Abstract
Purpose
For nearly 2 decades, the push-pull-mooring (PPM) model has been used frequently by scholars to explain consumers’ service switching intention and behavior. However, heterogeneity and incomparability between PPM model studies are prevalent issues: The chosen predictor variables, their categorization, their measurement, reported effect sizes, and effect directions vary considerably. By addressing these issues, the present meta-analytical review enables future researchers applying the PPM model to identify relevant variables and use valid measurements.
Design/methodology/approach
Based on 148 empirical studies employing the PPM model, the variables used to predict consumers’ service switching intention and behavior, their frequency of use, their categorization into push, pull, and mooring factors, and their measurement are assessed. The effect sizes and directions of the relationships between these variables and consumers’ service switching intention and behavior are analyzed using meta-analytic structural equation modeling. Additionally, the predictive capacity of this model and the influence of moderators are assessed.
Findings
Among the 148 empirical studies, 382 different independent variables were used. The three most frequently used and distinctly categorized independent variables are dissatisfaction (push), alternative attractiveness (pull), and switching costs (mooring). Overall, 152 unique sources were cited to measure these variables and the dependent variables. Dissatisfaction and alternative attractiveness increase switching intention, which positively affects switching behavior, while switching costs decrease switching intention. The model explains 30% of the variance in switching intention and 31% of the variance in switching behavior.
Originality/value
This study provides the first meta-analytical review of the PPM model to guide future research systematically.
Keywords
Citation
Marx, T. (2025), "The push-pull-mooring model of consumer service switching: a meta-analytical review to guide future research", Journal of Service Theory and Practice, Vol. 35 No. 7, pp. 1-29. https://doi.org/10.1108/JSTP-06-2024-0201
Publisher
:Emerald Publishing Limited
Copyright © 2024, Tobias Marx
License
Published by Emerald Publishing Limited. This article is published under the Creative Commons Attribution (CC BY 4.0) licence. Anyone may reproduce, distribute, translate and create derivative works of this article (for both commercial and non-commercial purposes), subject to full attribution to the original publication and authors. The full terms of this licence may be seen at http://creativecommons.org/licences/by/4.0/legalcode
1. Introduction
For nearly 2 decades, the push-pull-mooring (PPM) model—as introduced to service research by Bansal et al. (2005)—has been used by scholars to explain consumers’ service switching intention and behavior. Adapted from migration research (Bogue, 1969; Lee, 1966; Moon, 1995; Ravenstein, 1885), the PPM model posits that negative factors at the origin push individuals away (push) and positive factors at the destination pull them towards it (pull). Furthermore, personal and social factors can inhibit or facilitate migration decisions (mooring). When applying the PPM model to service research, the analogy is straightforward: Consumers (migrants) switch (move) from one service provider (place) to another. The PPM model provides a sound theoretical framework to include different predictor variables commonly associated with service switching intention and behavior by categorizing them as push, pull, or mooring factors (Bansal et al., 2005). Extensive studies on the PPM model in service research have resulted in a substantial body of knowledge. However, heterogeneity and incomparability between PPM model studies are prevalent issues: The chosen predictor variables, their categorization, their measurement, reported effect sizes, and effect directions vary considerably. This is problematic for a number of reasons: The lack of comparability across studies undermines the ability to draw conclusions about the efficacy of the PPM model. Additionally, conflicting results regarding effect sizes and directions make it difficult to identify clear patterns. Furthermore, inconsistent conceptualization and measurement hinder the theoretical development of the PPM model. Collectively, these issues diminish the utility of the PPM model for researchers and practitioners alike. To resolve this, this study provides the first comprehensive overview and meta-analytical test of the PPM model.
The contributions to the literature are numerous: First, the different variables used to predict consumers’ service switching intention and behavior in PPM model studies are assessed. Based on this, an analysis is conducted on how these variables are categorized into push, pull, and mooring factors and which ones are used most frequently. Second, differences in the measurement of these variables across the underlying studies are investigated. Third, the effect sizes and directions of the relationships between the most frequently used push, pull, and mooring variables (i.e. dissatisfaction, switching costs, alternative attractiveness) and consumers’ service switching intention and behavior are assessed. Thereby, this meta-analytical review provides a benchmark, allowing researchers to identify anomalous results and to design studies with appropriate sample sizes. Moreover, this allows for empirical generalizations about which aspects practitioners should focus on in their acquisition and retention strategies. Fourth, the average predictive capacity of the PPM model using these variables is assessed. Finally, the influence of moderators on the effects of dissatisfaction, switching costs, and alternative attractiveness on consumers’ switching intention is examined. This allows for conclusions on which variables are particularly relevant under specific conditions. Thus, the following research questions are answered:
- (1)
Which variables are used to predict consumers’ service switching intention and behavior in PPM model studies?
- (2)
How are these variables categorized into push, pull, and mooring factors?
- (3)
Which of these variables are most frequently used?
- (4)
Which measurement scales are most frequently used to operationalize these variables?
- (5)
What are the effect sizes and directions of the relationships between the most frequently used push, pull, and mooring variables and consumers’ service switching intention and behavior?
- (6)
What is the predictive capacity of the PPM model using these variables in explaining the variance in consumers’ service switching intention and behavior?
- (7)
What moderators affect the relationships between the most frequently used push, pull, and mooring variables and consumers’ service switching intention?
By addressing these questions, this paper offers guidance for future researchers applying the PPM model to improve their studies. The key issues in PPM model studies and how this meta-analytical review addresses them are summarized in Table 1. To achieve this, the rest of this paper is organized as follows: Chapter 2 briefly reviews the PPM model. Chapter 3 outlines the methodological approach. Chapter 4 presents the conceptual framework and hypotheses. Chapter 5 reveals the meta-analytical findings. Chapter 6 discusses the findings, highlighting their theoretical and practical significance. Lastly, Chapter 7 outlines the limitations of this study and proposes directions for future research.
2. Push-pull-mooring model
The PPM model, initially developed in migration research (Bogue, 1969; Lee, 1966; Moon, 1995; Ravenstein, 1885), explains why individuals leave their place of origin to move to a new destination. The model has three components: push, pull, and mooring factors. Push factors are negative factors at the origin that motivate individuals to leave (Stimson and Minnery, 1998). Pull factors are positive factors at the destination that attract individuals to migrate there (Moon, 1995). Migration research on push factors (e.g. economic recession, natural catastrophe) and pull factors (e.g. favorable climate, superior infrastructure) has a long history, dating back as early as the nineteenth century (Ravenstein, 1885). Throughout this extensive period, researchers have primarily focused on these factors as they induce migration, with comparatively less attention given to their counterparts—positive factors at the origin and negative factors at the destination (Bansal et al., 2005; Lee, 1966). Starting with key articles by Wolpert (1965, 1966), there was a shift in focus toward understanding how migrants perceive push and pull factors, realizing that it is not the actual factors themselves but the perceptions of these factors that result in migration (Lee, 1966). This led researchers to consider personal (e.g. personal anxiety) and social factors (e.g. family attachments) in influencing migration decisions (Bansal et al., 2005). Lee (1966) introduced the concept of “intervening obstacles,” which Jackson (1986) refined to “intervening variables,” acknowledging that these factors can either facilitate or inhibit migration. This idea was further developed into the concept of mooring factors, which are personal and social factors that can hold potential migrants to their place of origin or facilitate migration to the new destination (Longino, 1992; Moon, 1995). In migration research, a key distinction among these factors is the level of control individuals have over them. Push factors and pull factors are conceptualized as being outside the individual’s control (Bogue, 1969; Moon, 1995), whereas mooring factors are conceptualized as being, to some extent, within the individual’s control (Moon, 1995). Another important distinction is that individuals generally weigh push and pull factors against each other in a direct comparison, while mooring factors are assessed separately, providing the context for this comparison (Lee, 1966). Furthermore, mooring factors can moderate the relationships between push and pull factors and the migration decision (Lee, 1966).
Bansal et al. (2005) first applied the PPM model in service research to explain consumers’ switching intention and behavior between auto repair and hair styling services. In doing so, they used low quality, low satisfaction, low value, low trust, low commitment, and high price perceptions to conceptualize the push factor; alternative attractiveness to conceptualize the pull factor; and unfavorable attitude towards switching, unfavorable subjective norms, high switching costs, infrequent prior switching behavior, and low variety seeking to conceptualize the mooring factor. They found that the PPM model was highly effective in predicting the dependent variables and called for further research in this field. Since then, the PPM model has been widely used in various contexts to explain switching intention and behavior, reflecting its growing popularity in service research. For example, Zengyan et al. (2009) applied the model to social networking services, Ye and Potter (2011) used it to study web browsers, and Wu et al. (2017) applied it to examine cloud storage services. However, all of these studies differ in their selection of variables to conceptualize the push, pull, and mooring factors, as well as in their measurement of these variables.
In more detail, when applying the PPM model, Bansal et al. (2005) model the push, pull, and mooring factors as higher-order constructs. Specifically, they model them as reflective-reflective second-order constructs (Sarstedt et al., 2019) consisting of lower-order subdimensions such as subjective norms, switching costs, and alternative attractiveness. Bansal et al. (2005) demonstrate that the model with higher-order constructs explains more variance in switching intention (68%) than the model with direct effects, in which the lower-order subdimensions directly influence switching intention (61%). Nevertheless, most studies using the PPM model use the model with direct effects. Of the 148 studies identified as eligible in the third step of the data collection process in Chapter 3, 101 use the direct effects approach, while only 47 use the model with higher-order constructs.
As a result, a meta-analysis of the higher-order constructs is not feasible, not least because the higher-order constructs in each study are composed of different lower-order constructs. In this respect, the typical criticism of inappropriate comparisons in meta-analyses would apply (Cortina, 2003). Thus, this meta-analysis of the PPM model will focus on the lower-order constructs. The methodology is first reported in Chapter 3, based on which the specific variables for the conceptual model were selected. The conceptual model is then presented in Chapter 4.
3. Methodology
3.1 Selection and coding of studies
3.1.1 Data collection
In accordance with established guidelines for systematic literature reviews and meta-analyses (Liberati et al., 2009), this study’s data collection process is illustrated in Figure 1. Various search strategies were employed to build a comprehensive database (Eisend, 2017), including both published and unpublished studies from January 2005 to September 2023. First, a broad keyword search (search term: “push pull mooring”) was conducted to identify studies using the PPM model by searching the following electronic databases: Ebsco, Econbiz, Google Scholar, Jstor, Proquest, Scopus, Social Science Research Network, and Web of Science. Additionally, the references in all appropriate studies were reviewed to find additional studies not yet included in the database. The search yielded a total of 270 unique studies. To address the file drawer problem (Rosenthal, 1979), the authors of the 148 PPM model studies identified as eligible in the data collection process were contacted via e-mail to inquire about any unpublished work. However, no additional studies could be obtained through this approach.
3.1.2 Exclusion criteria
Consistent with other meta-analyses (e.g. Eisend, 2006; Feng et al., 2021; Lütjens et al., 2022), studies using the same dataset as already included studies were excluded. Due to translation barriers, studies that were not published in English were excluded. Studies that appeared in the keyword search but, upon further inspection, did not use the PPM model (e.g. Bölen, 2020) were also excluded. Furthermore, all qualitative studies (e.g. Liu et al., 2021) were excluded. Studies with dependent variables that were inconsistent with the PPM model (i.e. not switching intention or behavior or synonymous variables) (e.g. Fu, 2011) and studies using methods (e.g. machine learning) unsuitable for this meta-analysis (e.g. Al-Mashraie et al., 2020) were also excluded. Moreover, studies on geographical migration (e.g. Kaur and Kaur, 2023) and studies from a B2B context (e.g. Suh and Kim, 2018) were excluded. The correlation coefficient was used to measure effect size, which is typical in meta-analyses in service research (e.g. Black et al., 2014; Blut et al., 2016; Feng et al., 2021; Ranjan et al., 2015). Consequently, studies that did not report correlation coefficients or where the correlation coefficient could not be derived from other reported statistics were excluded.
