Dr. Mitja Pirc
A.T. Kearney Slovenia
mitja.pirc@atkearney.com
USING INTENTIONS AND ExPECTATIONS PERSPECTIVES TO EXPLORE THE INFLUENCE
of determinants
OF LOYALTY
Abstract: The purpose of this paper is to extend our understanding of what are the roles of determinants of loyalty. A proposition is made that there is a difference in whether the research questions are asked with intentions or expectations perspective, which trigger different cognitive processes about the future state of exchange relationship. We find that the intentions questions create an inward-looking perspective and focus respondents on factors such as satisfaction and attitude towards switching. The expectation questions however trigger a more outward-looking perspective and create a focus on the factors beyond customers' control, such as trust and switching costs. Implications are two-fold. First, researchers interchangeably use intentions and expectations measures, which can lead to misleading findings about which determinant is important for maintaining loyalty. Second, satisfaction is an inward-looking determinant, meaning that loyal customers are typically satisfied; however satisfied customers are not necessarily loyal. Trust is an outward-looking determinant and therefore high levels of trust are actually the ones that keep clients loyal. Key words: customer behavior and decision making, customer loyalty, determinants of customer loyalty
1. INTRODUCTION
The field of marketing has come to a consensus that understanding and maintaining customer loyalty is critical for companies' financial performance (Reichheld, 1996; Zeithaml et al., 2006). However, there is still a lack of agreement about the influence of determinants of loyalty and various have been proposed as central, for example, satisfaction, trust, and switching costs (Szymanski and Henard, 2001; Morgan and Hunt, 1994; Burnham, Frils, and Mahajan, 2003). When managers aim to maintain customer loyalty, they are faced with many candidates for the focal determinant. For example, they could decide to invest in keeping customers satisfied or could decide to design contracts and procedures to increase customer switching costs. However, the determinants chosen by managers differ in how they influence loyalty. Therefore, a misunderstanding, about what roles determinants play, can lead managers to focus on the less relevant determinants and use resources without achieving desired results. In this paper we propose a way to distinguish how different determinants influence customer loyalty, which could represent a competitive advantage for companies.
When studying customer loyalty and its determinants, researchers frequently measure loyalty by asking customers to forecast their future behavior. There are two ways of forming questions about customer forecasts that are commonly used: intentions (e.g., "Do you intend to be loyal to a specific provider?") and expectations (e.g., "Do you expect to be loyal to a specific provider?"). Previous research about individual forecasting has suggested that asking about intentions or expectations differs in terms of how individuals think about future events and behaviors (Warshaw and Davis, 1985). Intentions measure a conscious intention based on determinants which are under the respondent's control, such as motivation, attitudes, abilities, or beliefs. Expectations on the other hand measure a self-prediction of one's own future, which takes into account factors beyond the individual's control. These external factors include the ease or difficulty of performing the behavior as well as the anticipated obstacles.
The following example further illustrates the difference between intentions and expectations: "I intend to lose weight" and "I expect to lose weight". Individuals might have a motivation or a reason to lose weight which is expressed in the intentions form. However, they are also aware that there are influential factors beyond their
control, which are expressed in the expectations form. These external factors additionally prevent individuals from losing weight and include eating out as well as choices of other people. Using intentions or expectations question triggers different cognitive processes and consequently consumers use different sources of information to construct forecasts of future behavior (Sheppard, Hartwick, and Warshaw, 1988; Bettman, Luce, and Payne, 1998).
We build on this difference between intentions and expectations to explore the roles of determinants in forecasting loyalty. A study is designed to measure loyalty and its determinants. As measures of customer loyalty we use both the loyalty intentions questions and the loyalty expectations questions. The selected determinants are four commonly used predictors of loyalty: customer satisfaction, trust, attitude toward switching, and financial switching costs. We hypothesize that these determinants differ in terms of what control consumers have over them, with inward-looking determinants having a higher level of control and outward-looking determinants having a lower level of control. Exposing consumers to the intentions and expectations perspectives enables us to test the inward or the outward-looking nature of determinants.
In the next section, we present in more detail the conceptual difference between the intentions and expectations perspective. Further, we discuss selected determinants (satisfaction, trust, switching costs, and attitude towards switching) and their effects on customer loyalty. Based on previous findings we develop hypotheses about how these determinants influence loyalty intentions and loyalty expectations. Next, the research design is presented with a survey as a means of collecting multiple items for each of the variables used. Based on the hypotheses, a structural equation model is proposed and tested with other models for the effects of the determinants. In the end, conclusions, managerial implications, and limitations are discussed.
