A Study of Market Structure: Brand Loyalty and Brand Switching Behaviours for Durable Household Appliances

Chinho Lin
ZhiFeng Wang
Department of Industrial Management Science
WannYih Wu
Department of Business Administration
National Cheung Kung University, Taiwan


As the product life cycle and the price increase for most durable products, consumers tend to be highly involved and become more rational in their purchase (Anirudh, 1994). These factors, combined with others, such as the prevalence of global marketing and the reduction of international trade barriers, have resulted in highly competitive markets for durable products. In todays free and competitive markets, consumer preferences and key attributes of products have become so diversified that companies that can meet these customer expectations will prosper and grow, while those that fail to do so will decline. Allenby (1989) and Kannan & Wright (1991a,b) suggest that the design of successful marketing strategies requires a thorough understanding of the structure of the product market and the patterns of competition within those markets. Thus, it is crucial to collect market data to analyse market structures, especially for durable goods that have entered their mature stage, characterised by maturing product technologies, lack of product differentiation and similarity in product qualities. The increasing competition, brand image and attitude toward brands will greatly impact on the purchase decisions of customers. In order to position its products and brands, the manufacturer needs to comprehend the consumers attitude towards these factors. Therefore, it is essential to have a better understanding of brand loyalty.

It is also useful to segment the market based on brand loyalty and to understand the needs of both loyal customers and potential brand switchers, as well as their attitude towards key brand attributes. Based on such findings, corporations can find ways to increase their market share by fulfilling the needs of repeat buyers. In addition, corporations can convert brand switchers into loyal customers by focusing on certain key factors.

Thus, the intent of this study is to provide insights of the needs of two groups of customers, repeat buyers and brand switchers, in order to help corporations develop appropriate marketing strategies. This study was carried out in two stages:

  1. segmenting the market into the two consumer groups; and

  2. analysing the differences of the two groups based on the key factors that affect consumption behaviours.


Definition of brand and brand loyalty

The American Marketing Association defines brand as 'a name, term, sign, symbol, or design, or a combination of them, intended to identify the goods or services of one seller or group of sellers and to differentiate them from those of competitors,' (Kotler, 1995). Aaker (1995) defines a brand on different levels, stating that a brand is not merely the physical product, but is also composed of brand attributes, symbols, brandconsumer relationships, benefits of self expression, customer profiles, associations with the culture of the country of origin, and corporate identity. In essence, the brand provides a simple means for the customer to distinguish it from its peers. Padberg et al. (1974) stated that in the marketing process, a brand provides a means of communicating economic information; it facilitates product recognition and protects the customer from the risks associated with buying an unknown brand.

De Chernatony & McDonald (1998) stated that 'A successful brand is an identifiable product, service, person or place augmented in such a way that the buyer or user perceives relevant, unique added values which match their needs most closely. Furthermore, its success results from being able to sustain these added values in the face of competition.' Successful brands deliver benefits to satisfy customer needs. These needs include rational needs (such as features, packages or the price of a brand) and emotional needs (such as prestige, distinctiveness, style or the social reassurance of a brand).

Brand loyalty refers to the consumers behaviour of repeatedly purchasing a specific brand over a certain period of time. This is based on past behaviour, and the local consumer is highly likely to purchase the products of a specific brand currently and in the future. According to Aaker (1995), a powerful brand enjoys a high degree of brand loyalty. Related brand choice theories claim that, in order to increase the sales volume or marketing shares of a particular brand of products, it is necessary to either strengthen the brand loyalty of existing customers or try to persuade the consumers of other brands to switch. The former is called inertia or brand loyalty, and the latter, brand switching.

Although it is an important concept, brand loyalty measurement has not flourished in the marketing literature (Chaudhrui, 1999). Previous studies (CobbWalgren et al. 1995; Baldinger & Rubinson 1996; Dyson et al. 1996) concluded that a persons attitude toward a brand is relevant to the degree of their brand loyalty. Brand awareness and brand association are linked to consumers brand preferences. Ehrenberg et al. (1997) asserted that salient brands are high in both intention to buy the brand and brand loyalty. Ehrenberg (1997) stated that since the selling brand tends to be copied quickly by competitors, competitive brands lack variation between each other. Similar brands have different market shares because of the different number of people to whom each brand is salient. Ehrenberg et al. (1997) further argued that the main function of advertising is to reinforce an existing consumers propensity to buy a particular brand.

Research on brand buying behaviour and brand switching

There are many theories that are available to explain how consumers make product and/or brand choices. The 'expectancevalue model' argues that consumers assign scores to two parameters and make a mental calculation before making a decision. The first parameter is the degree to which consumers expect a pleasurable outcome. The second parameter is the value the consumers ascribe to a favourable outcome. This model is insufficient to explain the phenomenon because people have limited brand information and limited mental processing capabilities (Yi & Vanden Abeele 1989; Dabholkar 1994; Shoham et al. 1998).

