Using a Markov-switching approach to modelling value co-creation
Using a Markov-Switching Approach to Modelling Value Co-creation Yen-Hao Hsieh
Wei-Ting Chen
Tamkang University No.151, Yingzhuan Rd., Danshui Dist., New Taipei City 25137, Taiwan 886-2-26215656 ext. 2744
Tamkang University No.151, Yingzhuan Rd., Danshui Dist., New Taipei City 25137, Taiwan 886-2-26215656 ext. 2744
[email protected]
[email protected]
ABSTRACT Value co-creation within a service system has been an important issue for enterprises in order to increase their competition and advantages. By recognizing value variations, enterprises effortlessly figure out the difficulty of resource allocation and make the right service strategy. Consequently, this study employed a call center as an example to investigate value variation. This study collected real data from the call center of Kaohsiung city government between January/2012 and December/2015. Meanwhile, this study also adopted a Markovswitching model approach to analyze value co-creation and codestruction within the call center. The results revealed that both value co-creation and co-destruction can be modelled by the Markov-switching model approach.
CCS Concepts Theory of computation → Theory and algorithms for application domains→ Markov decision processes
Keywords Markov-switching; value co-creation; service system; call center
1. INTRODUCTION Services have been a key role to influence economic activities nowadays [1]. That is, service-dominant logic becomes an important concept for enterprises to take into account [16]. Enterprises pay attention to their services which not only fulfill customer’s needs but also increase customer’s positive feelings and emotions. The process of service delivery can be regarded as a service system. Hence, offering appropriate services with high quality service operations and environments enable customers to acquire valuable merits within the service system [6]. Consequently, enterprises can design interactive activities with services for customers to take part in the service system. According to service provisions that are defined by enterprises, different values would be created by configuring existing resources. Both enterprises and customers co-create values through the interactions of services [17]. However, values are very abstract and not easily to be observed and measured. Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for components of this work owned by others than the author(s) must be honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. Request permissions from [email protected]. ICEC '16, August 17 - 19, 2016, Suwon, Republic of Korea Copyright is held by the owner/author(s). Publication rights licensed to ACM. ACM 978-1-4503-4222-3/16/08…$15.00
DOI: http://dx.doi.org/10.1145/2971603.2971619
Understanding the levels of values within a service system becomes an essential topic which helps enterprises to examine if delivered services successfully suffice customers [11]. A service system can be regarded as a complex process that values could dynamically enhance or decrease [11]. Moreover, values could generate both positive and negative effects based on different situations (e.g., service environments, operation strategies or different types of customers). Service delivery is a continuous process that results in value variations. When enterprises optimize their resource allocation, value co-creation (i.e., the positive effect) must be generated between enterprises and customers [9]. If enterprises and customers waste existing resources, value co-destruction (i.e., the negative effect) will emerge from the relationship between enterprises and customers [13]. Hence, values of service delivery in a service system could change from a high (low) level to a low (high) level. Enterprises have pay attention to value variations while they arrange resources to provide customers with services. By recognizing value variations, enterprises effortlessly figure out the difficulty of resource allocation and make the right service strategy. Based on the research background and motivation, this study tries to explore the following research question: how do enterprises recognize and predict the trend of value variations within a service system? This study aims to apply Markov-switching model to build a reference model of value variations in service-dominant logic.
2. LITERATURE REVIEW 2.1 Value Co-creation and Value Codestruction Value co-creation is the key element within service systems based on the notion of service science and the concept of servicedominant logic. Value results from the beneficial applications of operant resources which are transmitted through operand resources or goods [17]. The co-creation of value is a desirable purpose as it can enable enterprises to focus on customers’ viewpoints and to improve the front-end process of identifying customer needs and wants [17]. Ramaswamy (2011) proposed that value co-creation is a procedure by which mutual value is expanded together. Value co-creation is regarded as a cooperative process including service interactions between businesses and customers [14]. GrÖ nroos (2008) investigated that customers play the role of creators to perform a value foundation through facilitation process [3]. Besides, the connotation of “value” is to not only distinguish what kind of value but also verify the roles of value [15]. The implication of “co-” is to understand what kind of resources can
be used for creating values [15]. “Creation” means that what kind of the approach can help enterprises and customers to produce values. Furthermore, GrÖ nroos (2008) mentioned that enterprises and customers play different but important roles in the value cocreation process [3]. An enterprise is a value facilitator who offers customers a foundation for value creation by employing existing resources. A customer is a value creator within service systems where existing resources available to customers and skills held by them are added.
viewpoint. In the role layer, there are main key roles including service providers and customers within the service system. In the resource layer, both enterprises and customers should properly utilize their resources to deliver and acquire services respectively. The resources include people, information, technology and organization. When service providers and customers can optimize the usage of existing resources (that is, existing resources are interacting with each other), the service system must result in value co-creation.
According to the service-dominant logic, value creation is from the utilization of resources. LoÏc Plé and Cáceres (2010) noted that service interactions within service systems could result in positive effects when enterprises and customers effectively utilize existing resources. Moreover, negative effects also could be generated within service systems while they misuse existing resources [13]. The definition of value co-destruction is that enterprises inappropriately use their resources or implement incorrect service operation processes to offer customers dissatisfied services.
