THE EFFECT OF MULTI-CHANNEL INTEGRATION ON U.S. RETAILERS' PERFORMANCE

THE EFFECT OF MULTI-CHANNEL INTEGRATION

ON U.S. RETAILERS' PERFORMANCE

A DYNAMIC CAPABILITIES PERSPECTIVE

MASTER THESIS

Master of Science Business Administration

Specialization Organizational & Management Control

University of Groningen, Faculty of Economics and Business

June 20, 2016

MICHEL VISSER

Maagjesbolwerk 91

8011LL Zwolle

m.visser.23@student.rug.nl

s1288369

Supervisor – drs. D.P. Tavenier

Co-Assessor – prof. dr. H.J. ter Bogt

WORD COUNT:

16.637 (including references and appendices) 12.829 (main text)

ABSTRACT

Consumers' changing shopping behavior and Internet's rapid and wide diffusion have disruptively impacted the retail market in the last two decades. To adapt to these changes in the market, and in search for increased performance, many retailers have adopted integrated multi-channel retail strategies. From a dynamic capabilities perspective, the relationship between multi-channel integration and firm performance is analyzed. This study empirically tests this relationship and, as such, contributes to the thin existing empirical evidence on this subject. An innovative measurement tool is used to obtain the level of multi-channel integration from annual reports of 98 publicly listed retailers in the U.S between 2011 and 2014. The main effect of interest is the expected positive relation between the level of multi-channel integration and firm performance. Using a regression model for panel data analysis, the main effect is estimated on several performance indicators. Furthermore, a positive moderating effect of IT capability on the main effect is expected and tested. The results provided no support for both effects. Hence, no positive relation between the level of multi-channel integration and any performance indicator was found. Several reasons for these results are discussed and directions for future research are offered.

KEYWORDS

Table of Contents

1 Introduction ... 1

2 Theoretical background ... 3

2.1 Evolution of retail toward omni-channel ... 3

2.1.1 Single channel ... 3

2.1.2 Multi-channel ... 4

2.1.3 Cross-channel ... 5

2.1.4 Omni-channel ... 6

2.2 Key drivers for multi-channel retailing ... 7

2.2.1 The Internet ... 7

2.2.2 Business benefits ... 7

2.2.3 Consumers' changing shopping behavior ... 8

2.3 Theoretical framework ... 8 2.3.1 Multi-channel literature... 9 2.3.2 Multi-channel taxonomy ... 9 2.3.3 Dynamic capabilities ... 10 2.3.4 Multi-channel capability ... 11 2.3.5 IT capabilities ... 12

2.4 Hypotheses development and conceptual model ... 13

3 Methodology ...16

3.1 Research setting ... 16 3.2 Data collection ... 16 3.3 Measurement ... 17 3.3.1 Dependent variable ... 17 3.3.2 Independent variable ... 18 3.3.3 Moderating variable ... 20 3.3.4 Control variables ... 22 3.4 Data analysis ... 23

3.5 Reliability and validity ... 24

4 Results ...26

4.1 Descriptive statistics... 26

4.2 Regression results ... 28

4.2.1 Dependent variable sales growth... 28

4.2.2 Dependent variable abnormal sales growth ... 30

4.2.3 Dependent variable ROA ... 31

4.2.4 Other results ... 32

5 Discussion and conclusions ...34

5.1 Discussion ... 34

5.1.1 Theoretical implications ... 34

5.1.2 Managerial implications ... 35

5.1.3 Limitations and future research ... 36

5.2 Conclusion... 37

References ...38

Appendix A.

Score card example ...42

Appendix B.

Stata commands ...43

Appendix C.

Regression results for 'firm sales growth' ...46

Appendix D.

Regression results for 'abnormal sales growth' ...48

Appendix E.

Regression results for 'ROA change' ...50

Appendix F.