The economists view of consumer behaviour hypothesises that consumers seek information until the marginal value that is gained is less than the cost of securing knowledge of the product. This model is also not acceptable since in many cases consumers are unable to acquire 'perfect' information (Girish 1989; Russo et al. 1998)

De Chernatony and McDonald (1998) propose a more accepted model for brand buying behaviour. It argues that the making of a brand purchase is determined by 'consumers seeking and evaluating small amounts of information. Consumers rely only a few piece of information with which they feel confident to help them decide how the brand might perform.' The amount of information that consumers seek may be determined by various factors such as time pressure, previous experience, advice from friends and the level of involvement in the brand purchase.

In recent years many researchers have studied brand choice and switching. The scope of these studies not only includes the analysis of the factors that affect the consumers brand choice and switching behaviour, but also brand switching behaviour itself and the actual process of switching brands. The studies can be divided into two categories:

  1. Reasons for brand switching. The goal of the study is to identify the key factors that lead to repetitive purchasing or brand switching of consumers. For example, Kannan & Wright (1991a, b) and Keaveney (1995) explore the major reasons for brand switching between services.

  2. Analysis of brand switching models. According to McCarthy et al. (1992), brand switching models generally adopt one of two approaches. In the first approach, individual brands are grouped into submarkets according to the frequency that brand switching occurs, with high frequency indicating intense competition. Within any of these submarkets, the competition is more intense than that between different submarkets. See, for example, Gregory & Lehmann (1985), Rao & Sabavala (1981), Rubinson et al. (1980), and Kalwani & Morrison (1977). In the other approach, consumers are divided into heterogeneous groups which establish the market structure. Examples are McCarthy et al. (1992) and Colombo & Morrison (1989) who, based on their on studies of brand loyalty and brand switching in the American automobile market, divided consumers into loyalists and potential consumers. Weerahandi & Moitra (1995) studied the brand switching behaviours of heterogeneous consumer groups in the services market. Others, including Blumen et al. (1955), Jeuland et al. (1980), Grover & Srinivasan (1992) and Jain et al. (1990) provided methods for analysing market structures based on various brand switching models.

These brand switching models can be classified into three types based on the factors that affect the choice process:

  1. Zero step model. This states that the consumers current choice in making a purchase is unaffected by past behaviour or experience. Jain et al. (1990) points out that the choice process involved in buying a house is a zero step model. In addition, scholars such as Bass et al. (1984) believe that the choice processes of buying frequently purchased consumer goods also follow a zero step model.

  2. Markovian stochastic process. In a Markovian stochastic process, a consumers past choice affects his or her current choice. The brand switching model provided by Blumen et al. (1955), in which the consumers current choice is affected by his or her impression of the last choice, is a Markovian stochastic process. In this model, the consumers degree of satisfaction affects his or her current choice of product. If having previously chosen a brand increases the chance of choosing the same brand, this is referred to as brand loyalty or as inertia (Jeuland et al. 1980; Kapil 1990). If, however, the consumers previous choice of a brand does decrease the chance that he or she will choose the same brand currently, this is called 'varietyseeking' (McAlister 1982). Other studies for constructing the brand switching model can be found in papers by Kalwani & Morrison (1977), Gregory & Lehmann (1985), Rubinson et al. (1980) and Rao & Sabavala (1981).

  3. Three choice model. McCarthy et al. (1992) expanded on the two choice model proposed by Blumen et al. (1955), creating the three choice pattern. As in the two choice model, the consumers past choice of product influences his or her current choice of product in a three choice model. However, in a three choice model, the consumers level of satisfaction with the current choice of product affects substitute choice behaviour, as influenced by brand recognition and attitude towards brand. Thus, in the three choice model it is possible that when the consumer becomes dissatisfied with current choices, substitute or future choices will result.

Brand marketing

Brand marketing is an important marketing strategy. If successfully implemented, it can result in substantial profits for the company. According to Aaker (1995), when the consumer is confronted with a variety of product choices and related information, he or she may not have the time or the patience to develop an in depth understanding of each product. Thus, opportunities exist for manufacturers who can shorten the decision making process and reduce the nonmonetary cost to the customer. This objective can be achieved through brand marketing in which the brand can embody complex product information and increase the level of recognition. The result is a shortened choice process for the consumer because he or she, by recognising the product, can efficiently compare it with other products. A company can increase its profitability through brand marketing strategies by increasing brand equity, systematically managing the brand and implementing line extensions to foster a powerful brand. Researchers agree that a powerful brand has a lasting influence on a companys sales and, according to Aaker (1995) and Alsop (2000), a powerful brand commands high brand equity.

In addition, brand awareness is tied to brand associations, which then may increase the perceived value of the brand. Finally, brands that have a high degree of loyalty can enjoy high power and value. According to Alsop (2000), brand value and brand power are perhaps two of the most important elements in the marketing mix. Brand power is characterised by the distinctive nature of brand personality brand value means different things to different consumers.


Definition of terms

Brand loyal customers

Loyalty to a particular brand refers to a characteristic of customers who will only buy products of that brand rather than switching to an alternative. In Figure 1, the area inside the inner dotted circle represents loyalty. More precisely, a consumer who is loyal indicates his or her previous, current and substitute (future) choices of the same brand with a probability of one (McCarthy et al. 1992).

