80% 90% 100% Model (one sample) Separate samples
Loyalty to Favourite Store
Figure 2: As Loyalty to Favourite Store Increases so Too Does the Repeat Rate for the Leading Brand
70% Repeat Rate for Leading Brand 60% 50% 40% 30% 20% 30% 40% 50% 60% 70% 80% 90% 100% Model (one sample) Separate samples
Loyalty to Favourite Store
4. Second Example: Diversified Factors of Brand Loyalty We now consider all top six brands in the same category. They ranged from 15.7% down to 5.6% market share. For each of the brands, an RBL analysis was conducted, with five covariates (Table 1): Number of persons in the household. Average spend, by the household, in supermarkets, per week. Average number of visits per week to supermarkets for members of the household. Loyalty to favourite supermarket (SCR of favourite supermarket). Total purchase rate for the product category
After exploratory analysis the models were found to fit best by taking the natural logs of all covariates except loyalty to the favourite supermarket. We use statistical tests to measure the impact of covariates. In each case, we compare the fit of the unconstrained model (including all five covariates) to that of a constrained model, in which one of the covariates has been removed. This is done separately for each brand and each covariate. The resulting significance tests are in Table 2. A key result is that the significant covariates are not the same for all brands. For the leading brand, only one covariate is significant, loyalty to favourite supermarket. This is the result discussed in the first example. The covariate has a major impact on loyalty but no significant impact on market share.
30
For the third and fourth brands no covariate is significant. For these brands there is no difference in the purchase behaviour between shoppers from different sized households, with different household spends, etc. As a general result this is not overly surprising. A brand may well have category wide and market wide appeal. Generally, when searching for demographic differences between brands, it is the exception that a covariate will be significant and relevant, rather than the rule. Consequently, the interesting results are where there is a significant effect, i.e. brands 2, 5 and 6. The impact of the covariates on the market share and repeat rate is given in Table 3 and Table 4. For each covariate the impact is calculated for a change from one standard deviation below its mean to one standard deviation above its mean. The results are only shown for the significant covariates. The changes shown in Table 3 may look small, but they are very relevant. As household expenditure changes
from the mean minus one standard deviation to the mean plus one standard deviation, the market share for brand 6 reduces by 2.3% share points. Since this brand has overall 5.6% market share, the change is from 6.7% to 4.5%, a very important change. Table 3 displays results in terms of market shares. For example, it shows that bigger households which spend less are more likely to purchase brands 5 and 6. In a classic marketing approach based on market shares, the target for these brands might be defined as such households. However, Table 4 provides a different analysis, focused on changes in repeat rate. In Table 4 the absolute level of change in the repeat rates is far greater than the absolute level of change in market shares (shown in Table 3). This reflects the fact that the typical levels are higher for repeat rates than for market shares. Given that repeat rates are higher there is greater potential for them to vary. Keeping this in mind, several of the results in Table 4 are still of particular relevance to
Table 1: Descriptive Statistics for the Covariates
Table 2: Each Covariate is Significant for At Least One of the Top Six Brands (Chi Square LR test, each with 2 df)
31
Table 3: Changes in Market Share (For example, as household spend increases from the mean minus one standard deviation to the mean plus one standard deviation, the market share for brand 2 increases by 3.8% share points)
Table 4: Changes in Repeat Rate (For example, as household spend increases from the mean minus one standard deviation to the mean plus one standard deviation, the repeat rate for brand 2 increases by 8.4% points; i.e. from 49% to 57%)
marketing management. For brand 1, we see again the earlier result, where households with more loyalty to their favourite supermarket are also more loyal to the leading brand. This also shows up for brands 5 and 6. Smaller households show greater loyalty with regards to brand 6. Smaller spending households show more loyalty with regards to brand 5. Households shopping less-often show more loyalty with regards to brands 2 and 6. How are these results to be interpreted? Brand 6 has a loyalty segment that is smaller households, which shop less often and are more loyal to their favourite
supermarket. From Table 3 we see that in this segment there is not a strongly relevant change in market shares. This segment is not likely to purchase more, but it is more loyal in their choice to buy or not buy Brand 6. In this segment Brand 6 has less elasticity. It will have less sensitivity to deals. It will switch less. Once buyers are won in this market, they are more likely to stay. This will have an impact on the selection of the marketing mix for this brand. Amongst these shoppers it will focus more on attracting new buyers. The brand manager may well be able to devise ways of doing this. What can be done to attract new buyers amongst smaller households who shop less often and are more loyal to their favourite supermarket? The market analysis tool has helped
32
identify the marketing objectives. The task is now for to the brand manager to find the appropriate marketing tactics. The analysis shows that targeting on the basis of loyalty is a possible alternative to the traditional targeting on the basis of market share. One can identify high loyalty segments where there is low elasticity and larger returns from attracting new buyers. Conversely one can identify low loyalty segments where there is considerable switching. As buyers are often switching, small initiatives can generate relevant results but they are likely to be short lived. Marketing initiatives can quickly win gains but these gains can also be quickly lost. Strategies should focus on regularly implementing inexpensive activities.
of the data the less the trend, so the problem may be solved, for example, by analysing quarterly data rather than annual. But generally RBL is a model for more established or mature markets. Two examples have been presented above. The first, with only one covariate, demonstrates how a characteristic of shoppers does not relate to their likelihood to purchase the brand but it still relates to their purchase behaviour. A simple descriptive analysis comparing the profile of the shoppers of the brands would miss this point. Nevertheless a descriptive analysis that included statistics, such as penetrations or purchase frequencies, could identify the important variations in behaviour. The RBL model is not necessary when there is only one covariate. Descriptive analyses will suffice. However, in revealed preference data there are often multiple covariates that are also often correlated. In the second example, not surprisingly, there is a correlation between household size and category purchase rate (r = 0.5). As the model analyses both covariates simultaneously the relative importance of each is determined. This type of pattern in the data, particularly when there are many covariates, cannot be easily investigated with descriptive analyses. This is one of the great strengths of the RBL model. It can analyse multiple covariates. 6. Comparison of RBL with other Discrete Choice Models 6.1 Discrete Choice Models The RBL model is an extension of the beta distribution. However, it has structural links to some of the major discrete choice models: the NBD/BBD, the Dirichlet model, logistic regression and multinomial logit. NBD/BBD. The NBD/BBD model is a simple extension, albeit without covariates, of the model described above. The BBD is applied to the binary choice, the selection between purchasing the specified alternative and all others, conditional on the total purchase rate for all alternatives (often, all brands in a category) over some time period. The negative binomial distribution (NBD) is used to model this total number of purchases made by each of the shoppers in the population. The NBD/BBD model serves to demonstrate that there are two forms of heterogeneity: (1) the variance in the NBD which reflects how much difference there is between heavy and light buyers of the category and (2) the variance in the BBD which reflects behavioural loyalty.
The market analysis tools presented here have the potential to open up a whole new array of marketing strategies based on loyalty segmentation. 5. Data Specification The data required for RBL is repeated choices from the same binary choice set. The data must record the repeated choices for each decision maker; at least two choices but more is better. All decision makers must be presented with the same dichotomous choice set. This is the form of panel data that is often encountered in purchase records, sales databases, scanner data and media surveys. There is a substantial supply of data to which RBL could be usefully applied. The covariates must be constant over the period during which the choice are made. This is typical of panel surveys where there is one baseline survey and then repeated follow-up surveys. The baseline collects the data that doesnt change or which the researcher accepts can only be measured once. This is the data for the covariates. The follow-up surveys record purchases, viewing etc. This is the repeated choice data. RBL is for stationary markets. Over the term of the data collection there can and almost always will be random fluctuations in discrete choice and in the influences on this discrete choice. A stationary market can still fluctuate. It is the nature of these fluctuations which RBL analyses. Many established markets are sufficiently stationary. However, the model is not intended for markets where there are substantial trends such as major evolving brands. Where trends exist, the smaller the term
33
Dirichlet. The Dirichlet Model is an extension of the NBD/BBD to cover more than binary choice. The BBD is replaced by its multivariate alternative, the Dirichlet multinomial distribution (DMD). Thus the Dirichlet is an NBD/DMD model. The NBD depicts the distribution for the category purchase rate. The DMD depicts the distribution for the repeated choices of the brands in the category conditional on the category purchase rate. It may sound attractive to generalize the Dirichlet Model to incorporate covariates. Both the NBD (Cameron and Trivedi 1998; Winkelmann 1997) and the DMD can be expanded in a similar manner as has been done in this paper with the BBD. Thus, covariates can be introduced to each of the two parts of the NBD/DMD. The outcome is a model which fits covariates simultaneously to all the brand performance measures for all the brands in the product category. Conceptually this is simple, but the devil is in the computing code. Besides, while the approach appears to be potentially useful, there is a self defeating dimension to adding covariates to the Dirichlet Model without a critical perspective. This approach requires fitting covariates to the NBD and DMD separately but then combining the results. It is better to keep the analyses separate: fit covariates to the NBD to investigate changes in the category purchase rates; fit covariates to the DMD to investigate changes in the brand preferences. This allows analysis of the covariates directly and separately on each of the heterogeneities; (1) variance in NBD and (2) variance in DMD. Logistic Regression. RBL can be seen as an extension of logistic regression. In logistic regression, only one binary choice is observed for each respondent, vs. a series of choices in RBL. If data are available on a single choice per respondent, the likelihood functions for logistic regression and RBL are identical. This is not just coincidence. The theoretical foundations of logistic regression are in loglinear modelling and random utility theory whereas the foundation of RBL is in the beta binomial distribution (BBD). These bodies of theory are related, and from the theoretical and practical perspective, it is both encouraging and not surprising that they generate the same functional form and likelihood function. Multinomial Logit. Random utility theory, as used in multinomial logit (MNL), is an excellent model for fitting covariates to the conditional choice between alternatives such as the brands in a product category. It models the mean choice probability. It has an excellent ability to model varying choice sets and the attributes of
each alternative. Obviously, RBL cannot do this as it analyses binary choice only. However, RBL incorporates repeated purchases, and it can therefore analyse variations in heterogeneity and loyalty directly. Expansion of the RBL model to incorporate more than two alternatives, the attributes of the alternatives and varying choice sets is conceptually possible, but seems exceedingly complex. 6.2 Identification of Stochastic Processes The comparison of the RBL and MNL models can be further developed by examining their stochastic forms. While the models have strong similarities in their likelihood functions they differ in their conceptualisation of the random influences on discrete choice. RBL identifies that each shopper has a fixed probability of selecting A or B over the choice occasions. Consider a subset of shoppers who all have the same values for each of the covariates; i.e. a common x. These shoppers do not all have the same probability. Each shopper i has a probability, pA,i, which is a separate observation from a beta process; i.e. pA,i is an observation of a random variable P with a beta distribution with parameters A(x) and B(x). Within the subset with common x all shoppers will have the same parameters for this beta process and the same expected value as in Equation 5 but they will not have the same probability pA,i. The observed repeated discrete choice generates a count of the number of times shopper i selects Alternative A. This count, rA,i, is an observation from a binomial process; i.e. rA,i is an observation of a random variable R with a binomial distribution based on k trials and the probability pA,i. Thus, in the RBL model there are two stochastic processes. There are two terms which reflect the unobserved influences on shoppers discrete choices. The first is reflected in the variation between shoppers in their latent probability, pA,i, even between shoppers with the same x. In the beta process this error is the difference between P and E[P]. The second is reflected in the variation within shoppers in their succession of k choices. Most shoppers do not always select the same alternative. In the binomial process this error is the difference between R/k and E[R/k]=P. Now it is also possible to discuss how RBL represents randomness in decision making. Each shoppers discrete
34
choices are affected by a range of influences. These might include households requirements and the marketing mix for the alternatives. Each shopper has a probability pA,i. This is nothing more than his/her long run stationary chance, or proportion of times, A would be selected. It is an aggregation of all the influences on the shopper. The differences between shoppers in the aggregation of these influences can be (1) observed and captured through x, or (2) unobserved and captured through the beta process. However, at successive choice occasions each shopper varies in his/her choice. This is because of fluctuations in some of the influences and is captured through the binomial process. The nature of the RBL model is that the observed influences must be constant over the period of the data collection and any fluctuation in the unobserved influences must be stationary. For the user of RBL this stochastic structure delivers an additional benefit which is theoretically a little complex. MNL is constrained because the variances of the error terms are not identified and consequently parameter values from similar studies on different data sets cannot be directly compared. With RBL this problem does not exist. The combination of the beta and binomial processes described above leads to the beta binomial process. The RBL model identifies the variance of the beta and the beta binomial processes as functions of x. Thus, all the variances of the stochastic processes are identified. This is in stark contrast to the MNL model where the stochastic error term in the utility has an unidentified Gumbel distribution. There is no ambiguity, no lack of identification, in the RBL model. Consequently, the parameter values for comparative studies using RBL and different data sets can be directly compared. 6.3 Revealed Preference RBL has been presented as a model for revealed preference data. This is data which records shoppers actual purchases. In the analysis of discrete choice, stated preference data is often used and is generated through respondents in surveys making hypothetical choices. The benefits of revealed preference data include (1) it can be inexpensive to obtain as it is automatically collected in customer databases and (2) it reflects shoppers actual choices and is not hypothetical. The benefits of stated preference data include (1) it can examine new attributes currently not available on the market and (2) it can be used to conduct experiments.
This last point is critical. Revealed preference data is usually observational, stated preference is usually experimental. Regardless of the type of data, be it discrete choice or interval/ratio data, the differences between observational and experimental data must be kept in mind. Firstly experimental data is much stronger in implying causation. Secondly with observational data there are limitations arising from omitted variables. Observational data can and will contain correlations in the covariates; i.e. colinearity. In stated preference data factorial designs can remove most colinearity. These correlations in revealed preference data can lead to biased results particularly when there are omitted covariates. For example, Table 2 to Table 4 show that for two brands visits to supermarkets is significant and relevant. It may be that there is a brand targeted on this basis so the marketing mix for that brand is correlated with the visits to supermarket. By excluding marketing mix as covariates and including visits to supermarket the tables might be interpreted incorrectly. The optimum solution is to select covariates intelligently, observe as many covariates as possible and to use a multivariate model. RBL is one such model as it allows for many covariates. When applying RBL to revealed preference data, as with all observational data, care must be taken to account for missing covariates. 6.4 Further Research RBL uses repeated binary choices, and covariates which are constant, to examine variations in market shares and repeat rates. It has been presented as a tool for revealed preference data. On this basis alone it is very useful for segmenting on repeat rates. However, there are clear benefits in expanding the model to include more than binary choice, covariates which change and experimental data. Already RBL can be used for comparisons between revealed preference data sets. Eventually the method will be expanded to stated preference where the replications involve different choice sets. Then it will be possible to compare revealed and stated preference, building on the work of Louviere, Hensher and Swait. This is the subject of further research. 7. Conclusion The paper introduces RBL, a model for repeated binary discrete choice data. RBL identifies the characteristics of decision makers that have an impact on classic measures of loyalty such as repeat purchase. Thus, it provides a
35
methodology for targeting on the basis of loyalty. It has the potential to identify empirical generalizations regarding the roles of covariates in loyalty, or to show that these roles vary between brands. It is a surprisingly simple model to use. Finally, RBLs applications potentially cover many forms of binary choice, not just brand choice. References Ben-Akiva, M. and Lerman, S.R., 1985. Discrete Choice Analysis: Theory and Application to Travel Demand. The MIT Press, Boston MA. Cameron, A.C. and Trivedi, P.K., 1998. Regression Analysis of Count Data. Cambridge University Press, Cambridge and New York. Edwards, A.W. F., 1976. Likelihood - An Account of the Statistical Concepts of Likelihood and its Application to Scientific Inference. Cambridge University Press, Cambridge and New York. Eliason, S.R., 1993. Maximum Likelihood Estimation Logic and Practice, Sage Publications, California. Guadagni, P.M. and Little, J.C., 1983. A logit model of brand choice calibrated on scanner data. Marketing Science 2 (Summer), 203-38. Johnson, N.L., Kotz, S. and Balakrishnan, N., 1994. Continuous Univariate Distributions, Volume 2 (2 ed.). John Wiley & Sons, New York. Johnson, N.L., Kotz, S. and Balakrishnan, N., 1997. Discrete Multivariate Distributions. John Wiley & Sons, Inc., New York. Kotz, S., Balakrishnan, N. and Johnson, N.L., 2000. Continuous Multivariate Distributions, Volume 1: Models and Applications (2 ed.). John Wiley & Sons, Inc., New York. Louviere, J.J., Hensher, D.A. and Swait, J., 1999. Stated Choice Methods Analysis and Application. Cambridge University Press, Cambridge and New York. McCullagh, P. and Nelder J.A., 1989. Generalized Linear Models (2nd ed.). Chapman & Hall/CRC, London. McFadden, D., 1984. Conditional logit analysis of qualitative choice behavior. In: Griliches, Z., Intriligator, M.D. (Eds.), Handbook of Econometrics. NorthHolland.
McFadden, D. 1974. Conditional logit analysis of qualitative choice behaviour. In: Zarembka, (Ed.), Frontiers in Econometrics. Academic Press, New York. Winkelmann, R., 1997. Econometric Analysis of Count Data (2nd ed.). Springer-Verlag, Berlin. Acknowledgements The authors wish to thank MarketingScan and GfK, and especially Laurent Battais, Raimund Wildner, and Grard Hermet, for providing the data and for their assistance in this study. Biographies Cam Rungie is a Senior Lecturer in Marketing at the University of South Australia. He completed his PhD on Discrete Choice at the University of South Australia in 2000, having previously completed an honours science degree in mathematics at the University of Adelaide and a Masters of Business Administration at Queens University, Canada. He spent 15 years in commercial market research and consulted to WHO on survey methods throughout Asia. His research interests include the analysis of purchase and audience data. He is consulting with OzTAM and ACNielsen and won an ESOMAR best paper award in 2000 with ACNielsen. Gilles Laurent is Carrefour Professor of Marketing at HEC School of Management, Paris. His Ph.D. is from MIT (1978). He has published papers in a variety of journals (JM, JMR, MkgS, MgtS, JCR, IJRM, MkgLet, etc.) on somewhat unrelated topics, including: the structure and measurement of consumer involvement, choice by older consumers, brand awareness, evoked sets, sales promotion, the reliability of brand attitude measures, consumers immediate memory for prices, linear programming, etc. He is a former editor of the International Journal of Research in Marketing, and a former academic trustee of the Marketing Science Institute. Correspondence Addresses
Cam Rungie, Senior Lecturer, School of Marketing, University of South Australia, GPO Box 2471, Adelaide SA 5001, Australia. Telephone: +61 (0)8 8302 0768. Facsimile: +61 (0)8 8302 0442. Email: cam.rungie@unisa.edu.au.; Gilles Laurent, Carrefour Professor of Marketing, Marketing Department, HEC School of Management, 78350 Jouy-en-Josas, France. Telephone: +33 (0)1 39 67 74 80. Facsimile: +33 (0)1 3967 7087. Email: laurent@hec.fr.
36
Environmental Attitudes
Environmental Attitudes and their Relation to the Dominant Social Paradigm Among University Students In New Zealand and Australia
William E. Kilbourne & Michael J. Polonsky
Abstract This paper develops a causal model of environmental attitudes using measures of the dominant social paradigm of Western industrial societies. Four components of the DSP framework are examined with regard to environmental attitudes and perception of change using a sample of university students from Australia and New Zealand. The results indicate that ones belief in the DSP has a negative effect on both environmental attitudes and perception of change necessary to ameliorate degradation of the environment. Thus, while public policy favors increasing awareness of and interest in the environment, policy instruments may remain ineffective in producing lasting change if the components of the DSP remain unchanged. It is argued that public policy ought to be directed at changing the DSP so that its negative effects will be minimized. Keywords: DSP, Environment, Attitudes, Socio-economic, Cosmology
1. Introduction Environmental concern across countries has been a topic of study for the past thirty years. Interest has waxed and waned over the period, but it has not disappeared. Globally, consumers environmental concern has evolved differently. For example, 31% of people in the US thought environmental issues were exaggerated compared with 24% in New Zealand, 23% in Australia and 22% in the UK. In the same poll it was suggested that 83% of Australians participate in recycling, followed by 71% in NZ, 65% in the US, and only 61% in the UK (Levine, 2002). The increasing pace of globalization means that environmental degradation will be an important issue long into the future. There have been many different approaches taken in examining the marketing and environment interface during that period. Kilbourne and Beckmann (1998) surveyed the marketing literature relating to the environment and concluded that there were three major streams of research evident. The first sought to conceptualize green marketing and green consumers. The underlying purpose in this type of research was to develop target markets for green products, but no common denominator emerged from the