Based on the remaining 148 studies, a dataset containing all predictor variables used and their categorization as push, pull, or mooring factors was created. Figure 2 shows the ten most frequently used independent variables and their respective categorization. As Figure 2 illustrates, categorization is ambiguous for some variables in the literature. In contrast, categorization for the three most frequently used variables (i.e. switching costs, alternative attractiveness, dissatisfaction) is unequivocal. Thus, these variables were included in the conceptual framework (Figure 3). The methodological reasons for not including additional predictor variables in the conceptual framework are explained in detail in Chapter 3.2.
3.1.3 Coding
In line with previous meta-analyses (e.g. Palmatier et al., 2006; Lütjens et al., 2022), many variables with definitions and operationalizations related to the determinants in the conceptual framework were found, albeit using different names. Following common practice in meta-analyses (e.g. Eisend, 2017), these variables were coded under the corresponding label of the focal determinants (e.g. synonyms for switching intention include migration intention and shifting intention).
An extensive coding template, which included all required definitions and coding instructions, was created for the coding process. Using this template, the author and a trained research assistant carried out the coding. The initial intercoder reliability was 0.94. All inconsistencies were resolved through discussion.
3.1.4 Final dataset
The final dataset to evaluate the conceptual framework includes (1) all empirical-quantitative PPM model studies that (2) report correlation coefficients or metrics that could be transformed into correlation coefficients (e.g. beta coefficients; Peterson and Brown, 2005), and (3) contain at least two variables from the conceptual framework. Ultimately, 264 effect sizes were obtained from 109 independent samples, reported in 107 published or unpublished studies conducted between January 2005 and September 2023. A list of all studies included in this meta-analysis can be found in the Supplementary material.
3.2 Meta-analytic structural equation modeling
Meta-analytic structural equation modeling (MASEM) was used to test the conceptual framework. MASEM refers to fitting structural equation models to meta-analytic data using correlation matrices (Jak et al., 2021). MASEM offers substantial advantages over meta-analyzing each effect in a model separately (Jak, 2015). First, it allows for the simultaneous evaluation of multiple predictors while accounting for dependencies between them (Jak and Cheung, 2020). Second, it provides an overall evaluation of model fit, which cannot be achieved through separate analyses of individual effects (Valentine et al., 2022). Modern multivariate MASEM methods use correlation matrices and sample sizes from each primary study as input (Jak and Cheung, 2020). In contrast, univariate MASEM methods pool correlation coefficients separately based on bivariate data and do not account for the dependency between correlations within studies (Jak, 2015). Because of these limitations, univariate MASEM methods are not recommended (Becker, 2000; Cheung and Chan, 2005; Jak, 2015).
The decision not to include additional variables in our conceptual framework stems from the fact that conducting multivariate MASEM with more than the three chosen independent variables is not possible with the data currently available. Including another variable would increase the number of non-redundant, off-diagonal correlations from ten to 15 in each correlation matrix. Since the other predictor variables are used considerably less frequently (e.g. habit is included in only 27 of the 148 studies, Figure 2), the resulting correlation matrices would be too sparse for the MASEM model to converge.
Further analysis was conducted using the metaSEM package in R (Cheung, 2015). When using MASEM, the five main considerations are (1) whether to use a fixed-effects or random-effects approach, (2) how to deal with possible study dependence, (3) how to deal with possible publication bias, (4) whether to correct estimates for attenuation and (5) how to incorporate judgments of study quality (Valentine et al., 2022).
3.2.1 Fixed-effects vs random-effects
The fixed-effects approach assumes that all studies have the same population effect sizes, meaning any differences between observed effect sizes are solely attributed to sampling error (Hedges and Vevea, 1998). In contrast, the random-effects approach assumes that population effect sizes vary from study to study (Jak and Cheung, 2020), meaning differences between observed effect sizes may arise for reasons other than sampling error (Jak, 2015). For example, observed effect sizes might differ between studies due to variations in participants’ age, education, or income (Borenstein et al., 2021; Valentine et al., 2022). The primary aim of the random-effects approach is to generalize the results of a meta-analysis beyond the included studies (Jak and Cheung, 2020). Since it is very likely that the population effect sizes differ between studies, the random-effects approach is generally preferred (Borenstein et al., 2021; Hedges, 2016). Therefore, the random-effects approach was used in this meta-analysis. Specifically, the one-stage MASEM (OSMASEM) approach, a multivariate random-effects MASEM method that does not involve two strictly separated stages, was used (Jak and Cheung, 2020). Among the various MASEM methods available, OSMASEM is the most versatile, as it can evaluate the effects of continuous and categorical moderators without creating subgroups (Jak and Cheung, 2022). Furthermore, it has been shown to perform well with incomplete data (Jak and Cheung, 2020), making it well-suited for this analysis.
3.2.2 Dependent estimates
In meta-analytic correlation matrices, the estimates are assumed to be independent (Valentine et al., 2022). Dependent estimates occur when one estimate offers insights about the magnitude or direction of another estimate (Van den Noortgate et al., 2013). Dependence may arise within studies when multiple operationalizations of the same variables are used, resulting in multiple estimates for a specific correlation (Van den Noortgate et al., 2013). Since none of the studies included in this meta-analysis provided more than one operationalization for the same variable, it was unnecessary to account for dependence by averaging the dependent effect sizes (Valentine et al., 2022).
3.2.3 Publication bias
Publication bias occurs when a study remains unpublished due to a lack of statistically significant findings on its main outcomes (Valentine et al., 2022). Publication bias severely threatens the validity of conclusions derived from systematic reviews or meta-analyses. In the context of MASEM, publication bias results in overestimated relationships between variables and underestimated heterogeneity estimates (Valentine et al., 2022). Preventing publication bias is best achieved through an exhaustive literature search for relevant studies (Jak and Cheung, 2020; Vevea et al., 2019). This is crucial for all systematic reviews and meta-analyses but holds even greater significance in MASEM due to the lack of consensus on addressing potential publication bias in MASEM (Valentine et al., 2022). A comprehensive literature search was conducted in this meta-analysis, and authors were contacted to inquire about unpublished work. Furthermore, the presence of publication bias in this meta-analysis is unlikely because the PPM model typically encompasses various independent variables to predict switching intention or behavior. Thus, the probability of a study not being published due to the non-confirmation of one or more predicted relationships is diminished.
3.2.4 Attenuation corrections
Even if two variables are perfectly correlated, the observed correlation of their measures will not be 1 in the presence of measurement error. This phenomenon, known as attenuation, is often corrected in meta-analyses to account for measurement error (Hunter and Schmidt, 2004; Valentine et al., 2022). To correct for measurement error, researchers must select a suitable reliability coefficient. Cronbach’s alpha is the predominant measure of internal consistency reliability (McNeish, 2018). Therefore, Cronbach’s alpha was employed as the reliability coefficient in this meta-analysis. However, not all studies included in this meta-analysis reported Cronbach’s alpha values for all measures, making it impossible to correct the correlations on an individual study level. Instead, like other meta-analyses in consumer research (e.g. Hogreve et al., 2017), the distribution of the reported Cronbach’s alpha values was used (Hunter and Schmidt, 2004). On average, Cronbach’s alpha values were consistently high for all measures (>0.8). Correcting for measurement error using the distribution-based approach resulted in some non-positive definite meta-analytic correlation matrices at the individual study level. This issue arose because individual correlations exceeded 1 after the correction. To address this, these correlations were removed (Valentine et al., 2022). Table 2 presents the uncorrected results, whereas Table A1 (see Appendix) shows the corrected results. The corrected and uncorrected results are similar, aligning with the findings of Michel et al. (2011). Given the high average Cronbach’s alpha values (>0.8), this outcome was expected. Due to the aforementioned limitations regarding the loss of information from non-positive definite correlation matrices and the small impact of the correction, uncorrected correlations were used for further analyses, in line with Valentine et al. (2002).
3.2.5 Incorporating judgments on study quality
In the context of systematic reviews and meta-analyses, study quality can be defined as the extent to which the methods used to answer a research question align with the goals of the review (Valentine, 2019). Assessing study quality presents multiple challenges: First, study quality is always context-dependent. Second, study quality is a multidimensional construct. Therefore, attempting to assign a single number, score, or unitary judgment to assess the quality of a specific study is misguided (Valentine, 2019). Given these considerations, study quality was strictly assessed empirically (Valentine et al., 2022), using multiple indicators of publication medium quality as proxies for study quality and including them as moderators (see Chapter 4.2.2.7).
3.2.6 Moderation analysis in MASEM
As previously mentioned, among the various MASEM approaches, OSMASEM is the most versatile, as it can evaluate the effects of continuous and categorical moderators without creating subgroups (Jak and Cheung, 2022). As recommended by Jak and Cheung (2020), only one moderator was incorporated at a time in the model, and all continuous moderators were standardized to achieve convergence and more stable results. Moreover, switching behavior had to be excluded from the moderation analysis due to an insufficient amount of correlations, in line with the recommendations by Jak and Cheung (2020).
3.2.7 Effect size metric and conversion
As stated earlier, this meta-analysis used the correlation coefficient as the effect size metric. The formula proposed by Peterson and Brown (2005) was used to convert beta coefficients to correlation coefficients. Other effect size conversions were conducted using equations from Borenstein et al. (2021).
4. Conceptual framework
4.1 Independent and dependent variables
4.1.1 Dissatisfaction
Dissatisfaction refers to a psychological state that arises when consumers’ experiences are coupled with disconfirmed expectations (Oliver, 1981). Satisfaction leads consumers to form repurchase intentions (Anderson and Sullivan, 1993) or continuation intentions (Bhattacherjee, 2001), while dissatisfaction deters them from future purchases or continued use. According to expectation-confirmation theory, consumers’ intention to discontinue using a service is primarily determined by their dissatisfaction (Anderson and Sullivan, 1993; Oliver, 1980). Thus, it is predicted that dissatisfaction is positively related to switching intention (H1).
4.1.2 Switching costs
Switching costs can be defined as consumers’ perceived, anticipated, or experienced costs associated with switching from one alternative to another (Burnham et al., 2003; Jones et al., 2002). Switching costs include different types of costs, particularly financial, procedural, and relational costs, as well as combinations of these types (Burnham et al., 2003). In line with Pick and Eisend (2014), the focus of this meta-analysis is on overall switching costs. To explore how the results vary with different types of switching costs, a moderating variable that differentiates between monetary and non-monetary switching costs is included. Switching costs are seldom explicitly assessed but become salient when consumers face reasons to consider switching (Burnham et al., 2003). As such, switching costs can be considered barriers that discourage defection (Jones et al., 2000, 2002). Most researchers agree that switching costs reduce switching intention (e.g. Heide and Weiss, 1995; Morgan and Hunt, 1994; Wathne et al., 2001). Therefore, it is hypothesized that switching costs are negatively related to switching intention (H2).
4.1.3 Alternative attractiveness
Alternative attractiveness refers to consumers’ perceptions regarding the positive characteristics of competing alternatives (Bansal et al., 2005; Jones et al., 2000; Ping, 1993). A lack of alternative attractiveness encourages retention (Ping, 1993). On the other hand, when consumers perceive higher alternative attractiveness, the likelihood of switching increases as the perceived benefits of making a switch increase (Jones et al., 2000). Thus, it is hypothesized that alternative attractiveness is positively related to switching intention (H3).
4.1.4 Switching intention and behavior
Switching intention can be defined as consumers’ intention to switch between alternatives, whereas switching behavior refers to consumers’ actual switching behavior (Ajzen, 1991; Bansal et al., 2005). Intentions are generally considered the best predictor of behavior (Ajzen, 1991). Therefore, it is predicted that switching intention is positively related to switching behavior (H4).
4.2 Moderators
Several moderator variables are explored to explain the variation observed in the empirical studies included in this meta-analysis. While all moderators serve the statistical function of capturing variation in empirical findings (Brown et al., 1998), this study follows the approach of other meta-analyses (e.g. Okazaki et al., 2020; Lütjens et al., 2022) by categorizing them into two types: substantive moderators, which are directly related to the study’s subject and for which hypotheses are proposed, and control moderators, which are used to adjust for potential methodological or study-related differences (Eisend, 2017; Lütjens et al., 2022).
