Channel Choice: A Literature Review

. The channel choice branch of e-government studies citizens ’ and businesses ’ choice of channels for interacting with government, and how government organizations can integrate channels and migrate users towards the most cost-ef ﬁ cient channels. In spite of the valuable contributions offered no systematic overview exist of channel choice. We present a literature review of channel choice studies in government to citizen context identifying authors, countries, methods, concepts, units of analysis


Introduction
Although the digitization of the public sector has taken place for decades [1] there is still a gap between the availability and uptake of online public services [2]. Even in the countries which are front runners in terms of citizens' adoption of electronic public services citizens keep using traditional channels in addition to online channels either as a supplement or as primary channels [3,4]. The continued use of traditional channels where the interaction takes place between individual citizens and government employees is costly compared to interaction through a website or other forms of self-service applications.
Several literature reviews within e-government have presented and synthesized the findings of studies of citizens' adoption of online services [5][6][7]. However, these studies tend to focus on citizens' intention to adopt an individual e-government service in isolation [7]. The channel choice (CC) literature studies citizens' choice of channels, and the interplay that takes place between citizens' use of channels for interacting with public authorities [8]. In spite of the valuable contributions the CC literature offers, no systematic review of the CC literature exists.
To cover this gap we present a literature review of the CC field in e-government. Our review analyzes 36 papers which study government to citizen interaction (G2C) through more than one type of channel. We combine and expand Webster & Watson's [9] and Schlichter and Kraemmergaard's [10] methods for finding, classifying and analyzing papers.

Scope of Review
Webster and Watson [9, p. xv] recommend that only one level of analysis is included in a literature review unless there is a strong rationale to include several levels. However, the studies within the CC field take place at several levels; insights from the analysis of how citizens choose channels for interaction with public authorities are used to make recommendations to these organizations on how to manage their channels. Due to this connection in the literature we include both levels in our review.
In the CC literature the terms channel or service channel are used to describe the various forms of communication available to citizens to interact with public authorities [11]. Reddick and Anthopolous [4, pp. 400-401] divide these channels into three types: traditional channels (face-to-face, telephone conversations and physical letters), e-government channels (web and e-mail) and new digital media (text messaging, social media and mobile apps). To focus our review, we only include papers which study at least two of these channel types. Further, only papers which study CC in a government to citizen context (G2C) are included. Results from studies of employees' CC may not be transferrable to citizens, as businesses' policies, structures and means of communication can affect employees' behavior. We want to study the managerial aspects of CC in e-government [1], and papers focusing on CC in relation to e-democracy or e-participation are omitted. Finally, due to the rapid technological development of online services, only papers published within the last decade (2005-2014) are included.
The papers are classified according to authors, country and methods based on a framework by Schlichter and Kraemmergaard [10]. Previous literature reviews of the e-government field have criticized scholars for not leaving their offices to collect data, for conducting cross-sectional rather than longitudinal studies, and for not studying what happens inside government organizations [12][13][14]. To find out if this criticism is applicable to the CC literature we expand the method classification to include researchers' involvement in the data collection process, the use of longitudinal studies, and practioners' involvement in the studies. As our topic is CC we also examine if the papers include data on channel traffic. For analyzing the papers we apply Webster and Watson's [9] conceptual analysis matrix identifying objects and level of analysis, conceptual models, and the theoretical frameworks used.
The next section present the methods used to find and analyze the papers in our review. In the third section we present a classification of the papers found, while section four presents the analysis of the papers. In section five we discuss the results with the aim of identifying gaps in the CC literature for future studies. Section six contains concluding remarks and limitations.

Method
The method section is divided into three parts. First we present the search for papers. We then present Schlichter and Kraemmergaard's [10] framework for the classification and Webster and Watson's [9] method for the concept-centric analysis.