research. A second stream related to conservation related initiatives. These focused on engendering conservation behaviors and legislative initiatives relating to such behaviors as recycling and reducing consumption. Again, no clear agenda evolved from the research. The third stream of research, however, began to focus on the more general area of sustainability. Here, the managerial orientation of previous environmental research was set aside for a more macro-oriented approach to the integration of the natural environment in marketing. It was this stream that introduced the study of environmental values and the role of the dominant social paradigm (DSP) (Dunlap and Van Liere, 1978; Pirages and Ehrlich, 1974) into the academic literature. The DSP is defined here as the shared beliefs and values that make up a cultures worldview and that functions as ideology. It is at this level that the present paper will focus. 2. Literature Review Cotgrove (1982) and Milbrath (1984) were the first to examine the DSP from an empirical perspective, and they concluded that there were several relevant dimensions in the construct. Among them could be discerned political, economic and technological components that were later confirmed by Dunlap and
37
van Liere (1984). The approach taken by all three studies was however, empirically driven and none developed a substantive theoretical conceptualization of the DSP or its components. The conceptualization and expansion of the DSP was begun by Kilbourne (1995) and Kilbourne, et al. (1997). They argued that the connection between marketing and the environment is through the DSP, and that to understand the marketing-environment relationship, more focus on the DSP was necessary. The DSP has since been conceptualized as a two dimensional construct with three components in each dimension (Kilbourne and Beckmann, 1998b). The first dimension is the socio-economic containing the political, technological, and economic components. The second dimension is the cosmological containing organization (human position in nature), structure (atomism-holism), and function (cooperation-competition) as the predominant character of the cosmos in which humanity finds and orients itself. We will examine the conceptualization of each of the dimensions briefly. Because there has been more research on the economic and human position dimensions, we will examine them more thoroughly than the others. 2.1 Political Dimension The political dimension is based on the political liberalism of Locke (1963 [1690]). The first essential element is possessive individualism arguing that individuals are in possession of themselves. As a result, all that is created by their labor rightfully becomes their property. This leads to the second aspect of political liberalism which is private property. In addition, Locke adds the rights of exclusivity to property and extends protection to it by governmental action if necessary. This leads to the third relevant aspect of political liberalism which is limitations on government. In Lockes theory, the only functions of government are the protection of individuals and their private property and the enforcement of contracts which acts as a further protection of property. Possessive individualism, private property, and limited government are three of the basic institutions of Western liberal society, and in Lockes time, they set the stage for the development of economic liberalism and capitalism that were to follow a century later. 2.2 Technological Dimension The development of science as we know it today began
in the seventeenth century most notably with Bacon (1944 [1620]) in whose hands the objective was transformed from the Aristotelian conception of science. In particular, Bacon saw science as a singular road to the betterment of mans estate (Leiss, 1972). Thus, science became, not the way of knowing the world, but the way to change it to lessen the burdens of human existence. The handmaiden of science became technology, the bridge between scientific knowledge and the transformation of nature (Merchant, 1980). The project of the enlightenment, at least in this cause, was more successful than even Bacon might have imagined. In the succeeding four centuries, technology has become the sine qua non of material progress and is so firmly embedded in the institutions of Western culture that it is virtually invisible (Winner, 1986). Technological optimism is the mode of thought that is most characteristic of industrial society (Postman, 1993), and this has earned it a fundamental position in the DSP of Western societies. Because technological progress is equated with progress per se, and progress, in general, has been equated with material progress (Bury, 1932), we see the clear link between the technological and the economic. The superordinate goal of Western societies is the institutionalization of economic growth, and the means to effect this growth is technology. The items in the measurement instrument used in this study are designed to reflect technological optimism. 2.3 Economic Dimension While a complete analysis of the relationship between neo-classical economics and the environmental problem is beyond the scope of this study, its essential aspects will be addressed. This can be separated into the micro and macro dimensions of the problem, and we will examine each briefly. In the micro dimension, with the transformation from classical to neo-classical economics, the ethical and political aspects were separated from the discipline as they were not easily empirically quantifiable and, thereby, were considered non-science (Kassiola, 1990; Mirowski, 1988; Mirowski, 1989). The basic result of Smiths (1937 [1776]) economic reductionism was a society of atomistic individuals with each pursuing their own self-interest in impersonal markets. It was then argued that the greatest good would accrue to society through the operation of the invisible hand (Smith, 1937 [1776]). Questions of morality and the distribution of wealth that had occupied classical economists were
38
expunged from economic discourse, and environmental considerations were only tangentially integrated. The abstract concept of utility replaced production and consumption of material goods further separating economics from nature. The pursuit of economic growth became the summum bonum of progress and, through it, all the needs of society could be met. The meaning of life was "desacralized" and became the pursuit of material gain. How to effect this economic process in the most technically efficient way became the province of marketing as we find it today. The over-arching goal of main-stream marketing remains, as suggested earlier, the maximization of consumption suggesting that it still follows the direct track of the neo-classical paradigm. Within the macro dimension, the primary aspects of the economic contribution to environmental degradation are derivatives of the micro dimension and include what the literature has often referred to as externalities (Mundt, 1993). It is generally assumed within economics that the maximum social good is achieved as the sum of individual goods, i.e., it is a product of the neo-classical assumptions. This has lead to a view that continuous economic growth is positive because it is assumed that the more we have, the better off we are. This is what Hetrick (1989) refers to as the ideology of consumption. The environmental consequences of this strategy have generally never been questioned because, as suggested earlier, economics has been separated from nature. Nature has traditionally been considered a free good to be used in the pursuit of growth (Mundt, 1993). Economic transactions focus on extraction costs and ignore any possible payments to nature. This effectively under-prices the real cost of goods and results in inefficient allocations that appear to be economically efficient. Whether it is possible to price resources appropriately is problematic and has been examined by O'Neill (1993) and Sagoff (1988) among others. The consequence of these historical developments in economics has been to transform conceptualizations of the good from the ethical to the material. Maximizing economic growth with the efficient allocation of resources through impersonal market mechanisms is the essence of neoclassical economics. The items in the measurement instrument reflect these economic goals. The socio-economic dimension has been examined empirically and found to be related to environmental attitudes and willingness to change consumption patterns in consumers (Kilbourne, Beckmann and Thelen, 2002).
Kilbourne and Alsem (1997) indicated that the economic dimension was the most influential in determining both environmental attitudes and consumers willingness to change, while the political and technological dimensions were sporadically relevant. To date there has been no empirical research on the structure and function dimensions of the cosmological domain. Beckmann, et al. (1998) have, however, examined the organizational dimension and concluded that it was related to value systems that were then predictive of ecological concern variables. We will first provide a brief explanation of the structure and function dimensions of the DSP and then a more thorough explanation of the organization dimension because more empirical research has been done on it than on the others. 2.4 Structural Dimension Structure here refers to beliefs about the composition of the world in which we live. It is not unrelated to the technological in that the scientific attitude that prevailed at the time was based on atomism suggesting that nature was composed of fundamental building blocks that were to be discovered. The prevailing belief is still that the universe can be benignly deconstructed in an effort to reduce it to these building blocks so that it can be understood and manipulated in the service of humanity. While this view has been subject to question in quantum mechanics for decades now, it is still the dominant view and pervades human thought and action (Capra, 1982). This belief effected the shift from the organic view of nature as a living, growing entity to the mechanical view in which all the elements of nature are like parts of a machine. The relationship between this view and political and economic liberalism is evident. Market and political relations have been considered impersonal competitive relations between independent individuals having no connection to each other, i.e., atomized individualism. Again, the elements of the DSP are mutually reinforcing. This leads to the next dimension of the DSP. 2.5 Functional Dimension With scientific development applied to biology, nature was reduced to the competitive struggle for existence in the Origin of Species. However, it was Herbert Spencers (1969 [1851]) concept, later known as Social Darwinism, that firmly established the role of interpersonal competition among independent individuals as the sine qua non of social progress. Darwin provided no universally desirable goal and was
39
therefore, non-teleological. Spencer clearly defined competition between individuals as the only means through which progress could be consistently insured. Andrew Carnegie expressed this law of competition and its role in material progress most succinctly in his statement, for it is to this law that we owe our wonderful material development (Donaldson and Werhane, 1993, p. 226). He added that it is only through competition between individuals that future progress is insured. This is, of course, a reflection of Hobbes (1950 [1651]) ubiquitous war of all against all in political philosophy preceding Locke. It was in Hobbes possessive market society that the sovereign power served the purpose of keeping competition within peaceful bounds (MacPherson, 1962). The concept of social competitiveness passed from Hobbes to Adam Smith through Locke and has since become the standard in assessing the ability of markets to best serve the march of material progress. Individual competitiveness and the human competition with nature are fundamental beliefs within the DSP, and they motivate the prevailing conception of the role of humans in nature. 2.6 Human Position Dimension The only aspect of cosmology that has, as yet, been included in empirical research examining marketing interactions with the natural environment has been the position of humans in the cosmos, i.e. the human position dimension as used here (Beckmann et al., 1998). While there is disagreement on the dimensionality of this construct, we have adopted the perspective that there is a single dimension with anthropocentrism at one end and ecocentrism at the other. Eckersley (1992) provides the most complete conceptual development of the human position dimension and our approach here is consistent with hers. We begin with our conceptualization of the anthropocentrism and ecocentrism poles of the construct and then examine the implications of empirical research extant. The essential feature of the human dimension of the cosmological domain is the belief that humans are separate from and morally superior to the rest of nature. As a result, humans consider themselves to be the masters of nature subduing it for their own instrumental purposes. With the demystification of nature (Lewis, 1947), through scientific and technological development, its manipulation and exploitation were assured and resulted in the death of nature (Merchant, 1980). This has been referred to as anthropocentrism. The antipodal position to anthropocentrism is
ecocentrism which considers nature to have inherent value regardless of its usefulness to humans (Merchant, 1980; Purser, Park, and Montuori, 1995; Shrivastava, 1995). There are two perspectives from which to examine the ecocentric position. The first is the position that the objective is "human emancipation and fulfillment in an ecologically sustainable society (Eckersley, 1992, p. 26)." This position has been described as human welfare ecology (Kilbourne, 1995; O'Riordan, 1976). The second acknowledges the same objective but with recognition of the moral standing of the nonhuman world and its right to continue evolving. The primary point of departure between the two views is the position of humans in the biosphere. Within the ecocentric view, there is no basis for assuming that humans have rights superseding or negating those of other life forms, which are considered to have inherent value in their own right. Ecocentric theorists postulate that the current ecological crisis stems from this over inflated sense of human value, or, as Ehrenfeld (1978) calls it, the arrogance of humanism. This view is pointed out most forcefully by Singer (2004) in his extensive work on animal rights, or what he refers to a speciesism. He argues that all sentient beings deserve a moral standing. It is further argued by ORiordan (1976) that even the weaker forms of anthropocentrism such as conservationism and human welfare ecology are not sustainable because, in the presence of human crisis, they would be sacrificed for the more humanist perspectives. Thompson and Barton (1994) examined the effect of anthropocentric and ecocentric beliefs on attitudes toward the environment and concluded, albeit weakly, that ecocentric beliefs affected environmental attitudes as measured by three scales constructed for their study, conservation behaviors, self-reported actions, and environmental apathy. However, they defined anthropocentrism and ecocentrism as two different constructs rather than opposite ends of a continuum as would have been more consistent with the literature. This leads to the intuitively inadequate reflection that one might be an anthropocentric ecocentric if the two constructs are defined as independent of each other. It would also possibly lead to measurement difficulties if the content of one scale is contained within the other. This could explain Thompson and Bartons (1994) weak and somewhat conflicting results suggesting that anthropocentrism has no effect but ecocentrism does. In
40
the current study, we are taking the approach tied more directly to Eckersleys (1992) formulation that characterizes the two constructs as opposite ends of a continuum rather than independent. Beckmann, et al. (1998) also examined the anthropocentrism component of the cosmological domain. They concluded that anthropocentrism was negatively related to self-transcendent values, which were then positively related to environmental attitudes. Thus, as individuals values focus beyond themselves, they tend to be less anthropocentric, and, consequently, have more positive environmental attitudes. Kilbourne (2002) included anthropocentrism as part of a structural equation model of environmental attitudes across cultures and found that it contributed significantly to the DSP latent variable. The DSP was then shown to be negatively related to environmental attitudes and willingness to change consumption behavior. The empirical evidence to date suggests a structural model of DSP and environmental attitudes exists and can
be empirically tested. The primary purpose of this paper is to test the validity of the model and to determine if it holds across the two similar developed countries, Australia and New Zealand. The model is presented in Figure 1. Included in the figure are the hypothesized influences. As can be seen, the model suggests that the DSP is negatively related to both environmental attitudes and change. Change as used here has two dimensions, one individual and the other social. Individual change refers to ones environmental concern and their expressed willingness to change consumption behavior. The social dimension of change refers to the degree of social and political change necessary to ameliorate declining environmental conditions. The relationship between environmental attitudes and willingness to change has been examined and shown to be consistently positive (Kilbourne, Beckmann and Thelen, 2002). We propose the causal model shown in Figure 1 based on this conceptualization of the environmental problem. The figure also indicates the observed variables the measurement of which is explained in the next section.
Figure 1: Proposed Causal Model
Pollution Extinction Ozone Ecology Political (-) Technological Economic DSP Human Position Consume Competition Atomism (-) Concern Change Social Individual Trade-off (+) Warming Consumption Water
DSP: Dominant social paradigm
Political Social Condition
41
The specific hypotheses to be tested are the following: H1: The path from DSP to ecology will be negative indicating an inverse relationship between the DSP and environmental attitudes. H2: The path from DSP to change will be negative indicating an inverse relationship between the DSP and perceived change. H3: The path from ecology to change will be positive indicating that as environmental attitudes improve, perceived change increases. H4: The structural model will be invariant across the two countries indicating similar relationships between the constructs. 3. Methodology 3.1 Sample The respondents for the study were selected as a convenience sample of university students from Australia and New Zealand. The final sample consisted of 148 from Australia and 122 from New Zealand. The sample contained 61% males and the approximate median age was 21. The use of students was considered acceptable here because the variables of interest are basic attitudinal variables that are stable within individuals and would not change as a consequence of being a student. An additional advantage was that the procedure controlled for age because most students fall in a fairly narrow age range, and the age group used tends to be more sensitive to environmental issues (Diamantopoulos et al., 2003). Questionnaires were distributed in class, and students were asked to complete and return them the next week. Participation was voluntary, and class credit was not given for participation. The response rate for Australia was approximately 70% and for New Zealand 78%. 3.2 Measurement instrument The questionnaire consisted of six sets of five questions measuring the six components of the DSP. Fifteen questions derived from Milbrath (1984) and Cotgrove (1982) were included to measure environmental attitudes. There were also 15 questions examining perceptions of the environment, willingness to change consumption behavior, and social change necessary to be more environmentally sensitive. The last four questions asked about demographic characteristics. Because of the large number of observed items relative to the sample sizes for the two countries, we chose to use
composite indices derived from the observed measures. The use of such composite measures as indicators of latent variables is recommended by Bagozzi and Heatherton (1994) and reduces random error and degrees of freedom allowing a smaller sample size to stay within acceptable limits. Because of the exploratory nature of the study, the 30 DSP items were subjected to factor analysis to determine the appropriate number of items for each measure and to verify that there were six sets of measures, one for each dimension of the DSP as described here. The results of this analysis suggested that the six dimensions were evident in the sample, but that they did not all contain the five original items. Anthropocentrism (alpha = .82) and economic (alpha = .76) each retained all five items. Atomism (alpha = .75) retained four items and technology (alpha = .74), political (alpha = .60), and competition (alpha = .67) each retained three items. The means of each of the DSP dimensions was used as the composite measure for the analysis. Following this reduction strategy, the DSP latent construct with six observed variables was analyzed to determine if the construct held up for both the Australian and New Zealand samples. It was found that configural invariance was not supported with one item in each sample being non-significant. Specifically, for the Australian sample the political variable was not significant and for the New Zealand sample, the competition variable was not significant. While it is not essential that the models be identical, to make meaningful comparisons across groups, it is preferable that the groups contain the same observed variables (Byrne, 2001). Consequently, we removed the two items for the remainder of the analysis. The four item construct proved to be invariant and was used in the model as the DSP latent construct. The p-value for the Chi Square statistic was 0.199, RMSEA was 0.043, the adjusted goodness of fit index (AGFI), the Tucker-Lewis index (TLI), and the comparative fit index (CFI) were all well above the 0.90 level recommended by Hair, et al. (1998) indicating a good fit for the construct. The environmental latent construct was derived by starting with the 15-attitudinal items and successively reducing them to achieve an acceptable fit for both countries. We used modification indices and correlations with the construct as the criteria for eliminating variables. The final environmental attitude latent construct was measured using seven items, and the fit was within acceptable parameters for both countries. The
42
items used related to pollution, species extinction, ozone depletion, global warming, levels of consumption, rainforest depletion, and water shortages. Coefficient alpha for the scale was 0.76. The RMSEA for this component of the model was 0.034, CFI, TLI, and AGFI were above 0.90. The p-value for Chi Square was 0.147 indicating a very good fit for both countries. The perception of change is also a latent construct and consisted of six items. The first two related to individual change and measured willingness to consume less and willingness to trade ecology for economy. The second two were social and measured the perceived social and political change necessary to ameliorate environmental degradation. The final two were measures of concern. A factor analysis of these six items revealed a two dimensional construct with concern, trade-off consumption, and limit consumption as the three items in the first dimension. The second dimension contained political change, social change, and condition of the environment as the items. These six items were shown to constitute a second order factor model with three items each for individual change and social change. The pvalue for the Chi square statistic was 0.18, RMSEA was 0.034, CFI, TLI, and AGFI were above 0.90 indicating a very good fit. After the three latent constructs were derived and validated separately as suggested by Joreskog and Sorbom (1993), they were combined in the structural equation model presented in Figure 1. The measurement model was then tested for each country separately to establish structural similarity as suggested by Byrne
(1999). The results of this analysis are presented in the first two rows of Table 1. After the measurement model was assessed separately for both countries, then we followed the procedures suggested by Steenkamp and Baumgartner (1998) for establishing the cross-cultural invariance of the measurement model before examining the structural model. The necessary steps in this procedure are to first establish configural invariance indicating that the factor structures of the models are similar. Next, metric invariance is established indicating that the same pattern of significant factor loadings exists for both models. Finally, the invariance of both variances and covariances is established. This indicates that the correlations between factors can be compared, but it is not strictly required for comparing the different structural models (Steenkamp and Baumgartner, 1998). To begin the process, we first established the configural invariance and then that model was used as a baseline model against which to compare succeeding models. To assess the increasingly constrained models we again used the AGFI, the TLI, the CFI, and the RMSEA. To compare successively constrained models to the baseline model, we used sequential Chi-square difference tests (SCDT) recommended by Anderson and Gerbing (1988). The statistical results of each successively constrained model are presented in Table 1. Configural invariance was established indicating that the factor structure for both countries was the same. The statistical result is presented in line 3 of Table 1. This was then used as the baseline for the SCDT on succeeding
Table 1: Results of Invariance Tests
* Values reported in this column are the p-values for the SCDT results. Values greater than 0.05 indicate that the constrained model fits the data as well as the unconstrained model.
43
Table 2: Correlations Between Constructs
Figure 2: Standardized Regression Coefficients