Potential switchers (or shoppers)

This term refers to all the customers who are not loyal in a particular product market. These customers may be influenced by various factors to switch brands. It is also possible that potential switchers will make repeated purchases of the same brand. In Figure 1, the area outside the inner dashed circle represents these shoppers. More precisely, a consumer who is a potential switcher indicates his or her previous, current and substitute (future) choices of the same brand with a probability of less than one (McCarthy et al. 1992).

Repeat buyers

Some shoppers make the same product choices in the previous, current, and substitute (or future) choice categories. This group, combined with the brand loyal group, are called repeat buyers. In Figure 1, the area inside the solid circle represents repeat buyers.

Brand switchers

A portion of the shoppers will switch products at least once when they make their current or subsequent choices. That is, at least one of the three choices will differ from the remaining choices. The area outside the solid circle represents brand switchers.

Model construct

A model is presented here to evaluate competitive market structures. This model can serve as a basis for designing marketing strategy. It also uses previous and current consumption behaviour to discover the factors relating consumers and brand loyalty, and predict a customers subsequent brand choice. The key factors of brand awareness are identified and the importance of the factors in heterogeneous consumer groups is determined. Then, we derive estimates of the proportion of loyal and brand switching consumers for individual brands and for the entire market using empirical data of previous, current, and future choice decisions in the refrigerator market in Taiwan. Subsequently, we determine how the key factors and their degree of influence differ for repeat buyers and brand switchers. The structure of the model is shown in Figure 2.


Brand parameters

We describe a model that uses three choice data concerning previous, current and future brand choices. Of these three, the first two represent the products actually purchased by the consumer, while the third is a product preference indicated by the consumer due to possible dissatisfaction with the first and second choices, or the impression of a more powerful brand. Next, we utilise the concepts of brand loyalty and brand switching to define the parameters for the research structure, as shown in the appendix. From the calculation process, which is illustrated in the appendix (equation (A.1) to equation (A.10)), we can derive Pi,jk (the proportion of those consumers who currently purchase brand j, and indicate a substitute choice in brand k, given they previously purchased brand i). This is shown in equation (1) below (McCarthy et al. 1992), describing the three choice model.


Pi,jk = { ai + bii,ipi,i(1 ai) for k = j = i
bij,kpi,i(1 ai) for any other case


ai = the proportion of consumers who make their current and substitute choices in submarket i with probability one, given that they previously purchased in the same submarket i;

pi,j = the proportion of shoppers who currently purchase in submarket j, given they previously purchased in submarket i;

bij,k = the proportion of shoppers who indicate their substitute choice in submarket k, given that their previous and current purchases were in submarkets i and j, respectively.

In this equation the values of ai, pi,j and bij,k can be estimated using maximum likelihood.


The maximum likelihood method (MLM) is a general method for estimating a large variety of parameters. There are n random variables, x1, x2, , xn, with the joint density function, f(x;q ). The maximum likelihood method of estimation suggests that we choose as our estimates the values of the parameters that maximise the following likelihood function:


L (q ) = g(X1,X2,...,Xn;q )
= f(x1;q)f(x2;q)...f(xn;q )

where g is a function of q . However, since q is unknown, we estimate it by q  = h(X1, X2, , Xn). q  is the maximum likelihood estimate of q  when q  = q  and L(q ) are maximised. Since the parameters, ai, pi,j and bij,k in equation (1) cannot be calculated from our data, we will estimate them using MLM. Let q  = f(a, p, b) include all the parameters. We can write the likelihood function as follows:


L (q  ) = N N N
P P P P nijk
i=1 j=1 k=1

Using equation (3) as our objective function, we have:


Maximises:  L (q  ) = N N N
P P P P nijk
i=1 j=1 k=1


Constraints:  Spi,j = 1, for all i
  Spi,j = 1, for all i, j; = (a, p, b)

This is a nonlinear programming model. GAMS (General Algebraic Modeling System, Anthony et al. 1992), was used to solve it. The solutions of these estimates are useful to understand the market structure, brand loyalty and brand switching behaviour in the appliances market.

Estimating the proportion of loyalty for the appliance market

We estimate that the proportion of loyalty in the overall market is based on the proportion of loyalty of individual brands. Let LP denote the proportion of loyalty in the entire market; PS, the proportion of potential brand switchers whose previous, current, and future choices are the same, and RP as the proportion of repeat buyers. RP includes LP as well as potential brand switchers. This process of inference is shown in equations (A.12) to (A.15) of the appendix.

Survey design

When the parameters are unclear or can not be directly estimated, the effective sample size can be determined using the Pvalue (Anderson et al. 1990).


n =
Z 2 p (1 p)
(1 a)

where Z =1.96, and  a = 0.05. To be conservative, we set P =0.5, and e = 0.03.

Accordingly, the effective sample size is 1067. We further assume that the response rate is 90% and the effective response rate is 80%. Thus, the minimum number of telephone interviews needed to be conducted is 1482.

We number our population individually based on the order (page number and order) of appearance of names in the telephone directory. The actual samples are selected using a table of random numbers.

The survey questionnaire contains two parts. Part 1 obtains information about consumers previous, current and substitute choices of refrigerators, their opinions on each choice, and the degree of importance of various factors which affect their choice. These seventeen factors derived from the research on brand choices are listed. Part 2 focuses on understanding demographic information such as geographical location, average household monthly income, and the effects of these factors on brand loyalty.