4.2.1 Substantive moderators
4.2.1.1 Provider switch vs technology switch
Research on consumers’ service switching behavior using the PPM model can be broadly categorized into two distinct types: provider switching vs technology switching. Grasping the differences between these two types of switching is essential for a comprehensive understanding of consumer switching behavior. Provider switching refers to a consumer’s choice to switch from one service provider to another. Examples include switching between personal cloud storage services (e.g. Cheng et al., 2019) or mobile instant messaging services (e.g. Sun et al., 2017). The technology remains consistent in these instances, but the service provider changes. In contrast, technology switching refers to a consumer’s choice to switch from one technology to another. Examples include switching from cash payments to mobile payments (e.g. Hsieh, 2021) or from petrol cars to electric cars (e.g. Sajjad et al., 2020). In these instances, the technology changes, and the provider is irrelevant. A few studies investigated technology switching but also explicitly mentioned the different providers (e.g. Fang and Li, 2022). In these cases, the technology switch takes precedence. In many industries, providers offer similar core services (e.g. airlines, hair stylists, hotels), making it difficult for consumers to distinguish one as considerably more attractive than another. In contrast, the attractiveness of a new technology stands out more clearly from existing alternatives. Thus, it is hypothesized that the positive effect of alternative attractiveness on switching intention is weaker for provider switches than for technology switches (H5).
4.2.1.2 Monetary switching costs vs non-monetary switching costs
Additionally, similar to other meta-analyses in this field (e.g. Pick and Eisend, 2014), this analysis reveals another essential distinction regarding studies that included switching costs as a variable: monetary vs non-monetary switching costs. Monetary switching costs refer to the loss of financially quantifiable resources (Burnham et al., 2003). Non-monetary costs refer to all perceived costs that cannot be quantified in financial or monetary terms and refer to psychological costs in terms of time and effort (Ping, 1993; Dick and Basu, 1994). Studies that did not include switching costs as a variable or did not provide adequate information to identify whether the switching costs were monetary or non-monetary were excluded from the analysis of this moderator. When switching entails monetary costs, the potential losses become more immediate and concrete for consumers. In line with the concept of loss aversion, this heightened focus on monetary losses likely increases switching costs and diminishes the attractiveness of the alternative (Kahneman and Tversky, 1979, 1984; Samuelson and Zeckhauser, 1988). Therefore, it is hypothesized that when switching costs are monetary, the negative effect of switching costs on switching intention is stronger (H6a), and the positive effect of alternative attractiveness on switching intention is weaker (H6b).
4.2.1.3 Contractual relation vs non-contractual relation
Furthermore, a distinction is made between contractual and non-contractual relations. A contractual relation involves a formal agreement that establishes a recurring payment structure and outlines conditions for potential termination or breach. Examples include video streaming services, where consumers pay annual or monthly subscription fees, and banking services, where consumers pay monthly fees and potential account closure fees in the event of a switch. In contrast, non-contractual relations, such as visiting a hair stylist or booking a hotel room, involve one-time transactions without ongoing contractual duties. Contractual relations create barriers that can elevate switching costs (Pick and Eisend, 2014). Moreover, consumers might experience a sense of obligation to uphold the contract (Pick and Eisend, 2014). Hence, consumers are likely to perceive higher switching costs in contractual relations, which makes them more reluctant to switch. Thus, it is hypothesized that when relations are contractual, the negative effect of switching costs on switching intention is stronger (H7).
4.2.2 Control moderators
4.2.2.1 Student sample vs non-student sample
Student samples are often more homogeneous than non-student samples (Agadullina and Lovakov, 2018). This sample homogeneity can affect the size of observed effects, as demonstrated by various meta-analyses (e.g. Brown and Stayman, 1992; De Matos and Rossi, 2008; Maseeh et al., 2021; Peterson, 2001). Furthermore, when considering student samples in the context of consumer switching behavior, it is important to recognize the unique demographic characteristics that distinguish students from non-student populations. Student samples typically consist of younger individuals predominantly engaged in their education rather than full-time employment. These unique characteristics suggest that students may respond differently to factors influencing consumer switching behavior. Numerous studies included in this meta-analysis (e.g. Chou et al., 2016) provide comprehensive details about their samples. This granularity enables the identification of samples that, while not exclusively composed of students, predominantly consist of this demographic. To attain a more balanced distribution in the analysis of this moderator, a distinction was also made between samples comprising mostly students and those that do not.
4.2.2.2 Information system vs no information system
An information system (IS) can be defined as a set of interrelated components that collect (or retrieve), process, store, and distribute information (Laudon and Laudon, 2021). For many ISs (e.g. social networking services), switching can be relatively easy and inexpensive, with alternatives being as little as a mouse click away (Chen and Hitt, 2002). Thus, in the context of ISs, consumers might react differently to factors that influence switching behavior.
4.2.2.3 Tangible vs intangible
In this study, all marketing offerings, both tangible and intangible, are considered as services, which are defined as the application of specialized competencies (knowledge and skills) through deeds, processes, and performances for the benefit of another entity or the entity itself (Vargo and Lusch, 2004a). This approach aligns with the service-dominant logic, recognizing that the traditional division between goods and services is outdated (Gummesson, 1995; Vargo and Lusch, 2004a, b). Nevertheless, whether the switch is between tangible or intangible goods is included as a control moderator in this meta-analysis.
4.2.2.4 Age
Age is a common sociodemographic control variable in the studies included in this meta-analysis. Therefore, it is logical to include age as a moderator in this study. Specifically, the (approximated) median age was examined, as age was most commonly reported in intervals in the studies included in this meta-analysis (e.g. Mohd-Any et al., 2023), making it more accurate to approximate the median age rather than the mean age. Studies that provided no information on age were excluded from the analysis of this moderator.
4.2.2.5 Gender
Gender is also a common sociodemographic control variable in the studies included in this meta-analysis, making it straightforward to include as a moderator. Specifically, the percentage of males in each sample was examined. Studies that provided no information on gender were excluded from the analysis of this moderator.
4.2.2.6 Region
Previous meta-analyses in consumer research have examined region as a moderator (e.g. Lütjens et al., 2022). Therefore, the region in which a study was conducted was also included as a moderator in this study. The studies included in this meta-analysis provided enough information to identify geographical differences at the country level. However, due to the absence of sufficient observations at the country level, this moderator was categorized at the regional level, namely (1) Asia and (2) Other (including studies from America, Africa, and Europe), because most of the studies included in this meta-analysis were from Asia. If the region could not be determined, those studies were excluded from the analysis of this moderator.
4.2.2.7 Publication medium quality
Publication media have different standards of quality. High-quality publication media typically require more rigorous research designs than lower-quality ones, for example, by requiring designs with control variables, larger sample sizes, or more sophisticated analyses (Palmatier, 2016). Such factors that increase the quality of research can affect the findings of an empirical study (Lee and Baskerville, 2003) and thus influence meta-analytic results (e.g. Carrillat et al., 2018; Floyd et al., 2000). Therefore, whether a study was published in a journal, usually indicating a diligent peer-review process, or appeared in other formats, such as books, conference papers, master’s or doctoral theses, and preprints, was assessed. Moreover, two measures of publication medium quality were included as moderators in this meta-analysis: (1) Scopus Cite Score from Harzing (2022) and (2) Citations per document (2022, 4 years) from https://www.scimagojr.com/.
4.2.2.8 Hofstede’s cultural dimensions
Research has shown that Hofstede’s cultural dimensions effectively predict and explain consumption differences across various services (De Mooij and Hofstede, 2002). Furthermore, Hofstede’s cultural dimensions (e.g. individualism) have proven to be important moderators in previous meta-analyses in consumer research (e.g. Pick and Eisend, 2014). Therefore, this meta-analysis also included Hofstede’s six cultural dimensions as control moderators. Hofstede’s cultural dimensions are particularly useful because data is readily available for many countries. The data was obtained from https://www.hofstede-insights.com/country-comparison-tool. If the country could not be determined, those studies were excluded from the analysis of this moderator.
5. Results
5.1 Overview of PPM model studies
Among the 148 empirical PPM model studies on consumer service switching included in this meta-analysis, 382 different independent variables were used. Figure 2 shows the ten most frequently used independent variables (Switching Cost: 86, Alternative Attractiveness: 67, Dissatisfaction: 53, Habit: 27, Subjective Norm: 26, Service Quality: 25, Inertia: 20, Perceived Usefulness: 19, Perceived Risk: 19, Trust: 17) and their categorization into push, pull, or mooring factors.
Furthermore, the investigated research contexts show considerable variety, with the three most commonly studied contexts being social networking services (9 studies), shopping channels (9 studies), and payment systems (9 studies). Other frequently studied contexts include financial services (7 studies), messaging services (6 studies), learning (6 studies), cloud services (6 studies), blogging (4 studies), accommodation services (4 studies), and delivery services (3 studies).
In terms of measurement, a total of 44 different sources were cited to measure dissatisfaction. The most frequently cited sources were Bhattacherjee (2001), with 11 citations, followed by Bansal et al. (2005) with 5 citations, and Sun et al. (2017) with 4 citations. Furthermore, a total of 51 different sources were cited to measure switching costs. The most frequently cited source was Bansal et al. (2005) with 12 citations, followed by Burnham et al. (2003) and Jones et al. (2000), each with 7 citations. To measure alternative attractiveness, a total of 43 different sources were cited. The most frequently cited source was Bansal et al. (2005), with 12 citations, followed by Hou et al. (2011), Jones et al. (2000), and Rezvani et al. (2015), each with 4 citations. Similarly, switching intention was measured citing 59 different sources. The most frequently cited source was Kim et al. (2006), with 18 citations, followed by Bansal et al. (2005) with 12 citations, and Jung et al. (2017) with 6 citations. Finally, switching behavior was measured citing 12 different sources. The most frequently cited source was Bansal et al. (2005), with 5 citations, followed by Hsieh et al. (2012) with 3 citations. Overall, 152 unique sources were cited to measure these five variables. Table 3 shows the most frequently cited scales for each variable.
5.2 Descriptive statistics
Table 4 presents the descriptive results for the variables included in the conceptual framework. The large differences between the minimum and maximum correlations, the Q-statistics (Cochran, 1954), and the I2-statistics (Higgins et al., 2003) indicate high levels of heterogeneity in the data. If correlation coefficients in MASEM are heterogeneous across studies, applying a random-effects approach and conducting moderation analysis is recommended (Jak, 2015). Therefore, OSMASEM is a fitting choice for conducting further analysis (Jak and Cheung, 2022). Additionally, the statistical power of the combined set of studies was calculated for each relationship (Muncer et al., 2003). A value greater than 0.5 indicates that the meta-analysis has sufficient power to detect a meaningful effect size (Van Vaerenbergh et al., 2014), which holds true for all relationships except the one between alternative attractiveness and switching costs. Finally, the skewness statistic was calculated for each relationship to assess the symmetry of the distribution of correlation coefficients. A value within the range of −1 to +1 generally indicates an acceptable level of skewness (Hair et al., 2013), which holds true for all relationships.
5.3 Main effects
Table 2 shows the results for the main effects of the conceptual framework. The overall model fit can be considered adequate based on fit indices (χ2/d.f. = 1.021, RMSEA = 0.001, SRMR = 0.078, TLI = 0.999, CFI = 0.999). The results reveal that dissatisfaction (β = 0.229, p < 0.001, H1) and alternative attractiveness (β = 0.418, p < 0.001, H3) increase switching intention, which has a positive effect on switching behavior (β = 0.557, p < 0.001, H4). Moreover, the results demonstrate that switching costs decrease switching intention (β = −0.159, p < 0.001, H2). Thus, all hypotheses regarding the main effects can be confirmed.
Table A2 (see Appendix) reports the results for the main effects of the conceptual framework, excluding switching behavior. Since this model has zero degrees of freedom, the fit statistics are not informative. The results reveal only minimal variation in the parameter estimates. As previously mentioned, in line with recommendations by Jak and Cheung (2020), this model is used for the moderation analysis due to an insufficient number of correlations including switching behavior.
5.4 Moderation analysis
Table 5 displays the results of the moderation analysis. Provider switch vs technology switch was assessed as a categorical moderator. The omnibus test indicates no significant effect overall (χ2(3) = 4.038, p = 0.257). However, the individual moderating effects reveal that provider vs technology switch significantly moderates the effect of alternative attractiveness on switching intention. Specifically, the coefficient equals −0.093, meaning that, on average, the effect of alternative attractiveness on switching intention is 0.093 smaller for provider switches.