Australia New Zealand
Ecology -0.60 (.001) -0.71 (.001)
Ecology
DSP
0.31 (.030)
DSP
0.26 (.030)
-0.74 (.001) Change
-0.72 (.001) Change
Significant levels in parentheses
* Values reported in this column are the p-values for the SCDT results. Values greater than 0.05 indicate that the constrained model fits the data as well as the unconstrained model.
models. For comparing nomological models it is necessary to establish metric and variance invariance but not to establish scalar invariance or covariance invariance if the correlations between constructs are not of interest (Steenkamp and Baumgartner, 1998). We first constrained the factor loadings to be equal for both countries. The results of the test (line four of Table 1) indicated that full metric invariance was supported. This allows the comparison of difference scores on the items so that differences observed in the latent variables are really between countries and cannot be attributed to measurement differences. The SCDT of 0.44 indicates that the constrained model predicts as well as the unconstrained model.
We next constrained the variances to be equal for both countries. The results of this test (line five of Table 1) indicated that variance invariance was supported as well. With these invariance tests completed, we then could test the structural model. To do this, we first allowed all parameters to vary freely to establish the validity of the structural model proposed. This is the model against which the constrained model is tested. Following this, we constrained the path coefficients in the model to be equal. If the constrained model predicts as well as the unconstrained model, it indicates that the path coefficients are the same for both countries. Having established the invariance of the model across the two countries, we could then determine if the coefficients
44
were in the hypothesized direction, and we could meaningfully compare them across countries. Correlations between the final constructs in the measurement model are presented in Table 2. 4. Results In the first stage of this analysis, we allowed the structural paths to vary freely to determine the significance of the structural model. We then used this model as the base line against which to compare the constrained model. The results, shown in Table 1, indicated that the structural model was valid. All fit statistics were within the accepted range. In this case, the only coefficient that was not significant was the path from ecology to change for the New Zealand group, though the coefficient was positive as predicted. We next constrained the paths of the structural model to be equal and reassessed the fit. For this analysis, shown in the last line of Table 1, the constrained model was shown to fit the data well. The SCDT of 0.42 indicates that the constrained model predicts as well as the free model. Thus, the proposed structural model was shown to be invariant between the two countries, Australia and New Zealand. In this constrained model, all structural paths were significant at less than 0.03 for both countries, and it was also shown that all indicator items for the latent constructs were statistically significant (p<.05). A summary of the results for the constrained structural model is provided by Figure 2. It was hypothesized that the latent constructs DSP, environmental attitudes, and perception of change would be statistically related. Specifically, the DSP would be negatively related to both environmental attitudes and perception of change. The relationship between environmental attitudes and perception of change was hypothesized to be positive. The results of the analysis confirmed all three of the hypotheses. Hypothesis 1 suggested a negative relationship between the DSP and ecology. The regression coefficients for Australia and New Zealand were 0.60 and 0.71 respectively and both were significant at p < 0.001. This confirmed hypothesis 1. Hypothesis 2 indicated a negative relationship between the DSP and change. The results were again confirmed with coefficients of 0.74 and 0.72 respectively. Both were significant at p < .001. For hypothesis 3, the results were partially confirmed. The path from ecology to change was positive and significant (p < .05) for the Australian sample, and the path was positive but not significant (p < .27) for the New Zealand
sample. However, when the paths were constrained to be equal to test hypothesis 4, the path from ecology to change for both the samples was positive and significant at p < .03 with coefficients of 0.31 for Australia and 0.26 for New Zealand. Structural invariance was supported indicating that the coefficients did not vary across the samples. This suggests that, while the path from ecology to change was positive for both countries, the effect was stronger for the Australian sample. 5. Discussion The results of the study have both theoretical and practical significance from a public policy perspective. Most research into the relationship between marketing and the environment that has focused on the link between environmental attitudes and specific behaviors or intentions (Kilbourne and Beckmann, 1998a) has found the relationship to be positive. It has been assumed that if environmental attitudes or knowledge can be improved, then individuals would change their behavior and become more environmentally friendly. While environmental attitudes and knowledge have, over the last thirty years, increased in the US, there has not been a significant increase in pro-environmental behavior. This has been referred to as the attitude behavior gap (Beckmann and Kilbourne, 1997). The results of this research offer a suggestion as to why this is the case. The argument presented here suggests that there are antecedent variables that need to be considered. This includes the DSP and its constituent components. Their role has not been extensively examined within the consumer behavior literature in regards to the environment - marketing relationship. The complexity of the constituent components makes it potentially difficult to empirically examine consumers underlying values in regards to the role of the consumer and consumption in the wider cosmos. In many situations, consumers do not in fact even consider the underlying connections they have to the natural environment. In this way, researchers are possibly asking consumers to consider issues that are not normally considered in a systematic process. Similar criticisms are frequently made with other research such as conjoint analysis that seeks artificial distinctions to be made by consumers. From a theoretical perspective, the results suggest that additional research should focus on defining the DSP and understanding its relationship to how the environment is viewed by consumers. This would include an integration of how national cultures view consumers
45
in the broader ecosystem as well as how they view the specific types of consumption behaviors. For example, in a society where the ownership of goods defines the self, there will most likely be a different set of interconnections between consumption and the environment as compared to a situation where consumption is seen more functionally as a means to support existence. Any such differences will, of course, have wider implications for policy development as well. However, it might be the case that these cultural differences in views partly explain the environmental anomaly indicating that, while citizens in the developed world have become more aware and concerned about the environment (Dunlap, 1991), the environmental movement in general is failing, at least in the US (Alperovitz, 2004; Dowie, 1995). The lack of explicit consideration of the DSP within consumers thinking also has significant public policy implications. The model presented suggests that traditional efforts on the part of public policy and consumer policy officials to increase environmental awareness will continue to be less effective than anticipated, because the four dimensions of the DSP examined in this study militate against changes in consumer attitudes and perceptions necessary to bring about meaningful environmental improvements. As such, public policy might not seek to effect changes in consumers attitudes directly, but rather policy makers might focus more attention on consumers more widely held beliefs in regards to aspects of the DSP as well. While changing individual, and therefore societal, commitment to the DSP could have a positive effect on both environmental attitudes and willingness to change, it is unclear which governmental body would take responsibility for seeking to bring about such changes. In most cases, environmental policy makers within a country are in fact extremely fragmented, focusing on narrowly defined issues of interest (Chilton 2000). The type of change in consumer attitudes that is needed would require policy makers transcending particular issues, as shifting the DSP is not simply about adopting recycling behavior, purchasing more environmentally responsible products, or even adapting a simplified lifestyle. Shifting the DSP might not fit within the traditional public policy framework of any one organizational unit. As such, it would be something that governments would have difficultly implementing, as no one policy area would see changing the DSP as their core area of responsibility. If this is the case, it may mean that all socially oriented programs are destined to fail without
some overarching integration of the need to reconsider the role of the consumer in the wider ecological system, i.e. shifting the DSP. 6. Limitations There are several limitations on this study that should be considered before generalization of the results is warranted. The first is that the sample was a convenience sample of university students. While the results are valid for this group, caution should be exercised in generalizing beyond them. In addition, the two samples were from Australia and New Zealand, and these countries are socio-economically and industrially similar. Because of this, generalization to other cultures can be problematic if they have different socio-economic characteristics and may not apply to other developed Western countries. Future research is warranted by the results, however. This research should expand the diversity of the cultures sampled to include a wider variety of characteristics. In addition, random samples of the populations should be used to allow for better comparisons and external validity. While we have sampled the full domain of the DSP construct, in these particular samples, two of the dimensions were not consistent across the samples. For the Australian sample, the political dimension was not significant, and for the New Zealand sample, competition dimension was not significant. It would have been preferable to be able to assess the effect of all six dimensions to get the full implications of the construct. The model assessed here contained four of the six dimensions so that comparisons of structural coefficients would be meaningful. While we did limit the study, here, our results are consistent with previous research indicating the relevance of the four dimensions in environmental attitudes and change. References Alperovitz, G., 2004. America beyond capitalism: Reclaiming our wealth, our liberty, and our democracy, John Wiley & Sons, New York, NY. Anderson, J.C. and Gerbing, D.W., 1988. Structural equation modeling in practice: A review and recommended two-step approach. Psychological Bulletin 103 (May), 411-23. Bacon, F., 1944 [1620]. Advancement of learning and Novum Organum, Wiley Book Co, New York. Bagozzi, R.P. and Heatherton, T.F., 1994. A general
46
approach to representing multifaceted personality constructs: Application to state self-esteem. Structure Equation Modeling 1 (1), 35-67. Beckmann, S.C. and Kilbourne, W.E. 1997, The interplay between the dominant social paradigm and value systems: Influences on Danish business students' environmental concern. Copenhagen, Denmark: Copenhagen Business School. Beckmann, S.C., Kilbourne, W.E., van Dam, Y., and Pardo, M., 1998. Anthropocentrism, value systems, and environmental attitudes: A multi-national comparison. In Andersson, P. (Ed.), Proceedings of the 27th European Marketing Academy Conference. Stockholm, Sweden, 1807-13. Bury, J.B., 1932. The idea of progress, The Macmillan Company, New York. Byrne, B.M., 2001. Structural equation modeling with AMOS, Lawrence Erlbaum Associates, Mahwah, NJ. Byrne, B.M. and Campbell, T.L., 1999. Cross cultural comparisons and the assumption of equivalent measurement and theoretical structure: A look beneath the surface. Journal of Cross-Cultural Psychology 30 (5), 555-74. Capra, F., 1982. The turning point, Fontana Flamingo Series, London. Chilton, K., W., 2000. Reengineering U.S. Environmental Protection, Business Horizons 43 (2), 7-16. Cotgrove, S., 1982. Catastrophe or cornucopia: The environment, politics, and the future, Wiley, New York. Diamantopoulos, A., Schlegelmilch, B.B., Sinkovics, R.R., and Bohlen, G.M., 2003. Can socio-demographics still play a role in profiling green consumers? A review of the evidence and an empirical investigation. Journal of Business Research 56 (6), 465-80. Donaldson, T. and Werhane, P.H. (Eds.), 1993. Ethical issues in business. Prentice-Hall, Englewood Cliffs, NJ. Dowie, M., 1995. Losing ground: American environmentalism at the close of the twentieth century, MIT Press, Cambridge, MA. Dunlap, R.E., 1991. Public opinion in the 1980's: Clear consensus, ambiguous commitment. Environment 33 (8), 10-15. Dunlap, R.E. and Van Liere, K.D., 1984. Commitment to the dominant social paradigm and concern for
environmental quality. Social Science Quarterly 65 (4), 1013-28. ----, 1978. The "new environmental paradigm": A proposed instrument and preliminary results. Journal of Environmental Education 9 (1), 10-19. Eckersley, R., 1992. Environmentalism and political theory: Toward an ecocentric approach, State University of New York Press, Albany. Ehrenfeld, D., 1978. The arrogance of humanism, Oxford University Press, New York. Hair, J.F., Anderson, R.E., Tatham, R.L., and Black, W.C., 1998. Multivariate data analysis, Prentice Hall, Upper Saddle River, NJ. Hetrick, W., 1989. The ideology of consumerism: A critique. In Bagozzi, R. and Peter, J.P. (Eds.), Proceedings of the American Marketing Association Winter Conference. Orlando, FL, 407-20. Hobbes, T., 1950 [1651]. Leviathan, E. P. Dutton and Sons, New York. Joreskog, K. and Sorbom, D., 1993. Lisrel 8: Structural equation modeling with the simplis command language, Lawrence Erlbaum Associates Publishers, Hillsdale, NJ. Kassiola, J., 1990. The death of industrial civilization, State University of New York Press, Albany, NY. Kilbourne, W., McDonagh, P., and Prothero, A., 1997. Sustainable consumption and the quality of life: A macromarketing challenge to the dominant social paradigm. Journal of Macromarketing 17 (1), 4-24. Kilbourne, W.E., 1995. Green advertising: Salvation or oxymoron? Journal of Advertising 24 (2), 7-19. ----, 2002. A multi-national examination of the role of the DSP in environmental attitudes: A structural equation modeling approach. In Wilson., J.W. (Ed.), Proceedings of the Atlantic Marketing Association. Savannah, GA, 102-06. Kilbourne, W.E. and Alsem, K.J., 1997. Environmental attitudes and marketing: Attitudes of business students in The Netherlands. In Arnott, D., Bridgewater, S., and Dibb, S. (Eds.), Proceedings of the 26th EMAC Conference. Warwick, England, 1805-13. Kilbourne, W.E. and Beckmann, S.C., 1998a. Review and critical assessment of research on marketing and the environment. Journal of Marketing Management 14 (6), 513-32.
47
----, 1998b. Socio-economic and cosmological dimensions of the dominant social paradigm: A multinational exploratory study of their role in environmental concern. In Andersson, P. (Ed.), Proceedings of the 27th Annual European Marketing Academy Conference. Stockholm, Sweden, 512-17. Kilbourne, W.E., Beckmann, S.C., and Thelen, E., 2002. The role of the dominant social paradigm in environmental attitudes: A multi-national examination. Journal of Business Research 55 (3), 193-204. Leiss, W., 1972. The domination of nature, George Braziller, New York. Levine, M. 2002, Threats to the environment, In www.roymorgan.com/news/polls/2002/. Lewis, C.S., 1947. The abolition of man, The Macmillan Company, London. Locke, J., 1963 [1690]. Two treatises on government, Cambridge University Press, Cambridge. MacPherson, C., B., 1962. The political theory of possessive individualism, The Clarendon Press, Oxford. Merchant, C., 1980. The death of nature: Women, ecology and the scientific revolution, Wildwood House, London. Milbrath, L., 1984. Environmentalists: Vanguards for a new society, University of New York Press, Albany, NY. Mirowski, P., 1988. Against mechanism, Rowman & Littlefield Publishers, Inc., Lanham, MD. ----, 1989. More heat than light, Cambridge University Press, New York. Mundt, J., 1993. Externalities: Uncalculated outcomes of exchange. Journal of Macromarketing 13 (2), 46-53. O'Neill, J., 1993. Ecology, policy and politics, Routledge, London. O'Riordan, T., 1976. Environmentalism, Pion Limited, London. Pirages, D.C. and Ehrlich, P.R., 1974. Ark ii: Social response to environmental imperatives, Freeman, San Francisco. Postman, N., 1993. Technopoly: The surrender of culture to technology, Vintage Books, New York. Purser, R.E., Park, C., and Montuori, A., 1995. Limits to anthropocentrism: Toward an ecocentric organization
paradigm? Academy of Management Review 20 (4), 1053-89. Sagoff, M., 1988. The economy of earth, Cambridge University Press, Cambridge. Shrivastava, P., 1995. Ecocentric management for a risk society. Academy of Management Review 20 (1), 118-37. Singer, P., 2004. Taking humanism beyond speciesism. Free Inquiry 24 (6), 19-22. Smith, A., 1937 [1776]. An inquiry into the nature and causes of the wealth of nations, Random House, New York. Spencer, H., 1969 [1851]. Social statics, Augustus Kelley, New York, NY. Steenkamp, J.-B.E.M. and Baumgartner, H., 1998. Assessing measurement invariance in cross-national research. Journal of Consumer Research 25 (June), 78-90. Thompson, S.C.G. and Barton, M.A., 1994. Ecocentric and anthropocentric attitudes toward the environment. Journal of Environmental Psychology 14 (2), 149-57. Winner, L., 1986. The whale and the reactor: A search for limits in an age of high technology, University of Chicago Press, Chicago. Biographies William E. Kilbourne, Ph. D. received his degree from University of Houston. He is a Professor of Marketing at Clemson University, and his research interests are in ethics, globalization, and environmental issues in marketing. He has published 40 articles in refereed journals and is currently the Globalization and Environmental section editor for the Journal of Macromarketing. Michael Jay Polonsky is the Melbourne Airport Chair in Marketing within the School of Hospitality Tourism and Marketing at Victoria University in Melbourne. His research interests are in the areas of environmental, social ethical issues in marketing; stakeholder theory; marketing education and cross cultural issues in marketing. He has published over 80 journal articles in referred journals. He is also on the editorial board of the amj. Correspondence Address
William E. Kilbourne, Department of Marketing, 245 Sirrine Hall , Clemson University, Box 341325, Clemson, SC 296341325. Facsimile: 864 656 0138 email: kilbour@clemson.edu
48
Evaluating a Model of Industrial Relationship Performance
Evaluating a Model of Industrial Relationship Performance: A Comparison of European and Chinese results using the IMP Data Base
James Wiley, Ian Wilkinson & Louise Young
Abstract Firms are not isolated economic actors. Their performance depends on more than their own efforts, capabilities and resources. It is also a function of the efforts and resources of others and on the relations a firm has both directly and indirectly with these counterparts. Based on these premises, we examine whether a model of international business relationship performance developed in a Western business context generalises to a Chinese context. Contrary to expectations, given the differences in business cultures, our results suggest that the same model applies in Swedish and Chinese samples of international supplier-customer relations. The exception is that connected business relations have a significant indirect effect on performance in the Chinese but not the Swedish sample. Differences also exist in the way some items used to measure two relationship dimensions operate in the two contexts. Management implications of our results and future research directions are discussed. Keywords: Business-to-business marketing, Cross cultural comparisons, Business relationships, Commitment, Understanding, Network connections
1. Introduction Firms are interdependent economic actors. They are embedded in networks of directly and indirectly linked firms and organisations, which have both professional and social dimensions. These include cooperative and competitive relations with suppliers, customers, distributors and complementors as well as with rivals. There has been much research and discussion in academic research and in practice about the potential role and importance of such relations in business, how they develop, and whether (or how) they may be managed (e.g. Hakansson and Snehota 1995, Christopher et al 2002, Sheth and Parvatiyar 2000, Wilkinson 2001). Business relations are the means by which key resources are accessed and co-developed, nonvalue added costs are reduced, and value is co-created and delivered to customers and other stakeholders. Relations and networks are potentially valuable resources for firms and a means of establishing and maintaining competitive advantage (e.g. Dyer and Singh 1998, Kanter 1994)
The development and management of such relations pose particular problems for firms engaged in international business activities because they involve dealing with counterparts from different social and business cultures with the attendant problems of communication and understanding. But in an ever more globalised world economy, such relations and networks become more frequent and necessary. The purpose of the research reported here is to compare the drivers of international supplier-customer relationship performance in business markets in a Western and Chinese context to see the extent to which the same relationship performance model applies in both. To do this we compare results from a sample of interviews conducted with international suppliers in Sweden and China. The data we use are from a largescale multi-national study of supplier-customer relations in business markets conducted by members of the Industrial Marketing and Purchasing (IMP) Group. The question of whether models of relationship performance generalise across countries has important
49
implications for research and practice. Most research on buyer-seller relations in business has been done in Western contexts and much of it focused on domestic rather than international relations (Wilkinson 2001), although this has changed somewhat in recent years. It is not clear whether the measures and models developed and tested in Western contexts work in the same way in contexts such as China. The paper is organised as follows. First we describe a model of relationship performance. We briefly review literature concerning the differences between Chinese and Western business practice and the role and importance of business relations. This work suggests that the model of relationship performance may not generalise to a Chinese context. Next, the methodologies used to collect the data in China and Sweden are described, followed by a description of the way the model was tested and compared between the two samples. Finally, we discuss the implications of our results for theory generalisation, research and practice.
2. A Model of Business Relationship Performance In order to compare the way Chinese and Western business relationships function we use a model originally developed and tested on a sample of European suppliers by Blankenburg-Holm, Eriksson, and Johanson (1996) (hereafter, BHEJ) using data drawn from the IMP study. The model is shown in Figure 1. Relationship performance is linked directly to two relationship dimensions Commitment and Understanding. Performance also is linked indirectly to Understanding and to Business Network Connections. Dwyer, et. al. (1987, p. 19) define Relationship Commitment as an implicit or explicit pledge of relational continuity between exchange partners. Relationship Commitment is hypothesised to affect performance because it leads to cooperation and compromise, which over time result in adjustment of business activities and resources to accommodate each partys way of doing business. These adjustments are hypothesised to result in reduced costs and/or increased
Figure 1: Blankenburg-Holm, et al. (1996) Model
Business Network Connection
H2
Relationship Commitment
H1 H3 H4
Relationship understanding
H5
Relationship Performance
50
value, which lead to higher performance. Greater performance in turn contributes to greater Commitment; which leads to greater willingness to engage in cooperative behaviour and compromise and thence to further performance enhancement in a feedback process. Hence, the model hypothesises a path between Relationship Performance and Commitment as well as between Commitment and Performance. Business Network Connections may be defined as a set of two or more connected business relationships, in which each exchange relation is between business firms that are conceptualized as collective actors (Emerson 1981; cited by Anderson, et. al. 1994, p. 2). Connected relations arise because the focal supplier and customer are involved in other relations than with each other and these can provide bases of comparison as well as have impacts on links between people, resources, activities and schemas brought to relationships (Anderson et al 1994, Wiley et al 2006). BHEJ argue as follows: When two firms cooperate in a focal business relationship they bring to the focal relationship their connected relationships. The evaluative criteria on which they base their engagement in the focal relationship are influenced, at least indirectly, by their business network contexts. The more the connected relationships support cooperation in the focal relationship, or can be adapted so that they do so, the easier it is for partners to coordinate their activities, increase the investment in the relationship, and thereby increase the joint productivity of the partner firms. (p. 1036). BHEJ focus on the positive impacts on commitment in the focal relation of relations connected and complementary to the focal relation, including the suppliers relations with their own suppliers and to other customers, and the customers relations to their own customers and to complementary suppliers. It should be noted that connected relations might also have deleterious effects on focal relationships such as relations with rivals, or when effort in one relation limits effort in another (Anderson, et al. 1994; Wiley, et. al. 2006). Here we assume the net effects of connected relations on commitment to the focal relation are positive. Relationship Understanding refers to working together to a common end based on mutual understandings of the parties involved. Relationship Understanding is hypothesised to be an antecedent to Commitment and plays a similar role to Business Network Connection in
that it reflects range and ability of firms to engage in cooperative behaviour. It is similar to Trust in the Morgan and Hunt (1994) commitment-trust theory of business relations. However, it is a cognitive based concept, while Trust is an affect based construct in most formulations, including the Morgan and Hunt framework. Understanding may be viewed as providing pre-conditions enabling firms to see opportunities for cooperation and for facilitating effective cooperative behaviours. BHEJ propose an indirect as well as direct relation between Understanding and Performance because understanding is hypothesised to have a direct link to Commitment. Such understandings are the basis for cooperating and working together i.e. for committing resources and effort to the relation, which is consistent also with Hunt and Morgans model. In sum, BHEJ advance five hypotheses regarding the models parameters: H1: Relationship Understanding will have a positive effect on Relationship Commitment. H2: Business Network Connection will have a positive effect on Relationship Commitment H3: Relationship Commitment will have a positive effect on Relationship Performance. H4: Relationship Performance will have a positive effect on Relationship Commitment. H5: Relationship Understanding will have a positive effect on Relationship Performance Based on their analysis of the IMP European sample of suppliers perceptions of their international business customer relations, the authors accepted hypotheses 1-3 and rejected hypotheses 4 and 5. BHEJ do not engage in cross group comparisons. A body of literature, however, suggests that differences may exist between European and Chinese settings for some of the relationships BHEJ hypothesise. Specifically, we evaluate the following two additional hypotheses. H6: Relationship Understanding will have a weaker effect on Relationship Commitment for Chinese compared to Swedish suppliers H7: The effect of connected relations will be stronger in the Chinese sample. We turn now to the rationale for the additional hypotheses.
51
3. Business Relations in China versus Western Contexts. There is reason to believe that the model may not generalize from European to Chinese settings as the role and importance of some dimensions may not be the same and additional dimensions of business relations may be important in China (Yeoh and Wilkinson 2005). The most commonly referred to concept about the way relations function in China is that of guanxi, which is seen as a pervasive and important element of Chinese society and business (e.g. Ambler and Styles 2003, Davies 1995, Kao 1993, Luo 1997, Redding 1990). Guanxi refers to the personal connections established with others on the basis of family ties, school, the army or region (Jacobs 1982, Kipnis 1997). Research suggests that such personal relations, play a more important role for the Chinese and this is likely to affect the role and importance of some behavioural dimensions of relations such as trust and understanding and perhaps networks and networking generally. Guanxi networks have been identified as a key reason for the success of overseas Chinese business empires building on the Chinese Diaspora (Chu 1996; Redding 1990, Wong et al 1999). Guanxi in turn creates indebtedness, social obligations or renqing, which can be regarded as a form of social capital (Yang 1994, Yeung and Tung 1996) and Chinese business relations have been characterised as a type of obligational contracting, whereby personal and professional debts are built up and acquitted over time in relations (Redding 1990). Another closely linked dimension which can play an important role in interpersonal interactions involving Chinese is that of face, a form of personal respect, that people seek to establish, preserve, protect and give through their interactions with others (Mavondo and Rodrigo 2001). Lastly, political connections can also play a more important role at times than economic or legal forces (Alston 1989). We should not exaggerate these differences, as their counterparts exist to some extent in Western contexts. Much research on business relations in the West reveals the importance of the personal and social dimensions of business relations in developing trust and cooperation (Bjorkman and Kock 1995, Kirkbride, Tan and Westwood 1991; Wu 1994). Davies (1995) has compared guanxi to the old-boy network, although such personal networks are more likely to be seen as problematic in the West, reflecting nepotism, pork-barrel politics or jobs for the boys (MacInnis 1993).