2. THEORY AND HYPOTHESES
2.1. CUSTOMER LOYALTY AND ITS DETERMINANTS
A number of different determinants of loyalty have been proposed in the literature, grounded in theory and supported with empirical studies. Determinants of loyalty can be divided into two groups, the perceptual determinants (e.g., satisfaction, trust) and the behavioral determinants (e.g., the number of items
purchased). The group of perceptual determinants has recently received an increased attention compared to the behavioral determinants. There is also a consensus in the field of marketing, that the effects of the behavioral determinants are mediated through the perceptual ones. In addition, hypotheses about the effects of perceptual determinants can be generalized across product and service categories. However, in many situations the behavioral determinants are the only data managers have about their customers, collected either through loyalty programs or information systems supporting operational processes.
Despite numerous studies, we still do not have a full understanding of how determinants influence loyalty and which are the central ones that managers should focus on (Brady et al., 2005). In this study, we have selected a group of four commonly used perceptual determinants: customer satisfaction, trust, customer switching costs, and attitude towards switching (Szymanski and Henard, 2001; Morgan and Hunt, 1994; Burnham et al., 2003; Bansal and Taylor, 1999). The aim is to further explore the nature and influence of these selected determinants. In the following sections, the intentions and expectations perspectives are presented, based on which we discuss each of the determinants and develop related hypotheses.
2.2. THE INTENTIONS AND ExPECTATIONS PERSPECTIVE
As mentioned earlier, we use two different forms of questions to measure consumers' loyalty forecasts: intentions and expectations. The purpose is to expose consumers to different perspectives on loyalty, which are hypothesized to influence the process of constructing forecasts. The differing effects of these two forms are based on a Warshaw and Davis's (1985) proposition that intentions and expectations differ as measures of forecasts. Intention is defined as a statement of conscious intention, while expectation is a self-prediction of one's own future behavior. Apart from the weight loss example in the previous section, further examples can be found with regards to consumer decisions. When consumers say that they intend to switch their internet service provider or change to a new car, they focus on their current situation and the reasons for the intended action. However, when they express their expectation about switching the internet service provider or a changing to a new car, they take into account additional factors. An example of a positive external factor could be a potential attractive
offer from a competitive provider; while an example of a negative external factor might the information about the additional cost (e.g., taxes) associated with purchasing a new car.
Individuals can therefore take two different perspectives on their future behavior. Within the intentions perspective they tend to focus on a conscious intention to perform future behavior. The resulting forecast is mainly based on the factors under the respondent's control, such as the reasons (motivation) to perform the behavior and the attitude towards performing the behavior (Warshaw and Davis, 1985; Ajzen, 1991). We shall call these the inward-looking factors. The expectations perspective is however based on a cognitive appraisal of the factors which are beyond respondent's control. Examples of these outward-looking factors include the obstacles and risk associated with performing the behavior (Warshaw and Davis, 1985; Ajzen, 1991).
The loyalty intentions (e.g., "I intend to be a loyal customer.") and the loyalty expectations (e.g., "I expect to be a loyal customer.") are frequently and interchangeably used to measure customers' forecast of their future loyalty behavior. Based in earlier discussion, the two measures of loyalty forecasts are proposed to be different constructs. H1: Loyalty intentions and loyalty expectations are different constructs.
2.3. THE ROLES OF DETERMINANTS OF
loyalty
As discussed in the previous section, loyalty intentions and loyalty expectations cause consumers to view future behavior in different ways. When forming a loyalty intention (LI) forecast, consumers focus more on the inward-looking determinants over which they have a higher level control. On the contrary, when forming a loyalty expectation (LE), forecasts focus more on the outward-looking determinants with a lower level of control. By acknowledging the inward and the outward nature of determinants, we can develop hypotheses for each of them about how well they predict loyalty intentions and expectations.
Customer satisfaction has been extensively studied in the marketing literature (Fornell, 1992; Oliver, 1997; Garbarino and Johnson, 1999) and has been shown to be an important determinant, positively related to loyalty. These findings have achieved a high level of awareness among practitioners and it is frequently claimed that satisfied customers are more loyal. We base the definition of customer satisfaction on Johnson and Fornell (1991) as the customer's overall
evaluation of the performance of an offering to date. Satisfaction is a judgment, constructed and internalized based on consumers' personal experience with a service or a product. It is an evaluation of the past performance and as such it does not take into account future external factors. Bansal and Taylor (1999) as well as Soderlund and Ohman (2005) have proposed that satisfaction is an entity with a relatively high level of control and thus an inward-looking determinant. Based on the discussion in the preceding section, customer satisfaction should have more weight in the intentions perspective compared to the expectations perspective. H2a: Satisfaction predicts loyalty intentions better than loyalty expectations.