This study also defines two core values including the utilitarian value and the hedonic value in the value layer [10]. These values will result from the interactions between service providers and customers in the service system. In the value variation layer, values can be changed and presented as the predictable status by using the Markov-switching model approach given values are unobserved data. According to this framework, this study attempts to define suitable indicators for evaluating values co-creation within service systems.
Hence, how to effectively and properly allocate existing resources to delivery services for enterprises would possibly result in the positive effect (value co-creation) and the negative effect (value co-destruction). That is, values within service systems could dynamically change due to the different situations of resource allocation even though enterprises ideally aim to propose the positive effect to customers. Managing values variation within service systems is an essential topic for enterprises to take into account.
Value variation
P=
P11 p12
St1= expansion duration
Value co-creation
P21 p22
St2= contraction duration
Value co-destruction
Value
value
Utilitarian value
2.2 Markov-Switching Model The Markov-switching model is an autoregressive model to identify the different statuses of data that are regarded from different populations. Based on the concept of the Markov chain, the trend and the variations of data become predictable. Meanwhile, irregular data can be represented as suitable statuses through the Markov-switching model approach. Hamilton (1989) proposed the Markov-switching model to figure out the non-linear problems and analyze the variations of parameters in the autoregressive model by using discrete statuses, especially in the economic field [4]. Schwartz (2011) investigated the relationships between the labor markets and the economic business cycle in America [15]. Krolzig (1997) extended the Markov-switching model to study the economic issues by analyzing the transformation information of different statuses [7]. Lee et al. (2013) employed the Markovswitching model approach to study the relations between the real estate index and the economic business cycle in Taiwan [8]. Consequently, the Markov-switching model is an appropriate way to identify and recognize data as a representable status, especially for the unobservable data. This study also would like to investigate value variations in service systems. It is difficult for enterprises to clearly ascertain what values present. This study adopts the Markov-switching model to examine the variation levels of values and clarifies the transformation statuses of values within different situations.
3. RESEARCH METHODOLOGY 3.1 Research Framework Figure 1 shows our research framework which includes four layers. The description of the framework is from the bottom-up
useful
Hedonic value
effective
delight
Resource interaction Provider
customer
People(Pp)
People (Pc) Technology (Tc)
Technology(Tp)
Share information (Ic)
Share information(Ip)
Organization (Oc)
Organization(Op)
Role Provider
customer
Expect state
Resource
Resource
Figure 1. Research Framework
3.2 Research Method The Markov-switching model approach is based on the concept of time series that needs to examine the features of data. That is, data can be divided into two types including stationary and nonstationary. It is feasible to predict and recognize the potential pattern and trend of stationary data that is with a stable structure. The common tests are Augmented Dickey-Fuller test (so-called
ADF test) and Phillips-Perron test (so-called PP test). Unit-root test is to ensure if data belongs to stationary data. There are three types of unit-root test as follows. i. ∆𝑌𝑡 = γ𝑌𝑡−1 + ∑𝑃𝑖=2 𝛽𝑖 ∆𝑌𝑡−𝑖+1 + 𝜀𝑡 ii. ∆𝑌𝑡 = 𝑎0 + γ𝑌𝑡−1 + ∑𝑃𝑖=2 𝛽𝑖 ∆𝑌𝑡−𝑖+1 + 𝜀𝑡 iii. ∆𝑌𝑡 = 𝑎0 + γ𝑌𝑡−1 + 𝑎1 𝑡 + ∑𝑃𝑖=2 𝛽𝑖 ∆𝑌𝑡−𝑖+1 + 𝜀𝑡 When the result of unit-root test belongs to non-stationary time series, there is need to implement the first order difference for the data. If the result of the first order difference still belongs to nonstationary time series, the second order difference is required. That is, the data should be suitable for stationary time series. Second, optimal lag selection is the key for time series to effectively predict the future trend of data. When the value of lag section is large, the performance of the Markov-switching model is poor. If the value of lag section is small, the bias of the Markovswitching model could be brought about [8]. Hence, this study would like to adopt Akaike Information Criterion and Schwarz Information Criterion to find out the most appropriate lag for the Markov-switching model. After unit-root test and optimal lag selection, the data can be applied to the Markov-switching model. This study uses the Markov-switching model for two statuses which was built by Hamilton (1989). The formula is as follows. yt − μst = ∅1 (yt−1 − μst−1 ) − ⋯ − ∅k (yt−k − μst−k ) + ϵ where yt means the original data, μst presents the variations of yt , and ϵ ~ i.i.id.(0, σ2). The Markov-switching model is in accordance with the principle of Markov chain. Pr[St = j|St−1 = i, St−2 = j, ⋯ , ] = Pr[St = j|St−1 = i] = Pij Pij presents the transformation probability of two statuses, ij represents the transformation direction of two statuses.