In this section, we apply empirical data from the Taiwanese household refrigerator market to the model. When analysing the market structure we assume that there exist only two types of consumer behaviours: loyalty and potential brand switching. The major refrigerator brands in Taiwan are as follows: National, Kolin, Sampo, Tatung, Sanyo, Tongyuan, Synco, Whirlpool, Westinghouse and Hitachi.

Consumers are asked to identify the key factors which led to loyalty in their previous and current choices. Based on information about these factors, brand parameters are calculated using the maximum likelihood objective function. Brand switching and repeat purchase ratios are then derived for the three choice model. Two properties of the loyal and potential brand switching groups are tested:

  1. the key factors for decision making in each group;

  2. how the factors differ for heterogeneous groups.

Lastly, implications for management are derived.

Key factors affecting consumer brand loyalty

In all, 1600 households were interviewed by telephone, of which 1262 (78.9%) responded. Of these, 1209 (75.6%) were effective. The major factors affecting appliance brand choices are listed in Table 1 in decreasing order of importance. All these factors have a positive effect. After sales service had the greatest influence on brand choice, while advertising had the least. The speed of repairs was more important to the repeat purchasing group than it was to the brand switching group. The period of effective use of the product was the factor on which the two groups differed the most. Based on these findings, we summarise that if firms cannot provide speedy repair services when needed, the degree of brand loyalty will diminish. As one would expect, to increase the level of brand loyalty, service programs must be improved.


Factors for repeat purchase Average value SD
After sales service 4.136 0.800
Useful life 4.072 0.781
Speed of repairs 4.045 0.824
Quality 3.940 0.764
Energy saving features 3.940 0.820
Design excellence 3.893 0.934
Brand reputation 3.846 0.904
Ease of use 3.670 0.888
Aesthetics 3.630 0.867
Extra features 3.501 0.934
Diversity of product line 3.499 0.899
Innovative design 3.439 0.966
Selection of sizes 3.419 0.887
Discounts or promotions 3.404 0.865
Popularity 3.330 0.937
Recommendation by sales staff 3.181 0.943
Advertisement 3.027 1.037

Factor analysis was used to identify the key factors required for a repeat purchase and the results of the factor analysis are summarised in Table 2. The major factors affecting loyalty and switching behaviours were: product attributes, after sale service, marketing capabilities, perceived quality/aesthetics, depth of product line, and brand recognition. These factors explained approximately 68.73% of the total variance. In order to investigate the differences on the perception of factors and variables for the two groups of consumers, a ttest was used, as shown in Table 2. It can be seen that repeat purchasers tended to give significantly higher scores for ease of use, extra features, speed of repairs, recommendation by sales staff, and reliability of quality. The implication for marketing strategy is that improving those factors can increase brand strength.


Factors Subfactors Repeat
Eigenvalue Cumulative t
Product attributes 5.67 33.33% 2.77 0.006**
Ease of use
Innovative design
Design excellence
Extra features
After sales service/usage 1.60 42.76% 0.62 0.537
Useful life
After sales service
Speed of repairs
Marketing capabilities 1.47 51.40% 1.03 0.302
Recommendation by sales staff
Discounts or promotions


Perceived quality/aesthetics 1.09 57.84% 0.68 0.497
Depth of product line 0.94 63.38% 0.47 0.638
Diversity of models
Diversity of sizes
Brand recognition 0.91 68.73% 0.11 0.913
Popularity 3.34 3.32 0.21 0.84

Analysis of brand switching behaviour

Once the key factors that affect brand loyalty are understood, we can estimate brand parameters from the information about previous, current and substitute choices obtained in the interviews. First, the estimation of brand switching is used to derive the related parameters. Then other parameters are estimated using equation (A.11) for q and substituting the known values of nijk into the objective function. Next the parameter estimates are solved by maximising the likelihood function using GAMS software (Anthony et al. 1992). The results are shown in Tables 46.

The market share and the loyalty for each brand are shown in Table 3. We observe that the proportion of loyal customers was high for all brands except Whirlpool (26.3%) and Synco (16.9%). Loyalty was the highest for Kolin (47.1%) and Tatung (46.3%). In addition, the proportion of loyalty for brands with the lowest market share (Westinghouse and Hitachi) were no lower than some brands with higher market share. Thus, we can infer that residents of Taiwan have a high degree of brand loyalty, and that the brand loyalty is not necessarily tied to market share.


Brand National Kolin Sampo Tatung Sanyo Tongyuan Synco Whirlpool Westinghouse Hitachi Others Total
Sample size 132 81 147 153 129 90 84 75 69 63 186 1209
Market share (%) 10.9 6.7 12.2 12.7 10.7 7.4 6.95 6.2 5.96 5.22 15.3 100
Proportion of loyalty (%) 38.1 47.1 30.1 46.3 37.6 30.2 16.9 26.3 31.7 35.2 22.7

In the following section the following topics are explored. How do previous choices affect current choice? How do previous and current choices affect their next choice? How do consumers make their next choice? What is the consumers attitude towards a brand?

How do previous choices affect current choice?