Similarly, monetary switching costs vs non-monetary switching costs were assessed as a categorical moderator. The omnibus test again indicates no significant effect overall (χ2(3) = 5.442, p = 0.142). Nevertheless, the individual moderating effects reveal that the effect of alternative attractiveness on switching intention is significantly moderated by monetary vs non-monetary switching costs. The coefficient equals −0.100, which means that, on average, the effect of alternative attractiveness on switching intention is 0.100 smaller when monetary switching costs are present.
Furthermore, the moderation analysis examined contractual relation vs non-contractual relation as a categorical moderator. Neither the omnibus test (χ2(3) = 2.040, p = 0.564) nor the individual effects reveal significant findings.
Student sample vs non-student sample was also assessed as a categorical moderator. The omnibus test indicates no significant effect overall (χ2(3) = 3.828, p = 0.281), and the individual moderating effects show no significant effects.
Similarly, mostly student sample vs not mostly student sample was assessed as a categorical moderator. The omnibus test reveals a significant effect overall (χ2(3) = 13.339, p = 0.004). Furthermore, the individual moderating effects show that the effect of dissatisfaction on switching intention is significantly moderated by sample type. The coefficient is 0.135, which means that, on average, the effect of dissatisfaction on switching intention is 0.135 greater if the sample consists mostly of students. Moreover, the effect of switching costs on switching intention is significantly moderated by sample type. In detail, the coefficient is equal to −0.185, meaning that, on average, the effect of switching costs on switching intention is −0.185 smaller if the sample consists mostly of students. Considering the overall negative average effect of switching costs on switching intention, this reduction implies an intensification of the negative impact, suggesting a stronger negative influence of switching costs on switching intention if the sample consists mostly of students.
The moderation analysis also examined tangible vs intangible as a categorical moderator. Neither the omnibus test (χ2(3) = 0.823, p = 0.844) nor the individual effects show significant findings.
Moreover, IS vs no IS was assessed as a categorical moderator. The omnibus test indicates a significant overall effect (χ2(3) = 8.808, p = 0.032). Furthermore, the individual moderating effects demonstrate that the effect of dissatisfaction on switching intention is significantly moderated by IS vs no IS. The coefficient is 0.146, which means that, on average, the effect of dissatisfaction on switching intention is 0.146 larger in the context of ISs.
Age (median, standardized) and gender (% male, standardized) were assessed as continuous moderators. For both variables, the omnibus tests show no significant effects overall (Age: χ2(3) = 2.919, p = 0.404, Gender: χ2(3) = 1.523, p = 0.677), and the individual moderating effects also indicate no significant findings.
Region (Asia vs other) was assessed as a categorical moderator. The omnibus test shows no significant effect overall (χ2(3) = 1.824, p = 0.610), and the individual moderating effects reveal no significant effects.
Similarly, journal vs no journal was assessed as a categorical moderator. Neither the omnibus test (χ2(3) = 4.295, p = 0.231) nor the individual effects suggest significant effects.
Scopus Cite Score (standardized) and citations per document (standardized) were assessed as continuous moderators. While the omnibus test for Scopus Cite Score (χ2(3) = 8.189, p = 0.042) indicates a significant overall effect, the individual moderating effects reveal no significant findings. For citations per document, neither the omnibus test (χ2(3) = 0.399, p = 0.940) nor the individual moderating effects show significant results.
Finally, Hofstede’s six cultural dimensions (standardized) were assessed as continuous moderators. The omnibus tests suggest no significant effect overall (Power distance: χ2(3) = 1.688, p = 0.640, Individualism: χ2(3) = 2.262, p = 0.520, Masculinity: χ2(3) = 0.651, p = 0.885, Uncertainty avoidance: χ2(3) = 2.154, p = 0.541, Long-term orientation: χ2(3) = 0.488, p = 0.921, Indulgence: χ2(3) = 2.024, p = 0.567), and the individual moderating effects reveal no significant effects for any of the six dimensions.
Chapter 6 offers an interpretation of the results and discusses their theoretical and managerial implications.
6. Discussion and implications
Heterogeneity and resulting incomparability between PPM model studies are prevalent issues. After nearly 2 decades of research on the PPM model, this is the first study to provide a comprehensive overview and meta-analytical test of the PPM model to address this problem. The findings reveal several important theoretical (Chapter 6.1) and managerial implications (Chapter 6.2).
6.1 Theoretical implications
First, this study reveals widespread heterogeneity and resulting incomparability in the selection, categorization, and measurement of push, pull, and mooring variables. Among the 148 PPM model studies identified as eligible in the data collection process, 382 different independent variables were used. As shown in Figure 2, the three most frequently used and unequivocally categorizable variables within the top ten are dissatisfaction (push), alternative attractiveness (pull), and switching costs (mooring). Moreover, there is considerable heterogeneity in the measurement of these variables. Given that appropriate construct measurement is essential for the validity and comparability of research results, and that even small variations in measurement can have a significant impact, this finding is very concerning (Bergkvist and Langner, 2017). Thus, future research should use the conceptual framework, which includes dissatisfaction, alternative attractiveness, switching costs, switching intention, and switching behavior, as a baseline model for PPM model studies to enhance comparability. Additionally, the measurement of these variables should be harmonized by appropriately using the most frequently cited scales, as shown in Table 3. The use of these scales is recommended not only due to their prevalent adoption in existing PPM model studies, but also because they demonstrate content validity, ensuring that the measurements accurately reflect the definitions of the corresponding variables (Bergkvist and Langner, 2017).
Second, the results emphasize the predictive power of the PPM model. Previously, anecdotal evidence suggested that the PPM model explained more variance in the dependent variables switching intention and switching behavior than comparable models. For example, Lim and Choi (2017) and Chang et al. (2014) examined switching intention in the context of social networks. Lim and Choi (2017) explained 37% of the variance in switching intention using a conceptual model based on the stressor-strain-outcome framework. In contrast, Chang et al. (2014) achieved a variance explanation of 58% using the PPM model. Similarly, Goode (2015) and Wu et al. (2017) investigated switching intention in the context of cloud storage services. Goode (2015), using a model based on the technology acceptance model, explained 49% of the variance in switching intention. Wu et al. (2017), using the PPM model, explained 59% of the variance, demonstrating the superior predictive power of the model. This study substantiates these findings with meta-analytical evidence: The model, which consists of only the three most frequently used PPM variables, explains 30% of the variance in switching intention and 31% of the variance in switching behavior (see Table 2). Therefore, future PPM model studies should include dissatisfaction, alternative attractiveness, and switching costs as baseline independent variables. Implementing this recommendation should not only improve predictive capacity, but also reduce the previously mentioned prevailing arbitrariness and heterogeneity in selecting independent variables in PPM model studies. While this provides a strong foundation, researchers are encouraged to explore and evaluate the inclusion of additional independent variables to determine their ability to explain additional variance. The inclusion of affective variables (e.g. anxiety, regret, affective commitment) would be particularly interesting, as dissatisfaction, alternative attractiveness, and switching costs are cognitive variables.
Third, this study substantially advances the evaluation of nomological validity within the PPM literature. Nomological validity refers to the degree to which predictions in a formal theoretical network containing a construct of interest are confirmed (Bagozzi, 1981). In the context of nomological validity, it is advisable to conduct multiple replications of model tests across large samples to reduce the likelihood of encountering false positive results due to chance (Hagger et al., 2017). Once sufficient evidence is available, these replications should be subjected to MASEM, which serves as a powerful tool to assess the collective evidence supporting a nomological network, while also correcting for methodological inadequacies (Hagger et al., 2016; Cheung and Hong, 2017). MASEM provides a rigorous test of nomological validity by drawing on the aggregated evidence from numerous replications. This approach is particularly robust because it is based on multiple replications and incorporates corrections for potential biases that could lead researchers to erroneous conclusions regarding the acceptance or rejection of the nomological network (Hagger et al., 2017). Currently, the PPM literature encompasses a substantial number of replications for the most commonly observed effects, but it lacks a comprehensive MASEM analysis. Consequently, this study is notably valuable in addressing this deficiency. In this regard, the unique contribution of this study relative to existing meta-analyses becomes clear. While there have been meta-analyses exploring specific relationships—such as Pick and Eisend (2014), who investigated the relationship between switching costs and switching—none have examined the comprehensive nomological network that encompasses dissatisfaction, alternative attractiveness, switching costs, switching intention, and switching behavior collectively. Moreover, this meta-analysis is distinct in that it only includes studies using the PPM model.
Fourth, the results show that dissatisfaction and alternative attractiveness positively influence switching intention, which increases switching behavior. Moreover, the results show that switching costs decrease switching intention. These effect directions align with expectations derived from theoretical considerations, confirming the hypotheses. Regarding effect sizes, the results show that switching intention exerts a large positive influence on switching behavior, alternative attractiveness exerts a large positive influence on switching intention, dissatisfaction exerts a medium positive influence on switching intention and switching costs exert a small negative influence on switching intention (Cohen, 1988). These findings provide a benchmark for future studies, offering reliable estimates of the true effect sizes. This helps researchers conduct a priori analyses, enabling them to calculate the necessary sample size based on the required significance level, the desired statistical power, and the to be detected effect size (Cohen, 1988). Additionally, by establishing these effect sizes, future research can move beyond merely confirming relationships and instead focus on exploring moderators and boundary conditions, thereby advancing theory development. The fact that alternative attractiveness exerts the largest influence on switching intention is congruent with migration research, as factors of the destination dominate the decision-making process for geographical migration decisions (Walmsley et al., 1998). This is because the choice of a destination often precedes the decision to move, as a visit to the destination triggers a reassessment of the conditions at the place of origin (Stimson and Minnery, 1998).
Fifth, the findings reveal a significant effect of the first substantive moderator, provider vs technology switch. Specifically, the positive effect of alternative attractiveness on switching intention is weaker for provider switches than for technology switches, confirming H5. This finding suggests that consumers perceive switching between technologies as a more significant upgrade than switching between providers. Thus, researchers should differentiate between provider and technology switches in future studies, as this distinction enables the development of more precise theoretical frameworks, enhancing the predictive capacity of future research.
Sixth, the findings show a significant effect of the second substantive moderator, monetary vs non-monetary switching costs. Concretely, the positive effect of alternative attractiveness on switching intention is weaker when switching costs are monetary, confirming H6b. This finding suggests that consumers perceive alternatives as a less significant upgrade when switching costs are monetary. H6a could not be confirmed, indicating no significant differences between monetary and non-monetary switching costs in their capacity to reduce switching intention. While unexpected, this finding aligns with previous meta-analytical evidence (Pick and Eisend, 2014).
Seventh, the findings show no significant effect of the third substantive moderator, contractual vs non-contractual relation. Namely, the effect of switching costs on switching intention does not differ between contractual and non-contractual relations, rejecting H7. As hypothesized, some consumers might experience heightened switching costs in contractual relations. However, others may anticipate and accept these costs from the outset, as they are predetermined in contractual relations, reducing the likelihood that these costs will influence future switching decisions (Pick and Eisend, 2014).
Eighth, the results indicate two significant effects of the control moderator, mostly student sample vs not mostly student sample. Specifically, the positive effect of dissatisfaction on switching intention is stronger for samples consisting mostly of students, and switching costs have a stronger negative influence on switching intention if the sample consists mostly of students. Students tend to have more limited budgets, making switching costs (especially monetary) more significant for them. This heightened financial sensitivity could increase the negative impact of switching costs on their switching intention. Moreover, dissatisfaction among students could more easily induce a reassessment of spending priorities, leading to increased switching intention.
Ninth, the results show a significant effect of the control moderator, IS vs no IS. Concretely, the positive effect of dissatisfaction on switching intention is stronger in the context of IS. Research has shown that dissatisfaction is the primary reason users discontinue their incumbent IS (Parthasarathy and Bhattacherjee, 1998). ISs frequently involve rapidly evolving technologies, where user expectations are constantly increasing. Therefore, any perceived shortcomings may lead to increased dissatisfaction due to these high expectations and, thus, increased switching intention.