4. Data Analysis and Discussion 4.1 Data Collection
Table 1. Unit-root test of service level UnitRoot test
t value
1%Threshold
5% Threshold
10% Threshold
P value
ADF PP
-5.151591*** -5.256507***
-4.165756 -4.165756
-3.508508 -3.508508
-3.184230 -3.184230
0.0006 0.0004
Table 2. The lag length of service level
Lag 1 2 3 4 5
AIC -1.582924 -1.584002* -1.584002 -1.584002 -1.584002
SIC -1.503418 -1.503418 -1.503418 -1.503418 -1.503418
This study defines two statuses including the expansion status (socalled status 1) and the contraction status (so-called status 2) in order to present the levels of value variation. Status 1 shows that values can be co-created by service providers and customers. Status 2 represents value co-destruction within service systems. Two statuses can be presented in probabilities. The matrix of transition probability is as follows. P P 0.934078 0.065922 P = [ 11 12 ] = [ ] P21 P22 0.831905 0.168095
Figure 2. Durations of two statuses
This study tries to use a call center as an example that can be regarded as a service system to examine its value variation. A call center is to offer customers customer services that composed of key roles (including service providers and customers) and important resources (including information, technology, organization and people). This study collected open data of the call center of Kaohsiung city government from January/2012 to December/2015. This study used Eviews tool to analyze the collected data. Meanwhile, this study also defines critical values of the call center including service level (utilitarian value), first call resolution (utilitarian value) and customer satisfaction (hedonic value).
4.2 Utilitarian Value Analysis As mentioned earlier, the first step is to ensure if the data belongs to the stationary. Table 1 shows that the results of unit-root test are significant by using ADF test and PP test. That is, t values (5.151591 and -5.256507) are lower than all thresholds (i.e., 1%, 5% and 10% thresholds). Hence, the data of service level can be regarded as the stationary. Then, the lag length is needed to select to find optimal prediction results. This study uses AIC and SIC methods to find out the optimal lag selection. Table 2 shows the AIC and SIC values in different lag lengths. Since the SIC values (-1.503418) are constant, the AIC value (-1.584002) of the second lag length is small. Hence, the optimal lag selection is 2. Figure 3. Transition probability of service level
The probability of status 1 that keeps in itself is 0.934078. The probability of status 2 that keeps in itself is 0.168095. Besides, the durations for status 1 that keeps in itself are about 15.16947 months. The durations for status 2 that keeps in itself are only about 1.20206 months (as depicted in Figure 2). Figure 3 shows the transition probability of two statuses of service level. The results present that service providers and customers can properly employ existing resources to deliver and perceive services. Hence, value co-creation can be resulted from the call center (i.e., a service system). Table 3. Unit-root test of first call resolution rates UnitRoot test ADF PP
t value
1%Threshold
-5.170556*** -5.256459***
-3.577723 -3.577723
5% Threshold -2.925169 -2.925469
10% Threshold -2.600658 -2.600658
P value 0.0001 0.0001
Table 4. The lag length of first call resolution rates
Lag 1 2 3 4 5
AIC -2.891113* -2.869363 -2.869363 -2.869363 -2.869363
SIC -2.811607 -2.811607 -2.811607 -2.811607 -2.811607
The other value is defined as first call resolution rates. The indicator mainly reveals the efficiency of a call center to deal with customers’ problems. Table 3 shows that the results of unit-root test are significant by using ADF test and PP test. That is, t values (-5.170556 and -5.256459) are lower than all thresholds. The data of first call resolution rates can be regarded as the stationary. Table 4 shows the AIC and SIC values in different lag lengths. Since the SIC values (-2.811607) are constant, the AIC value (2.891113) of the second lag length is small. Hence, the optimal lag selection is 1. The probability of status 1 that keeps in itself is 0.978098. The probability of status 2 that keeps in itself is 0. Besides, the durations for status 1 that keeps in itself are about 45.65704 months. The durations for status 2 that keeps in itself are only about 1 month (as depicted in Figure 4). Figure 5 represents the transition probability of two statuses of first call resolution rates. The results show that value co-creation can be also resulted from the call center. 𝑃 𝑃12 0.978098 0.021902 𝑃 = [ 11 ]=[ ] 𝑃21 𝑃22 1.000000 0.000000
Figure 4. Durations of two statuses
Figure 5. Transition probability of first call resolution rates
4.3 Hedonic Value Analysis Customer satisfaction is a critical indicator to represent hedonic value, especially in a call center. Table 5 shows that the results of unit-root test are significant by using ADF test and PP test. That is, t values (-5.060793 and -5.018586) are lower than all thresholds. The data of first call resolution rates can be regarded as the stationary. Table 6 shows the AIC and SIC values in different lag lengths. Since the SIC values (-5.291863) are constant, the AIC value (-5.383747) of the second lag length is small. Hence, the optimal lag selection is 6. Table 5. Unit-root test of customer satisfaction Unit-Root test ADF PP
t value -5.060793 -5.018586
1% Threshold -3.577723 -3.577723
5% Threshold -2.925169 -2.925169
10% Threshold -2.600658 -2.600658
Table 6. The lag length of customer satisfaction
Lag 1 2 3 4 5 6 7 8
AIC -5.383747 -5.383747 -5.383747 -5.383747 -5.383747 -5.383747* -5.370593 -5.370593
SIC -5.291863 -5.291863 -5.291863 -5.291863 -5.291863 -5.291863 -5.291863 -5.291863
The probability of status 1 that keeps in itself is 0.723104. The probability of status 2 that keeps in itself is 0.447005. Besides, the
P value 0.0001 0.0001
durations for status 1 that keeps in itself are about 3.611466 months. The durations for status 2 that keeps in itself are only about 1.808335 months (as shown in Figure 6). Figure 7 shows the transition probability of two statuses of first call resolution rates. The results show that the time for customer satisfaction is longer than customer dissatisfaction. The call center has to maintain customer relationships to increase the durations of their satisfaction. P P 0.723104 0.276896 P = [ 11 12 ] = [ ] P21 P22 0.552995 0.447005
the call center. The results revealed that both value co-creation and co-destruction can be modelled by the Markov-switching model approach. That is, values can be represented in a measurable way given values’ nature is the abstract. By analyzing the two statuses, enterprises can effectively figure out if service providers and customers properly use their existing resources and closely interact with other.