Attitude towards switching behavior has been frequently proposed as a determinant of loyalty (Ajzen and Fishbein, 1980; Bansal and Taylor, 1999). It is defined as the degree to which a person has a favorable or an unfavorable evaluation or appraisal of the behavior in question (Ajzen and Fishbein, 1980). Bansal and Taylor (1999) have found support that attitude towards switching behavior is conceptually different from perceived behavioral control. Attitude towards future behavior therefore represents the individual's own attitude and is less based on external factors. Ajzen (1991) has proposed that attitude toward this behavior is more inward-looking and a better predictor of intentions. The hypothesized distinction in predicting the intention and the expectation forecast is thus similar to the one expressed for satisfaction. H2b: Attitude towards switching predicts loyalty intentions better than loyalty expectations.
Trust is another determinant of loyalty which has recently received much attention and has been claimed to be a better predictor of loyalty than satisfaction (Morgan and Hunt, 1994; Burnham, Frels, and Mahajan, 2003). It is defined as the expectation held by customers that the provider is dependable and can be relied on to deliver on promises (Morgan and Hunt, 1994). When asked about the level of trust in a specific company, customers imagine future events and how the company would keep its promises or resolve problems that could occur. Trust is therefore an evaluation of the company's future performance and explicitly includes factors that go beyond the control of the respondent. We can consider trust as a positive determinant of loyalty with an outward-looking nature, which is more closely linked to the expectations perspective. H2c: Trust predicts loyalty expectations better than loyalty intentions.
Switching costs are related to the act of switching the provider and have been shown to influence loyalty positively (e.g., Burnham et al., 2003). Various forms of switching costs have been discussed in the literature: procedural, psychological, financial, relational, and legal (Burnham et al., 2003). In this study, the focus is on the perceived financial switching cost, which acts as a barrier for customers, decreases their switching behavior, and therefore increases their loyalty. Examples of financial switching costs are penalty fees or costs related to setting up a new service or purchasing a new product. Similarly, when switching a retailer, consumers can face an increase in the transportation cost. Financial switching costs are therefore hypothesized to be an outward-looking factor which carries more weight in the expectations perspective. H2d: Financial switching costs predict loyalty expectations better than loyalty intentions.
In this paper, we are interested in the relation between the loyalty intention, the loyalty expectation, and their relations with determinants of loyalty. The link between loyalty forecasts and actual behavior is not studied here. This has been explored and empirically tested before by Sheppard, Hartwick and Warshaw (1988), who have shown in their meta-analysis that expectations are better predictors (in terms of R2) of actual behavior than intentions.
3. STUDY ANALYSIS AND RESULTS
This section presents the study developed to test for the proposed hypotheses. First, the research design is presented together with the measures used for the constructs. Next, the measurement model is tested for convergent validity, dimensionality, and discriminant validity. Finally, we test whether determinants differ in how well they predict loyalty intentions and expectations.
3.1. RESEARCH DESIGN
The data were collected using a paper and pencil questionnaire distributed to undergraduate students at a university in a metropolitan area. The students were approached when leaving the classroom and were asked to fill out the questionnaire (found in Appendix at the end of the document) for academic purposes. The participants were offered a financial incentive in form of a lottery, in which 5 participants were randomly chosen to receive a prize of 10 euros each. There were in total 187 questionnaires distributed and 117 were returned. Three of the returned questionnaires were not completed and thus the final dataset had 114 observations. The
Table 1: Items used for constructs and tests of the measurement model
Items		Loading	Error	C	a	VE	VS
Switching Intention							
I1	I plan to change my mobile services provider.	.81	.24	.77	.88	.73	.56
I2	I intend to start using another provider.	.69	.35	.65			
I3	I will change my provider within the next year.	.80	.23	.78			
Switching Expectation							
E1	I expect to change my provider in the future.	.78	.22	.77	.90	.73	.47
E2	How likely is that you will stay with your current mobile phone services provider?	.76	.24	.76			
E3	How likely is that you will change it in the future?	.76	.28	.71			
Satisfaction							
S1	How satisfied or dissatisfied are you with your current mobile phone services provider?	.83	.16	.84	.93	.80	.71
S2	Overall, how do you feel about it?	.86	.11	.89			
S3	How well does it meet your needs at this time?	.80	.29	.70			
Trust							
T1	I feel can trust my mobile phone services provider.	.82	.23	.77	.88	.71	.71
T2	My provider is responsive to customers' problems.	.70	.34	.65			
T3	I feel my mobile services provider is reliable.	.75	.26	.75			
Attitude towards switching							
A1	It would take me a lot of effort to choose another mobile services provider.	.64	.29	.56	.84	.61	.56
A2	I do not feel like going through the whole process of changing the mobile provider.	.76	.31	.70			
A3	Going through the process of changing to another provider would be an unpleasant experience.	.67	.32	.60			
Perceived financial switching cost							
F1	Changing my mobile provider would cost me money.	.83	.21	.84	.90	.79	.31
F2	If I choose another services provider my costs will go up.	.80	.19	.77			
F3	Changing my mobile service provider would have significant financial impact for me. *	.34	.11	.32			
Satorra-Bentler scaled %2 = 161.17 (104 df); p-value = .000; NFI = .88; NNFI = .93; CFI = .95; RMSEA = .071 (90 % interval: .048-.091); * - item F3 is not used in the measurement model
C= Communality; a = Cronbach Reliability; VE=Variance Extracted; VS= Maximum Variance Shared (Fornell & Larcker test)
survey questions were about mobile phone services, which were used by all the respondents. A pre-study had been done with 32 graduate students in order to ensure the clarity of items used.