The Search for Papers
The papers were found in a three step process following Webster and Watson [9].
1. Search for papers in selected journals and conference proceedings 2. Database search 3. Backwards and forwards searches We began our search for papers in selected journals recognized as core e-government journals by scholars [15] and in the proceedings of EGOV. The first round of searches was conducted in January 2015 using keywords found through an iterative process. An initial series of keywords were supplemented as papers with new keywords were found. Further, inspired by Hofmann et al. [7] we contacted eight experts within the CC field for additional keywords, of which five replied. 13 keywords were used; CRM, channel behavior, channel choice, channel ict architecture, channel integration, channel management, channel marketing, channel strategy, customer relationship management, integrated service delivery, multichannel, multi-channel and orchestrating service delivery.
The keyword search included titles, abstracts, and keywords. After removing duplicates we ended with 239 papers. Papers were included if they focused on CC in a G2C context, included at least two types of channels, were published no later than 2005, and written in English. After reading the abstracts 212 papers were omitted as they only studied one type of channel or were outside the G2C domain. This left 27 papers of which two were omitted as they were inaccessible from the university libraries we had access to. After reading the remaining 25 papers 17 were included in the review.
Webster & Watson recommend that a database search is conducted as the second step to find additional papers. Following the recommendation of an expert in the field, we used the E-government Reference Library (EGRL). We downloaded EGRL version 10.0 (July 2014) to Mendeley Reference Manager for Windows (version 1.13.3) and conducted keyword searches in titles, abstracts and keywords using the 13 keywords. 56 papers were found of which 31 had been found in step 1, two were inaccessible, and one was written in Dutch. This excluded 34, leaving us with 22 papers. After reading these four papers were added to the pool bringing the total to 21.
The third step consisted of using Google Scholar to find papers that either referenced or were referenced to by the 21 papers. 68 papers were found which initially seemed relevant according to our selection criteria. We omitted four conference papers which were earlier editions of journal papers already found. Four papers were unavailable. After reading either the abstracts or the whole papers we were left with 15 relevant papers. These 15 papers were added to the final pool, bringing the total to 36. Appendix A presents an overview of the 36 papers.

Classification of Papers
For the analysis and coding we created a one page template for each paper which contained bibliographical information, abstracts, coding results and notes. This data was entered into a spreadsheet (MS Excel) and analyzed at an aggregated level. The classification of methods follow the framework by Schlichter and Kraemmergaard which they developed for a literature review of the enterprise resource planning field [10]. We removed one method category, archival, as it overlapped with other categories in the papers found. Table 1 presents the classification.
Researchers' involvement in data collection ( Table 2) was coded following Blaikie [16, p. 161]. Longitudinal studies followed Blaikie's definition 'a study extended in time' [16, p. 201]. Practioners' involvement was coded if the authors had direct contact with government organizations' employees through workshops, interviews, surveys etc. Channel traffic was coded if it was presented in numerical form.

Concept-Centric Analysis of the Papers
To synthesize the CC literature we conducted a concept-centric analysis following Webster and Watson [9]. As we read the papers we created a template with the primary concepts covered, and the units of analysis. A pattern quickly emerged; part of the papers studies factors impacting CC at the individual level, while another part studies processes related to multichannel management (MCM) at the organizational level. Webster and Watson state that the conceptual analysis should be supplemented with information on the variables examined, and a conceptual and theoretical analysis of how and why the variables are related [9]. We therefore coded factors, processes and theories applied as well.

Classification of the CC Literature
In this section we present the classification of the papers according to authors, countries and methods applied. We also discuss practioners' involvement and the use of channel traffic in the papers. Table 3 presents an overview of the most prolific authors, while Table 4 presents the papers according to first author's country.

Authors and Country
The majority of the papers were written by a small group of authors from only a few countries. Three scholars have authored or co-authored 21 of the 36 papers.
The papers in the pool are written by first authors from nine different countries. Authors from The Netherlands have published 40 percent of the papers and authors from the Netherlands, US and Canada have published 29 of the 36 papers.  Table 5 presents the papers according to the primary method applied. Four papers are labeled as 'combined' as they rely on several methods.