6. ACKNOWLEDGMENTS This work was sponsored by the Ministry of Science and Technology of Republic of China under grant No. MOST 1032410-H-032 -038. The advice and financial support of the MOST are gratefully acknowledged.
7. REFERENCES Figure 6. Transition probability of customer satisfaction
[1] Galetzka, M., Joost W.M. Verhoeven, & Ad Th.H. Pruyn. (2006). Service validity and service reliability of search, experience and credence services: A scenario study. Int J of Service Industry Mgmt, 17(3), 271-283. [2] Gronroos, C. (2006). Adopting a service logic for marketing. Marketing Theory, 6(3), 317-333. Retrieved from [3] Grönroos, C. (2008). Service logic revisited: Who creates value? and who co-creates? European Business Review, 20(4), 298-314. [4] Hamilton, J. D. (1989). A new approach to the economic analysis of nonstationary time series and the business cycle. Econometrica, 57(2), 357-384. [5] Hannu Saarijärvi, P.K. Kannan, & Kuusela, H. (2013). Value co‐creation: Theoretical approaches and practical implications. European Business Review, 25(1), 6-19. [6] Heinonen, K., Strandvik, T., Karl-Jacob Mickelsson, Edvardsson, B., Sundström, E., & Andersson, P. (2010). A customer-dominant logic of service. Journal of Service Management, 21(4), 531-548. [7] Krolzig, H. (1997). The markov-switching vector autoregressive model. In H. Krolzig (Ed.), Markov-switching vector autoregressions: Modelling, statistical inference, and application to business cycle analysis (pp. 6-28). Berlin, Heidelberg: Springer Berlin Heidelberg.
Figure 7. Transition probability of customer satisfaction
5. CONCLUSIONS Value co-creation within a service system has been an important issue for enterprises in order to increase their competition and advantages [17]. Enterprises can make the right decisions and plan the correct strategies, when they can figure out the values of delivered services. Although values are well-known and popular terms, it is difficult for enterprises and scholars to clearly define what values represent. Consequently, this study employed a call center as an example to investigate value variation. This study collected real data from the call center of Kaohsiung city government between January/2012 and December/2015. Meanwhile, this study also adopted a Markov-switching model approach to analyze value co-creation and co-destruction within
[8] Lee, C., Liang, C., & Chou, H. (2013). Identifying taiwan real estate cycle turning points- an application of the multivariate markov-switching autoregressive model. Advances in Management and Applied Economics, 3(2), 123. Retrieved from [9] Maglio, P. P., Nusser, S., & Bishop, K. (2010). A service perspective on IBM's brand. Marketing Review St.Gallen, 27(6), 44-48. [10] Overby, J. W., & Lee, E. (2006). The effects of utilitarian and hedonic online shopping value on consumer preference and intentions. Journal of Business Research, 59(10–11), 1160-1166. [11] Plé, L., & Rubén Chumpitaz Cáceres. (2010). Not always cocreation: Introducing interactional co-destruction of value in service-dominant logic. The Journal of Services Marketing, 24(6), 430-437. [12] Ramaswamy, V. (2011). It's about human experiences… and beyond, to co-creation. Industrial Marketing Management, 40(2), 195-196.
[13] Robertson, N., Polonsky, M., & McQuilken, L. (2014). Are my symptoms serious dr google? A resource-based typology of value co-destruction in online self-diagnosis. Australasian Marketing Journal (AMJ), 22(3), 246-256. [14] Romero, D., & Molina, A. (2011). Collaborative networked organisations and customer communities: Value co-creation and co-innovation in the networking era. Production Planning & Control, 22(5-6), 447-472. [15] Saarijärvi,H., Kannan, PK. and Kuusela, H., (2013) "Value co-creation: theoretical approaches and practical
implications", European Business Review, Vol. 25 Iss: 1, pp.6 – 19 [16] Schwartz, J. (2011). Labor market dynamics over the business cycle: Evidence from markov switching models. Empirical Economics, 43(1), 271-289. [17] Vargo, S. L., & Lusch, R. F. (2004). Evolving to a new dominant logic for marketing. Journal of Marketing, 68(1), 117.