As shown in Table 4, past choices greatly impact current choices. This was especially the case for Tatung, which had the highest proportion of repeat purchasing, but was not true for Synco and Sampo. Among the brand switchers who previously chose Synco, 21.4% switched to Kolin and 17.9% switched to Tatung for their current choice. Among the brand switchers who chose Sampo previously, 16.6% indicated an intention to switch to National. We observed little interbrand switching occurring among the brands with the lowest market shares, such as Westinghouse and Hitachi. For example, Hitachi only has a 5.22% market share, but its proportion of loyalty was 35.2%. It appears that consumers whose favourite brand belongs to the group with the lowest market share are more loyal than consumers of products with higher market shares.


Current choice (j)
pi,j National Kolin Sampo Tatung Sanyo Tong yuan Synco Whirl pool Westing house Hitachi Others



National 0.457 0.023 0.091 0.045 0.068 0.091 0.000 0.068 0.023 0.000 0.023
Kolin 0.114 0.508 0.037 0.114 0.037 0.114 0.000 0.037 0.000 0.000 0.037
Sampo 0.166 0.082 0.347 0.143 0.061 0.061 0.000 0.061 0.000 0.020 0.061
Tatung 0.050 0.120 0.059 0.630 0.039 0.020 0.040 0.040 0.000 0.000 0.000
Sanyo 0.214 0.051 0.028 0.075 0.473 0.028 0.028 0.000 0.028 0.000 0.075
Tongyuan 0.056 0.109 0.0109 0.178 0.000 0.360 0.040 0.000 0.074 0.074 0.000
Synco 0.047 0.214 0.000 0.179 0.000 0.071 0.310 0.036 0.036 0.000 0.107
Whirlpool 0.090 0.040 0.160 0.080 0.000 0.000 0.040 0.350 0.160 0.080 0.000
Westinghouse 0.100 0.043 0.043 0.087 0.000 0.043 0.000 0.130 0.465 0.000 0.087
Hitachi 0.052 0.000 0.052 0.107 0.052 0.052 0.000 0.182 0.000 0.450 0.052
Others 0.130 0.147 0.114 0.147 0.000 0.016 0.000 0.016 0.032 0.032 0.365

How do previous and current choices affect next choice?

From Table 5, we infer that if a customer has made a repeat purchase as a past or current choice, then the probability that he or she will make the same repeat purchase is very high. Among potential switchers, current Whirlpool users had the highest (18%) proportion attracted by National, followed by Tongyuan, with 16% attracted to 'others,' National with 13% attracted to 'others,' and Sampo with 13% attracted to Tongyuan.


Substitute choice

Proportion of repeat purchasing Brand switching preference 1 Brand switching preference 2
Brands bii, i bii,k (k # i) bii,k (k # i)
National 75% Others (13%) Tatung (4%), Sanyo (4%),
Tongyuan (4%)
Kolin 91% Others (9%) No
Sampo 69% Tongyuan (13%) National  (9%)
Tatung 84% Tongyuan (6%), National (6%) Sampo (3%)
Sanyo 80% National (5%), Kolin (5%)
Westinghouse (5%), others (5%)
Tongyuan 84% Others (16%) No
Synco 56% Kolin (11%), Tatung (11%),
Westinghouse (11%), others (11%)
Whirlpool 82% National (18%) No
Westinghouse 82% National (9%), others (9%) No
Hitachi 88% Sanyo (12%) No
Others 72% Kolin (8%), Sampo (8%),
Whirlpool (8%)
Sanyo (4%)

How does a consumer make their next choice?

Let bj,k denote the proportion of consumers who currently buy brand j and indicate their substitute choice to be brand k. The bj,k of our study are shown in Table 6. For consumers whose current brand choice is National, their next brand choices, in order of preference, are 'others' and Tongyuan. The proportion of repeat purchasers is only 0.154. For those whose current brand choice is Kolin, their next brand choices, in order of preference, are National and 'others.' For consumers whose current brand choice is Sampo, their next brand choices, in order of preference, are Whirlpool and Tongyuan. Those whose current brand choice is Tatung tend to choose National next. For those whose current brand choice is Sanyo, their next brand preference is Sampo. Those whose current brand choice is Tongyuan will choose Kolin next. Lastly, those whose current brand choice is Synco have a preference for Kolin next. Customers of brands with low market shares, such as Whirlpool, Westinghouse and Hitachi, indicate a low preference for brand switching. The customers of brands other than Synco and Westinghouse demonstrate inferior brand awareness for those two brands.


Current choice (k)
pi,j National Kolin Sampo Tatung Sanyo Tong yuan Synco Whirl pool Westing house Hitachi Others