Finally, while no further significant moderating effects were identified at the individual level, the omnibus test of all three moderating effects reveals a significant overall effect for the Scopus Cite Score, an indicator of journal quality. A closer examination of the individual moderating effects that approached significance, together with the other indicators of publication medium quality, suggests that studies published in high-quality media tend to report smaller main effects than studies published in low-quality media. This finding is consistent with other meta-analyses in consumer research (e.g. Lütjens et al., 2022). One possible explanation is that high-quality media tend to require more rigorous research designs (Palmatier, 2016), which help minimize methodological deficiencies that can otherwise lead to exaggerated effect sizes (Fleming et al., 2014).
6.2 Managerial implications
Switching behavior in a business context inherently involves two entities: the one that loses a customer and the one that gains a customer. This dual nature of switching underlines that the managerial implications are always twofold, affecting both retention and acquisition strategies. This meta-analysis shows that companies can proactively motivate customers to switch by (1) demonstrating and communicating the benefits and competitive advantages of their service (alternative attractiveness), (2) highlighting the shortcomings of the competing service (dissatisfaction), and by (3) making it as easy and cheap as possible for customers to switch to their service (switching costs).
To enhance alternative attractiveness, companies should clearly demonstrate and communicate the unique benefits and competitive advantages of their offerings. This may include showcasing unique features, superior quality, or better pricing. Furthermore, freemium business models (Gu et al., 2018) or offering free trials seem to be highly promising initiatives for companies, especially considering that, according to the results, alternative attractiveness exerts the strongest influence on switching intention. This is because when customers experience a service from a different company for the first time, it may cause them to reconsider the adequacy of their currently used service. To capitalize on dissatisfaction with the competing offers, companies should identify gaps in their competitors’ offerings and position their services as solutions to these gaps. Finally, to reduce switching costs, effective strategies may include offering incentives to new customers, simplifying the switching process, reimbursing customers for financial expenses associated with switching (Bergel and Brock, 2018), and providing excellent customer support.
Furthermore, the findings illustrate the benefits of a combined marketing approach, aligning with insights from previous research (Chuah et al., 2017). Rather than focusing on just one of the determinants of switching behavior, companies should aim to address all of them simultaneously. For example, a retail bank could enhance its marketing communication by moving beyond a one-sided advertising message such as “Open a savings account with us and enjoy a 5% interest rate (alternative attractiveness)”. Instead, a holistic advertising message could be “Open a savings account with us and enjoy a 5% interest rate (alternative attractiveness), significantly higher than the 2% interest rate offered by your current bank (dissatisfaction), and take advantage of our free account transfer service to ease the transition (switching costs)”. In applying this, managers should acknowledge that customers vary widely in their responsiveness to marketing approaches, and that not all customers are equally advantageous to the company (Piha and Avlonitis, 2015).
In addition to these recommendations, companies should also consider the impact of various moderators on the main effects, as outlined in Chapter 6.1 of this study. Understanding these factors can guide more nuanced and effective managerial decisions. For example, companies targeting students should prioritize the management of switching costs, as these play a particularly important role in influencing students’ switching decisions. By tailoring their approaches to specific contexts, companies can more accurately address the unique challenges and opportunities presented by different customer segments and scenarios.
7. Limitations and future research directions
The results and implications of this meta-analysis should be considered in light of some important limitations, which also suggest new avenues for future research. First, as mentioned above, the PPM model is a sound theoretical framework to include different predictor variables commonly associated with switching intention and behavior by categorizing them as push, pull, or mooring factors (Bansal et al., 2005). Given the wide range of predictor variables within the PPM literature (see Figure 2), this meta-analysis focused on the three most commonly used variables due to methodological constraints. Future meta-analyses in this area could expand the model to include additional variables as the volume of relevant studies increases. Moreover, due to the methodological constraint of using correlation matrices as input and not having access to the underlying data, this meta-analysis was unable to model the interaction effects required to test switching costs as a moderator of the effects of dissatisfaction and alternative attractiveness on switching intention. According to migration research, mooring factors can moderate the relationships between push and pull factors and the migration decision (Lee, 1966), and their moderating role has been empirically demonstrated in service research by Bansal et al. (2005). Further investigation into the moderating role of mooring factors in future research would contribute to a more detailed understanding of switching and thus holds significant theoretical and managerial value.
Second, the number of PPM model studies examining actual switching behavior is relatively small, with most research focusing on switching intention due to the easier study design. The PPM literature would benefit greatly from more empirical results on actual switching behavior.
Third, the PPM model has been applied in various contexts, with some receiving more attention than others (see Chapter 5.1). The PPM literature would be enriched by additional studies in lesser-explored contexts.
Fourth, the accuracy of the meta-analytical results is affected by suboptimal methodological reporting in some of the included studies. Comprehensive reporting should particularly include a correlation matrix, a complete description of the sample, the measurement instruments used, their sources, and the reliability coefficients of these instruments. Moreover, as noted above, construct measurement differs across the studies included in this meta-analysis, affecting the accuracy of the results.
Finally, this meta-analysis is among the first in service research to use the OSMASEM approach developed by Jak and Cheung (2020). This new MASEM approach has several advantages over other methods (Jak and Cheung, 2022; Steinmetz and Block, 2022). Thus, it should be used more frequently in future meta-analyses when applicable. Table 6 outlines key research directions within themes for future PPM model studies.
8. Conclusion
The aim of this study was to address the prevailing heterogeneity and the resulting incomparability in the application of the PPM model and, based on this, to develop proposals to harmonize future research. Among the 148 PPM model studies identified as eligible in the data collection process, 382 different independent variables were used. The three most frequently used and unequivocally categorizable variables within the top ten are dissatisfaction (push), alternative attractiveness (pull), and switching costs (mooring). Moreover, considerable heterogeneity in the measurement of these variables was found. Through MASEM, dissatisfaction, alternative attractiveness, and switching costs were found to significantly influence switching intention, explaining 30% of the variance in switching intention, which explains 31% of the variance in switching behavior, thus making them well suited predictor variables in their intended domain. Additionally, these relationships are influenced by moderators, revealing interesting new insights for researchers and practitioners. Future research should use the conceptual framework, which includes dissatisfaction, alternative attractiveness, switching costs, switching intention, and switching behavior, as a baseline model for future PPM model studies to enhance comparability and achieve high predictive power. Additionally, the measurement of these variables should be harmonized by appropriately using the most frequently cited scales, as shown in Table 3.
Figures
Figure 1
Data collection process
Figure 2
Most frequently used PPM variables
Figure 3
Conceptual framework
Issues in PPM model studies and meta-analytical insights to improve future research
| Issue | Status quo | Future research based on meta-analytical insights |
|---|---|---|
| Predictor variable choice | Predictor variables are selected arbitrarily across PPM model studies without a clear rationale | This meta-analytical review identifies the ten most frequently used predictor variables in PPM model studies, providing a reference point. This enables researchers to justify their variable selection based on cumulative evidence |
| Predictor variable categorization | Predictor variables are categorized inconsistently across PPM model studies | The meta-analytical review identifies the categorization for the ten most frequently used predictor variables in PPM model studies, providing a reference point. This allows researchers to justify their variable categorization based on cumulative evidence |
| Measurement | Predictor variables and dependent variables are measured inconsistently across PPM model studies | This meta-analytical review provides an overview of the most frequently used valid measurement scales, encouraging standardization. This enables researchers to select measurement scales that improve comparability across studies |
| Effect sizes | Effect sizes differ across PPM model studies with no benchmark for comparison | The meta-analytical review provides a benchmark for effect sizes. This allows researchers to identify anomalous results and ensure appropriate sample sizes in their study designs |
| Effect directions | Effect directions differ across PPM model studies with no benchmark for comparison | This meta-analytical review provides a benchmark for effect directions. This enables researchers to identify anomalous results |
| Variance explained | Variance explained differs across PPM model studies with no benchmark for comparison | This meta-analytical review provides a benchmark for variance explained. This enables researchers to identify anomalous results |
| Moderators | Individual level moderators are investigated | This meta-analytical review provides insights into study level moderators. This provides researchers with a further explanation for differences in effect sizes |
Source(s): Author’s own work
OSMASEM results
| Parameters | K | N | Estimate | Standard error | z-value | p-value |
|---|---|---|---|---|---|---|
| Hypothesized paths | ||||||
| DS → SI | 46 | 17,466 | 0.229 | 0.036 | 6.429 | <0.001 |
| SC → SI | 74 | 25,779 | −0.159 | 0.046 | −3.480 | <0.001 |
| AA → SI | 53 | 20,138 | 0.418 | 0.032 | 12.855 | <0.001 |
| SI → SB | 15 | 5,978 | 0.557 | 0.045 | 12.253 | <0.001 |
| Correlations | ||||||
| DS ↔ AA | 18 | 7,477 | 0.271 | 0.046 | 5.889 | <0.001 |
| DS ↔ SC | 30 | 10,913 | 0.030 | 0.053 | 0.556 | 0.578 |
| AA ↔ SC | 28 | 9,296 | 0.001 | 0.060 | 0.010 | 0.992 |
Note(s): R2SI = 30%, R2SB = 31%, k: Number of correlations per relation, N: Total number of respondents across k samples, DS: Dissatisfaction, SC: Switching Costs, AA: Alternative Attractiveness, SI: Switching Intention, SB: Switching Behavior
Source(s): Author’s own work
Most frequently cited measurement scales
| Scale | Items | Source |
|---|---|---|
| Dissatisfaction | How do you feel about your overall experience with [service]?