Wei-Ting Chen
Tamkang University No.151, Yingzhuan Rd., Danshui Dist., New Taipei City 25137, Taiwan 886-2-26215656 ext. 2744
Tamkang University No.151, Yingzhuan Rd., Danshui Dist., New Taipei City 25137, Taiwan 886-2-26215656 ext. 2744
[email protected]
[email protected]
ABSTRACT Value co-creation within a service system has been an important issue for enterprises in order to increase their competition and advantages. By recognizing value variations, enterprises effortlessly figure out the difficulty of resource allocation and make the right service strategy. Consequently, this study employed a call center as an example to investigate value variation. This study collected real data from the call center of Kaohsiung city government between January/2012 and December/2015. Meanwhile, this study also adopted a Markovswitching model approach to analyze value co-creation and codestruction within the call center. The results revealed that both value co-creation and co-destruction can be modelled by the Markov-switching model approach.
CCS Concepts Theory of computation → Theory and algorithms for application domains→ Markov decision processes
Keywords Markov-switching; value co-creation; service system; call center
1. INTRODUCTION Services have been a key role to influence economic activities nowadays [1]. That is, service-dominant logic becomes an important concept for enterprises to take into account [16]. Enterprises pay attention to their services which not only fulfill customer’s needs but also increase customer’s positive feelings and emotions. The process of service delivery can be regarded as a service system. Hence, offering appropriate services with high quality service operations and environments enable customers to acquire valuable merits within the service system [6]. Consequently, enterprises can design interactive activities with services for customers to take part in the service system. According to service provisions that are defined by enterprises, different values would be created by configuring existing resources. Both enterprises and customers co-create values through the interactions of services [17]. However, values are very abstract and not easily to be observed and measured. Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for components of this work owned by others than the author(s) must be honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. Request permissions from [email protected]. ICEC '16, August 17 - 19, 2016, Suwon, Republic of Korea Copyright is held by the owner/author(s). Publication rights licensed to ACM. ACM 978-1-4503-4222-3/16/08…$15.00
DOI: http://dx.doi.org/10.1145/2971603.2971619
Understanding the levels of values within a service system becomes an essential topic which helps enterprises to examine if delivered services successfully suffice customers [11]. A service system can be regarded as a complex process that values could dynamically enhance or decrease [11]. Moreover, values could generate both positive and negative effects based on different situations (e.g., service environments, operation strategies or different types of customers). Service delivery is a continuous process that results in value variations. When enterprises optimize their resource allocation, value co-creation (i.e., the positive effect) must be generated between enterprises and customers [9]. If enterprises and customers waste existing resources, value co-destruction (i.e., the negative effect) will emerge from the relationship between enterprises and customers [13]. Hence, values of service delivery in a service system could change from a high (low) level to a low (high) level. Enterprises have pay attention to value variations while they arrange resources to provide customers with services. By recognizing value variations, enterprises effortlessly figure out the difficulty of resource allocation and make the right service strategy. Based on the research background and motivation, this study tries to explore the following research question: how do enterprises recognize and predict the trend of value variations within a service system? This study aims to apply Markov-switching model to build a reference model of value variations in service-dominant logic.
2. LITERATURE REVIEW 2.1 Value Co-creation and Value Codestruction Value co-creation is the key element within service systems based on the notion of service science and the concept of servicedominant logic. Value results from the beneficial applications of operant resources which are transmitted through operand resources or goods [17]. The co-creation of value is a desirable purpose as it can enable enterprises to focus on customers’ viewpoints and to improve the front-end process of identifying customer needs and wants [17]. Ramaswamy (2011) proposed that value co-creation is a procedure by which mutual value is expanded together. Value co-creation is regarded as a cooperative process including service interactions between businesses and customers [14]. GrÖ nroos (2008) investigated that customers play the role of creators to perform a value foundation through facilitation process [3]. Besides, the connotation of “value” is to not only distinguish what kind of value but also verify the roles of value [15]. The implication of “co-” is to understand what kind of resources can
be used for creating values [15]. “Creation” means that what kind of the approach can help enterprises and customers to produce values. Furthermore, GrÖ nroos (2008) mentioned that enterprises and customers play different but important roles in the value cocreation process [3]. An enterprise is a value facilitator who offers customers a foundation for value creation by employing existing resources. A customer is a value creator within service systems where existing resources available to customers and skills held by them are added.
viewpoint. In the role layer, there are main key roles including service providers and customers within the service system. In the resource layer, both enterprises and customers should properly utilize their resources to deliver and acquire services respectively. The resources include people, information, technology and organization. When service providers and customers can optimize the usage of existing resources (that is, existing resources are interacting with each other), the service system must result in value co-creation.
According to the service-dominant logic, value creation is from the utilization of resources. LoÏc Plé and Cáceres (2010) noted that service interactions within service systems could result in positive effects when enterprises and customers effectively utilize existing resources. Moreover, negative effects also could be generated within service systems while they misuse existing resources [13]. The definition of value co-destruction is that enterprises inappropriately use their resources or implement incorrect service operation processes to offer customers dissatisfied services.