On the basis of the foregoing discussion we expect that affect-based trust may play a more prominent role in relations involving Chinese suppliers. If affect-based trust plays a more important role in business relations involving Chinese suppliers, we might expect that the cognitive-based Understanding dimension included in the BHEJ model may play a lesser role in driving performance when compared to European suppliers. Therefore, we hypothesise that the role and importance of Relationship Understanding will play a less significant role in the Chinese context. The impact on other aspects of the model are not clear except that, if Chinese business people are generally more sensitive to the role and impact of relations and networks, we expect there to be differences in the role and importance of Business Network Connections. In addition, Connected Relations may play a more important role in China because of a history of government planning of supplier customer relations in the Chinese economy and the greater involvement of government in trade. 4. Methodology 4.1 Data Bases The empirical analysis is based on a database established in the second Industrial Marketing and Purchasing (IMP) group international research project (IMP2) and a follow up study using the same methodology in China. The IMP group is a loose collection of researchers primarily based in Europe with interests in the study of business markets. Their research has resulted in numerous publications and research studies. For an overview of its development see Turnbull, et al. (1996) and Wilkinson (2001). The IMP2 study, initially, was a pan European collaborative study of inter-firm relations in business markets. Locally based researchers conducted in depth personal interviews with industrial firms in a number of European countries. These interviews focused on the characteristics and development of a self-selected, important relation with a counterpart in other countries. The data used from Europe are from a sample of supplier firms in Sweden. The samples of suppliers in other countries were insufficient to allow meaningful analysis and, because the questionnaires used were translated into the local languages, it was not considered appropriate to aggregate them, even though BHEJ aggregated them for their analysis. The supplier companies in the study belong to different industries, ranging from raw materials to equipment and include relations with customers in a mix of other European countries.
52
Interviews were conducted with marketing executives, who were asked to select one of the firm's most important customers in a specific country so that an even distribution of relationships across the customer countries could be achieved. The respondents were asked to select a customer relationship they were responsible for and of which they had personal experience. Thus, the respondents play a key role in the firm's enactment of the relationship. Questions about the focal customer relationship were answered in a personal interview that followed a standardized questionnaire. The IMP2 research was later extended to China and other Asian countries. For China, the questionnaire was translated into Chinese and back translated by an independent native bilingual Chinese speaker. For details of this see Dawson et al. (1997). The sample of Chinese suppliers with international business customers was developed in cooperation with the Chinese Bureau of Statistics. The sampling frame was based on the database of the Third Industrial Census and the database of the First National Basic Business Census developed by the Chinese Bureau of Statistics. For purpose of sampling, the frame was stratified into Northern (Beijing as the centre), Eastern (Shanghai as the centre), Southern (Guangdong as the centre) and Middle western (Sichuan and Chongqing as the representatives) regions and a target sample size of 100 was set based on interview costs. Approximately 200 suppliers in each region were selected for the initial sample, including Sino-Foreign joint ventures, large-sized industrial enterprises, and those involved in import and export. The specific respondent was determined by telephone pre-interview, to establish they satisfied relevant conditions and if they were willing to cooperate in the study. Officers from the Chinese Bureau of Statistics conducted personal interviews using the full IMP2 questionnaire. Interviews continued until the target sample size of 100 was achieved. The resulting sample comprised 50% state owned firms, 35% joint ventures, 4% Chinese owned and 3% foreign owned. Eight percent did not specify their ownership structure. The customers were spread over 32 different countries, the most numerous being USA 14%, Japan 10% and Hong Kong 11%. Forty-nine percent of customers were from the Asian region, 37% from Western countries (including 19% from Europe), 6% were from South America and 8% were not specified.
The Swedish and Chinese samples vary in their mix of types of relations because Sweden and China are at different stages of development of international trade links. Thus it is not possible to match industries or countries between the two samples and the average duration of the supplier-customer relations in the European sample is much longer (mean = 14.8 years since first delivery) compared to the Chinese sample (mean = 4.2 years). We take this difference in mean duration into account when interpreting our results. 4.2 Measure Development Following BHEJ, multi-items scales were used to measure each of the constructs in the model. Analysis is based on the 84 Swedish and 93 Chinese respondents who responded to at least two items of the multi-item scales. Two Swedish and seven Chinese respondents were excluded following this criterion. The items used are shown in Table 1. Five point scales were used for measuring Relationship Understanding and Commitment with 1 = strongly disagree and 5 = strongly Agree. For Business Network Connection items five point scales were used to rate the impact of each type of connected relation from 1 = not at all to 5 = very much. Relationship Performance is measured using a single item five point scale rating profit performance over the last 5 years from 1 = very bad, to 5 = very good. We follow Steenkamp and Baumgartner (1998) (S&B) to establish measure invariance across countries. The regression coefficient of one item (the marker) is set equal to one and its intercept to zero to identify a base model. Further restrictions impose configural, metric, and scalar invariance of scales. Measures may be completely or only partially invariant. All the scales discussed satisfy complete metric invariance. S&B require that the intercept of at least one item besides the marker item must be invariant in order to satisfy partial scalar invariance. At least partial scalar invariance is necessary for means of latent variables to be comparable across countries. All the scales discussed satisfy at least partial scalar equivalence. In addition, the scales satisfy at least partial error invariance and complete variance invariance. Table 1 shows the standardized regression coefficients and intercepts for each item in the two samples and relevant fit statistics. One of the items used by BHEJ in their Business Network Connection scale was excluded because including the item resulted in a significant
53
Table 1: Measurement Model Results
lack-of-fit for the measurement model. Excluding the item results in the scale having complete metric, scale, variance, and error invariance across countries. Using the second item (motive) as the marker item, Relationship Commitment satisfies complete metric and partial scalar invariance, 2 (10) = 13.6, p = .19 (IFI = .94, CFI = .93). Partial scalar invariance is achieved by restricting the intercepts for the fourth item to be equal (in addition to the intercepts of the second item being restricted to equal zero). The intercepts for the first and third items have significantly different values in the two countries. Interpretations of these differences are provided in the discussion section. Using the second item (misunderstanding) as the marker item, Relationship Understanding satisfies complete metric and partial scalar invariance (imposed by constraining the intercept for the first item to be equal
across countries), 2 (13) = 22.73, p = .05 (IFI = .94, CFI = .94). The intercepts for the third and fourth items have significantly different values in the two countries The third item (customers customer relationships) is the marker item for the Business Network Connections scale. The fit of an invariance model that includes metric, scalar, variance, and error invariance yields a 2 (8) = 7.61, p .47 (IFI = 1.00, CFI = 1.00). Constraining all parameters to be equal did not result in a significant increase in chi-square over the configural invariance model (2 (6) = 7.44, p = .38). We conclude that the BNC scale exhibits complete invariance across countries. Lastly, because a single item measure of Relationship Performance is used, no test of invariance across samples is possible in this case.
54
5. Results Following Anderson and Gerbing (1984), a two-step approach was adopted for testing Hypotheses 17. First, as discussed above, measurement model parameters for the invariance models were estimated. Secondly, these parameters were entered as fixed values in the measurement parts of the BHEJ structural model. As a result, all measurement model parameters are fixed and only structural model parameters are estimated for the BHEJ structural model. The lack-of-fit for the model is significant with a chisquare of 262.1, d.f. = 167, p = .000. Although the chisquare measure of exact fit is unsatisfactory, the RMSEA measure of close fit is satisfactory, RMSEA = .06, p = .19. Alternative measures of fit, however, are less than the widely accepted guideline of .90 (IFI = .67, CFI = .68). The IFI and CFI indices indicate that there is substantial room for improvement in the model. Furthermore, since the measurement parts of the structural model are known to fit, the observed lack-offit may be attributed to the structural part of the model. We provide additional comments in the discussion section. Consistent with BHEJ, the paths between Performance and Commitment and between Understanding and Performance are not significant; p = .39 and p = .90, respectively. Thus hypotheses H4 and H5 are not supported. Hypotheses H1 and H3 are supported. The paths between Commitment and Performance and between Understanding and Commitment are both positive and differ significantly from zero at beyond the 0.000-level. Furthermore, restricting the paths between Understanding and Commitment and between Commitment and Performance to be equal across countries does not result in a significant increase in lackof-fit, p = .23 and p = .98, respectively. The standardized coefficient for the path between Understanding and Commitment is .49, between Commitment and Performance is .90. We conclude that there is no support for hypothesis H6 as the relationships between the constructs do not differ across countries. Although the paths between Understanding and Commitment do not differ across countries, the countries do differ in their average levels of Understanding. The average level of Understanding is greater for Sweden than for China; 3.88 versus 3.51, respectively, values which differ significantly from one another at beyond the 0.01-level.
The results for the paths between Business Network Connections and Relationship Commitment are equivocal. Restricting the two country coefficients to be equal to zero does not result in a significant increase in lack-of-fit (p = .09), which implies no relationship between these dimensions. Closer examination indicates that the path is insignificant in Sweden (p = .22) but significant in China (p = .02). The apparent paradox for the paths stems from the pooling of an insignificant path with a significant one. The standardized coefficient in Sweden is .13, while for China it is .43. That is, support exists for Hypothesis H7 in that Business Network Connections are more strongly linked to Relationship Commitment in China, but do not result in greater Commitment in Sweden. The average values for BNC in the two countries do not differ (p = 1.00). The intercepts for Commitment do not differ across countries. However, intercepts for Performance do significantly differ. For equivalent levels of Commitment, Chinese Profitability ratings are on average less than their Swedish counterparts or, put another way, equally Committed respondents judge their 5-year profitability to be lower than do Swedish counterparts. The Swedish R2 values for Commitment and Performance are .28 and .23, respectively. For the Chinese sample they are .40 and .27. Two additional paths were evaluated. The regression coefficient for a path between BNC and Performance was insignificant in both countries. The coefficient between BNC and Relationship Understanding was insignificant in both countries. 6. Discussion Our results support the conclusion that measurement invariance obtains across countries. For a given level of Understanding, the average levels of agreement to the statement Misunderstandings between our two companies are rare are equal. However, Chinese respondents provide on average higher levels of agreement to the statement It is easy to agree about how to handle the various issues that arise in this relationship and lower levels of agreement with the statement The customers motives are generally clear to us. For Commitment, average levels of agreement to We consider the exchange part of a wider relationship are equal. However, Chinese suppliers provide higher levels of agreement with the statement We would not supply another customer at the expense of this current
55
Figure 2: Standardised Direct and Indirect Effects on Performance
customer and lower levels of agreement with the statement that This customer believes our firm is committed to the relationship. The differences may reflect substantive issues. Given the historic mix of relatively undifferentiated products, with purchase decisions dominated by price considerations, it is not surprising that Chinese suppliers perceive their profitability to be less than European counterparts. Likewise, given the relatively recent entry of Chinese firms into international trade, and the concomitant lower average duration of their relations with foreign customers, it is not surprising that Chinese suppliers believe that their customers perceive them to be less committed to relationships, when compared with longer established European supplier-customer relationships. For similar reasons Chinese suppliers may be more likely to agree that their customers motives are less clear, when compared with Swedish suppliers. Chinese suppliers also may be less willing to supply others, compared to the Swedes, because of their dependence on the customer. Regarding the structural portion of the BHEJ model, there are differences between countries in their mean Understanding and the intercept for the Understanding/Commitment Performance path. For equal levels of Understanding and Commitment Chinese suppliers on average rate performance less than do Swedish suppliers.
Consistent with BHEJ we do not find support for a feedback path from Performance to Commitment, or for a direct path between Understanding and Performance. The ability of path modeling to detect two-way effects is especially difficult with cross sectional data, where identification issues make estimation problematic. Aside from identification issues, the absence of the path could reflect lower levels of variance in the performance measure. The sample is restricted to self-designated important relations, which would tend to bias the sample to better performing relations. Such issues need to be explicitly accounted for in future research. One way of overcoming this is to ask respondents to identify relations that differ in their performance or importance and to select one at random to be the focal relation (e.g. Kim 2000, Young and Wilkinson 1997). The absence of a path between Relationship Understanding and Performance suggests that Commitment acts as a mediator variable between these two dimensions. The significance of the path between Business Network Connections and Commitment in the Chinese but not the Swedish sample lends support to hypothesis H7. The greater impact of connected relations in China is consistent with the view that connected relationships can facilitate cooperation and compromise in focal relationships, which over time lead to greater commitment to the focal relationship and in turn to greater performance. Given the less mature nature of
56
Chinese relationship, there plausibly are greater opportunities for cooperation, compromise and commitment and for this to increase Performance. An alternative explanation is that Chinese firms are more sensitive to relationship issues because of the pervasive nature and importance of these in Chinese society and business, as reflected in concepts such as guanxi. Figure 2 summarises the standardized direct and indirect effects on Performance for the Swedish and Chinese samples. Since Business Network Connection (BNC) does not have a significant effect on Commitment in the Swedish sample, only Commitment (directly) and Understanding (indirectly) may affect Performance. In both samples, Commitment and Understanding have comparable standardized effects on Performance (.48 and .25 respectively in Sweden and .52 and .24 respectively in China). BNC has a standardized indirect effect on Performance of .22 in China. Overall, the total effects of BHEJ model constructs on Performance are greater in China than in Sweden. The total effects of Understanding and Commitment combined are virtually identical. The entire difference in total effect on Performance is accounted for by the indirect effect of BNC on Performance in China. The fit statistics indicate that there is much variance unaccounted for in the dependent variable that is still to be explained. This is to be expected, as it is seldom the case that researchers believe their model accounts for most of the variation in a dependent variable. Here, the model would need to be extended to include macro economic factors such as the state of relevant economies as well as market and industry specific variables such as competition. In addition, the hypothesized drivers of cooperation and compromise are not measured and included in the model. 7. Conclusions and Future Research The management of market and business relations is becoming a key area of attention in both research and practice (Dyer and Singh 1998, Pham et al 2004, Ritter et al 2004). The management of supplier-customer relations across international borders represents a particular challenge because different cultures and business practices encounter each other. This creates both problems and opportunities both of which firms cannot afford to miss as the world economy becomes more globalised and China becomes a major participant. Our research has shown that a model of international
business relationships generalizes across China and a European country. This is an encouraging result for both scholars studying international business relations and for international managers. Because the model is consistent with aspects of Morgan and Hunts Commitment-Trust model, it also provides some support for its applicability in a Chinese international business context. One exception to the generalizability of paths across countries is the non-significance of a path between Connected Business Relations and Relationship Commitment for Swedish focal relations. The types of connected relations included in the model are suppliers suppliers, customers customers and complementary suppliers, which are involved in creating value and supporting the focal relation. These affect commitment in the focal relation by providing opportunity and motivation for cooperation and compromise. As noted, this difference is open to both cultural and business interpretations. It may reflect the greater role and importance of and hence sensitivity to connections in Chinese business and society, as many researchers have pointed to. Alternatively, the differences between path coefficients could reflect the less mature stage of development of Chinese suppliers international customer relations and hence their greater vulnerability to the effects of connected others. To determine which explanation is correct requires additional research. More generally, the result indicates that relationship management in international business is not just about developing and managing coordination and cooperation within a relation but also involves recognising, valuing, linking and coordinating activities across connected relationships. Relationship management is part of a broader capability that influences firms performance in international business, which some have referred to as network competence (e.g. Ritter et al 2002, 2004). One potential reason for not finding more differences between the two contexts is the focus on international supplier customer relations rather than domestic business relations. International business relations are less likely to reveal any cultural differences in the way business relationships function because China has to engage with major and more experienced international customers who are from the West or are more attuned to Western habits. In addition, the results could reflect the growing Westernisation of Chinese business because of the
57
impact of Western firms establishing operations in China. We note that as Chinese international trade expands it involves, increasingly, dealing with business customers from Western contexts. It is unclear how this affects the way relationship function do the Chinese tend to adopt Western practices or do Western customers of Chinese suppliers become more similar to the Chinese in terms of the way business relations operate? Thus, even in China, Chinese businesses are increasingly dealing with non-Chinese firms, which may undermine traditional ways of doing business.The business discipline of a more competitive market serves to reduce the role and importance of guanxi and personal obligations in business relations as the search for differential advantage in terms of cost, value and quality become more prominent. Finally, it should be noted that it is likely that firms having greater international experience may be more likely to appear on sampling frames and, hence, tend to be over represented in national samples. Some measurement issues arise that suggest areas for additional future research. First is the translation of commitment in terms of obligation, which accounts for differences in the way some items function as part of multi-item scales. Measures need to be developed and tested of relationship obligation as well as commitment, in order to determine whether they are part of a more general underlying construct, separate dimensions and whether obligation is a dimension specific to Chinese and perhaps other non-western business cultures. Second is the issue of the equivalence of some of the items used in the measures and their response categories, when presented in Chinese versus other languages. We have detected systematic differences in some items. Research on this type of measurement issue becomes more important as China develops as an increasingly important market and trading partner, and because firms wish to compare study results between Western and Chinese contexts. One possibility is to compare the ratings of multilingual and multicultural people answering the same questions in different languages. Finally, future research should attempt to develop more robust and comprehensive measure of relationship performance and value, including both direct and indirect functions such as information, innovation and reputation effects (Walter, et. al. 2001, Walter et al 2003, Wiley et al 2006).
References Alston, J.P., 1989. Wa, Guanxi, and Inhwa: Managerial Principles in Japan, China, and Korea. Business Horizons 32 (March-April), 26-31. Ambler, T., Styles, C., 2003. The Coexistence of Transaction and Relational Marketing. Industrial Marketing Management 32 (8), 633-642. Anderson, J.C., Gerbing, D.W., 1984. Structural Equation Modeling in Practice: A Review and Recommended Two-Step Approach. Psychological Bulletin 103 (3), 411-423. Anderson, C., Hakansson, H., Johanson, J., 1994. Dyadic Business Relationships within a Business Network Context. Journal of Marketing 58 (October), 1-15. Bjorkman, I., Kock S., 1995. Social Relationships and Business Networks: The Case of Western Companies in China. International Business Review 4 (4), 519-555. Blankenburg-Holm, D., Eriksson, K., Johanson, J., 1996. Business Networks and Cooperation in International Business Relationships. Journal of International Business Studies 27 (5), 1033-1053. Christopher, M., Payne, A., Ballantyne, D., 2002. Relationship Marketing: Creating Stakeholder Value, Butterworth Heinemann, Oxford. Chu, P., 1996. Social Network Models of Overseas Chinese Entrepreneurship: The Experience in Hong Kong and Canada. Canadian Journal of Administrative Sciences 13 (4), 358-365. Davies, H., 1995. Interpreting Guanxi: the Role of Personal Connections in a High Context Transitional Economy. In: Davies H., (Ed.), China Business: Context and Issues, Longman Asia, pp. 155-169. Dawson, B., Wilkinson, I., Young, L., 1997. Conceptual and Methodological Issues in Translation of a Business Questionnaire into Chinese. In: Smith, S.M., (Ed.), Proceedings of Sixth Symposium on Cross Cultural Consumer and Business Studies, Brigham Young University, Utah, pp. 183-187. Dwyer, F.R., Schurr, P.H., Oh, S., 1987, Developing Buyer-Seller Relationships. Journal of Marketing 51, (April), 11-27. Dyer J., Singh, H., 1998. The Relational View: Cooperative Strategy and Sources of Interorganizational Competitive Advantage. Academy of Management Review 23 (4), 660-679.
58
Emerson, R.M., 1972. Exchange Theory, Part I: Exchange Relation and Network Structures. In: Berger, M., Zelditch Jr., M., Anderson, B., ed., Sociological Theories in Progress, Vol 2, Houghton Mifflin, Boston, pp. 2-65. Hakansson, H., Snehota, I., 1995. Developing Relationships in Business Networks, Routledge, London. Jacobs, J. B., 1982. The Concept of Guanxi and Local Politics in a Rural Chinese Cultural Setting. In: Greenblatt, S., Wilson, R., Wilson A., (Eds.), Social Interaction in Chinese Society, Praeger, New York pp. 209-236 Kanter, R., 1994. Colaborative Advantage: The Art of Alliances. Harvard Business Review 72 (July-August) 96-108. Kao, J., 1993. The Worldwide Web of Chinese Business. Harvard Business Review, 71 (March-April), 24-36. Kirkbride, P.S., Tang, S.F.Y., Westwood, R.I., 1991. Chinese Conflict Preferences and Negotiating Behaviour: Cultural and Psychological Influences. Organizational Studies, 2 (3), 365-386. Kim, K., 2000. On Interfirm Power, Channel Climate, and Solidarity in Industrial Distributor-Supplier Dyads. Journal of the Academy of Marketing Science, 28 (3), 388-405. Kipnis, A., 1997. Producing Guanxi: Sentiment, Self, and Subculture in a North China Village, Duke University Press, Durham, North Carolina. Luo, Y., 1997. Guanxi and Performance of Foreigninvested Enterprises in China: An Empirical Inquiry. Management International Review. 37 (1), 51-70. Mavondo, F.T., Rodrigo E.. 2001. The effect of relationship dimensions on interpersonal and interorganizational commitment in organizations conducting business between Australia and China. Journal of Business Research 52 (2), 111-121. MacInnis, P., 1993. Guanxi or Contract: A Way to Understand and Predict Conflict between Chinese and Western Senior Managers in China-based Joint Ventures. In: McCarty, D., Hille, S., (Eds.), Research on Multinational Business Management and Internationalisation of Chinese Enterprises, Nanjing University Nanjing, China, pp. 345-351.
Morgan, R.M., Hunt, S. D., 1994. The CommitmentTrust Theory of Relationship Marketing. Journal of Marketing 58 (July), 20-38. Redding, S. G., 1990. The Spirit of Chinese Capitalism, W. de Gruyter, Berlin, New York. Ritter, T., Wilkinson, I.F., Johnston, W.J., 2002. Measuring Network Competence: Some International Evidence. Journal of Business and Industrial Marketing 17 (2/3), 119-138. Ritter, T., Wilkinson, I.F., Johnston, W.J., 2004. Firms Ability to Manage in Business Networks: A Review of Concepts. Industrial Marketing Management 33 (April), 175-183. Sheth J., Paravatiyar A.,2000. The Handbook of Relationship Marketing, Sage, Newbury Park California. Steenkamp, J.E.M., Baumgartner H., 1998. Assessing Measurement Invariance in Cross-National Consumer Research. Journal of Consumer Research 25 (1), 78-90. Turnbull, P., Ford, .I.D., Cunningham, M., 1996. Interaction, Relationships and Networks in Business Markets: an evolving perspective. Journal of Business and Industrial Marketing 11 (3/4), 44-62. Walter, A., Ritter, T., Gemunden, H.G., 2001. Value Creation in Buyer-Seller Relations. Industrial Marketing Management 30 (4) 365-377. Walter, A., Muller, T.A., Helfert, G., Ritter, T., 2003. Functions of Industrial Supplier Relationships and Their Impact on Relationship Quality. Industrial Marketing Management 32 (2) 365-377. Wiley, J., Wilkinson, I. F., Young, L.C., 2006. The Nature, Role and Impact of Connected Relations: A Comparison of European and Chinese Perspectives Journal of Business and Industrial Marketing (in press). Wilkinson, I.F., 2001. History of Channels and Network Thinking in Marketing in the 20th Century. Australasian Marketing Journal 9 (2) 23-53. Wong, Y.H., Chan R.Y., 1999. Relationship Marketing in China: Guanxi, Favouritism and Adaptation. Journal of Business Ethics 22 (2), 107-118. Wu, W.P., 1994. Guanxi and its Managerial Implications for Western Firms in China: A Case Study. Paper presented at the International Conference on Management Issues for China in the 1990s, University of Cambridge. Cambridge, U.K.
59
Yang, M.M., 1994. Gifts, Favors and Banquets: The Art of Social Relationships in China. Cornell University Press, New York. Yeoh, K., Wilkinson, I.F., 2005. Value Creation in Chinese and European Business Relationships. In: Brown, D., MacBean, A., (Eds.) Challenges for Chinas Development: An enterprise perspective Routledge: London (in press). Yeung, I.Y., Tung, R. L., 1996. Achieving Business Success in Confucian Societies: the Importance of Guanxi, Organizational Dynamics 25 (2) 54-65. Young, L.C., Wilkinson, I.F., 1997. The Space Between: Towards a Typology of Interfirm Relations. Journal of Business to Business Marketing 4 (2) 53-97. Biographies James Wiley is Professor of Marketing and Chair of Marketing Science at the Victoria University, Wellington. He received his PhD from the University of Washington and has taught at universities in the USA, Canada, Australia, and Hong Kong. During 1995, he was a Fulbright Fellow at Cominius University, Slovakia. His current research focuses on Choice Theory; Multinational survey research, especially those based on repeated samples and panels; and Complexity Theory applications to Marketing.
Ian Wilkinson is Professor of Marketing at the University of New South Wales. He was educated in the UK and Australia and has held academic posts at various American, European as well as Australian universities. His current research focuses on the evolution and management of interfirm relations and networks in domestic and international business. Louise Young is Professor of Marketing at the University of Technology, Sydney. She was educated in the USA and Australia and has various permanent and visiting academic posts at Australian, European and American universities. Her current research focuses on the nature and impact of psychological drivers upon relationships and networks in business and their evolution. Correspondence Addresses James Wiley, Professor and Chair of Marketing Science, Faculty of Commerce, Victoria University of Wellington, School of Marketing and International Business, Wellington, New Zealand, email: james.wiley@ vuw.ac.nz; Ian Wilkinson, Professor School of Marketing, University of New South Wales, Kensington, NSW, 2050, email: i.wilkinson@unsw.edu.au; Louise Young, Professor of Marketing, School of Marketing, University of Technology Sydney, P.O. Box 123, Broadway, NSW, 2007, email: l.young@uts.edu.au
60
Research Notes
Designing and Implementing Internet Questionnaires Using Microsoft Excel