Methodology
Case studies and surveys are the most frequently applied methods. Eighteen papers include results from surveys, but only eleven use surveys as a primary method; ten study the factors that influence citizens' choice of channels and one studies the adoption of multiple channels in organizations. Twelve paper present individual or multiple case studies, based on documentary material and interviews, workshops or other forms of collaborations with practitioners. Four papers develop theory, and focus mainly on exploring and explaining government organizations' strategies for multichannel management through various theoretical lenses. One paper presents a role-playing game as a method for involving case-workers multichannel management, and the results from applying this method in practice. None of the 16 of the papers which apply qualitative methods relies on one method. Rather, interviews or focus groups discussions are combined or conducted preliminary to a survey. Table 6 presents the highest level of data in the papers according to researchers' involvement. Primary data has been collected for 22 of the 36 papers. Secondary data is used in seven papers which use survey results on individuals' CC and channel satisfaction for statistical modeling. Six papers present only tertiary data, while one paper does not present any data. Times series are used frequently, but only one paper presents a longitudinal study, with six months between data collection points.
There is a high level of practitioner involvement in the papers, largely due to the many case studies based on interviews with employees. Of the 36 papers, 21 include involvement or collaboration with practioners. The authors' biographies reveal that four   [3] authors behind three of the papers [18,19,34] have worked in government organizations, in three cases at the top level. Seven papers presents channel traffic, of which three presents the same data [11,38,40]. Three papers contain a single table or paragraph with channel traffic [8,19,23]. Finally one paper analyzes channel data as a part of a field experiment conducted in 2008 [37]. This data only concerns transactions; however, information inquiries related to the transactions are not presented. Further, except for the field experiment, the latest data on channel traffic is from 2006.

Concept-Centric Analysis of the CC Literature
This section present the concept-centric analysis of the pool of papers following Webster and Watson [9]. During coding we focused on the two overall concepts in the papers; CC which focuses on the factors that influence citizens' choice of channel and MCM which focuses on the processes and issues related government organizations management of multiple channels. Table 7 presents the result of this analysis.
Of the 36 papers 14 study CC at the individual level, while 15 study MCM at the organizational level. There are five papers which overlap these levels, of which two presents the results of field experiments and three are theoretical. One paper, presenting the results of a MCM design study takes place at the group level. One paper [11] does not fit into either level, but focuses on the channels and services delivered, and the development of channel traffic over time. None of the papers study CC at the group level, although a few briefly mention that citizens can also influence each other, or ask each other for help in dealings with public authorities.

Studies at the Individual Level
Of the 19 papers which study citizens' CC for interaction with public authorities three are theoretical and 11 use survey data either for descriptive analysis and/or to test the factors that influence this choice. Four papers explore the factors through qualitative methods, two of which also use surveys. Three papers study the effects of organizations' instruments for channel integration and migration, and how these instruments are Legend: O = organization, G = group, I = individual, * = service channel perceived by citizens. These studies are noteworthy as they cross the boundaries between the individual and organizational unit of analysis.
Most of the studies at the individual level apply variance models to test the impact of independent variables on citizen channel and/or source choice. Nine study citizens' satisfaction with a channel and/or interaction. Satisfaction is both studied as a dependent variable, based on channel chosen, and as an independent variable, where satisfaction with a previous encounter influence future interactions. The factors influencing channel choice have been found through qualitative studies, informed by previous studies, adoptions studies such as TAM [49], marketing theory, and theoretical frameworks from media and communication theory especially Media Richness Theory (MRT) [50], Channel Expansion Theory [51] and Uses and Gratifications research [52]. The papers test a number of different factors. To provide a simple overview we clustered the independent variables into four groups during coding. Note that satisfaction was studied both as an independent and dependent variable. Table 8 presents the factors studied, and the papers which study them.
An alternative to the variance models is presented by Teerling and Pieterson [30] who use a process model to illustrate how governments' marketing efforts and a person's previous experiences also influence channel choice. This model is interesting as it acknowledges that channel choice is not just a psychological process taking place within citizens, but also a social process where citizens can be influenced by external factors. This is important as government organizations can then impact citizens' CC before an interaction takes place.