National 0.154 0.042 0.173 0.076 0.024 0.208 0.030 0.056 0.000 0.000 0.238
Kolin 0.156 0.101 0.108 0.023 0.103 0.114 0.049 0.023 0.099 0.099 0.123
Sampo 0.107 0.125 0.071 0.054 0.099 0.215 0.008 0.220 0.031 0.031 0.043
Tatung 0.186 0.155 0.100 0.123 0.079 0.030 0.055 0.056 0.006 0.101 0.110
Sanyo 0.076 0.106 0.193 0.072 0.146 0.041 0.041 0.041 0.045 0.132 0.106
Tongyuan 0.190 0.220 0.069 0.099 0.099 0.115 0.054 0.008 0.008 0.039 0.099
Synco 0.050 0.151 0.050 0.060 0.140 0.050 0.191 0.140 0.060 0.050 0.060
Whirlpool 0.214 0.107 0.047 0.017 0.017 0.017 0.017 0.182 0.077 0.047 0.259
Westinghouse 0.132 0.033 0.079 0.033 0.033 0.079 0.124 0.101 0.107 0.033 0.246
Hitachi 0.050 0.050 0.050 0.050 0.106 0.095 0.050 0.140 0.050 0.129 0.231
Others 0.085 0.123 0.154 0.055 0.059 0.025 0.025 0.032 0.070 0.146 0.226

Probability of repeat purchasing

If customers, whether loyal or potential switchers, purchase the same brand again, it is called repeat purchase behaviour. It can be summarised by equation (A.11) (see appendix).


LP + PS = ai + bii,ipi,i(1 ai) (6)

where LP denotes the proportion of loyal customers, PS denotes the proportion of brand switchers, ai is the proportion of loyal customers who will buy the brand again, and bii,ipi,i(1 ai) is the proportion of brand switchers who will buy the same brand again. The proportions of repeat purchasers are shown in Table 7.


National Kolin Sampo Tatung Sanyo Tong yuan Synco Whirl pool Westing house Hitachi Others
PS 0.212 0.245 0.167 0.284 0.241 0.211 0.143 0.212 0.260 0.26 0.20
RP 0.593 0.715 0.468 0.747 0.617 0.513 0.312 0.475 0.577 0.61 0.43
PS: the proportion of brand switchers
RP: the proportion of repeat purchasers

The proportion of repeat purchasing for the entire market exceeds 50% for nearly all brands while that for potential switchers is only 20%.


LP = S m ai msi

Next we used the above equation (7) to calculate the proportion of repeat purchasing for loyal customers. For example, if LP=33.12%, it means that for a market with 100 consumers, 33.12% of the consumers are loyal. We can then calculate PS from the sum of the weighted proportions of LP for the available brands. In our study, PS=21.99% for the market. PS of brand i is the proportion of nonloyal consumers who currently purchase brand i and indicate a preference for brand i as the next choice.

Finally, we calculated the proportion of repeat purchasers using equation (A.15) (see appendix) and obtained RP=55.11%. This indicates that over 50% of all customers in the market have either repurchased the same brand or plan to repurchase it in the future. This also indicates that the refrigerator market has reached its mature stage, where there is little differentiation in product attributes. Consumer brand choice is thus greatly influenced by the level of satisfaction with a brand. This information is useful for planning marketing strategies.

From the above results, we deduced that the degree of brand loyalty among refrigerator buyers in Taiwan is high. Brand image and attitude towards a brand will influence future brand choice substantially. In addition, the proportion of consumers repurchasing the same brand is quite high.


In the appliance market, consumers are paying increasing attention to the after sales service and quality of different brands. In response to such changes, firms need to refocus their marketing efforts. In particular, manufacturers need to deliver more service values, providing speedy responses to consumers questions and requests for repairs. As for quality, we find that consumers perceived value consists of price versus quality and image.

In our analysis of repeat purchasing and brand switching behaviours (Table 2), it is shown that the most important factors influencing repeat purchasing are the following: ease of use, innovative design, design excellence, and extra features. By paying attention to these factors increasing product value through innovative technologies, thoughtful design, and unusual features manufacturers can create a good brand image. Moreover, it was found that the only key factor that brand switchers pay more attention to than repeat buyers is the depth of product line. Although the difference is not significant, a lack of a diversity of sizes and models could lead to brand switching. In addition to offering a range of models, manufacturers need to communicate the features of their products through advertising.

In the brand switching analysis (see Tables 3 and 7) it was found that while there is a moderate degree of brand loyalty in the Taiwan refrigerator market, there is also a considerable amount of repeat purchasing behaviour exhibited by nonloyal consumers. Firms need to focus on ways of retaining these potential converts.

Another interesting finding is that a weaker brand image usually correlates with a low degree of brand loyalty and that brand choice is affected by past experience (see Tables 3 and 5). The implication here is that firms need to build a powerful brand image and increase consumer satisfaction in order to increase brand loyalty. This may be achieved through prompt response to consumer feedback.

Although the data were collected from southern Taiwan, the results of the study may be useful formulating marketing policy in developing countries or countries whose DNP is close to Taiwan, especially those in the AsiaPacific Rim. Furthermore, the approach implemented here can be adapted to study brand loyalty and switching behaviour in any country.


We describe a model that uses a three choice data structure: previous, current and next choices. Of these, the first two are the products actually purchased by the consumer, while the third is a product preference indicated by the consumer due to possible dissatisfaction with the first two choices or the impression of a more powerful brand (McCarthy et al. 1992).