Iain R. Black, Alejandra Efron, Christina Ioannou & John M. Rose
Abstract Over the past ten years, the use of the Internet and e-mail as communication tools has become ubiquitous. In the survey arena, the rising costs of gathering data have been partly compensated by the use of the internet and e-based technologies which offer a range of new, relatively cost effective survey design and delivery options. This paper reports on two studies where Microsoft Excel was used to design and gather data without the additional investment associated with specialist programs. Study one examines the development of a multi-attribute survey conducted to create a new scale using a local (Australian) population of students. The second describes the use of Excel in a stated choice experiment that was sent to an international sample of museum managers. These studies show that it requires minimal programming skill on behalf of the researcher whilst offering the many of the cost, administrative and questionnaire design benefits seen with specialist software and Internet delivery. We conclude that Microsoft Excel can be considered when designing online surveys as it provides a wide range of features and benefits that allow for flexible, rich instrument design and fast, potentially accurate, data collection, checking and entry. Keywords: Microsoft Excel, Online Data collection, Survey design,
1. Introduction It is an important goal for market researchers to improve methods for designing and implementing questionnaires in order to reduce costs, reduce research timescales and improve data quality. These methods should be capable of being used in a wide range of research types, constructing a comprehensive range of questionnaire design features and using a full range of stimulus materials. Implementation should be charecterised by fast data collection, easy access to respondents and allow fast, accurate coding, checking and data entry. In this paper we report on two studies where Microsoft Excel was used to design Internet questionnaires. In both cases email was used to return the surveys after being delivered either by email or by posting onto a bulletin board. The ubiquitous and familiar nature of Microsoft Excel promises cost, speed and usability benefits for market researchers if it can be used successfully to design and implement survey research. It is the aim here to demonstrate Excels ability to do this.
Studies into the use of internet technologies in research have found both advantages and disadvantages. A number of authors have found they provide speed and cost advantages over paper and pencil methods (Ilieva et al., 2002; Best and Kreuger 2002; Dommeyer and Moriarty, 2000; Schaefer and Dillman, 1998). In addition, there thought to be are presentational and multimedia benefits (Couper, 2001 and Forrest, 1999). However, reservations over data quality and comparability (Couper, 2000; McDonald and Adam, 2003) have been raised. Cost savings fall into a number of categories; stationary savings come from the lack of paper, envelopes, ink and postage (Sheehan and McMillan, 1999), administration savings include reduction in photocopying, letter fulfilment, typing and scanning and labour savings are seen with reduced data entry, cleaning, and coding (McDonald and Adam 2003). However, Thomson et al. (2003) cautions us by detailing the high start up cost linked to soft and hardware purchases and staff training.
61
Research Notes
Internet technologies also increase the speed with which surveys are delivered, returned and the data entered, checked and cleaned (Dommeyer & Moriarty, 2000; Sheehan, 2001). This combination of benefits lead these authors to conclude that internet based surveys are more cost effective than paper alternatives. These technologies allow the researcher to use a full range of colour, sound, art and video as stimulus material or to improve the presentation of the survey and allows data to be collected on a wide range of research questions. Whilst Dillman (2000) found that visual presentation improves response rates and is important in survey design, Deutsken et al., (2004) found that visual versions of surveys were associated with lower response rates. This apparent discrepancy can be explained in terms of the longer download times involved with large files and this must be considered. In addition presentation can be affected by local computer setting (browser configuration, user preferences, monitor settings) and this leads to the concern that the researchers can lose control of this important area (Thomson et al., 2003). A number of respondent, sample and data issues exist with online surveys and Internet data collection. It is clear that not everyone has convenient access to the Internet, making these methods appropriate only for a restricted (though growing) range of demographics (Ilieva, 2002). In addition, connection problems and network outages are a well-known feature and may frustrate attempts to access and successfully complete online surveys (Thomson et al., 2003). McDonald and Adam (2003) found a lower response rate to online surveys compared to traditional methods and that respondents were different from non-respondents. Concerns over response rates are exacerbated by the growing problem of unsolicited email (Spam) that has led to greater respondent resistance, legal restrictions and software barriers. For example, it is now illegal in Australia to recruit respondents without a precommercial existing relationship. Anti-virus software and firewalls make it harder to recruit respondent and a resistance to pop ups has seen click through rates fall dramatically (Dreze and Hussherr, 2003). These combine to make it more difficult to reach samples using email. Questions over the identity of respondents have also been raised as well as the need to ensure that surveys are completed only once per respondent. This is a particular problem when incentives are used (Tierney 2000). Finally, a lack of trust in technology and has been
found to exacerbate typical respondent concerns of anonymity, confidentiality, data security and delivery (Ilieva et al., 2002; Kraut & Saari, 1999). To use such technologies, software is needed to design the survey in an electronic format. For example, SNAP, Websurveyor or SSI Web can be used to design surveys for implementation via WWW, email and Bulletin boards (e.g. Deal, 2002; Rapoza, 2003). However, the cost benefits of using the Internet are eroded by having to purchase these packages and then train people to use them. Surveys can be designed as HTML files (Dillman, 2000; Jackson and DeCormier, 1999) though this is more complex and requires specific programming skill. The aim of this paper is to demonstrate that Microsoft Excel can be used to design internet based questionnaires and to collect the results from them. Furthermore, due to the ubiquitous nature of Excel and the transferable skills gained from using it, it is able to do so and avoid many of the cost disadvantages of specialist software. We will illustrate a wide range of design features which allow a full range of question and scale types to be created in an aesthetically pleasing and creative manner. The benefits and issues of using this program in combination with email delivery will then be assessed and recommendations as to Microsoft Excels suitability as a survey design and delivery platform will discussed. First, however we outline the questionnaire and research design features that Excel must be capable of creating. We follow this with an introduction into the features and tools that allow questionnaires to be designed. It is not the aim of this research to provide the reader with a stepby-step guide on how to use these functions, readers requiring this should look to the comprehensive handbooks such as Walkenbach (2003). 2. Survey Design Features In order to assess the use of Microsoft Excel survey design platform we must first provide criteria against which to compare it. Schaefer and Dillman (1998) discuss the importance of well-crafted introductory letters and invitations as these can increase the response rate. In addition to the numerous requirements regarding wording, questionnaires should be respondent friendly (Dillman 1978; Dillman, Sinclair and Clark, 1993), this describes a form that is easy for respondents to complete, avoids confusion about what or how to answer it and results in respondents feeling neutral or positive (Dillman, Sinclair and Clark, 1993, p2). The use of
62
Research Notes
instructions, clear layout of questions and answer formats, appropriate font size and type and appropriate number of questions per page addresses these requirements. A design platform must allow a full range of scale properties contained in nominal, ordinal, interval and ratio scales to be used. It should also be capable of providing open, closed, graphic and written response formats and allowing randomisation of response options so to combat order effects (Green and Tull, 1978). It is also important to minimise respondent burden by effectively including skip patterns and question sequencing and should be capable of including a wide range of stimulus material including multimedia effects (Dillman, 2000). There should be no restrictions on the number of questions that can be asked and the length of the questionnaire. It is also important for internet data collection methods to be able to code answers automatically into a data file (Ioannou and Black, 2004). 3. Microsoft Excel Functions and Tools This review of functions is designed to familiarise the reader with the parts of the programme we used when designing our data collection instruments. The core features used are found in either the Forms Toolbar or Control Tool Box (figure 1 ). These features allow for the design of a wide range of response formats such as check and option boxes or answer lists, which, when combined with question text and other features, produce the survey. When designing the questionnaire the
worksheet is treated as the blank canvas upon which the questions and answers are written. The text of the survey (questions, responses and instructions) is written into individual cells on a worksheet in the same way that text is entered into any Excel spreadsheet. The response options designed using the Forms Toolbar or Control Tool Box are then placed near the cells containing the text (see Appendix 1). Either method can be used and as with many Microsoft functions, the choice is down to personal preference. Commonly used functions such as cut and paste, paste link, cell protection, font options and formulas are combined within the worksheet to link the respondents choices to the datasheet and to improve the presentation, sequencing and respondent friendliness of the questionnaire. 4. Training Prior to combining the above features into a questionnaire, a short period of training was required. The authors possess a range of research experience ranging from novice to early career lecture staff and we would rate ourselves as having an intermediate level of knowledge of Excel functions. Specific training was conducted by one member of the team who demonstrated (for approximately 20 minutes) the Forms toolbar features and how to create macros and cell-link worksheets. Following this, additional time was invested in self-education in order to become familiar with the features and to experiment with the dynamics of the application. Overall learning to use Excel to design
Figure 1:
63
Research Notes
Figure 2: Example of Multi-item question formate 1
Figure 3: Example of ordinal response format
Figure 4: Command Button for Navigation Click here to proceed
questionnaires was felt to be a straightforward extension of our survey design and Excel skills. We now move on to detail two different data collection instruments designed using Excel. 5. Satisfaction Typology Survey 5.1 Background The first data collection instrument designed was a survey for research into consumer satisfaction typologies (Oliver, 1997) using a student sample. The use of Excel was driven by a lack of funds to purchase commercially available software and the need for fast data collection and preparation. 5.2 Design of the questionnaire The questionnaire design process was guided by the nine-step framework of Churchill and Iacobucci (2002). The questionnaire was transferred from a paper version written using Microsoft Word. There is nothing to stop the researcher designing the questionnaire directly using Excel but in this case, we only discovered the possibility of using Excel after the pre-test.
The questionnaire and data sheet were designed on individual worksheets within a workbook. The number of worksheets depends on the length of the questionnaire and how many questions you want to fit on each sheet. We found it useful to use a multiple worksheet format which showed one section of the questionnaire at a time rather than try and fit the whole questionnaire on a sheet as it allowed clear instructions to be written for each section and reduced the probability of respondents being intimidated by the length. The text of the survey (questions, responses and instructions) is written into individual cells on a worksheet. The response options designed using the Forms Toolbar. It is important to note that the worksheet tabs which appear at the foot of open Excel file (where the name of the sheet is placed) were hidden in order control navigation through the questionnaire. The questionnaire comprised predominantly of quantitative questions using five point Likert scales and seven point semantic-differential scales as the main response format. See Figures 2 and 3 for examples of the question and response formats used. The key feature
64
Research Notes
from the Forms toolbar used to design each of the response formats was the Option buttons inserted within a Group box. Each Option Box was inserted into the Group Box in a specific order and each Group Box was cell-linked to a cell in the answer sheet. For example, to construct the semantic differential scale shown in Figure 2, seven Option Boxes where inserted within a Group Box and text written into adjacent cells. The adjectives used in the scales were then written into adjacent cells in the worksheet. It is possible by adding cell protection to ensure that only one answer can be selected. By giving each Option Box a title, it is simple to design ordinal/interval scales as shown in Figure 3. These examples demonstrate the programs ability to design a range of questions easily and automate data entry and capture. 5.3 The answer sheet Each answer is paste linked to a password-protected, hidden worksheet within the Excel file. All the answers for each respondent were contained on one row Configuring the answer sheet is time-consuming during the design phase though it is rewarding overall because of time saved during data entry. 5.4 Navigation Navigation through the survey was controlled using three methods. First, simple macros were used. Instructions were created using the Command button so that when cells were clicked upon, the respondent moved to the next worksheet and hence through the survey. The relevant cells then had appropriate instructions written in to them so to guide the respondents actions (see figure 4). Navigation was also controlled by using the IF or VLOOKUP commands so to ensure that respondents cannot move forward unless all questions are answered. Finally, cells were locked so that only certain cells could have information inserted into them. 5.5 Presentation A number of simple features were used to ensure that the survey was easy to read and presentable. The FormatCell-Font function allows manipulation of size, font, effect and colour. This was used throughout the design process to create an aesthetically pleasing style and font scheme for the questionnaire. Respondent instructions and response formats were colour coded using FormatCells-Patterns. The Forms-Toggle-Grid" feature was used to create a mono-coloured background. In order to create a spacious layout and avoid crowding, the
questions/scales were spread over multiple spreadsheets within a single workbook. The canvas style quality of the application (as opposed to a template design) provides the researcher with the ability to insert their own creativity into the design of the questionnaire and to apply the variety of features at their own discretion. 5.6 Order Effects The ordering of questions is an important consideration in questionnaire design as bias may occur if responses are in a consistent order. Therefore, multiple versions of the original questionnaire were created by rearranging the order of questions/scale items in each of the questionnaires. In this case, Excel provided a simple and efficient way to do this. Once the master questionnaire was designed, scale items were randomised using the cut and paste feature to create five versions of the questionnaire (yielding a total of six questionnaires). The key benefit of using the cell link function was that questions in different orders can easily be coded in the same order within the answer sheet (i.e. the answers to the scale item positioned at number 5 in questionnaire version one and as item 14 in questionnaire version two would always appear in the same cell within the answer sheet). 6. Procedure 6.1 Administration of the Questionnaire Two versions of the surveys were designed (one for mobile phone and the other for cars) each comprising 136 questions split into 10 sections. These were distributed to 450 students at an Australian university, and a response rate of 74% was achieved. After successfully pre-testing the questionnaire on a sample of 70 students, the Excel version was designed and the file was password protected and uploaded onto the University bulletin board. Instructions were also provided on how to download, complete and return the questionnaire. It is common for students to download content from this site (lecture slides, tutorial information etc) and no difficulties were reported. As there were two versions of the questionnaire (based on Mobile phones or Automobiles ownership) a screening process was used to ensure that students were granted access only to a version relevant to their circumstances. This was achieved by instructing participants to download the version which was saved as their tutorial session time. The survey was then completed as part of a tutorial with each tutorial class randomly allocated one of the six
65
Research Notes
versions of the questionnaire. The respondents then saved the Excel file to their file space on the university system and emailed it as an attachment back to the researcher. Instructions were then given to delete the file from this location. In order to preserve respondent confidentiality, university email accounts were used as these do not disclose the full name of the user. 6.2 Data Editing and Entry As emailed responses were received, the Excel attachments were sequentially saved using unique filename. Once all the respondents had replied, the saved attachments were opened in turn and the row of data from the hidden answer sheet was copied-pasted into a master answer sheet. The Paste SpecialValues function available in Edit was used here to avoid data entry errors and to increase speed. On completion, the master file was then exported from Excel into SPSS for analysis. Once this was completed, data cleaning was performed and additional coding steps were taken to transform code reversed items and to replace missing values. An important point to highlight is the need to be systematic in your approach to opening each file and moving it to another location once the data has been removed. Simple eyeballing of the data as it added to the master file was found to be effective in guarding against duplication. This was also assessed by ensuring that at end of the process, the number of rows corresponds exactly to the number of email attachments received. Overall, this procedure is relatively time consuming and open to the possibility of errors when compared to the automated construction of a data file offered by specialist software. However, it was thought to save considerable time when compared with coding from paper surveys, and cut and paste errors were readily identified and fixed. Whilst in this case the data was transferred to SPSS, it should be mentioned that Excel has its own statistical package add-ins (e.g., PACE XL, Analysis ToolPak, PHstat2) that could have been used. 6.3 Summary The use of standard Excel functions in conjunction with the Forms toolbar allowed for fast, flexible questionnaire design. An intermediate level of knowledge with Excel is required by the researcher in order to know how to use these functions and to learn quickly additional uses. It does however depend on a strong knowledge of questionnaire design. In this case the decision to use Excel was taken late in the design process after a draft paper questionnaire had been drawn up. It was easy to learn how to convert this into a
spreadsheet and no changes had to be made to the question wording or response formats. It allowed for a more aesthetically pleasing survey that was faster and cheaper to distribute and considerably faster to code, edit and enter. In this case, Excel (distributed via Bulletin board and email) proved to be a useful design and implementation platform and one we could recommend. 7. Choice Modelling Experiment 7.1 Background The second case is based on the design and administration of an unlabelled choice based conjoint experiment distributed via email to an international sample of museum directors, curators and exhibition managers. Once again, the use of Microsoft Excel to design and implement the experiment was borne out of the lack of funds combined with the necessity of receiving answers from a widely dispersed international sample. 7.2 Design of the questionnaire The questionnaire for this study was divided into two different parts: Part One gathered information on the characteristics of the respondents and their respective institution by means of a traditional survey. Part Two displayed 16 of the 32 choice sets of the choice based conjoint experiment (for more information on this methodology, refer to Louviere et al., 2000 or Hensher et al., 2005). In a choice based conjoint experiment, the researcher presents the respondent with a number of hypothetical scenarios with a description of the levels of the attributes, and asks the respondents to choose among these scenarios (Louviere et al. 2000). Figure 4 shows an example of a choice profile used in this experiment, together with the contextual explanation given to the respondent. It also demonstrates again the presentation and flexibility possible with Excel. Each choice is paste linked to a password-protected, hidden worksheet using the same procedure detailed earlier in this paper. 7.3 Types of Questions and Response formats A wide range of scaling properties, scale type, question and response formats were used in this research. These include nominal, ordinally-interval and ratio scales and open, closed, multiple choice and multiple item Likert scales response formats. These features were designed using Control Toolbox toolbar where the Option buttons inserted within a Group box was used to provide answer categories. For example, appendix 1 shows the use of a closed response format (with nominal scaling properties) for answers to the question Have you spent more than six months in another country?
66
Research Notes
Figure 5: Demonstrating the use of additional information in Excel
The Combo Box feature allowed multiple choice answers to be displayed by using a drop-down type of feature. The analyst first writes the list of answers to be included in the box, in a column outside of the exhibited questionnaire. This can then be linked to a number of questions and is convenient for those answers that are repeated in many questions for example with Likert scale. See Figure 5 and Appendix 1 for an example of how the same answers are reused in different parts of the questionnaire. Open ended answer formats are created using a Text Box with Design Mode activated. This is the preferred method of creating these questions because it allows the researcher to restrict the respondent to writing answers in desired cells only. The international nature of the sample and the heterogenous level of fluency in English made it necessary to design a number of comments sections in the questionnaire containing definitions and additional information. Respondents who needed more explanation were able to position the cursor over the red marks the Insert Comment commands uses to indicate that additional information is available (See Figure 5 and 6).
7.4 Navigation Controlling navigation through survey was easier than in study one because all questions fitted onto two worksheets. Macros were used to control navigation between these in the same way described in study one for most respondents. However, as will be described in more detail later, where security levels on respondents machines did not allow macros to function, then worksheet tabs were unhidden and instructions on how to move forward were given. 7.5 Presentation Presentation of the questions was a particular issue as the respondents were art specialists and their positions did not permit them much time to complete the experiment. Excel allows for the analysts creativity to be displayed, enabling them to insert pictures, diagrams, charts, photographs, sound, video files and other multimedia features. Insertions are easy to use and benefit presentation. Figure 5 shows the use of a colourful world map including drop down boxes. The researcher should balance these possibilities with files final size. It should not be assumed that the
67
Research Notes
Figure 6: Example choice profile
respondents have broadband and consequently consideration should be given to likely download and file opening times. Furthermore, hotmail or yahoo types of account severely limit email file sizes. 7.6 Order Effects Literature in choice based conjoint (Hensher 2004; Louviere et al. 2000; Train 2003) recommends randomising the order in which the choice profiles appear to different respondents. Six versions of the experiment were created, (this number could easily have been higher
if required) using simple formulas to randomise the order in which the choice sets were presented. 8. Procedure 8.1 Administration of the Questionnaire The survey was aimed at key decision makers involved in negotiating displays of international travelling exhibitions. It was emailed to 266 potential respondents over five continents, and a total response rate of 38% was achieved. An extensive process was used to identify
68
Research Notes
potentially eligible institutes and the email address of the key personnel. Personal contact, conference attendance and museum websites were particularly useful sources of potential respondents and snowball sampling was also used. Whilst we did not purchase or use a specific mailing list, doing so is likely to reduce respondents concerns over SPAM and virus threats. An e-mail invitation was used including a very brief explanation of the research aims. After opting in a second e-mail with Excel file attached. The use of email allowed a negotiation process to be entered into and a number of respondents were persuaded to complete the experiment when workloads eased. We had been concerned over problems that may have been caused by recipients running older versions of Excel, however Excel is designed to be backwards compatible and no issues of this sort were raised. This ability to communicate with the respondents proved important with a small number (six) who reported that they were unable to write their answers inside the text box for open-ended questions. This was due to the macros settings on their machines. As not all of these recipients had administration rights over their computers which would have allowed them to change these setting, they were instead sent instructions on how to unlock sheet protection and to write in adjacent cells. This is a more desirable suggestion than the other option which is to reduce the computers security settings. However, control over where respondents write answers was lost and additional time was spend locating these answers during data entry. Once the experiment had been completed, respondents sent the Excel file as an email attachment back to the researcher. The data from each file was then removed in the same answer sheet by answer sheet processed as used in study one. There was one additional step however which demonstrates further the flexibility of Excel. The NLogit 3.0 software used to model the answers to the experiment requires data to be entered in a text format. Once all data had been entered into the master answer sheet, a simple formula was used to convert the numerals to text. Finally the resultant worksheet was then saved in the .txt format also required by Nlogit 3.0. 9. Summary Again, Excel was found useful in designing and implementing a complex questionnaire. A wide range of response formats was possible and it placed no restrictions on the data we wanted to collect. The use of
macros caused minor problems for some respondents but these were quickly and easily addressed. Email once again proved a fast method of distribution and a small number of respondents commented in replies that it had allowed them to choose where they completed the questionnaire. The literature review provided criteria against which to assess Microsoft Excel as a survey design and implementation platform. For Excel to be used in research, it must be capable of incorporating a number of questionnaire design features to ensure that it allows accurate communication of answers from the desired sample. We found it straightforward to include a full range of scales types over both studies and these were presented in a respondent friendly fashion. Cognitive burden was reduced by being able to include additional instruction, hide worksheets and skip patterns. A full range of stimulus material was used, allowing it to be used for a wide range of marketing research objectives and data quality was improved by randomising scale items and choice profiles. It was found to be a cost-efficient method of design as it eliminated software, printing, delivery and coding costs. There were however there are costs associated with the time to become familiar with the Forms Toolbar and Control Tool Box features. Researchers who are not intermediate level users can expect to have to commit to additional Excel training. A further cost is associated with the researchers time used to construct the master data file from the individual excel files. This took approximately one minute per respondent and therefore multiplied considerable considering the size of the samples in our two studies. However, it was still considerably faster than manual coding and less costly than machine reading. In larger surveys, there may be additional costs based on the size of the researchers email account and additional server capacity may have to be purchased. In summary whilst it is less expensive than purchasing specialist software and implementing paper and pencil methods, researchers must factor in training and additional researcher time in their overall assessment of cost. We believe though, that for our studies Excel provided a cost effective method. A number of features inherent in Excel improve the chances of obtaining a sufficiently large sample that answer the questions judiciously. Administering the research via internet (from email or hosted online) gave the respondents 24/7 access to the questionnaire from any geographic location with internet access. This is
69
Research Notes
convenient for respondents and this flexibility may have enhanced the response rates. We encountered no significant network or server problems, these issues are less likely with email rather than hosted surveys. We also found that the majority of completed questionnaires were returned quickly. Ensuring that only the desired sample complete your survey is vital. Whilst we worked hard to ensure respondents were eligible and that that only one survey per email address was submitted like all other data collection methods we cannot guarantee that identity of the respondent. There was no benefit in completing multiple surveys in our studies but the weakness remains for those using some form of incentive. We attempted to maintain anonymity in a number of ways for example by assigning a unique name to each respondent when the results were pasted into SPSS. However, there are some concerns where communication occurred to sort out problems. The use of macros and embedded multimedia files whilst easy to use, did cause us some issues regarding respondents computer settings. Email communication removed the impact of this in one sense but caused problems in another by breaking the barrier of anonymity. Data quality may therefore have been affected via demand and interviewer effects. Researchers should consider using a third party to manage this communication or during the invitation to participate respondent should be asked not reveal their name in any correspondence. Manipulating security levels in order to enable macros also highlights concerns over security and virus infection. All reasonable steps must be taken to ensure that the emails are clean of infections and that the host site is secure. This must be communicated clearly to the respondents. An important related issues are the challenges that firewall and anti-spam software are present not only the use of Excel but all email based questionnaires and emailed and pop-up invitations. Specifically here, the pre-existing relationships between the student sample and researchers in study one and the snowball sampling procedure used in study two, provided ways around common security devices. It is recognised that were these procedures are not used or relationships do not exist then email questionnaires are becoming more difficult to use. Overall, the process of learning to use Excel in this way was uncomplicated and difficulties were solved using the
help function. We would recommend that people with an intermediate knowledge of Excel can use it to design and administer survey questionnaires. There is then considerable flexibility in which data collection method to use. The researcher should not, however, take for granted respondents familiarity with the Excel program. In this research, some respondents with a post-graduate education and a high career position needed very simple instructions in basic Excel functions. Again, email communication with the respondents allowed all problems of this sort to be worked out. 10. Conclusion and Recommendations It is the conclusion of this research that Excel provides a useful, flexible and easy to use Internet survey design programme. It should be considered by academics (particularly research students), schools and commercial organisations (particularly smaller companies with limited resources) as a legitimate alternative to specific survey software . Its usability in these organisations is improved however where pre-existing relationships exist as this helps avoid problems of reaching the sample caused by internet security programs. Whilst we did not use it to design an online hosted survey, it could be developed for this purpose. This use may be where it is more suited to the commercial world because design and presentational issues can be more rigorously controlled. Internal company research is another area to consider. The slightly ad hoc nature of the email attachment survey, by this we mean it is clear that email attachments are clever use of a program primarily designed for other purposes, may work better for the academic. This association may reinforce the not-forprofit nature of the research and the goodwill this generates may support the higher academic response rates. References Best, S.J., Krueger, B., 2002. New approaches to Assessing Opinion: The Prospects for Electronic Mail Surveys. International Journal of Public Opinion Research, 14, 1, 73-93. Churchill, G.A. and Iacobucci, D., 2002. Marketing Research: Methodological Foundations. South-Western, Thomas Learning. Couper, M.P., 2000. Web surveys: A review of issues and approaches. Public Opinion Quarterly, 64, 4, 464-495.
70
Research Notes
Couper, M.P., Traugott, M. W., Lamias, M. J., 2001. Web survey design and administration. Public Opinion Quarterly, 65, 2, 230-254. Deal, K., 2002. Get your conjoint online, in several flavors. Marketing Research, 14, 4,44-46. Deutskens, E., de Ruyter, K., Wetzels, M. and Osterfeld, P., 2004. Response Rate and Response Quality of Internet-Based Surveys: An Experimental Study. Marketing Letters, 15, 1, 21-36. Dillman, D.A., 1978. Mail and telephone surveys: The total design method. New York, Wiley. Dillman, D.A., 2000. Mail and Internet surveys: The tailored design method. New York, Wiley. Dillman, D.A., Sinclair, M. D., and Clark, J. R., 1993. Effects of questionnaire length, respondent-friendly design, and a difficult question on response rates for occupant-addressed census mail surveys. Public Opinion Quarterly, 57, 3, 289-304. Dommeyer, C.J. and Moriarty, E., 2000.Comparing two forms of an e-mail survey: Embedded versus attached. International Journal of Market Research, 42, 1, 39-50. Dreze, X. and Hussherr, F. X., 2003. Internet advertising: Is anybody watching?" Journal of Interactive Marketing, 17, 4, 8-23. Forrest, E., 1999. Internet Marketing Research. Sydney, McGraw-Hill. Green, P.E. and Tull, D., 1978. Research for Marketing Decisions. 4th ed. Englewood Cliffs, N.J, Prentice Hall. Hensher, D.A., 2004. Information processing strategies in stated choice studies: the implications on willingness to pay of respondents ignoring specific attributes. University of Sydney Institute of Transport Studies Working Paper series, 04-1. Ilieva, J., Baron, S., Healy, N., 2002. Online surveys in marketing research: Pros and cons. International Journal of Market Research, 44, 3, 361-376. Ioannou, C. and Black, I. R., 2004. Measuring Customer Satisfaction Response Modes: Fournier and Mick revisited. European Marketing Academy Conference, Murcia 18-23 May. Jackson, A. and DeCormier, R., 1999. E-mail survey response rates: targeting increases response. Marketing Intelligence & Planning, 17, 3, 135-140.
Kraut, A.I., Saari, L.M., 1999. Organizational surveys: Coming of age for a new era. In Kraut, A.I., Korman, A.K. (Eds.). Evolving practices in human resource management: Responses to a changing world of work, 302-327, San Francisco: Jossey-Bass. Louviere, J.J., Hensher, D.A., and Swait, J.D., 2000. Stated choice methods: analysis and applications. Cambridge: Cambridge University Press. McDonald, H. and Adam S., 2003. A comparison of online and postal data collection methods in marketing research. Marketing Intelligence & Planning, 21, 2, 85-95. Oliver, R.L., 1997. Satisfaction: a behavioral perspective on the consumer. New York, McGraw-Hill. Rapoza, J., 2003. WebSurveyor 4.1 Gets More Out of Data ; When a complex, costly enterprise survey application isn't a high priority for deployment, a hosted online survey application is a good choice for departmental use. eWeek, 20, 37, 68. Schaefer, D.R. and Dillman, D. A. (1998) Development of a standard e-mail methodology: Results of an experiment. Public Opinion Quarterly, 62, 3, pp.378-398. Sheehan, K.B., 2001. E-mail survey response rates: A review. Journal of Computer Mediated Communication, 6, 2. Sheehan, K., B, McMillan S.J., 1999. Response variation in e-mail surveys. An exploration. Journal of Advertising Research, 39, 4, 45-54. Tierney, P., 2000. Internet-based evaluation of tourism Web site effectiveness: Methodological issues and survey results. Journal of Travel Research, 39, 2, 212-220. Train, K., 2003. Discrete Choice Methods with Simulation. Cambridge: Cambridge University. Thompson, L.F., Surface, E.A., Martin, D.L. and Sanders, M.G., 2003. From paper to pixels: Moving personnel surveys to the Web. Personnel Psychology, 56, 1, 197-228. Walkenbach, J., 2003. Excel 2003 Bible. Indianapolis, Wiley. Biographies Dr Iain Black is a lecturer in the discipline of marketing at the University of Sydney. His research interests include physicians decision making processes, sustainable development and construction of the self.
71
Research Notes
Alejandra Efron is a PhD student at the Institute for Transport and Logistics studies, University of Sydney. Christina Ioannou is a PhD student in the discipline of Marketing at the University of Sydeny. John M Rose is a lecturer in the Institute for Transport and Logistics studies, University of Sydney. His researches and has published widely in the field of choice modelling.
Correspondence Addresses
Dr Iain Black, Discipline of Marketing, University of Sydney, Sydney, NSW Australia, 2006, Telephone: +61 (2) 9036 9143, Facsimile: +61 (2) 9351 6872, Email: I.Black@econ.usyd. edu.au; Alejandra Efron, Institute for Transport and Logistic Studies, University of Sydney, Sydney, NSW Australia, 2006, Telephone: +61 (2) 9351 0168, Fax +61 2 9351 0088, Email: alejandra@its.usyd.edu.au; Christina Ioannou, Discipline of Marketing, University of Sydney, Sydney, NSW Australia, 2006; Christina Ioannou, Telephone: +61 (2) 9351 4627, Facsimile: +61 (2) 9351 6872, Email: C.Ioannou@econ. usyd.edu.au; John M. Rose, Institute for Transport and Logistic Studies, University of Sydney, Sydney, NSW Australia, 2006, Telephone: +61 (2) 9351 0168, Facsimile: +61 (2) 9351 0088, Email: JohnR@its.usyd.edu.au.
72
Book Reviews
The Leaky Funnel