| Bhattacherjee (2001) |
| Switching costs |
| Bansal et al. (2005) |
| Alternative attractiveness |
| Bansal et al. (2005) |
| Switching intention |
| Kim et al. (2006) |
| Switching behavior | Did you switch your [service] in the past 2 months? | Bansal et al. (2005) |
Note(s): Dissatisfaction: 7-point semantic differential scale; the endpoints were reversed from the original to measure Dissatisfaction rather than Satisfaction, as is commonly done in PPM studies; Switching Costs, Alternative Attractiveness, Switching Intention: 7-point Likert scale ranging from strongly disagree to strongly agree; Switching Behavior: Yes/No; measured 2 months later than the other variables; [service] needs to be replaced depending on the research context
Source(s): Author’s own work
Descriptive statistics
| k | N | r | rmin | rmax | Q | I2 | Power | Skewness | |
|---|---|---|---|---|---|---|---|---|---|
| DS ↔ SI | 46 | 17,466 | 0.334 | −0.190 | 0.680 | 931.109** | 95.167% | >0.999 | −0.458 |
| SC ↔ SI | 74 | 25,779 | −0.143 | −0.715 | 0.831 | 9692.828** | 99.247% | >0.999 | 0.499 |
| AA ↔ SI | 53 | 20,138 | 0.477 | 0.060 | 0.924 | 3948.301** | 98.683% | >0.999 | 0.275 |
| SI ↔ SB | 15 | 5,978 | 0.545 | 0.124 | 0.810 | 427.419** | 96.724% | >0.999 | −0.618 |
| DS ↔ AA | 18 | 7,477 | 0.270 | −0.035 | 0.646 | 473.803** | 96.412% | >0.999 | 0.225 |
| DS ↔ SC | 30 | 10,913 | 0.030 | −0.466 | 0.685 | 1713.256** | 98.307% | 0.879 | 0.402 |
| AA ↔ SC | 28 | 9,296 | 0.003 | −0.580 | 0.852 | 4946.656** | 99.454% | 0.057 | 0.515 |
Note(s): k: Number of correlations per relation, N: Total number of respondents across k samples, r: mean correlation, rmin: minimum correlation, rmax: maximum correlation, Q: Q-statistic, I2: I2-statistic, Power: Power test using N as sample size and α = 0.05, *p < 0.05, **p < 0.01
Source(s): Author’s own work
OSMASEM moderation analysis results
| Moderator | DS → SI | SC → SI | AA → SI |
|---|---|---|---|
| Substantive moderators | |||
| Provider switch (N = 49) vs technology switch (N = 51) | 0.024 (0.055) | −0.016 (0.072) | −0.093 (0.045) * |
| Monetary switching costs (N = 26) vs non-monetary switching costs (N = 46) | −0.024 (0.070) | −0.093 (0.075) | −0.100 (0.050) * |
| Contractual relation (N = 14) vs non-contractual relation (N = 84) | −0.094 (0.083) | −0.062 (0.093) | 0.054 (0.067) |
| Control moderators | |||
| Student sample (N = 13) vs non-student sample (N = 87) | 0.067 (0.069) | −0.127 (0.086) | 0.031 (0.081) |
| Mostly student sample (N = 28) vs not mostly student sample (N = 72) | 0.135 (0.055) * | −0.185 (0.070) ** | −0.031 (0.050) |
| Tangible (N = 10) vs intangible (N = 90) | −0.052 (0.109) | −0.073 (0.139) | −0.040 (0.076) |
| IS (N = 81) vs no IS (N = 19) | 0.146 (0.072) * | 0.143 (0.103) | 0.095 (0.061) |
| Age (N = 89) | −0.035 (0.025) | −0.008 (0.035) | −0.014 (0.026) |
| Gender (N = 98) | −0.020 (0.029) | 0.007 (0.036) | −0.019 (0.029) |
| Asia (N = 78) vs other region (N = 14) | −0.012 (0.069) | 0.095 (0.086) | 0.067 (0.081) |
| Journal (N = 85) vs no journal (N = 15) | −0.104 (0.066) | 0.118 (0.094) | 0.033 (0.066) |
| Scopus cite score (N = 35) | −0.064 (0.034) | −0.048 (0.49) | −0.058 (0.040) |
| Citations per document (N = 88) | 0.001 (0.026) | −0.011 (0.034) | −0.014 (0.024) |
| Power distance (N = 92) | −0.003 (0.026) | 0.042 (0.033) | 0.002 (0.023) |
| Individualism (N = 92) | 0.009 (0.027) | −0.032 (0.032) | −0.037 (0.031) |
| Masculinity (N = 92) | 0.018 (0.026) | −0.006 (0.034) | 0.003 (0.022) |
| Uncertainty avoidance (N = 92) | −0.027 (0.026) | −0.029 (0.035) | 0.019 (0.023) |
| Long-term orientation (N = 92) | 0.002 (0.032) | 0.004 (0.034) | 0.020 (0.030) |
| Indulgence (N = 92) | 0.022 (0.029) | −0.009 (0.034) | −0.034 (0.025) |
Note(s): N: Number of samples, *p < 0.05, **p < 0.01, DS: Dissatisfaction, SC: Switching Costs, AA: Alternative Attractiveness, SI: Switching Intention
Source(s): Author’s own work
Future research directions within themes
| Theme | Future research directions |
|---|---|
| Improve comparability among PPM studies | Use dissatisfaction, alternative attractiveness, and switching costs as baseline independent variables to predict switching intention (and switching behavior) |
| Use the most cited, valid measurement scales provided in this study (see Table 3) | |
| When measuring other independent variables, provide the exact measurement scales that were used | |
| In addition, reporting should include a correlation matrix, a complete description of the sample, the sources and the reliability coefficients of the measurement scales | |
| Expansion of PPM variables | Explore additional independent variables to determine their ability to explain more variance |
| The inclusion of affective variables (e.g. anxiety, regret, affective commitment) could be particularly interesting, as dissatisfaction, alternative attractiveness, and switching costs are cognitive variables | |
| Moreover, the inclusion of context-specific variables (e.g. digital literacy in digital switching contexts) could be interesting and help to explain additional variance | |
| When exploring additional independent variables, be sure to provide a sound theoretical argumentation of whether the variables are push, pull, or mooring variables | |
| Explore moderators | Table 4 offers meta-analytical insights into the effects of study level moderators, providing explanations for differences in effect sizes. To further advance theory development, primary studies should investigate the effects of individual level moderators |
| Examine the role of mooring factors in moderating the relationships between push and pull factors and switching | |
| More studies on actual switching behavior | Conduct studies on actual switching behavior, as there is currently a lack of such studies, to improve practical relevance |
| In doing so, investigate variables that influence the intention-behavior gap (e.g. habit) to better understand why switching intention does not always result in actual switching behavior | |
| When available, use objective data sources (e.g. transaction records, customer logs) to validate self-reported measures of switching behavior | |
| More studies in underexplored or new research contexts | Conduct more studies in underexplored or new consumer service switching research contexts (see Chapter 5.1) to enrich the PPM literature |
| Explore B2B switching contexts, where push, pull, and mooring factors, as well as their influence on switching, may differ significantly from those in consumer service switching |
Source(s): Author’s own work
OSMASEM results with attenuation corrections
| Parameters | k | N | Estimate | Standard error | z-value | p-value |
|---|---|---|---|---|---|---|
| Hypothesized paths | ||||||
| DS → SI | 46 | 17,466 | 0.235 | 0.042 | 5.559 | <0.001 |
| SC → SI | 74 | 25,779 | −0.182 | 0.054 | −3.371 | <0.001 |
| AA → SI | 51 | 19,182 | 0.457 | 0.037 | 12.412 | <0.001 |
| SI → SB | 15 | 5,978 | 0.638 | 0.052 | 12.202 | <0.001 |
| Correlations | ||||||
| DS ↔ AA | 18 | 7,477 | 0.307 | 0.053 | 5.842 | <0.001 |
| DS ↔ SC | 30 | 10,913 | 0.034 | 0.060 | 0.557 | 0.578 |
| AA ↔ SC | 28 | 9,296 | 0.001 | 0.070 | 0.016 | 0.987 |
Note(s): R2SI = 36%, R2SB = 41%, k: Number of correlations per relation, N: Total number of respondents across k samples, DS: Dissatisfaction, SC: Switching Costs, AA: Alternative Attractiveness, SI: Switching Intention, SB: Switching Behavior
Source(s): Author’s own work
OSMASEM results without switching behavior
| Parameters | K | N | Estimate | Standard error | z-value | p-value |
|---|---|---|---|---|---|---|
| Hypothesized paths | ||||||
| DS → SI | 46 | 17,466 | 0.227 | 0.037 | 6.163 | <0.001 |
| SC → SI | 74 | 25,779 | −0.150 | 0.046 | −3.257 | <0.001 |
| AA → SI | 53 | 20,138 | 0.416 | 0.032 | 12.829 | <0.001 |
| Correlations | ||||||
| DS ↔ AA | 18 | 7,477 | 0.271 | 0.046 | 5.886 | <0.001 |
| DS ↔ SC | 30 | 10,913 | 0.029 | 0.053 | 0.553 | 0.580 |
| AA ↔ SC | 28 | 9,296 | 0.001 | 0.060 | 0.015 | 0.988 |
Note(s): R2SI = 30%, k: Number of correlations per relation, N: Total number of respondents across k samples, DS: Dissatisfaction, SC: Switching Costs, AA: Alternative Attractiveness, SI: Switching Intention
Source(s): Author’s own work
The supplementary material for this article can be found online.
References
Agadullina, E.R. and Lovakov, A.V. (2018), “Are people more prejudiced towards groups that are perceived as coherent? A meta-analysis of the relationship between out-group entitativity and prejudice”, British Journal of Social Psychology, Vol. 57 No. 4, pp. 703-731, doi: 10.1111/bjso.12256.
Ajzen, I. (1991), “The theory of planned behavior”, Organizational Behavior and Human Decision Processes, Vol. 50 No. 2, pp. 179-211, doi: 10.1016/0749-5978(91)90020-T.
Al-Mashraie, M., Chung, S.H. and Jeon, H.W. (2020), “Customer switching behavior analysis in the telecommunication industry via push-pull-mooring framework: a machine learning approach”, Computers and Industrial Engineering, Vol. 144, 106476, doi: 10.1016/j.cie.2020.106476.
Anderson, E.W. and Sullivan, M.W. (1993), “The antecedents and consequences of customer satisfaction for firms”, Marketing Science, Vol. 12 No. 2, pp. 125-143, doi: 10.1287/mksc.12.2.125.
Bagozzi, R.P. (1981), “Attitudes, intentions and behavior: a test of some key hypotheses”, Journal of Personality and Social Psychology, Vol. 41 No. 4, pp. 607-627, doi: 10.1037/0022-3514.41.4.607.
Bansal, H.S., Taylor, S.F. and St James, Y. (2005), “‘Migrating’ to new service providers: toward a unifying framework of consumers' switching behaviors”, Journal of the Academy of Marketing Science, Vol. 33 No. 1, pp. 96-115, doi: 10.1177/0092070304267928.
Becker, B.J. (2000), “Multivariate meta-analysis”, in Tinsley, H.E.A. and Brown, S.D. (Eds), Handbook of Applied Multivariate Statistics and Mathematical Modeling, Academic Press, pp. 499-525.
Bergel, M. and Brock, C. (2018), “The impact of switching costs on customer complaint behavior and service recovery evaluation”, Journal of Service Theory and Practice, Vol. 28 No. 4, pp. 458-483, doi: 10.1108/JSTP-02-2017-0035.
Bergkvist, L. and Langner, T. (2017), “Construct measurement in advertising research”, Journal of Advertising, Vol. 46 No. 1, pp. 129-140, doi: 10.1080/00913367.2017.1281778.
Bhattacherjee, A. (2001), “Understanding information systems continuance: an expectation-confirmation model”, MIS Quarterly, Vol. 25 No. 3, pp. 351-370, doi: 10.2307/3250921.
Black, H.G., Childers, C.Y. and Vincent, L.H. (2014), “Service characteristics' impact on key service quality relationships: a meta-analysis”, Journal of Services Marketing, Vol. 28 No. 4, pp. 276-291, doi: 10.1108/JSM-12-2012-0261.
Blut, M., Wang, C. and Schoefer, K. (2016), “Factors influencing the acceptance of self-service technologies: a meta-analysis”, Journal of Service Research, Vol. 19 No. 4, pp. 396-416, doi: 10.1177/1094670516662352.
Bogue, D.J. (1969), Principles of Demography, John Wiley & Sons, New York.
Bölen, M.C. (2020), “From traditional wristwatch to smartwatch: understanding the relationship between innovation attributes, switching costs and consumers' switching intention”, Technology in Society, Vol. 63, 101439, doi: 10.1016/j.techsoc.2020.101439.
Borenstein, M., Hedges, L.V., Higgins, J.P.T. and Rothstein, H.R. (2021), Introduction to Meta‐analysis, 2nd ed., John Wiley & Sons, New York.
Brown, S.P. and Stayman, D.M. (1992), “Antecedents and consequences of attitude toward the ad: a meta-analysis”, Journal of Consumer Research, Vol. 19 No. 1, pp. 34-51, doi: 10.1086/209284.
Brown, S.P., Homer, P.M. and Inman, J.J. (1998), “A meta-analysis of relationships between ad-evoked feelings and advertising responses”, Journal of Marketing Research, Vol. 35 No. 1, pp. 114-126, doi: 10.1177/002224379803500111.
Burnham, T.A., Frels, J.K. and Mahajan, V. (2003), “Consumer switching costs: a typology, antecedents, and consequences”, Journal of the Academy of Marketing Science, Vol. 31 No. 2, pp. 109-126, doi: 10.1177/0092070302250897.
Carrillat, F.A., Legoux, R. and Hadida, A.L. (2018), “Debates and assumptions about motion picture performance: a meta-analysis”, Journal of the Academy of Marketing Science, Vol. 46 No. 2, pp. 273-299, doi: 10.1007/s11747-017-0561-6.
Chang, I.-C., Liu, C.-C. and Chen, K. (2014), “The push, pull and mooring effects in virtual migration for social networking sites”, Information Systems Journal, Vol. 24 No. 4, pp. 323-346, doi: 10.1111/isj.12030.
Chen, P.-Y. and Hitt, L.M. (2002), “Measuring switching costs and the determinants of customer retention in internet-enabled businesses: a study of the online brokerage industry”, Information Systems Research, Vol. 13 No. 3, pp. 255-274, doi: 10.1287/isre.13.3.255.78.
Cheng, S., Lee, S.-J. and Choi, B. (2019), “An empirical investigation of users' voluntary switching intention for mobile personal cloud storage services based on the push-pull-mooring framework”, Computers in Human Behavior, Vol. 92, pp. 198-215, doi: 10.1016/j.chb.2018.10.035.