Suppose there are n brands in the refrigerator market and x indicates the customer type: loyal customer (x = l) and potential brand switcher (x = s). A1 and A2 represent previous and current choices, respectively. Equation (A.1) describes the loyal customer and brand switcher proportions for the previous choice:


Loyal: p(x = l |A1= i) = ai

Potential switchers: p (x = s | A1= i) =1 ai 

For the current choice, the probabilities are:


Loyal customer:  p(A2 = j | A1 = i, X = l) = { 1    if i = j
0    if i j


Potential brand switchers: p(A2 = j | A1 = i, x = s) = pi,j

Combining equations (A.1A.3), we derive that the proportion of those customers who currently purchase brand j, given that they had previously purchased brand i:


pi,j { aipi,j (1 ai) if j = i
 pi,j (1 ai) if j i

This is the equation for the two choice model. Next, we describe the three choice model. In this model, the consumer becomes dissatisfied with the current choice of product and considers alternative products. First, let s represent the substitute choice and k (= 1,2,3,,n) denote the brand selected by the consumer. For the loyal customer, the probability of selecting brand k as the substitute choice is:


p(s = k | A1 = i, A2 = j, x = l) = { 1        if k = j = i
0        otherwise

The proportion of potential brand switchers who indicate their substitute choice as brand k, given that their previous and present choices are brands i and j, respectively, is:


p(S = k |A1= i, A2 = j, x = s) = bij,k 

Next, let aii denote the proportion of those customers who make brand i their previous and current choice:


p(x = l | A1 = i, A2 = j) = { aii if j = i
0 if j i

Symmetrically, the proportion of potential switchers who currently choose brand j, given that they had previously purchased brand i, is:


p(x = s | A1 = i, A2 = j) = { 1 aii if j = i
1 if j i

Combining equations (A.5), (A.6), (A.7) and (A.8), we derive that the proportion, pij,k, of those consumers who make brands i and j their previous and current choices and indicate brand k as their substitute choice is as follows:


pij,k = { aii + bii,i  (1 aii) for i = j = k
bii, (1 aii) for i = j k
bij,k for i j and k

According to our previous definition, ai can be written as ai = aii pi,i, and thus aii = ai/Pi,i. We can also substitute equation (A.4) for Pi, and obtain an equation for Pi,jk, which is the proportion of those consumers who previously purchased brand i, and would most likely make brand j and k their current and substitute choices as:


pij,k = { ai + bii,i pi,i  (1 ai) for k = j = i
bij,k pi,i (1 ai) otherwise

This is called the three choice model. In equation (A.10), the parameters on the right hand side, ai, pi,j and bij,k cannot be measured or calculated directly. Thus, we use the maximum likelihood method to estimate these parameters. We can write the likelihood function as equation:


Maximises: L (q ) = N N N
P P P P nijk
i=1 j=1 k=1


Constraints: Spi,j = 1, for all i
Spi,j = 1, for all i, j; q = (a, p, b)

The solution for the above equations, which are curvilinear, requires a nonlinear method. We used GAMS (General Algebraic Modelling System, see Anthony et al. 1992) to solve the parameter estimates. These estimates are useful for understanding the market structure, brand loyalty behaviour, and brand switching behaviour in the appliance market.

When the market contains m brands, and n is the sample size; let ni be the number of respondents whose previous choice is brand i, and let msi be the market share rate of brand i.


msi = ni , (i = 1, 2, ...,m)

LP, the proportion of loyal customers in the entire market, is given as:


LP  m ai msi

PS, the proportion of potential brand switchers whose previous, current and next choices are the same, is given as:


PS  m bii,i pi,j (1 ai) msi

RP, all the consumers whose previous, current and next choices are the same, is given as:


RP = LP + PS  m aimsi  m bii,i pi,j (1 ai) msi
i=1 i=1


Aaker, D.A. (1995). Building Strong Brands. New York: Free Press.

Aaker, D.A. & Joachimsthaler, E. (2000) Brand leadership. Brandweek, New York.

Allenby, G.M. (1989) A unified approach to identifying, estimating and testing demand structures with aggregate scanner data. Marketing Science, 8, 3, pp. 265280.

Alsop, R. (2000) Blue chips lose brand power to lowertier firms, survey says. Wall Street Journal, B8.

Anderson, D.R., Sweeney, D.J. & Williams, T.A. (1990) Statistics for Business and Economics. St. Paul: West Pub. Co.

Anirudh, D. (1994) Durable goods monopolists, rational consumers, and improving products. Marketing Science, 13, 1, pp. 100120.

Anthony, B., Kendrick, D. & Meeraus, A. (1992) GAMS: A Users Guide, Release 2.25.

Baldinger, A. & Rubinson, J. (1996) Brand loyalty: the link between attitude and behavior. Journal of Advertising Research, 36, 6, pp. 2234.

Bass, F.M., Givon, M.M., Kalwani, M.U., Reibstein, D. & Wright, G.P. (1984) An investigation into the order of the brand choice behavior. Marketing Science, 3, Fall, pp. 267287.

Blumen, I., Kogan, M. & McCarthy, P.J. (1955) The industrial mobility of labour as probability process. Cornell Studies of Industrial and Labor Relations, 6. New York: Cornell University Press.

Chaudhuri, A. (1999) Does brand loyalty mediate brand equity outcomes? Journal of Marketing Theory and Practice, pp. 136146.

CobbWalgren, F.J., Ruble, C.A. & Donthu, N. (1995) Brand equity, brand preference, and purchase intent. Journal of Advertising, 24, 3, pp. 2540.