The dependent variables were consumers' forecasts of switching their mobile provider in the future. The items for the constructs of switching intentions and switching expectations were measured on a ten-point scale and based on Warshaw and Davis (1985), Zeithaml, Berry, and Parasuraman (1996), and Soderlund and Ohman (2005). Satisfaction was measured using items from Fornell (1992). Attitude toward the switching was based on Bansal and Taylor (1999). Trust was measured in line with Morgan and Hunt (1994). Switching costs measured in financial terms were based on the work of Bansal and Taylor (1999), Burnham, Frels, and Mahajan (2003), and Jones, Mothersbaugh, and Beaty (2002). For all determinants of loyalty a seven-point scale was used. All the variables were cognitive, subjective variables and therefore each was measured using three items (Table 1). Table 1 also includes tests of the measurement model.
3.2. MEASUREMENT MODEL
The variables of interest in this study are perceptual determinants, collected through a survey and each based on multiple items. Testing the relations between variables is based on a structural equation modeling. The difference in the switching intentions and the switching expectations constructs (hypothesis H1) is explored by testing the measurement model: convergent validity test, dimensionality test, and discriminant validity test (Kline, 1998). Testing the hypotheses H2a to H2d is based on comparing how selected determinants influence different dependent variables (intentions and expectations). The comparison is based on the fit of structural models.
The correlations of the constructs are reported in Table 2. The exploratory factor analysis is done, using the principal component analysis and the varimax rotation. The Cronbach's alpha supports the reliability of the constructs with values between .84 and .93 (Table 1). The F3 item for the perceived financial switching cost construct has a low loading (.34) and a low communality (.32) compared with other items in the factor analysis. The measurement model without the item F3 has a significantly better fit and this item is removed from further analysis. By examining the items F1, F2, and F3, we can observe that the items F1 and F2 ask respondents about the absolute impact of switching costs, while F3 asks about the impact relative to overall wealth. The switching costs in absolute terms seem to matter for consumers, however compared to their overall wealth they could be perceived as not so relevant. This could explain low loadings of F3.
The convergent validity is tested with a confirmatory factor analysis including all the constructs (Bagozzi, Yi, and Philips, 1991). A common method factor is included as a means of accounting for random and systematic errors (Podsakoff et al., 2003). With regards to testing structural equation models, there is no consensus about which test is best for evaluating the fit of the model. Several different tests are therefore used jointly as advised by Marsh (1994). We use a chi-square test with a scaling correction to improve the approximation of the goodness-of-fit test statistics (Satorra and Bentler, 1988). An acceptable model fit is indicated when a chi-square value divided by degrees of freedom is less than 5. Next the following fit indices are also used with acceptable fits indicated by specific values: NNFI (Non-Normed Fit Index) and CFI (Comparative Fit Index) values exceeding .90 and a RMSEA (Root Mean Square Error) value less than .08 (Marsh, 1994). In Table 1 we can observe that the measurement model has a satisfactory
Table 2: Correlations of study constructs (the data are standardized)
	X1	X2	X3	X4	X5
X1 - Switching Intention					
X2 - Switching Expectation	.67**				
X3 - Satisfaction	-.83**	-.47**			
X4 - Attitude towards switching	-.20*	-.26*	-.43**		
X5 - Trust	-.78**	-.56**	.83**	-.32**	
X6 - Perceived financial switching cost	-.45**	-.65**	.11	.20*	.05
* - p < .05; ** - p < .01
fit: Satorra-Bentler scaled chi-square = 161.17 (104 df); NFI = .88; NNFI = .93; CFI = .95; RMSEA = .071 (.048-.091). Both the exploratory and the confirmatory factor analysis support the convergent validity of constructs.
The discriminant validity between variables exists when there is a low correlation between items measuring different variables. The following test, proposed by Fornell and Larcker (1981), is used. For each of the variables we calculate the variance extracted (VE) by the items used to measure this specific variable. Next, the variance shared (VS) of a specific variable is the highest variance shared with any other variable used in the analysis. When for a specific variable the variance extracted (VE) is larger than the variance shared (VS), then it is sufficiently different from other variables. From Table 1 we can conclude that all the variables satisfy the discriminant validity test.