Cheung, M. W.-L. (2015), “metaSEM: an R package for meta-analysis using structural equation modeling”, Frontiers in Psychology, Vol. 5, 1521, doi: 10.3389/fpsyg.2014.01521.
Cheung, M. W.-L. and Chan, W. (2005), “Meta-analytic structural equation modeling: a two-stage approach”, Psychological Methods, Vol. 10 No. 1, pp. 40-64, doi: 10.1037/1082-989X.10.1.40.
Cheung, M.W.L. and Hong, R.Y. (2017), “Applications of meta-analytic structural equation modeling in health psychology: examples, issues, and recommendations”, Health Psychology Review, Vol. 11 No. 3, pp. 265-279, doi: 10.1080/17437199.2017.1343678.
Chou, S.-Y., Shen, G.C., Chiu, H.-C. and Chou, Y.-T. (2016), “Multichannel service providers' strategy: understanding customers' switching and free-riding behavior”, Journal of Business Research, Vol. 69 No. 6, pp. 2226-2232, doi: 10.1016/j.jbusres.2015.12.034.
Chuah, S. H.-W., Rauschnabel, P.A., Marimuthu, M., Thurasamy, R. and Nguyen, B. (2017), “Why do satisfied customers defect? A closer look at the simultaneous effects of switching barriers and inducements on customer loyalty”, Journal of Service Theory and Practice, Vol. 27 No. 3, pp. 616-641, doi: 10.1108/JSTP-05-2016-0107.
Cochran, W. (1954), “The combination of estimates from different experiments”, Biometrics, Vol. 10 No. 1, pp. 101-129, doi: 10.2307/3001666.
Cohen, J. (1988), Statistical Power Analysis for the Behavioral Sciences, 2nd ed., Lawrence Erlbaum Associates, London.
Cortina, J.M. (2003), “Apples and oranges (and Pears, Oh My!): the search for moderators in meta-analysis”, Organizational Research Methods, Vol. 6 No. 4, pp. 415-439, doi: 10.1177/1094428103257358.
De Matos, C.A. and Rossi, C.A.V. (2008), “Word-of-mouth communications in marketing: a meta-analytic review of the antecedents and moderators”, Journal of the Academy of Marketing Science, Vol. 36 No. 4, pp. 578-596, doi: 10.1007/s11747-008-0121-1.
De Mooij, M. and Hofstede, G. (2002), “Convergence and divergence in consumer behavior: implications for international retailing”, Journal of Retailing, Vol. 78 No. 1, pp. 61-69, doi: 10.1016/S0022-4359(01)00067-7.
Dick, A.S. and Basu, K. (1994), “Customer loyalty: toward an integrated conceptual framework”, Journal of the Academy of Marketing Science, Vol. 22 No. 2, pp. 99-113, doi: 10.1177/0092070394222001.
Eisend, M. (2006), “Two-sided advertising: a meta-analysis”, International Journal of Research in Marketing, Vol. 23 No. 2, pp. 187-198, doi: 10.1016/j.ijresmar.2005.11.001.
Eisend, M. (2017), “Meta-analysis in advertising research”, Journal of Advertising, Vol. 46 No. 1, pp. 21-35, doi: 10.1080/00913367.2016.1210064.
Fang, Y.-H. and Li, C.-Y. (2022), “Does the sharing economy change conventional consumption modes?”, International Journal of Information Management, Vol. 67, 102552, doi: 10.1016/j.ijinfomgt.2022.102552.
Feng, C., Ma, R. and Jiang, L. (2021), “The impact of service innovation on firm performance: a meta-analysis”, Journal of Service Management, Vol. 32 No. 3, pp. 289-314, doi: 10.1108/JOSM-03-2019-0089.
Fleming, P.S., Koletsi, D., Seehra, J. and Pandis, N. (2014), “Systematic reviews published in higher impact clinical journals were of higher quality”, Journal of Clinical Epidemiology, Vol. 67 No. 7, pp. 754-759, doi: 10.1016/j.jclinepi.2014.01.002.
Floyd, D.L., Prentice-Dunn, S. and Rogers, R.W. (2000), “A meta‐analysis of research on protection motivation theory”, Journal of Applied Social Psychology, Vol. 30 No. 2, pp. 407-429, doi: 10.1111/j.1559-1816.2000.tb02323.x.
Fu, J.-R. (2011), “Understanding career commitment of IT professionals: perspectives of push-pull-mooring framework and investment model”, International Journal of Information Management, Vol. 31 No. 3, pp. 279-293, doi: 10.1016/j.ijinfomgt.2010.08.008.
Goode, S. (2015), “A note on service switching: evidence from cloud storage services”, Services Marketing Quarterly, Vol. 36 No. 2, pp. 153-172, doi: 10.1080/15332969.2015.1014240.
Gu, X., Kannan, P.K. and Ma, L. (2018), “Selling the premium in freemium”, Journal of Marketing, Vol. 82 No. 6, pp. 10-27, doi: 10.1177/0022242918807170.
Gummesson, E. (1995), “Relationship marketing: its role in the service economy”, in Glynn, W.J. and Barnes, J.G. (Eds), Understanding Services Management, John Wiley & Sons, pp. 244-268.
Hagger, M.S., Chan, D.K.C., Protogerou, C. and Chatzisarantis, N.L.D. (2016), “Using meta-analytic path analysis to test theoretical predictions in health behavior: an illustration based on meta-analyses of the theory of planned behavior”, Preventive Medicine, Vol. 89, pp. 154-161, doi: 10.1016/j.ypmed.2016.05.020.
Hagger, M.S., Gucciardi, D.F. and Chatzisarantis, N.L.D. (2017), “On nomological validity and auxiliary assumptions: the importance of simultaneously testing effects in social cognitive theories applied to health behavior and some guidelines”, Frontiers in Psychology, Vol. 8, pp. 1-14, doi: 10.3389/fpsyg.2017.01933.
Hair, J.F., Black, W.C., Babin, B.J. and Anderson, R.E. (2013), Multivariate Data Analysis, 7th ed., Pearson, London.
Harzing, A.-W. (2022), “Journal quality list”, 69th ed., available at: Harzing.com
Hedges, L.V. (2016), “Applying meta-analysis to structural equation modeling”, Research Synthesis Methods, Vol. 7 No. 2, pp. 209-214, doi: 10.1002/jrsm.1214.
Hedges, L.V. and Vevea, J.L. (1998), “Fixed- and random-effects models in meta-analysis”, Psychological Methods, Vol. 3 No. 4, pp. 486-504, doi: 10.1037/1082-989X.3.4.486.
Heide, J.B. and Weiss, A.M. (1995), “Vendor consideration and switching behavior for buyers in high-technology markets”, Journal of Marketing, Vol. 59 No. 3, pp. 30-43, doi: 10.1177/002224299505900303.
Higgins, J., Thompson, S.G., Deeks, J.J. and Altman, D.G. (2003), “Measuring inconsistency in meta-analysis”, British Medical Journal, Vol. 327 No. 7414, pp. 557-560, doi: 10.1136/bmj.327.7414.557.
Hogreve, J., Iseke, A., Derfuss, K. and Eller, T. (2017), “The service-profit chain: a meta-analytic test of a comprehensive theoretical framework”, Journal of Marketing, Vol. 81 No. 3, pp. 41-61, doi: 10.1509/jm.15.0395.
Hou, A.C.Y., Chern, C.-C., Chen, H.-G. and Chen, Y.-C. (2011), “‘Migrating to a new virtual world’: exploring MMORPG switching through human migration theory”, Computers in Human Behavior, Vol. 27 No. 5, pp. 1892-1903, doi: 10.1016/j.chb.2011.04.013.
Hsieh, P.-J. (2021), “Understanding medical consumers' intentions to switch from cash payment to medical mobile payment: a perspective of technology migration”, Technological Forecasting and Social Change, Vol. 173, 121074, doi: 10.1016/j.techfore.2021.121074.
Hsieh, J.-K., Hsieh, Y.-C., Chiu, H.-C. and Feng, Y.-C. (2012), “Post-adoption switching behavior for online service substitutes: a perspective of the push–pull–mooring framework”, Computers in Human Behavior, Vol. 28 No. 5, pp. 1912-1920, doi: 10.1016/j.chb.2012.05.010.
Hunter, J.E. and Schmidt, F.L. (2004), Methods of Meta-Analysis: Correcting Error and Bias in Research Findings, 2nd ed., Sage Publications, New York.
Jackson, J.A. (1986), Aspects of Modern Society: Migration, Longman, London.
Jak, S. (2015), Meta-Analytic Structural Equation Modelling, Springer, Berlin.
Jak, S. and Cheung, M. W.-L. (2020), “Meta-analytic structural equation modeling with moderating effects on SEM parameters”, Psychological Methods, Vol. 25 No. 4, pp. 430-455, doi: 10.1037/met0000245.
Jak, S. and Cheung, M. W.-L. (2022), “Can findings from meta-analytic structural equation modeling in management and organizational psychology be trusted?”, Preprint Article. doi: 10.31234/osf.io/b3qvn.
Jak, S., Li, H., Kolbe, L., de Jonge, H. and Cheung, M. W.-L. (2021), “Meta‐analytic structural equation modeling made easy: a tutorial and web application for one‐stage MASEM”, Research Synthesis Methods, Vol. 12 No. 5, pp. 590-606, doi: 10.1002/jrsm.1498.
Jones, M.A., Mothersbaugh, D.L. and Beatty, S.E. (2000), “Switching barriers and repurchase intentions in services”, Journal of Retailing, Vol. 76 No. 2, pp. 259-274, doi: 10.1016/S0022-4359(00)00024-5.
Jones, M.A., Mothersbaugh, D.L. and Beatty, S.E. (2002), “Why customers stay: measuring the underlying dimensions of services switching costs and managing their differential strategic outcomes”, Journal of Business Research, Vol. 55 No. 6, pp. 441-450, doi: 10.1016/S0148-2963(00)00168-5.
Jung, J., Han, H. and Oh, M. (2017), “Travelers' switching behavior in the airline industry from the perspective of the push-pull-mooring framework”, Tourism Management, Vol. 59, pp. 139-153, doi: 10.1016/j.tourman.2016.07.018.
Kahneman, D. and Tversky, A. (1979), “Prospect theory: an analysis of decision under risk”, Econometrica, Vol. 47 No. 2, pp. 263-291, doi: 10.2307/1914185.
Kahneman, D. and Tversky, A. (1984), “Choices, values, and frames”, American Psychologist, Vol. 39 No. 4, pp. 341-350, doi: 10.1037/0003-066X.39.4.341.
Kaur, A. and Kaur, P. (2023), “Rethinking international migration in Punjab: a push-pull-mooring framework”, Journal of International Migration and Integration, Vol. 25 No. 1, pp. 331-358, doi: 10.1007/s12134-023-01077-x.
Kim, G., Shin, B. and Lee, H.G. (2006), “A study of factors that affect user intentions toward email service switching”, Information and Management, Vol. 43 No. 7, pp. 884-893, doi: 10.1016/j.im.2006.08.004.
Laudon, K. and Laudon, J. (2021), Management Information Systems: Managing the Digital Firm, 17th ed., Pearson, New York.
Lee, E.S. (1966), “A theory of migration”, Demography, Vol. 3 No. 1, pp. 47-57, doi: 10.2307/2060063.
Lee, A.S. and Baskerville, R.L. (2003), “Generalizing generalizability in information systems research”, Information Systems Research, Vol. 14 No. 3, pp. 221-243, doi: 10.1287/isre.14.3.221.16560.
Liberati, A., Altman, D.G., Tetzlaff, J., Mulrow, C., Gøtzsche, P.C., Ioannidis, J.P.A., Clarke, M., Devereaux, P.J., Kleijnen, J. and Moher, D. (2009), “The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health care interventions: explanation and elaboration”, PLoS Medicine, Vol. 6 No. 7, e1000100, doi: 10.1371/journal.pmed.1000100.
Lim, M.S. and Choi, S.B. (2017), “Stress caused by social media network applications and user responses”, Multimedia Tools and Applications, Vol. 76 No. 17, pp. 17685-17698, doi: 10.1007/s11042-015-2891-z.