Bookman Media, Melbourne, Australia
Hugh Macfarlane (2003)
The Leaky Funnel uses the unusual technique of the business narrative to deliver its message on how fictional character Sue Hunt sets out as the new Chief Executive Officer of HardBits (a manufacturer of plastic beads) to improve the management and effectiveness of the sales and marketing function. There are two interesting processes that the reader should keep in mind as he or she reads through the narrative. The first process describes the interpersonal journey of Sue in reshaping the existing team of disparate functional silos into one where the management team operates with a common set of values and goals. The second process describes how the marketing mix strategies eventually adopted emerge through the often unpredictable interaction of the inter-functional team under Sues control. Very broadly, the book divides into five discrete stages of Sues journey. Stage one (covering the chapters A Beginning of Sorts and Hearing Voices) examines the search for information and clues that is vital to understanding the marketing context in which HardBits operates. Stage two (covering the chapter entitled Four Anchors) focuses on a more detailed analysis of the kinds of common problems and barriers faced by organisations as they seek to build new business. Stage three (known in the book as Mosaic) focuses on understanding the customer decision making process, the value of sales opportunities at each process stage and the gaps in the process where opportunities are lost. Stage four (known in the book as The Road and the Rubber) seeks to tie the previous sections together into a cohesive and integrated marketing strategy. The final stage (chapter title Come Monday) provides the epilogue and reveals the full version of the HardBits Marketing Plan. In Stage One (covering the chapters A Beginning of Sorts and Hearing Voices), Sue is set into a business context in which she has already formed a pre-conceived negative view of HardBits lack of customer-centric behaviour which evolved from her previous role as the
CEO of one of the companys customers. Through her own customer field visits Sue begins the search for information and clues as to the cause of HardBits poor performance. She soon realizes however, that HardBits is not such a poor performer after all and many of the problems are in fact due to the poor customer forecasting of requirements by customers themselves. To address this issue she takes on board the proposition that Vendor Managed Inventory could be used as a value-added mechanism to improve delivery performance and provide a means of moving closer to the customer. Sue also takes on board the proposition that Consultative Selling could be used as a method to become more customer-centric by being able to identify the different buying styles of customers. However, after four months of using these initiatives there was no significant improvement in sales. Sue then realizes that she and her team are focusing too much on the solutions rather than identifying and defining the problems. In other words she had answers, but did she know what the question was? She then set about interviewing all the senior mangers to ask what is stopping us from getting close to our customers and why we arent getting more new ones? The major barriers identified from the interview process were: a lack of qualified leads and a lack of understanding of the customer buying process; a lack of brand positioning and differentiation strategy; a lack of integration between marketing and the sales process; a lack of time effective sales measures. Some insights into potential answers to these problems began to emerge through the interaction process between individual team members. For example, ideas generated by the group included: (1) using the SPIN selling model (Rackham1998) to identify qualified leads; (2) using Michael Porters Competitive Advantage Theory (Porter 1980) to identify the basis for HardBits competitive advantage; (3) adapting the positioning approaches adopted in Iacobuccis edition of Kellogg on Marketing (2001) to position HardBits as an Innovation Partner in the customer value chain.
73
Book Reviews
Stage two (covering the chapter entitled Four Anchors) is essentially a more detailed elaboration of some of the barriers identified in Stage One and sets the scene for Stage Three. For example, it shows that there is no integration between the sales and marketing functions; that the buying process is ignored; that the marketing and sales tactics used are arbitrary and that current key performance indicators are either too late or measuring the wrong things. Stage three (known in the book as Mosaic) produces a detailed mapping of the sales process for HardBits by activity and the time duration of each activity in the process. The concept of the sales funnel is to measure each stage of the (customer) journey. How many make it from stage one to stage two? And how many of them make it to stage three, etcetera? This mapping process allows leakage rates at each stage of the sales process to be estimated and strategies to be devised to plug those leaks as they arise. It also allows the integration of the sales and marketing functions, with marketing incubating each sales lead until it is warm enough to be passed on to the sales department. Likewise, those prospects which have gone cold would be recycled back to the marketing department for further incubation. This kind of approach means that all the metrics need to be based around each stage of the sales process. Stage four (known in the book as The Road and the Rubber) addresses the companys overall differentiation strategy as one in which the marketing team will position HardBits as an Innovation Partner in the customers value chain. The targets for this strategy will be selected Sales or Marketing Directors of the 1000 companies who are currently using plastics as a key component of their product development, but lack the ability to innovate substantially in-house. Sue and her team then identify the sales process in terms of the kind of activities that HardBits will have to perform at each stage of the process. They also build a three-year sales forecasting model which is centered around the assumptions and the steps identified in the sales process. Overall, the main strength of this book is the conceptual treatment provided for the analysis of the complex, time protracted sale in a business-to-business marketing context. Of interest to academics and practitioners alike is the thinking process that Sue goes through as CEO of HardBits in building both her team and developing a strategy using the Sales Funnel technique. Marketing is shown to be a process which emerges through the
interaction of all the actors in Sues inter-functional management team. The major weaknesses of this book are: (1) the authors failure to address any of the vast arrays of relevant articles and books written on businessto-business marketing and sales management, including the work of the IMP Group (Industrial Marketing and Purchasing Group) and all the literature written in the area of relationship marketing. (2) The authors failure to provide any objective supporting evidence that the application of the sales funnel approach results in improved performance (both sales and profitability) over existing methods. This weakness is a general one relevant to many practitioner books on sales and marketing. Without objective evidence provided through either qualitative or quantitative research these kinds of books easily can be labeled as just an anecdotal experience relevant to one specific context. (3) There is a considerable amount of repetition and overlap between one section and another. (4) In terms of structure, the book would have been much easier to read if there had been a proper system of numerical chapter numbers and an alphabetical index at the end of the book. It is also worth noting that there are quite a number of software packages on the market which provide facilities for the measurement of sales opportunities using the sales funnel technique (e.g., ACT and Goldmine). On balance, however, the detailed investigation of the activities involved in the sales process, and the integration of marketing strategies and tactics with those sales process activities, presents a rewarding insight into how the Sales Funnel technique can be woven effectively into a companys marketing process. References Iacobucci, D., 2001 (editor). Kellogg on Marketing. John Wiley & Sons, USA. Porter, M., 1980. Competitive Strategy: Techniques for Analyzing Industries and Competitors. The Free Press, Simon and Schuster, USA. Rackham, N., 1998. SPIN Selling. McGraw-Hill, USA. www.imp.org This website provides access to articles covering the business-to-business interaction process developed by leading proponents of the Industrial Marketing and Purchasing Group
David M. Gray School of Marketing University of New South Wales
74
Book Reviews
Services Marketing in Asia: Managing People, Technology and Strategy (Second Edition)
Pearson Prentice Hall, Singapore
Christopher Lovelock, Jochen Wirtz, Hean Tat Keh & Xiongwen Lu (2004)
Services Marketing in Asia is a textbook designed for advanced undergraduate degree students or Master of Business Administration students working in an Asian business environment. This is the only textbook on services marketing that has been specifically written with a direct focus on Asian practice, viewed from multiple perspectives. The book is a successful attempt to provide a comprehensive knowledge of the service marketing and management concepts. The content of the book serves to introduce students to the latest research and developments in the services sector, ranging from customer relationship management, customer asset management, and six sigma quality to revenue (yield) management and customer feedback systems. It substantially covers consumer behaviour issues, service staff management issues, branding, business-to-business services, and technology-based services. The book has a four-part structure. Part one, Understanding Service Products, Consumers, and Markets, contains chapters 1 to 3 (An Introduction to Service Marketing, Consumer Behaviour in Service Encounters, and Positioning Services in Competitive Markets). Part two, Key Elements of Service Marketing, contains chapters 4 to7 (Creating the Service Product, Designing the Communication Mix for Services, Pricing and Revenue Management, and Distributing Services). Part three, Managing the Service Delivery Process, contains chapters 8 to 11 (Designing and Managing Service Processes, Balancing Demand and Capacity, Planning the Service Environment, and Managing People for Service Advantage). Part four, Implementing Service Marketing, contains chapters 12 to 15 (Managing Relationships and Building Loyalty, Customer Feedback and Service Recovery, Improving Service Quality and Productivity, and Organizing for Service Leadership).
As the authors note, the key features of this book include its strong managerial orientation and strategic focus, use of memorable conceptual frameworks, incorporation of key academic research findings, use of interesting examples to link theory to practice, and inclusion of carefully selected cases to accompany the text chapters. The book is especially useful at present because of the rapid growth of service sectors in Asian countries. The cases provide a broader coverage of service issues and application areas on firms operating across many Asian cultures (e.g., Banyan Tree, DHL, KFC, McDonalds and Giordano) as well as organizations operating in specific Asian countries and regions, including China, Hong Kong, Taiwan, India, Malaysia, Myanmar, the Philippines, and Singapore. Lets look more closely at a few of the chapters to get a feel for the book. Chapter 2, Consumer Behaviour in Service Encounters, analyses how the extent of customer contact in service encounters affects the nature of service encounters, shapes consumer behaviour and can impact strategies for achieving productivity and quality improvements. The authors take a range of service organisations to discuss how the nature of service emphasises customer interaction with equipment used in service creation and delivery or with service personnel. Using illustrations based on the National University of Singapore (NUS) and China Unicom Ltd. (a telecommunication service in China), the authors contrast consumer behaviour in high- and low-contact service situations. Another section in this chapter focuses on customer expectations. Giving the example of Daniel Bethany of American Express that consumers want memorable experiences, not gadgets, the authors state that service companies must understand this shift in consumer behaviour and attitude to create opportunities for differentiation. Expectations are likely to vary in relation to differently positioned service providers in the
75
Book Reviews
same industry. While travellers expect no frills service for a short domestic flight on a budget carrier, they would undoubtedly be very dissatisfied with that same level of service on a full service airline flying from Kuala Lumpur to Hong Kong, even in economy class. Taking the case of the Sunway Lagoon in Malaysia, the authors succinctly illustrate that customers may base their prepurchase expectations on factors like news stories, the firms marketing efforts, word-of-mouth, etc. At several places the chapter presents illuminating insights from various magazines published in Asia. Chapter 3, Positioning Services in Competitive Markets, emphasises the need to examine levels of competition in the market for services firms and analyses various issues involved in developing a positioning strategy. Presenting the case of DHL No one knows Shanghai like we do. Heres your chance to get to know it better too, DHLs positioning strategy in China is based on an in-depth knowledge of Asia and hence DHLs ability to deliver on time, every time, is demonstrated. The authors discuss the topic of segmentation to develop appropriate positioning strategies. The examples of Shin Kong Life Insurance Company promoting the specially Designed for Women insurance policy targeted at young working women of China (p. 81), and the ICICIs example (p. 82) focusing on the rural population of India, explain the need to develop the right service concept for a particular target segment. Further, the
authors discuss that local knowledge is the most important factor for a service firm to succeed in an overseas market. This is illustrated through the example of HSBC (Hong Kong and Shanghai Banking Corporation) (p. 85) showing the different associations among people from three different countries of the meaning of bread (Philippines, India and Germany). Chapter 14, Improving Service Quality and Productivity, reviews the challenges involved in improving both productivity and quality in service organisations. The implementation of service marketing strategies for improving customer satisfaction should not prove costly and disruptive for the service firm. Chapter 15, Organizing for Service Leadership, discusses that marketing activities in service organizations extend beyond the responsibilities assigned to a traditional marketing department. The service leadership of INFOSYS Technologies, Indias premier software house, clearly shows that the underlying success factors are consistent focus and commitment to stakeholders, clients, employees, and innovation. In sum, Services Marketing in Asia contains adequate information to make a valuable contribution to bestpractice services management in Asia.
Neeru Sharma University of Western Sydney
76
Book Reviews
Marketing Communications: Theory and Applications