The validity of the construct aggregation is tested with the dimensionality test (Bagozzi and Edwards, 1998). Two structural models are compared; one with the switching intentions and the switching expectations modeled as a single construct (Figure 1) and another, where they are modeled as two different constructs (Figure 2). Both models also include all the other variables.
Comparing the fit of structural equation models is based on the standard test of a chi-square difference (Anderson and Gerbing, 1988; Brady et al., 2005). The chi-square difference is the chi-square fit statistic for one model minus the corresponding value for the second model. The difference in the chi-square statistic is evaluated against the corresponding difference in the degrees of freedom. If the chi-square difference is
Figure 1: One factor model
significant, then there is a difference in the model fit.
The single-factor model (Figure 1) has the following fit: Satorra-Bentler scaled chi-square = 250.80 (109 df); NFI = .79; NNFI = .83; CFI = .87; RMSEA = .11 (.090 -.125). The two-factor model (Figure 3.2) has the following fit: Satorra-Bentler scaled chi-square = 161.17 (105 df); NFI = .88; NNFI = .93; CFI = .95; RMSEA = .071 (.048-.091). The two-factor model fits the data better as the chi-square difference is significant (chi-square difference = 89.63; difference in degrees of freedom = 4; p < .01). The measurement model is therefore supported by all three tests: the convergent validity test, the dimensionality test, and the discriminant validity test.
3.3. STRUCTURAL PATH MODEL wITH DETERMINANTS OF LOYALTY
The difference between the switching intentions and the switching expectations constructs is further tested using determinants of loyalty as the explanatory variables. Three structural models are explored, the first one with a single dependent variable (Figure 3), the second one with two dependent variables (Figure 4), and the third model with two dependent variables, however including the effect that the intentions variable has on the expectations variable (Figure 5). In all three models four explanatory constructs are used: satisfaction, trust, perceived financial switching cost, and attitude towards switching.
In Table 1 the comparison of models is done with corresponding fit indices. As we can observe the best model is the simple two dependent variables model shown in Figure 4. Using the difference in chi-square and the difference in degrees of
Figure 2: Two factor model
E2
E3
Switching Expectations
Figure 5: The model with two dependent variables, where intentions influence expectations
Table 3: Testing models for the difference in the path coefficients
Models	Fit indices					Model fit comparison
	df	X 2	NNFI	CFI	RMSEA	
Single factor (Figure 3)	109	301.54	.76	.81	.110	A x 2 / A df = 22.88 **
Two factor (Figure 4)	104	187.13	.90	.93	.077	Best model
Two factor with interaction (Figure 5)	103	236.47	.85	.88	.092	A x 2 / A df = 49.43 **
** - significant at p < .01
Table 4: Direct effects of determinants of loyalty (p - reg. coefficient; s.e. - standard error)
	Dependent variables -	ß (s.e.)		
Independent variables	Switching Intention		Switching Expectation	
Satisfaction	- .57 (.23)	**	n.s.	
Attitude towards switching	- .52 (.19)	**	n.s.	
Trust	- .44 (.21)	*	- .73 (.23)	**
Perceived switching cost	- .24 (.09)	**	.49 (.14)	**
R2	.51		.32	
Overall fit indices: Satorra-Bentler scaled chi-square = 187.13 (104 df); p-value = .000; NFI = .88; NNFI = .90; CFI = .93; RMSEA = .077 (.061-.091); * - p < .05; ** - p < .01
Table 4 shows the specific results of testing the most appropriate model, the structural model with two dependent variables (Figure 4). For each of the dependent variables the coefficients of determinants are shown together with their standard errors. All the studied determinants have been previously found to have a positive effect on loyalty. The dependent variables are switching (disloyalty) forecasts and the obtained negative sign of coefficients corresponds to the literature.
3.4. COMPARING COEFFICIENTS BETWEEN THE SWITCHING INTENTIONS AND EXPECTATIONS
In hypotheses H2a to H2d we have proposed that the determinants of loyalty differ in how they influence the switching intentions and the switching expectations. In Table 3 we can observe that the coefficients between the intentions and the expectations appear to be different, which is in line with H2a to H2d. Both satisfaction and attitude towards switching have a stronger effect on switching intention than switching expectation. On the other hand, trust and perceived switching costs have a stronger effect on expectations than intentions. In order to test whether these differences in coefficients are significant, the two factor model based on Figure 4 is taken as the baseline model. Next, for each of the determinants a constrained structural model is constructed, where the coefficient of the specific determinant is held equal for both dependent variables. For example, the constrained model for satisfaction is based on Figure 4 with the constraint that the two coefficients for the switching intentions and the switching expectations are the same.