Studies at the Organizational Level
The 21 papers which take place at the organizational level are much more diverse in terms of topics studied than those at the individual level. Table 9 presents an overview of these topics. Channel integration and migration are the most frequently studied topics, followed by inter-and cross organizational cooperation related to MCM. Due to the limits of this review we only briefly cover the topics here.
Pieterson's studies of government organizations' channel positioning strategies stand out as they are presented in four papers [3,36,38,39]. He uses a process model to illustrate how public authorities can migrate citizens towards the most efficient channels to reduce administrative costs and increase citizen satisfaction. The studies are informed through theories from media science such as MRT, Bordewijk and van Kaam's [53] classification of tele-information services, a historical analysis of government organizations' channel strategies, and through a series of field experiments from the Dutch Channels in Balance project [24,30,36].
Kernaghan discusses the different types of MCM collaboration between government organizations and presents two models to visualize these variations. The first describes inter-and cross organizational partnerships, in terms of actors, services and channels involved [21]. The second model describes the degree to which organizations involved in MCM can be integrated, from informal cooperation, where they share information, to full consolidation, where they give up individual goals and policies and References e-government and e-commerce literature and institutional theory but no explicit theoretical framework [18,21,23,25,28,35] Intermediaries Intermediation theory, marketing theory, transaction cost theory, [19,20] Other (various) Technology adoption models [22,43] become fully harmonized [33]. This is reminiscent of the vertical and horizontal integration of government organizations which is frequently studied in e-government literature, such as Layne and Lee's [54] often cited e-government web-stage model. Kernaghan differs from Layne and Lee, however, in that he does not present consolidation as an inevitable last stage, but rather as one of several strategic options to consider depending on one's needs and resources. In this way Kernaghan avoids the technologic determinism which the web-stage models have been criticized for.
Kernaghan's studies are mostly informed through case studies, especially from Service Canada, rather than any explicit theoretical framework. Klievink and Janssen [25] categorize challenges related to MCM coordination based on a literature review from several fields including e-commerce and e-government. They identify three layers which cover the political, organizational, and information and technological aspects to MCM coordination and present these in an analytical framework. Kernaghan and Flumian discuss similar barriers [18,21] with a stronger emphasis on problems caused by changing political climates and power struggles.
In another study Klievink and Janssen focus on public and private intermediaries [20]. Based on case studies and transaction cost theory they discuss the positive roles intermediaries play in facilitating government to citizen interaction, and the strategies government organizations can employ in relation to them in the shape of a process model. Another perspective on intermediaries comes from Frey and Holden [19] who study the channel conflicts that can arise when private companies appear as intermediaries. The authors apply the theoretical concept of distribution channel management from marketing literature and two case studies to illustrate how government organizations can handle these conflicts. Like Janssen and Klievink they acknowledge the positive role intermediaries can play in MCM. However, Frey and Holden note the importance of protecting the interests of the private companies in addition to those of the government and citizens, while Janssen and Klievink are more concerned with ensuring that citizens have equal access to government services.