Pearson Education: Australia
John R. Rossiter & Steven Bellman (2005)
Overview Marketing Communications: Theory and Applications is a text aimed at both marketing managers and students (i.e., undergraduate and postgraduate marketing communication students). As the title suggests, the book promises to deliver theoretical frameworks and executional procedures that will help managers negotiate, what the authors rightfully contend, are a myriad of choices available to them. Central to the text is the notion of the brand and the role of marketing communications in presenting the brand to consumers. To this end, a six-step model of marketing communications guides the structure of the text. The steps being: (1) brand positioning, (2) campaign objectives, (3) creative strategy, (4) promotion strategy, (5) media strategy, and (6) campaign management. The text begins in a feisty style. As part of a discussion of the academic-practitioner divide, we are told that the academic literature "has remained straightjacketed in terms of marcoms ideas for at least 10 years and has pursued trivial sameness with just a few exceptions. Academic researchers, particularly in advertising, must learn to get real." While this sentiment probably resonates with the target market of practitioners, it is arguable. For readers who fiercely endorse the notion that academic research is out of touch I suggest they read Calder, Phillips and Tybout (1981) (specifically, their discussion of effects application and theoretical application). This article could be used as a vehicle to explain the apparent neglect of reality by academic advertising research, which presumably pursues theoretical application. Having said that, criticisms in the preface are also aimed at practitioners. The text draws upon a useful spread of journals and is not solely restricted to the Journal of Advertising and Journal of Advertising Research as one might expect for a marketing communications text. Indeed, top-tier
journals such as Journal of Consumer Research, Journal of Marketing Research, Marketing Science, and top social psychology journals receive a number of citations, as do journals as diverse as the Australasian Marketing Journal, International Journal of Research in Marketing, and Journal of Consumer Marketing. Rossiter and Bellman are to be congratulated for going to this level of effort. This is important given research that highlights the diversity of marketing journals (see for example Baumgartner and Pieters 2003). Text Structure and Content Marketing communications is a subject that students often equate with advertising. Hence, the challenge for authors is to write a text that addresses advertising, without demoting other aspects of marketing communications to insignificance. Yet at the same time, advertising needs to be recognized as a key, visible aspect of this area and addressed accordingly. The authors here present advertising and promotions as two major, distinct elements of marketing communications, with public relations and personal selling as related areas that are managed in a separate but consistent fashion. The structure of the text essentially follows the aforementioned six-step model with the addition of a section on other marcoms. Chapters 1 to 4 set the foundation. Chapter 1 clearly delineates between advertising and promotions, provides background figures, and emphasizes the assertion that brand equity is the main determinant of profit. In other words, despite the tactics suggested throughout to generate consumer response, managers should keep in mind that a beneficial effect on brand equity should be a key motivation for long-term success. Chapter 2 presents three levels of effects which I will discuss in a subsequent section. Chapters 3 and 4 outline a series of useful models that fit
77
Book Reviews
into the overall six-step framework; for example, the TCB model (i.e., Target customer, Category need, and Key Benefit) for brand positioning. Interestingly, even topics such as brain research (what the authors term "brain-levels theory") get an airing here. This clearly differentiates the text from many other marketing communication books. Also included are topics such as a brief nod to archetypes and the manner in which brand benefits should be communicated is discussed. Benefit positioning is addressed in detail in Chapter 4. A multiattribute approach for managers is presented with detailed examples. Throughout, the authors retain their focus on how this relates to the brand and bring in the importance of integrating marketing messages. Chapters 5 and 6 study campaign objectives. The authors discuss brand loyalty segmentation and present a questionnaire which can be used to classify a potential market. The notion of desired communication effects is addressed in relation to persuading consumers to take action. Consistent with previous parts of the text, the importance of the brand is again highlighted regarding, for example, brand recognition and brand preference. Chapters 7 to 10 consider creative strategies for advertising. The authors set the scene with figures indicating effectiveness issues, and then present components of an ad as a foundation for discussing such issues as the key benefit claim and the creative idea. The suggestions in Chapter 7 regarding brainstorming and the examples are informative. Chapters 11 and 12 examine advertising media strategy. The authors recommend that managers be directly involved in media planning, and provide a logical sequence of topics and decisions for managers to consider. In particular, reach and frequency are discussed in depth in an insightful way. Chapters 13 and 14 address the campaign budget and tracking. This includes the good use of data, examples and practical advice. Chapters 15 to 18 address other areas, such as sales promotions, corporate image advertising, personal selling and social marketing campaigns. This section was a nice surprise, as will be discussed in relation to personal selling. All chapters conclude with a summary and a set of discussion questions. Generally, there are seven questions for each chapter, with a surge to 14 questions for the chapter on Sales Promotion. While some teachers may seek more questions, there are some benefits to the approach Rossiter and Bellman have taken. Namely: (a)
The questions require some effort on the part of reader that is of an interesting nature (e.g., find an ad, design an ad). The authors generally do not ask the reader to recite terms from the chapter with minimal application. (b) The product contexts for the questions are varied, ranging from the expected (e.g., Coke) to the unexpected (e.g., a CD for the Dixie Chicks). Interestingly, given the authors awareness of the American target audience for the text the discussion questions are not specific to American culture. This can be a challenge when using Marketing Communications texts from America, where discussion questions - or indeed instructor manuals often focus on examples unfamiliar to Australasian students. The authors get around this issue nicely by referring readers to websites for further information. The internet exercises on the website that relates to the text are also useful and in some instances contain brief hints to assist people. The website also contains Media Mania, a useful program on media scheduling models written by Peter Danaher. In terms of presentation, the text is well-structured and easy to follow with models and formulae clearly presented. There are good signposts of where the text is going and also when areas will be discussed in greater detail. Personally I prefer sources of models to be fully cited under a figure or diagram, whereas this text tends to cite authors and year of publication, with a note at the end of the chapter with fuller details. This allows for commentary on the paper but requires flicking to the end of the chapter to get the information. Some terms are also not in the glossary or index. For example, the term "polyarchical" is mentioned in the text, the summary, and is the subject of a discussion question, but it is not defined in the glossary or referenced in the index. Yet the writing itself has a simple, clear, almost chatty style that reflects an appreciation of when things may be difficult for a novice reader. For example, "The diagram may look complicated but it's quite easy to understand when you work your way through it" (p. 24), or "The terminology with which we describe them below will look Martian at first but special terms are necessary and we'll define them carefully" (p. 235). Equally there are some nice touches of humour, such as, "and Newton (that's Derek, not Sir Isaac) recommends . . ." (p. 407), although at times the writing does lean towards stating the obvious. For example, "will produce a 0.3% (that's three-tenths of one percent) . . ." (p. 12). Indeed the preceding example shows a feature of the
78
Book Reviews
writing style of the text; namely, the frequent use of parentheses to elucidate some point. This results in longish sentences at times, but eliminates the need for footnotes or further chapter endnotes. The parentheses are also useful for referring to other chapters. The endnotes are generally informative, although at times are indulgent (e.g., how one of the authors had just had children and had a recent need for diapers, p. 41) or thanking individuals for providing examples (couldn't they have been thanked for this in the acknowledgements?). Still, these are minor points, and whether you view such things as adding colour to the text or as unnecessary diversions depends on the personal preference of the reader. Points of difference How then does this text differ from a standard marketing communications text beyond its brand focus? Some points of difference include: Specific recommendations. A strength of books by John Rossiter is that he offers recommendations. Unlike texts which say nothing more than "this issue is crucial" and present a model or two, Rossiter and Bellman offer their readers concrete rules. For example, commission levels in retail sales (p. 405). This is one of the key virtues of the text. Integrating the message. Contrary to some texts which revolve around the notion of Integrated Marketing Communications (IMC), the authors present a pragmatic argument for taking a more restricted focus. Rather than arguing that marketers use a broad variety of media to present a message, the authors suggest that advertisers tend to use one primary medium with one or two secondary media. They back this view up with data and tie it to notions of reach and frequency. Three levels of effects. The authors move away from citing the standard hierarchy of effects model, and instead present three simultaneous levels of effects, discussed in Chapter 2. The three levels are: first, ad or promotion processing (e.g., "attention"), second, brand communication effects (e.g., "category need"), and third, customer decision stages (e.g., "not in the market"). The authors discuss how some areas can act in a hierarchicial fashion, but recognize where other areas do not (e.g., for processing effects). They also include the role of emotions under processing which augments a cognitive perspective. Many of these elements will be familiar to consumer researchers who have an understanding of
memory, attitudes, involvement and processing of stimuli. The authors discuss the typical considered purchase versus impulse purchase (analogous to high and low involvement decisions), and then usefully bring together the customer response stages with the planning stages for a manager. Australian examples and media references. In contrast to American texts, this text includes data from other countries, including understandably enough, Australia, to supplement American data, and to engage the reader. Personal selling. Surprising perhaps, given the depth given to advertising and promotion, the chapter on personal selling is very interesting. Whereas some marketing communications texts simply give a definition of the selling process, Rossiter and Bellman present a variety of interesting features. For example, discussing how General Mental Ability (i.e., IQ level) relates to hiring. Consumer individual differences also get a workout in this chapter in relation to such issues as salesperson rejection and attribution style. A nice touch. While these examples of points of difference are a strength, at times one could argue that the stated contributions may be "new" but not necessarily "unique" as suggested in the preface. Specifically, one such contribution is the "Consistent approach carried through to other marcoms - sales promotion; personal selling and telemarketing; sponsorships, PR, and corporate image advertising; and social marketing campaigns." Yet the notion of consistency parallels that of IMC where a synthesized approach or uniform message is presented through multiple media. Thus, such a contribution would not appear to be unique, even if it is differentiated from the IMC approach. Conclusion Overall, Marketing Communications: Theory and Applications by Rossiter and Bellman presents a great deal of useful theories and concrete, actionable recommendations. The brand focus, and the recognition of brand equity, sets this book apart from texts that revolve around an IMC approach. For that alone it represents an interesting and engaging take on this everpopular topic. Personally, it is the sort of text I would consider for undergraduates, rather than for postgraduates. However this would depend on the nature of one's postgraduate programme. In summary, this text is a highly useful addition to the texts on offer to marketing educators.
79
Book Reviews
References Baumgartner, H. and Pieters, R., 2003. The structural influence of marketing journals: a citation analysis of the discipline and its subareas over time. Journal of Marketing 67 (2), 123-139. Calder, B.J., Phillips, L.W. and Tybout, A.M., 1981. Designing research for application. Journal of Consumer Research 8 (September), 197-207.
Brett A. S. Martin University of Auckland Business School
80
amj
Editorial Objectives
Australasian Marketing Journal
Notes for Contributors
The Australasian Marketing Journal is an academic journal written for both scholars and practitioners. The objective of the amj is to publish articles that enrich the practice of marketing while simultaneously contributing to the advancement of the discipline. Therefore, manuscripts accepted for publication must be conceptually and theoretically sound, offer significant research findings and insights, and also suggest meaningful implications and recommendations for practitioners. Given amjs diverse readership, preference is given to manuscripts that are generalisable across industries, nations and economies. Further, if the manuscript reports the findings of original research, the methodology and findings not only should be scientifically defensible but presented in a manner that readers with limited backgrounds in research methods and statistical analyses are not discouraged from reading the article. Further explanation of the Editorial objectives is to be found on pages 5-6 of volume 9, Number 1, 2001 and also on the journal web-site (www.marketing.unsw.edu.au/amj) General Principles It is our intention to publish well-written, readable articles with broad appeal and of international relevance. Contributors are encouraged to focus on either conceptual or empirical work and to outline practical implications for marketing. Topics should always relate to some aspect of marketing. New and first-time authors are particularly encouraged. At amj we are keen to publish more good papers which emanate from the Asia-Pacific region, or focus on the region in a global context. Editorial Scope It should be noted that research is not the only basis for an acceptable article. Case analyses, creative concepts and applications, book reviews, commentaries and other thoughtprovoking manuscripts will be considered. Particularly welcome are future-oriented manuscripts that offer visions of marketing in the twenty-first century and practical road maps for getting there. The Review Process Each paper is reviewed by the editor and, if it is judged as
potentially suitable for this publication, it is then sent to two or three referees for double-blind peer-review. Based on their recommendations, the editor then decides whether the paper should be accepted as is, revised or rejected. Copyright Articles submitted to the journal should be original contributions and should not be under consideration for any other publication at the same time. Authors submitting articles for publication warrant that the work is not an infringement of any existing copyright and will indemnify the publisher against any breach of such warranty. For ease of dissemination and to ensure proper policing of use, papers and contributions become the legal copyright of the publisher unless otherwise agreed. This extends to both hardcopy and electronic versions of the amj. Indexing amj is indexed by: Public Affairs Information Service; ANBAR and Emerald Review. Manuscript Requirements Three copies of the manuscript should be submitted in double or one-and-half line spacing with wide margins and numbered pages. All authors should be listed, with author details printed on a separate sheet. Authors should not be identified anywhere else in the article. Formatting of Manuscripts As a guide, articles should be between 3,000 and 6,000 words in length. A title of not more than eight words should be provided. A brief autobiographical note should be supplied including full name, affiliation, e-mail address and full international contact details. Authors must supply an abstract of 100-150 words. Up to six keywords should be included which capture the principal topics covered by the article. Where there is methodology, it should be clearly described under a separate heading. Headings and sub-headings must be short, clearly defined and numbered. Endnotes should be used only if absolutely necessary and must be identified in the text by consecutive numbers, enclosed in square brackets and listed at the end of the article. Figures, charts and diagrams should be kept to a minimum. They must be black and white with minimum shading and
82
numbered consecutively, using Arabic numbers with a brief title and labelled axes. In the text, the position of the figure should be shown by typing on a separate line the words take in Figure 2. Good quality originals must be provided. Particular attention should be paid to the legibility and clarity of figures, charts and diagrams. Tables must be numbered consecutively with roman numerals and a brief title. In the text, the position of the table should be shown by typing on a separate line the words take in Table 4. Tables should be carefully designed to communicate a clear and simple message. Numbers should be rounded to two significant figures. Photos and illustrations must be supplied as good quality black and white original half tones with captions. Their position should be shown in the text by typing on a separate line the words take in Plate 2. References to other publications must be in the following style Smith (1992) reported that or This problem has been studied previously (e.g. Smith et al., 1979). The author should make sure that there is a strict one-to-one correspondence between the names and years in the text and those on the list. The list of references should appear at the end of the main text (after any appendices, but before tables and legends for figures). It should be double or one-and-half spaced and listed in alphabetical order by authors name. References should appear as follows: For books: Kotler, P., Ang, S.H., Leong, S.M., Tan, C.T., 1996. Marketing Management: An Asian Perspective, Prentice-Hall, Singapore. For articles in collected volumes: Douglas, S.P., Morrin, M.A., Craig, C.S., 1994. Cross-national consumer research traditions. In: Laurent, G., Lilien, G.L., Pras, B. (Eds.),
Research Traditions in Marketing. Kluwer Academic Publishers, Boston, MA, pp. 289-306. For articles in journals: Brodie, R.J., Danaher, P.J., 2000. Building models for marketing decisions: Improving empirical procedures. International Journal of Research in Marketing 17 (2-3), 135-139. Note that journal titles should not be abbreviated. Final Submission of the Article Once accepted for publication, the final version of the manuscript must be provided, accompanied by a 3.5 disk of the same version labelled with: disk format; author name(s); title of article; journal title; file name. Alternatively, an e-mailed copy is acceptable. Author/s should also supply a short biography and 3/4 page executive summary outlining the implications for practitioners and/or public policy makers. This should not simply replicate the abstract. Implications should directly follow from the research and not be purely speculative. The manuscript will be considered to be the definitive version of the article. The author must ensure that it is complete, grammatically correct and without spelling or typographical errors, and with a correct record of references. In preparing the disk or e-mailed copy, please use Microsoft Word format. Page set-up should be in A4 standard (not Letter). All tables must be inserted in true Microsoft Word format. No heading styles should be used. Figures provided electronically must be in tif, gif or pic file extensions. All figures and graphics must also be supplied as good quality originals. Each article must be accompanied by a completed and signed Journal Article Record Form.
Submission requirements Manuscript must be clean, good quality, hard copy and in triplicate. Include an abstract and keywords; brief biography of authors and a separate 3/4 page executive summary. Follow style, reference & format guidelines. Include any figures, photos and graphics as good quality originals. Submissions should be sent to: The Editor Richard Varey Department of Marketing The Waikato Management School, University of Waikato Private Bag 3105, HAMILTON NEW ZEALAND
83