Liu, Z., Zhao, Y., Chen, S., Song, S., Hansen, P. and Zhu, Q. (2021), “Exploring askers' switching from free to paid social Q&A services: a perspective on the push-pull-mooring framework”, Information Processing and Management, Vol. 58 No. 1, 102396, doi: 10.1016/j.ipm.2020.102396.
Longino, C.E., Jr (1992), “The forest and the trees: micro-level considerations in the study of geographic mobility in old age”, in Rogers, A. (Ed.), Elderly Migration and Population Redistribution, Bellhaven, pp. 23-34.
Lütjens, H., Eisenbeiss, M., Fiedler, M. and Bijmolt, T. (2022), “Determinants of consumers' attitudes towards digital advertising – a meta-analytic comparison across time and touchpoints”, Journal of Business Research, Vol. 153, pp. 445-466, doi: 10.1016/j.jbusres.2022.07.039.
Maseeh, H.I., Jebarajakirthy, C., Pentecost, R., Ashaduzzaman, M., Arli, D. and Weaven, S. (2021), “A meta-analytic review of mobile advertising research”, Journal of Business Research, Vol. 136, pp. 33-51, doi: 10.1016/j.jbusres.2021.06.022.
McNeish, D. (2018), “Thanks coefficient alpha, we’ll take it from here”, Psychological Methods, Vol. 23 No. 3, pp. 412-433, doi: 10.1037/met0000144.
Michel, J.S., Viswesvaran, C. and Thomas, J. (2011), “Conclusions from meta-analytic structural equation models generally do not change due to corrections for study artifacts”, Research Synthesis Methods, Vol. 2 No. 3, pp. 174-187, doi: 10.1002/jrsm.47.
Mohd-Any, A.A., Sarker, M. and Hui, F.L.Z. (2023), “Understanding users' switching intention of cloud storage services: a push-pull-mooring framework”, Journal of Consumer Behaviour, Vol. 23 No. 2, pp. 748-768, doi: 10.1002/cb.2239.
Moon, B. (1995), “Paradigms in migration research: exploring ‘moorings’ as a schema”, Progress in Human Geography, Vol. 19 No. 4, pp. 504-524, doi: 10.1177/030913259501900404.
Morgan, R.M. and Hunt, S.D. (1994), “The commitment-trust theory of relationship marketing”, Journal of Marketing, Vol. 58 No. 3, pp. 20-38, doi: 10.1177/002224299405800302.
Muncer, S.J., Craigie, M. and Holmes, J. (2003), “Meta-analysis and power: some suggestions for the use of power in research synthesis”, Understanding Statistics, Vol. 2 No. 1, pp. 1-12, doi: 10.1207/S15328031US0201_01.
Okazaki, S., Eisend, M., Plangger, K., de Ruyter, K. and Grewal, D. (2020), “Understanding the strategic consequences of customer privacy concerns: a meta-analytic review”, Journal of Retailing, Vol. 96 No. 4, pp. 458-473, doi: 10.1016/j.jretai.2020.05.007.
Oliver, R.L. (1980), “A cognitive model of the antecedents and consequences of satisfaction decisions”, Journal of Marketing Research, Vol. 17 No. 4, pp. 460-469, doi: 10.2307/3150499.
Oliver, R.L. (1981), “Measurement and evaluation of satisfaction processes in retail settings”, Journal of Retailing, Vol. 57 No. 3, pp. 25-48.
Palmatier, R.W. (2016), “Improving publishing success at JAMS: contribution and positioning”, Journal of the Academy of Marketing Science, Vol. 44 No. 6, pp. 655-659, doi: 10.1007/s11747-016-0497-2.
Palmatier, R.W., Dant, R.P., Grewal, D. and Evans, K.R. (2006), “Factors influencing the effectiveness of relationship marketing: a meta-analysis”, Journal of Marketing, Vol. 70 No. 4, pp. 136-153, doi: 10.1509/jmkg.70.4.136.
Parthasarathy, M. and Bhattacherjee, A. (1998), “Understanding post-adoption behavior in the context of online services”, Information Systems Research, Vol. 9 No. 4, pp. 362-379, doi: 10.1287/isre.9.4.362.
Peterson, R.A. (2001), “On the use of college students in social science research: insights from a second-order meta-analysis”, Journal of Consumer Research, Vol. 28 No. 3, pp. 450-461, doi: 10.1086/323732.
Peterson, R.A. and Brown, S.P. (2005), “On the use of beta coefficients in meta-analysis”, Journal of Applied Psychology, Vol. 90 No. 1, pp. 175-181, doi: 10.1037/0021-9010.90.1.175.
Pick, D. and Eisend, M. (2014), “Buyers' perceived switching costs and switching: a meta-analytic assessment of their antecedents”, Journal of the Academy of Marketing Science, Vol. 42 No. 2, pp. 186-204, doi: 10.1007/s11747-013-0349-2.
Piha, L.P. and Avlonitis, G.J. (2015), “Customer defection in retail banking: attitudinal and behavioural consequences of failed service quality”, Journal of Service Theory and Practice, Vol. 25 No. 3, pp. 304-326, doi: 10.1108/JSTP-04-2014-0080.
Ping, R.A. Jr (1993), “The effects of satisfaction and structural constraints on retailer exiting, voice, loyalty, opportunism, and neglect”, Journal of Retailing, Vol. 69 No. 3, pp. 320-352, doi: 10.1016/0022-4359(93)90010-G.
Ranjan, K.R., Sugathan, P. and Rossmann, A. (2015), “A narrative review and meta-analysis of service interaction quality: new research directions and implications”, Journal of Services Marketing, Vol. 29 No. 1, pp. 3-14, doi: 10.1108/JSM-01-2014-0029.
Ravenstein, E.G. (1885), “The laws of migration”, Journal of the Statistical Society of London, Vol. 48 No. 2, pp. 167-253, doi: 10.2307/2979181.
Rezvani, Z., Jansson, J. and Bodin, J. (2015), “Advances in consumer electric vehicle adoption research: a review and research agenda”, Transportation Research Part D: Transport and Environment, Vol. 34, pp. 122-136, doi: 10.1016/j.trd.2014.10.010.
Rosenthal, R. (1979), “The file drawer problem and tolerance for null results”, Psychological Bulletin, Vol. 86 No. 3, pp. 638-641, doi: 10.1037/0033-2909.86.3.638.
Sajjad, A., Chu, J., Anwar, M.A. and Asmi, F. (2020), “Between green and gray: smog risk and rationale behind vehicle switching”, Journal of Cleaner Production, Vol. 244, 118674, doi: 10.1016/j.jclepro.2019.118674.
Samuelson, W. and Zeckhauser, R. (1988), “Status quo bias in decision making”, Journal of Risk and Uncertainty, Vol. 1 No. 1, pp. 7-59, doi: 10.1007/BF00055564.
Sarstedt, M., Hair, J.F., Cheah, J.-H., Becker, J.-M. and Ringle, C.M. (2019), “How to specify, estimate, and validate higher-order constructs in PLS-SEM”, Australasian Marketing Journal, Vol. 27 No. 3, pp. 197-211, doi: 10.1016/j.ausmj.2019.05.003.
Steinmetz, H. and Block, J. (2022), “Meta-analytic structural equation modeling (MASEM): new tricks of the trade”, Management Review Quarterly, Vol. 72 No. 3, pp. 605-626, doi: 10.1007/s11301-022-00293-6.
Stimson, R.J. and Minnery, J. (1998), “Why people move to the ‘sun-belt’: a case study of long-distance migration to the Gold Coast, Australia”, Urban Studies, Vol. 35 No. 2, pp. 193-214, doi: 10.1080/0042098984943.
Suh, C.J. and Kim, J.-H. (2018), “Buyers' switching intentions in a manufacturing supply chain: a migration theory perspective”, International Journal of Operations and Production Management, Vol. 38 No. 12, pp. 2246-2265, doi: 10.1108/IJOPM-02-2017-0118.
Sun, Y., Liu, D., Chen, S., Wu, X., Shen, X.-L. and Zhang, X. (2017), “Understanding users' switching behavior of mobile instant messaging applications: an empirical study from the perspective of push-pull-mooring framework”, Computers in Human Behavior, Vol. 75, pp. 727-738, doi: 10.1016/j.chb.2017.06.014.
Valentine, J.C. (2019), “Incorporating judgments about study quality into research syntheses”, in Cooper, H., Hedges, L.V. and Valentine, J.C. (Eds), The Handbook of Research Synthesis and Meta-Analysis, 3rd ed., Russell Sage Foundation, pp. 129-140.
Valentine, J.C., Cheung, M. W.-L., Smith, E.J., Alexander, O., Hatton, J.M., Hong, R.Y., Huckaby, L.T., Patton, S.C., Pössel, P. and Seely, H.D. (2022), “A primer on meta-analytic structural equation modeling: the case of depression”, Prevention Science, Vol. 23 No. 3, pp. 346-365, doi: 10.1007/s11121-021-01298-5.
Van den Noortgate, W., López-López, J.A., Marín-Martínez, F. and Sánchez-Meca, J. (2013), “Three-level meta-analysis of dependent effect sizes”, Behavior Research Methods, Vol. 45 No. 2, pp. 576-594, doi: 10.3758/s13428-012-0261-6.
Van Vaerenbergh, Y., Orsingher, C., Vermeir, I. and Larivère, B. (2014), “A meta-analysis of relationships linking service failure attributions to customer outcomes”, Journal of Service Research, Vol. 17 No. 4, pp. 381-398, doi: 10.1177/1094670514538321.
Vargo, S.L. and Lusch, R.F. (2004a), “Evolving to a new dominant logic in Marketing”, Journal of Marketing, Vol. 68 No. 1, pp. 1-17, doi: 10.1509/jmkg.68.1.1.24036.
Vargo, S.L. and Lusch, R.F. (2004b), “The four service marketing myths: remnants of a goods-based, manufacturing model”, Journal of Service Research, Vol. 6 No. 4, pp. 324-335, doi: 10.1177/1094670503262946.
Vevea, J.L., Coburn, K. and Sutton, A. (2019), “Publication bias”, in Cooper, H., Hedges, L.V. and Valentine, J.C. (Eds), The Handbook of Research Synthesis and Meta-Analysis, 3rd ed., Russell Sage Foundation, pp. 383-430.
Walmsley, D.J., Epps, W.R. and Duncan, C.J. (1998), “Migration to the new South Wales North coast 1986-1991: lifestyle motivated counterurbanisation”, Geoforum, Vol. 29 No. 1, pp. 105-118, doi: 10.1016/S0016-7185(97)00023-7.
Wathne, K.H., Biong, H. and Heide, J.B. (2001), “Choice of supplier in embedded markets: relationship and marketing program effects”, Journal of Marketing, Vol. 65 No. 2, pp. 54-66, doi: 10.1509/jmkg.65.2.54.18254.
Wolpert, J. (1965), “Behavioral aspects of the decision to migrate?”, Papers – Regional Science Association, Vol. 15 No. 1, pp. 159-169, doi: 10.1111/j.1435-5597.1965.tb01320.x.
Wolpert, J. (1966), “Migration as an adjustment to environmental stress”, Journal of Social Issues, Vol. 22 No. 4, pp. 92-102, doi: 10.1111/j.1540-4560.1966.tb00552.x.
Wu, K., Vassileva, J. and Zhao, Y. (2017), “Understanding users' intention to switch personal cloud storage services: evidence from the Chinese market”, Computers in Human Behavior, Vol. 68, pp. 300-314, doi: 10.1016/j.chb.2016.11.039.
Ye, C. and Potter, R.E. (2011), “The role of habit in post-adoption switching of personal information technologies: an empirical investigation”, Communications of the Association for Information Systems, Vol. 28, pp. 585-610, doi: 10.17705/1CAIS.02835.
Zengyan, C., Yinping, Y. and Lim, J. (2009), “Cyber migration: an empirical investigation on factors that affect users' switch intentions in social networking sites”, Proceedings of the 42nd Hawaii International Conference on System Sciences, Waikoloa, HI. doi: 10.1109/HICSS.2009.140.