Colombo, R.A. & Morrison, D.G. (1989) A brand switching model with implications for marking strategies. Marketing Science, 8, 1, pp. 8999.

Dabholkar, P.A. (1994) Incorporating choice into an attitudinal framework: analysing models of metal comparison processes. Journal of Consumer Research, 21, 1, pp. 100118.

De Chernatony, E. & McDonald, M.H.B. (1998) Creating Powerful Brands in Consumers, Service and Industrial Markets. Oxford: Butterworth and Heinemann.

Dyson, P., Farr, A. & Hollis, N.S. (1996) Understanding, measuring, and using brand equity. Journal of Advertising Research, 36, 6, p. 921.

Ehrenberg, A.S.C., Barnard, N. & Scriven, J. (1997) Differentiation or salience. Journal of Advertising Research, 37, 6, pp. 8291.

Girish, P. (1987) Presearch decision making in consumer durable purchases. The Journal of Consumer Marketing, 4, 1, pp. 7182.

Gregory, C. & Lehmann, D.R. (1985) A model of marketing mix, brand switching, and competition. Journal of Marketing Research, 22, 3, pp. 318329.

Grover, R. & Srinivsan, V. (1992) Reflections on a simultaneous approach to market segmentation and market structuring. Journal of Marketing Research, 29, 4, pp. 474476.

Jain, D., Bass, F.M. & Chen, Y.M. (1990) Estimation of latent class models with heterogeneous choice probabilities: an application to market structuring. Journal of Marketing Research, 27, 1, pp. 94101.

Jeuland, A.P., Bass, F.M. & Wright, G.P. (1980) A multibrand stochastic model compounding heterogeneous Erlang timing and multinomial choice process. Operation Research, 28, 2, pp. 253277.

Kalwani, M.U. & Morrison, D.G. (1977) A parsimonious description of the hendry system. Management Science, 23, 5, pp. 467477.

Kannan, P.K. & Wright, G.P. (1991a) Modelling and testing structured markets: a nested logic approach. Marketing Science, 10, 1, pp. 5882.

Kannan, P.K. & Wright, G.P. (1991b) On testing competitive market structure. Marketing Science, 10, 4, pp. 338347.

Kapil, B. (1990) Modeling inertia and variety seeking tendencies in brand choice behavior. Marketing Science, 9, 3, pp. 263287.

Keaveney, S.M. (1995) Customer switching behavior in service industries: an exploratory study. Journal of Marketing, 59, 2, pp. 7182.

Kotler, P. (1995) Marketing management analysis, planning, implementation, and control, 8th edn. New York: Prentice Hall.

McAlister, L. (1982) A dynamic attribute satiation model of variety seeking behavior. Journal of Consumer Research, 9, 2, pp. 141150.

McCarthy, P.S., Kannan, P.K., Chandrasekharan, R. & Wright, G.P. (1992) Estimating loyalty and switching with an application to the automobile market. Management Science, 38, 10, pp. 13711393.

Padberg, D.I., Walker, F.E. & Kepner, K.W. (1974) Measuring consumer brand preference. Journal of Farm Economics, 56, 2, pp. 3847.

Rao, V.R. & Sabavala, D.J. (1981) Inference of hierarchical choice processes from panel data. Journal of Consumer Research, 8, 1, pp. 8596.

Rubinson, J.R., Vanhonacker, W.R. & Bass, F.M. (1980) On a parsimonious description of the hendry system. Management Science, 26, 2, pp. 215226.

Russo, J.E., Meloy, M.G. & Medvec, V.H. (1998) Predecisional distortion of product information. Journal of Marketing Research, 35, 4, pp. 438452.

Shoham, A., Rose, G.M. & Kahle, L.R. (1998) Marketing of risky sports: form intention to action. Academy of Marketing Science Journal, 26, 4, pp. 307321.

Weerahandi, S. & Moitra, S. (1995) Using survey data to predict adoption and switching for services. Journal of Marketing Research, 32, 1, pp. 8596.

Yi, Y. & Vanden Abeele, P. (1989) An investigation of the structure of expectancy-vale attitude and its implications. International Journal of Research in Marketing, 6, 2, pp. 7192.


Chinho Lin

Chinho Lin is a Professor of Industrial Management Science at National Cheng Kung University in Taiwan. He holds a BS and MS in industrial management from national Cheng Kung University, an MS in Operations Research from Columbia University and a PhD in business administration from the City University of New York. His research interests include business research method simulations, quality management, inventory control, applications of AI and manufacturing strategy. He has published in Decision Sciences, Information and Managament, TQM, International Journal of Production Research, Journal of the Operational Research Society, Applied Mathematical Letters and elsewhere.

Zhi�Feng Wang

Zhi�Feng  Wang is a masters student in the Department of Industrial Management Science at National Cheung Kung University in Taiwan.

Wann�Yih Yu

Wann�Yi Yu is a Professor and the Chairman of the department of Business Administration at National Cheung Kung University. He received his PhD in Business Administration from Oklahoma University. His research interests include business research methods and international management. he has published in International Business Review, Cross Cultural Management, Journal of Global Marketing and Journal of Midwest Marketing.

Figure 1: market structure of repeat purchasing and brand switching