The baseline (unconstrained) model and the four constrained structural models are tested for fit and the fit indices are found in the Table 5. Next,
the fit of each of the four constrained structural models is compared with the fit of the baseline model. The chi-square difference test provides support that the baseline model fits the data significantly better than any of the constrained models. Therefore, support is found for each of the hypotheses H2a to H2d. Based on these empirical results we can conclude that the switching intentions and the switching expectations differ in their explanatory variables.
4. GENERAL DISCUSSION
4.1. CONTRIBUTION TO THE THEORY
The purpose of this paper is to extend our understanding of how determinants influence customer loyalty. We base our study on the difference between the intentions and expectations perspectives. A proposition was made that measuring customer loyalty with these two perspectives triggers different cognitive processes with regards to forecasting loyalty. When customers are asked about their intentions, they tend to take a more inward-looking perspective and focus on the factors that are under their control, such as their own motivation and reasons for future behavior. On the other hand, when customers are asked about their expectations, they tend to take a more outward-looking perspective and focus more on the factors beyond their control, such as the obstacles and risk associated with loyalty behavior.
We find support that determinants of loyalty differ in how they influence loyalty intentions and expectations. These differences in effects are based on the inward or outward-looking nature of determinants. Satisfaction and Attitude towards switching are more inward-looking determinants and are found to be better predictors of loyalty intentions as opposed to loyalty expectations. On the other hand, Trust and Perceived financial
Table 5: Testing models for the difference in the path coefficients
Constrained coefficient in the	Fit indices					Model comparison
model (hypothesis)	df	X 2	NNFI	CFI	RMSEA	(all A df = 1)
None (baseline model)	104	187.13	.90	.93	.077	Baseline model
Satisfaction (H2a)	105	257.34	.82	.85	.112	A x 2 = 70.21 **
Attitude towards switching (H2b)	105	223.78	.85	.88	.093	A x 2 = 36.65 **
Trust (H2c)	105	205.07	.87	.91	.088	A x 2 = 17.94 **
Switching cost (H2d)	105	217.67	.86	.90	.092	A x 2 = 30.54 **
** - significant at p < .01
switching costs are more outward-looking and are found to be better predictors of loyalty expectations as opposed to loyalty intentions.
This study therefore contributes to the discussion about the roles and influence of determinants of loyalty (e.g., Oliver, 1999; Brady et al., 2005). As mentioned previously, there is still no agreement about which determinants are central to loyalty. Our results suggest that satisfaction is the most important determinant for loyalty intentions, while trust is the most important determinant for loyalty expectations. In previous research, either intentions or expectations have been used to capture forecasts and therefore different empirical findings can be due to researchers using different measures.
Loyalty intentions and expectations have been previously proposed to be different constructs (Warshaw and Davis, 1985; Sheppard et al., 1988; Soderlund and Ohman, 2005). Here, a stronger support is provided, as they differ both in the measurement model as well as how they are influenced by determinants (Zaltman et al., 1982).
The question now remains, which of the two measures of loyalty should be used, when aiming to identify its determinants? The answer is that they can both be useful, given that it is understood that they evoke different perspectives. When loyalty behavior with few obstacles is explored, such as word-of-mouth, then the intentions perspective might provide a better insight about which determinants matter most. However, when the studied loyalty behavior is more complex and depends on external factors, such as switching the service provider, then the expectations perspective might be more suitable.
4.2. MANAGERIAL IMPLICATIONS
Marketing researchers interchangeably use intentions and expectations measures, which can lead to misleading findings about how determinants influence loyalty. When measuring loyalty with an intention question, inward-looking determinants, such as satisfaction, would be supported as central. However, if loyalty expectations were used, then outward-looking determinants, such as trust and switching costs, would emerge as key to customer loyalty. Using different measures would lead to different findings and therefore different focus and actions in loyalty programs.
The intentions and expectations measures provide different views on loyalty and can both be
of interest to managers. Loyalty intentions signal customer's motivation to remain loyal, which can lead to a positive word of mouth, willingness to pay more, or up-grade the service. However, companies would like their customers to stay loyal even in the case of a bad service or a competitive offer. The construct of loyalty expectation signals this deeper commitment to the current provider. In order to achieve a higher level of loyalty expectation, companies should refocus their attention and resources to build trust with their customers as well as to increase perceived switching costs (e.g., penalty fees).
Our findings are important both for companies that are defending their customer base, typically incumbents, as well as companies trying to expand their customer base, typically new entrants in the market. The first group is working heavily on increasing satisfaction, trust, switching costs, and attitude towards switching behavior. On the contrary, the new entrants try to win customers over by decreasing switching costs and changing attitude towards switching by making it easier for customers to switch. The new entrants can also try to undermine trust in the incumbent.