This study also defines two core values including the utilitarian value and the hedonic value in the value layer [10]. These values will result from the interactions between service providers and customers in the service system. In the value variation layer, values can be changed and presented as the predictable status by using the Markov-switching model approach given values are unobserved data. According to this framework, this study attempts to define suitable indicators for evaluating values co-creation within service systems.
Hence, how to effectively and properly allocate existing resources to delivery services for enterprises would possibly result in the positive effect (value co-creation) and the negative effect (value co-destruction). That is, values within service systems could dynamically change due to the different situations of resource allocation even though enterprises ideally aim to propose the positive effect to customers. Managing values variation within service systems is an essential topic for enterprises to take into account.
Value variation
P=
P11 p12
St1= expansion duration
Value co-creation
P21 p22
St2= contraction duration
Value co-destruction
Value
value
Utilitarian value
2.2 Markov-Switching Model The Markov-switching model is an autoregressive model to identify the different statuses of data that are regarded from different populations. Based on the concept of the Markov chain, the trend and the variations of data become predictable. Meanwhile, irregular data can be represented as suitable statuses through the Markov-switching model approach. Hamilton (1989) proposed the Markov-switching model to figure out the non-linear problems and analyze the variations of parameters in the autoregressive model by using discrete statuses, especially in the economic field [4]. Schwartz (2011) investigated the relationships between the labor markets and the economic business cycle in America [15]. Krolzig (1997) extended the Markov-switching model to study the economic issues by analyzing the transformation information of different statuses [7]. Lee et al. (2013) employed the Markovswitching model approach to study the relations between the real estate index and the economic business cycle in Taiwan [8]. Consequently, the Markov-switching model is an appropriate way to identify and recognize data as a representable status, especially for the unobservable data. This study also would like to investigate value variations in service systems. It is difficult for enterprises to clearly ascertain what values present. This study adopts the Markov-switching model to examine the variation levels of values and clarifies the transformation statuses of values within different situations.
3. RESEARCH METHODOLOGY 3.1 Research Framework Figure 1 shows our research framework which includes four layers. The description of the framework is from the bottom-up
useful
Hedonic value
effective
delight
Resource interaction Provider
customer
People(Pp)
People (Pc) Technology (Tc)
Technology(Tp)
Share information (Ic)
Share information(Ip)
Organization (Oc)
Organization(Op)
Role Provider
customer
Expect state
Resource
Resource
Figure 1. Research Framework
3.2 Research Method The Markov-switching model approach is based on the concept of time series that needs to examine the features of data. That is, data can be divided into two types including stationary and nonstationary. It is feasible to predict and recognize the potential pattern and trend of stationary data that is with a stable structure. The common tests are Augmented Dickey-Fuller test (so-called
ADF test) and Phillips-Perron test (so-called PP test). Unit-root test is to ensure if data belongs to stationary data. There are three types of unit-root test as follows. i. ∆𝑌𝑡 = γ𝑌𝑡−1 + ∑𝑃𝑖=2 𝛽𝑖 ∆𝑌𝑡−𝑖+1 + 𝜀𝑡 ii. ∆𝑌𝑡 = 𝑎0 + γ𝑌𝑡−1 + ∑𝑃𝑖=2 𝛽𝑖 ∆𝑌𝑡−𝑖+1 + 𝜀𝑡 iii. ∆𝑌𝑡 = 𝑎0 + γ𝑌𝑡−1 + 𝑎1 𝑡 + ∑𝑃𝑖=2 𝛽𝑖 ∆𝑌𝑡−𝑖+1 + 𝜀𝑡 When the result of unit-root test belongs to non-stationary time series, there is need to implement the first order difference for the data. If the result of the first order difference still belongs to nonstationary time series, the second order difference is required. That is, the data should be suitable for stationary time series. Second, optimal lag selection is the key for time series to effectively predict the future trend of data. When the value of lag section is large, the performance of the Markov-switching model is poor. If the value of lag section is small, the bias of the Markovswitching model could be brought about [8]. Hence, this study would like to adopt Akaike Information Criterion and Schwarz Information Criterion to find out the most appropriate lag for the Markov-switching model. After unit-root test and optimal lag selection, the data can be applied to the Markov-switching model. This study uses the Markov-switching model for two statuses which was built by Hamilton (1989). The formula is as follows. yt − μst = ∅1 (yt−1 − μst−1 ) − ⋯ − ∅k (yt−k − μst−k ) + ϵ where yt means the original data, μst presents the variations of yt , and ϵ ~ i.i.id.(0, σ2). The Markov-switching model is in accordance with the principle of Markov chain. Pr[St = j|St−1 = i, St−2 = j, ⋯ , ] = Pr[St = j|St−1 = i] = Pij Pij presents the transformation probability of two statuses, ij represents the transformation direction of two statuses.