Australasian Marketing Journal - Volume 13 - Number 2 - 2005

Hochgeladen von

Dokumentinformationen

Originalbeschreibung:

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Australasian Marketing Journal - Volume 13 - Number 2 - 2005

Hochgeladen von

Copyright:

Verfügbare Formate

Editorial

amj Progress and Prospects

Australasian Marketing Journal 13 (2), 2005

Australasian Marketing Journal 13 (2), 2005

Australasian Marketing Journal 13 (2), 2005

Australasian Marketing Journal 13 (2), 2005

Australasian Marketing Journal 13 (2), 2005

Consumer Loyalty: Singular, Additive or Interactive?

Consumer Loyalty: Singular, Additive or Interactive?

Australasian Marketing Journal 13 (2), 2005

Consumer Loyalty: Singular, Additive or Interactive?

Australasian Marketing Journal 13 (2), 2005

Consumer Loyalty: Singular, Additive or Interactive?

Word-of-mouth Search motivation Resistance to persuasion CONSEQUENCES

Figure 2. Dick and Basus (1994) Typology of Loyalty

Australasian Marketing Journal 13 (2), 2005

Consumer Loyalty: Singular, Additive or Interactive?

Australasian Marketing Journal 13 (2), 2005

Consumer Loyalty: Singular, Additive or Interactive?

Australasian Marketing Journal 13 (2), 2005

Consumer Loyalty: Singular, Additive or Interactive?

Australasian Marketing Journal 13 (2), 2005

Consumer Loyalty: Singular, Additive or Interactive?

Australasian Marketing Journal 13 (2), 2005

Consumer Loyalty: Singular, Additive or Interactive?

Australasian Marketing Journal 13 (2), 2005

Consumer Loyalty: Singular, Additive or Interactive?

-0.03 -1.05 0.34 -1.75 -0.30

.74 .000 .19 .001 .135

0.23 0.58 0.55 2.02 0.62

.03 .005 .08 .001 .01

Australasian Marketing Journal 13 (2), 2005

Consumer Loyalty: Singular, Additive or Interactive?

Australasian Marketing Journal 13 (2), 2005

Consumer Loyalty: Singular, Additive or Interactive?

Australasian Marketing Journal 13 (2), 2005

Consumer Loyalty: Singular, Additive or Interactive?

Cox & Estimate Snell R2 % 17.6 1.50 0.22

Australasian Marketing Journal 13 (2), 2005

Consumer Loyalty: Singular, Additive or Interactive?

Australasian Marketing Journal 13 (2), 2005

Consumer Loyalty: Singular, Additive or Interactive?

Australasian Marketing Journal 13 (2), 2005

Consumer Loyalty: Singular, Additive or Interactive?

Australasian Marketing Journal 13 (2), 2005

Consumer Loyalty: Singular, Additive or Interactive?

Australasian Marketing Journal 13 (2), 2005

Consumer Loyalty: Singular, Additive or Interactive?

Australasian Marketing Journal 13 (2), 2005

Repeated Binary Logit: Analysing Variation in Behavioural Loyalty

Repeated Binary Logit: Analysing Variation in Behavioural Loyalty

Australasian Marketing Journal 13 (2), 2005

Repeated Binary Logit: Analysing Variation in Behavioural Loyalty

Australasian Marketing Journal 13 (2), 2005

Repeated Binary Logit: Analysing Variation in Behavioural Loyalty

Australasian Marketing Journal 13 (2), 2005

Repeated Binary Logit: Analysing Variation in Behavioural Loyalty

Loyalty to Favourite Store

Loyalty to Favourite Store

Australasian Marketing Journal 13 (2), 2005

Repeated Binary Logit: Analysing Variation in Behavioural Loyalty

Table 1: Descriptive Statistics for the Covariates

Australasian Marketing Journal 13 (2), 2005