4.3. LIMITATIONS AND FUTURE RESEARCH
Previous research has identified that expectations are better predictors of actual behavior compared to intentions (Sheppard et al., 1988; Soderlund and Ohman, 2005). A limitation of this study is that an actual behavior was not measured. This would make the study richer and provide a further test of the relation between loyalty forecasts and behavior. Nevertheless, the focus of this study is on the relation between determinants of loyalty and forecasts of loyalty. Understanding this aspect of loyalty is important as many academics and practitioners use customer forecasts and not actual behavior to measure loyalty, draw conclusions, and make decisions.
The second limitation comes from the sample, both in terms of the small sample size as well as the student population. The hypotheses would need to be tested using a broader population and a bigger sample. Nevertheless, as the effects are found within such a homogenous group, it is proposed that they are also present in the broader population. Further studies done in other industries and countries would provide tests for generalizing the proposition.
Another important limitation is that only switching behavior was studied. Zeithaml et al. (1996) have proposed that in order to gain the full picture of
relevant factors in customer relationship management, one needs to study a broader set of behaviors: such as positive and a negative word of mouth, complaining, cross, and up-selling. In addition to this, effects of other determinants of loyalty would need to be explored, such as value for money, quality, cost, length of relationship, and customer characteristics (e.g. gender or age).
5. REFERENCES
1.	Ajzen, Icek (1991), "The Theory of Planned Behavior", Organizational Behavior and Human Decision Processes, 50 (1991), 179-211.
2.	Ajzen, Icek and Martin Fishbein (1980), Understanding Attitudes and Predicting Social Behavior, Englewood Cliff, NJ: Prentice Hall.
3.	Ajzen, Icek and Thomas J. Madden (1986), "Prediction of Goal-Directed Behavior: Attitudes, Intentions and Perceived Behavioral Control", Journal of Experimental Social Psychology, 22, 453-474.
4.	Anderson, James C. and David W. Gerbing (1988). "Structural equation modeling in practice: A review and recommended two step approach", Psychological Bulletin, 103 (May), 411-423.
5.	Bagozzi, Richard P. (1975), "Marketing as Exchange", Journal of Marketing; Oct. 1975; pg. 32.
6.	Bagozzi, Richard P. and Jeffrey R. Edwards (1998), "A general approach for representing constructs in organizational research", Organizational Research Methods, Vol. 1, No. 1, 45-87.
7.	Bagozzi, Richard P., Youjaer Yi and Lynn W. Phillips (1991), "Assessing Construct Validity in Organizational Research", Administrative Science Quarterly, 36 (1991): 421-458.
8.	Bettman, James R., Mary Frances Luce, and John W. Payne (1999). "Constructive Consumer Choice Processes", Journal of Consumer Research, Vol. 25, No. 3 (Dec., 1998), 187-217.
9.	Brady, Michael K., Gary A. Knight, J. Joseph Cronin Jr., G. Tomas, M. Hult, and Bruce D. Keillor (2005), "Removing the contextual lens: A multinational, multi-setting comparison of service evaluation models", Journal of Retailing, 81 (3, 2005), 215-230
10.	Burnham, Thomas A., Judy L. Frils, and Vijay Mahajan (2003), "Consumer Switching Costs: A Typology, Antecedents, and Consequences", Journal of the Academy of Marketing Science, Apr 2003; 31: 109 - 126.
11.	Fornell, Claes (1992), "A National Customer Satisfaction Barometer: The Swedish Experience", Journal of Marketing, Vol. 56, No. 1 (Jan., 1992), 6-21.
12.	Fornel, Claes and David F. Larcker (1981), "Evaluating Structural Equation Models with Unobservable Variables and Measurement Error", Journal of Marketing Research, Vol. 18, No. 1 (Feb., 1981), 39 - 50.
13.	Garbarino, Ellen and Mark S. Johnson (1999), "The Different Roles of Satisfaction, Trust and Commitment for Relational and Transactional Consumers", Journal of Marketing, Vol. 63 No. 2 (April, 1999), 70-87.
14.	Johnson, Michael D. and Claes Fornell (1991), "A framework for comparing customer satisfaction across individuals and product categories", Journal of Economic Psychology, Vol. 12, No. 2, pp. 267-286.
15.	Jones, Michael A., David L. Mothersbaugh, and Sharon E. Beatty (2002), "Why customers stay: measuring the underlying dimensions of services switching costs and managing their differential strategic outcomes", Journal
of Business Research, Volume 55, Number 6, June 2002, pp. 441-450 (10).
16.	Kline, Rex B. (1998), Principles and practice of structural equation modeling, Guilford Press, New York.
17.	Marsh, Hert W. (1994), "Confirmatory Factor Analysis Models of Factor Invariance: A Multifaceted Approach," Structural Equation Modeling, 1 (1), 5-34.