4. Data Analysis and Discussion 4.1 Data Collection
Table 1. Unit-root test of service level UnitRoot test
t value
1%Threshold
5% Threshold
10% Threshold
P value
ADF PP
-5.151591*** -5.256507***
-4.165756 -4.165756
-3.508508 -3.508508
-3.184230 -3.184230
0.0006 0.0004
Table 2. The lag length of service level
Lag 1 2 3 4 5
AIC -1.582924 -1.584002* -1.584002 -1.584002 -1.584002
SIC -1.503418 -1.503418 -1.503418 -1.503418 -1.503418
This study defines two statuses including the expansion status (socalled status 1) and the contraction status (so-called status 2) in order to present the levels of value variation. Status 1 shows that values can be co-created by service providers and customers. Status 2 represents value co-destruction within service systems. Two statuses can be presented in probabilities. The matrix of transition probability is as follows. P P 0.934078 0.065922 P = [ 11 12 ] = [ ] P21 P22 0.831905 0.168095
Figure 2. Durations of two statuses
This study tries to use a call center as an example that can be regarded as a service system to examine its value variation. A call center is to offer customers customer services that composed of key roles (including service providers and customers) and important resources (including information, technology, organization and people). This study collected open data of the call center of Kaohsiung city government from January/2012 to December/2015. This study used Eviews tool to analyze the collected data. Meanwhile, this study also defines critical values of the call center including service level (utilitarian value), first call resolution (utilitarian value) and customer satisfaction (hedonic value).
4.2 Utilitarian Value Analysis As mentioned earlier, the first step is to ensure if the data belongs to the stationary. Table 1 shows that the results of unit-root test are significant by using ADF test and PP test. That is, t values (5.151591 and -5.256507) are lower than all thresholds (i.e., 1%, 5% and 10% thresholds). Hence, the data of service level can be regarded as the stationary. Then, the lag length is needed to select to find optimal prediction results. This study uses AIC and SIC methods to find out the optimal lag selection. Table 2 shows the AIC and SIC values in different lag lengths. Since the SIC values (-1.503418) are constant, the AIC value (-1.584002) of the second lag length is small. Hence, the optimal lag selection is 2. Figure 3. Transition probability of service level
The probability of status 1 that keeps in itself is 0.934078. The probability of status 2 that keeps in itself is 0.168095. Besides, the durations for status 1 that keeps in itself are about 15.16947 months. The durations for status 2 that keeps in itself are only about 1.20206 months (as depicted in Figure 2). Figure 3 shows the transition probability of two statuses of service level. The results present that service providers and customers can properly employ existing resources to deliver and perceive services. Hence, value co-creation can be resulted from the call center (i.e., a service system). Table 3. Unit-root test of first call resolution rates UnitRoot test ADF PP
t value
1%Threshold
-5.170556*** -5.256459***
-3.577723 -3.577723
5% Threshold -2.925169 -2.925469
10% Threshold -2.600658 -2.600658
P value 0.0001 0.0001
Table 4. The lag length of first call resolution rates
Lag 1 2 3 4 5
AIC -2.891113* -2.869363 -2.869363 -2.869363 -2.869363
SIC -2.811607 -2.811607 -2.811607 -2.811607 -2.811607
The other value is defined as first call resolution rates. The indicator mainly reveals the efficiency of a call center to deal with customers’ problems. Table 3 shows that the results of unit-root test are significant by using ADF test and PP test. That is, t values (-5.170556 and -5.256459) are lower than all thresholds. The data of first call resolution rates can be regarded as the stationary. Table 4 shows the AIC and SIC values in different lag lengths. Since the SIC values (-2.811607) are constant, the AIC value (2.891113) of the second lag length is small. Hence, the optimal lag selection is 1. The probability of status 1 that keeps in itself is 0.978098. The probability of status 2 that keeps in itself is 0. Besides, the durations for status 1 that keeps in itself are about 45.65704 months. The durations for status 2 that keeps in itself are only about 1 month (as depicted in Figure 4). Figure 5 represents the transition probability of two statuses of first call resolution rates. The results show that value co-creation can be also resulted from the call center. 𝑃 𝑃12 0.978098 0.021902 𝑃 = [ 11 ]=[ ] 𝑃21 𝑃22 1.000000 0.000000
Figure 4. Durations of two statuses
Figure 5. Transition probability of first call resolution rates
4.3 Hedonic Value Analysis Customer satisfaction is a critical indicator to represent hedonic value, especially in a call center. Table 5 shows that the results of unit-root test are significant by using ADF test and PP test. That is, t values (-5.060793 and -5.018586) are lower than all thresholds. The data of first call resolution rates can be regarded as the stationary. Table 6 shows the AIC and SIC values in different lag lengths. Since the SIC values (-5.291863) are constant, the AIC value (-5.383747) of the second lag length is small. Hence, the optimal lag selection is 6. Table 5. Unit-root test of customer satisfaction Unit-Root test ADF PP
t value -5.060793 -5.018586
1% Threshold -3.577723 -3.577723
5% Threshold -2.925169 -2.925169
10% Threshold -2.600658 -2.600658
Table 6. The lag length of customer satisfaction
Lag 1 2 3 4 5 6 7 8
AIC -5.383747 -5.383747 -5.383747 -5.383747 -5.383747 -5.383747* -5.370593 -5.370593
SIC -5.291863 -5.291863 -5.291863 -5.291863 -5.291863 -5.291863 -5.291863 -5.291863
The probability of status 1 that keeps in itself is 0.723104. The probability of status 2 that keeps in itself is 0.447005. Besides, the
P value 0.0001 0.0001
durations for status 1 that keeps in itself are about 3.611466 months. The durations for status 2 that keeps in itself are only about 1.808335 months (as shown in Figure 6). Figure 7 shows the transition probability of two statuses of first call resolution rates. The results show that the time for customer satisfaction is longer than customer dissatisfaction. The call center has to maintain customer relationships to increase the durations of their satisfaction. P P 0.723104 0.276896 P = [ 11 12 ] = [ ] P21 P22 0.552995 0.447005
the call center. The results revealed that both value co-creation and co-destruction can be modelled by the Markov-switching model approach. That is, values can be represented in a measurable way given values’ nature is the abstract. By analyzing the two statuses, enterprises can effectively figure out if service providers and customers properly use their existing resources and closely interact with other.