Discussion
In this section we discuss the results of our literature review with the aim of identifying methodological and knowledge gaps in the CC literature. Table 10 presents six areas for future CC studies, which could bring the field further forwards.
The CC literature is dominated by a few authors and countries. Many of the papers study actual use and involve practitioners. This limits the places where the studies could have been carried out, as well as their generalizability. Studies from other countries and of specific services could offer valuable contributions to the literature.
Many methods are used to collect and analyze data, but two types of studies stand out; statistical analysis of survey data of citizens' CC, and case studies of MCM at the organizational level. The studies of CC appear more harmonized and coherent than those of MCM. Part of this may be because they are carried out by a small group of authors who cross-reference each other. However, these studies also revolve around one topicindividual's CCuse similar variance models and explicitly refer to the same theoretical frameworks to inform their analyses. There is a strong sense of progress and building on each other's work, and both empirical and theoretical contributions are offered. However, they are largely based on survey data from secondary data sets which the researcher cannot influence. Although a few studies use qualitative studies to inform the survey creation, CC scholars repeatedly state a need for supplementing surveys through qualitative methods [3,4,8,40,41,47].
Methods of direct observation are time consuming to conduct, but they provide valuable contextual information [16] and could inform areas which have only been slightly touched upon; situational constraints, habits and how the service in question and its importance to the citizens influence CC. Observations could study an entire service encounter from the citizens' point of view and the interplay that takes place between channels during such an encounter. This would enable CC scholars to explore citizen initiated requests and explain why these requests occur and gain insight into channel switching and supplementing behavior.
Data on channel traffic could update and supplement the existing knowledge on MCM. Longitudinal studies of channel traffic could be used to evaluate the effects of MCM instruments on citizens' channel behavior. Most of the existing analyses of channel traffic are based on data which is a decade old, and it is unknown if the conclusions based on this data still hold up today.
Future CC studies could examine how citizens influence and help each other when interacting with government organizations. It is striking that the papers in this review focus at only the individual or organization level. There are no studies of CC at the group level, although both private and public intermediaries are mentioned at organizational level, and several studies mention that friends and family members can be intermediaries [25,32,44]. Teerling and Pieterson's process model seems suitable for this task as it illustrates external parties' influence on citizens' CC [30]. A series of conceptual models have been presented to illustrate channel integration and migration, inter-and cross organizational collaboration and barriers to MCM. However, most of the authors seem to either build new models or improve their own. Having presented some of the overlaps in the MCM studies at organizational level here, we would suggest that the existing conceptual models are criticized, tested or synthesized before new models are created. This could lead to a more mature and coherent field. We also recommend that theoretical frameworks are used to inform these models to a higher extent.
Finally we recommend that new field experiments are conducted to study the effects of MCM instruments. The existing studies have been valuable to bridge the individual and organizational levels, but they have been carried out in one country by a small group of scholars. New experiments could contribute by including new service areas, target groups, and MCM instruments. Further they could examine the effects of MCM on all available channels, rather than a few isolated channels. Field experiments could also to examine the effects of MCM instruments on new digital media, which previous experiments have not covered.

Conclusion
This paper has presented an overview of 36 papers from the CC literature found and analyzed following Webster and Watson [9]. The classification of the papers expanded a framework by Schlichter and Kraemmergaard [10]. The importance of supplementing the search for papers with forwards and backwards searches has been demonstrated as new papers were added in each step. Hofmann's method of contacting authors to inquire about keywords proved fruitful [7]. Our analysis has revealed multiple gaps in the CC literature. We have suggested six areas which future studies could address to contribute to the theoretical and empirical development of the CC field.
There are several limitations to our study. Many of the papers were found due to authors citing themselves. This self-citing means the pool of papers revolve around a few authors and countries. The effect may have been strengthened by the sources searched, keywords used, and the fact that papers from certain publishers were inaccessible. It is possible that we may have missed papers for these reasons. Our conceptual analysis is limited to two main areas due to author resources and spatial limitations. A synthesis of results, recommendations for practioners, and a more in-depth discussion of suggestion for future studies were omitted for similar reasons.
Future literature studies could address these limitations by expanding the search, classification and analysis conducted here. An analysis of author keywords, citations, sources and disciplines could illuminate the relationships between the papers and to other fields. Future studies could synthesize and discuss results, suggestions for future studies and recommendations for practioners. We welcome input from scholars on these issues and will gladly share our data for further analysis upon request.