18.	Morgan, Robert M. and Shelby D. Hunt(1994), " The Commitment-Trust Theory of Relationship Marketing", Journal of Marketing, Vol. 58 (July 1994), 20-38.
19.	Oliver, Richard L. (1997), Satisfaction: a behavioral perspective on the consumer, McGraw Hill.
20.	Oliver, Richard L. (1999), "Whence Consumer Loyalty?", Journal of Marketing, Vol. 63, Fundamental Issues and Directions for Marketing (Special issue 1999), 33-44.
21.	Podsakoff, Philip M., Scott B. Mackenzie, Jeong-Yeon Lee, and Nathan P. Podsakoff (2003), "Common Method Bias and Behavioral Research: A Critical Review of the Literature and Recommended Remedies", Journal of Applied Psychology, 2003, Vol. 88, No. 5, 879-903.
22.	Reichheld, Frederick F. (1996), "Learning from Customer Defections", Harvard Business Review, March/April, pp. 56-69.
23.	Satorra, Albert and Peter M. Bentler (1988), "Scaling corrections for chi-square statistics in covariance structure analysis", Proceedings of the American Statistical Association, 1988.
24.	Sheppard, Blair H., Jon Hartwick, and Paul R. Warshaw (1988), "The Theory of Reasoned Action: A Meta-Analysis of Past Research with Recommendations for Modifications and Future Research", Journal of Consumer Research, Vol. 15, No. 3 (December, 1988), 325-343.
25.	Soderlund, Magnus (2003), "The retrospective and the prospective mind and the temporal framing of customer satisfaction", European Journal of Marketing, Vol. 37, No. 10, 2003, 1375-1390.
26.	Soderlund, Magnus and Niclas Ohman (2005), "Assessing behavior before it becomes behavior: An examination of the role of intentions as a link between satisfaction and repatronizing behavior", International Journal of Service Industry Management, Vol. 16, No.2, 2005, 169-185.
27.	Szymanski, David M. and David H. Henard (2001), "Customer Satisfaction: A Meta-Analysis of the Empirical Evidence", Journal of the Academy of Marketing Science, Volume 29, No.1, 16-35.
28.	Warshaw, Paul R. and Fred D. Davis (1985), "Disentangling behavioral intention and behavioral expectation", Journal of Experimental Social Psychology, (1985), 21:33, 213-228.
29.	Zaltman, Gerald, Karen LeMasters and Michael Heffring (1982), Theory Construction in Marketing, John Willey & Sons.
30.	Zeithaml, Valarie A., Leonard L. Berry, and A. Parasuraman (1996), "The Behavioral Consequences of Service Quality", Journal of Marketing, Vol. 60, No. 2 (April, 1996), 31-46.
31.	Zeithaml, Valarie A., Ruth N. Bolton, John Deighton, Timothy L. Keiningham, Katherine N. Lemon, and Andrew, J. Petersen (2006), "Forward-Looking Focus. Can Firms Have Adaptive Foresight?", Journal of Service Research, Vol. 9, No. 2, 168-183.
APPENDIx: Measures USED	
for study constructs	
Switching Intentions (ten-point scale, »do not agree at all« /	
» agree completely«)	
I1: I plan to change my mobile phone service provider.	
I2: I intend to start using another provider.	
I3: I will change my provider within the next year.	
Switching Expectations (ten-point scale)	
E1: I expect to change my provider in the future. ("do not	
agree at all" /" agree completely")	
E2: How likely is that you will change it in the future? ("very	
unlikely" /"very likely")	
E3: How likely is that you will stay with your current mobile	
service provider? (as above)	
Overall Satisfaction (seven-point scale) ("very dissatisfied" /	
"very satisfied")	
S1: How satisfied or dissatisfied are you with your current	
mobile phone service provider?	
S2: Overall, how do you feel about it? ("very unhappy" / "very	
happy")	
S3: How well does it meet your needs at this time?	
("extremely poor" / "extremely well")	
Trust (seven-point scale, "do not agree at all" / "agree	
completely")	
T1: I feel can trust my mobile phone service provider.	
T2: My provider is responsive to customers' problems.	
T3: I feel my service provider is reliable.	
Attitude towards switching (seven-point scale, "do not	
agree at all" / "agree completely")	
A1: It would take me a lot of effort to choose another mobile	
service provider.	
A2: I do not feel like going through the whole process of	
changing the mobile provider.	
A3: Going through the process of changing the provider would	
be an unpleasant experience.	
Switching cost (seven-point scale, "do not agree at all" /	
"agree completely")	
F1: Changing my mobile service provider would cost me	
money.	
F2: If I choose another service provider my costs will go up.	
F3: Changing my mobile service provider would have	
significant financial impact for me.