6. ACKNOWLEDGMENTS This work was sponsored by the Ministry of Science and Technology of Republic of China under grant No. MOST 1032410-H-032 -038. The advice and financial support of the MOST are gratefully acknowledged.
7. REFERENCES Figure 6. Transition probability of customer satisfaction
[1] Galetzka, M., Joost W.M. Verhoeven, & Ad Th.H. Pruyn. (2006). Service validity and service reliability of search, experience and credence services: A scenario study. Int J of Service Industry Mgmt, 17(3), 271-283. [2] Gronroos, C. (2006). Adopting a service logic for marketing. Marketing Theory, 6(3), 317-333. Retrieved from [3] Grönroos, C. (2008). Service logic revisited: Who creates value? and who co-creates? European Business Review, 20(4), 298-314. [4] Hamilton, J. D. (1989). A new approach to the economic analysis of nonstationary time series and the business cycle. Econometrica, 57(2), 357-384. [5] Hannu Saarijärvi, P.K. Kannan, & Kuusela, H. (2013). Value co‐creation: Theoretical approaches and practical implications. European Business Review, 25(1), 6-19. [6] Heinonen, K., Strandvik, T., Karl-Jacob Mickelsson, Edvardsson, B., Sundström, E., & Andersson, P. (2010). A customer-dominant logic of service. Journal of Service Management, 21(4), 531-548. [7] Krolzig, H. (1997). The markov-switching vector autoregressive model. In H. Krolzig (Ed.), Markov-switching vector autoregressions: Modelling, statistical inference, and application to business cycle analysis (pp. 6-28). Berlin, Heidelberg: Springer Berlin Heidelberg.
Figure 7. Transition probability of customer satisfaction
5. CONCLUSIONS Value co-creation within a service system has been an important issue for enterprises in order to increase their competition and advantages [17]. Enterprises can make the right decisions and plan the correct strategies, when they can figure out the values of delivered services. Although values are well-known and popular terms, it is difficult for enterprises and scholars to clearly define what values represent. Consequently, this study employed a call center as an example to investigate value variation. This study collected real data from the call center of Kaohsiung city government between January/2012 and December/2015. Meanwhile, this study also adopted a Markov-switching model approach to analyze value co-creation and co-destruction within
[8] Lee, C., Liang, C., & Chou, H. (2013). Identifying taiwan real estate cycle turning points- an application of the multivariate markov-switching autoregressive model. Advances in Management and Applied Economics, 3(2), 123. Retrieved from [9] Maglio, P. P., Nusser, S., & Bishop, K. (2010). A service perspective on IBM's brand. Marketing Review St.Gallen, 27(6), 44-48. [10] Overby, J. W., & Lee, E. (2006). The effects of utilitarian and hedonic online shopping value on consumer preference and intentions. Journal of Business Research, 59(10–11), 1160-1166. [11] Plé, L., & Rubén Chumpitaz Cáceres. (2010). Not always cocreation: Introducing interactional co-destruction of value in service-dominant logic. The Journal of Services Marketing, 24(6), 430-437. [12] Ramaswamy, V. (2011). It's about human experiences… and beyond, to co-creation. Industrial Marketing Management, 40(2), 195-196.
[13] Robertson, N., Polonsky, M., & McQuilken, L. (2014). Are my symptoms serious dr google? A resource-based typology of value co-destruction in online self-diagnosis. Australasian Marketing Journal (AMJ), 22(3), 246-256. [14] Romero, D., & Molina, A. (2011). Collaborative networked organisations and customer communities: Value co-creation and co-innovation in the networking era. Production Planning & Control, 22(5-6), 447-472. [15] Saarijärvi,H., Kannan, PK. and Kuusela, H., (2013) "Value co-creation: theoretical approaches and practical
implications", European Business Review, Vol. 25 Iss: 1, pp.6 – 19 [16] Schwartz, J. (2011). Labor market dynamics over the business cycle: Evidence from markov switching models. Empirical Economics, 43(1), 271-289. [17] Vargo, S. L., & Lusch, R. F. (2004). Evolving to a new dominant logic for marketing. Journal of Marketing, 68(1), 117.