Regression results for 'lagged sales growth' ...52

1 Introduction

The introduction of online shopping and the changing shopping behavior of consumers have had a disruptive impact on the retail market (Verhoef et al., 2015). The growing share of online sales causes many traditional retailers to close their shops, with many visible vacancies in the shopping streets as a result (Molenaar, 2016). In a reaction to adapt to these changes in the market, many retailers adopted a multi-channel strategy (Verhoef et al., 2015). The multi-channel literature assigns several performance enhancing benefits to adopting such a strategy: access to new markets and customers; access to multi-channel shoppers, who buy more and more often; and operational benefits through synergies between channels and departments (Lewis et al., 2014).

The goal of this study is to empirically examine if retailers gain performance benefits by adopting a multi-channel strategy and, hence, if it pays off to change their business model.

Multi-channel retailing has gained considerable attention in the retailing literature (e.g. Neslin et al., 2006; Neslin & Shankar, 2009; Verhoef et al., 2015). In the progression of the multi-channel retailing research, the article by Neslin et al. (2006) was of particular importance (Beck & Rygl, 2015; Verhoef et al., 2015). Current multi-channel literature provides a solid base of frameworks (e.g. Zhang et al., 2010), analytical models (e.g. Yan et al., 2010), and empirical studies based on different data sources (e.g. van Baal, 2014; Oh et al., 2012; Xia & Zhang, 2010). The leading theoretical stream in strategic management (Bharadwaj, 2000), the resource-based view is gaining importance in multi-channel retailing research as well, as several multi-multi-channel academics have adopted it (e.g. Oh et al., 2012; Zhu & Kraemer, 2005; Zhu et al., 2015).

For multi-channel retailers, IT-based resources are crucial in order to effectively integrate operations and information streams across channels and departments (Mollenkopf et al., 2007; Oh et al., 2012). The ability to effectively utilize these IT-based resources in combination with other resources to create a unique competitive advantage is referred to as IT capability (Bharadwaj, 2000). Higher levels of multi-channel integration require higher levels of coordination and integration of operations and information flows (Neslin et al., 2006), and hence, higher levels of IT capabilities are expected to support this.

To address this literature gap, this research will try to answer the following research questions:

Q1: What is the relationship between multi-channel integration and retailers' performance? Q2: How do IT capabilities moderate the relationship between multi-channel integration and firm

performance?

By addressing the literature gap, the current study contributes to the multi-channel retailing literature. First, by using a recently developed tool to measure multi-channel integration (Cao & Li, 2015), this research provides information on the usefulness of this measurement tool for multi-channel retailing research. Second, the quantitative study adds to the thin extant empirical evidence, resulting in a more mature multi-channel retailing research field (van Aken et al., 2012). From a management control perspective this research subject is interesting because in the process of developing and employing an integrated multi-channel strategy, managers face decisions on which channels to engage in, how to design them, to what level integrate them, how to distribute resources among them, how to obtain cooperation and coordination between channels and departments, and how to evaluate the performance of the different channels (Neslin et al., 2006; Neslin & Shankar, 2009).

This research builds on the study of Cao & Li (2015) who examined the effect of multi-channel integration on sales growth and found supporting evidence for this effect. This study is novel in three ways: (1) it uses a more recent time frame; (2) it introduces a new moderating effect (i.e. IT capabilities); and (3) in addition to the effect of multi-channel integration on sales growth, this study examines the effect on several other performance indicators.

The empirical research described in this thesis, tests the impact of multi-channel integration on publicly listed retailers in the U.S between 2011 and 2014. Multi-channel integration data is collected from secondary data sources. Using a regression model for panel data analysis, the main effect is estimated on several performance indicators, such as sales growth and abnormal sales growth. Furthermore, the theorized positive moderating effect of IT capability on the main effect is tested.

2 Theoretical background

In this chapter, the theoretical background is presented for the empirical research of this thesis. First, a description will be given on how the retail market has changed and how retailers have evolved in the last two decades. Second, the motivations for retailers to adopt a multi-channel strategy to counter the changing environment will be further elaborated. Lastly, the theoretical framework for this study will be presented, followed by the hypotheses development and the conceptual model.

2.1 Evolution of retail toward omni-channel

As presented in the introduction, the setting for this study is the retail market. Firms are considered retailers when they offer products or services to consumers (Berry et al., 2010). This definition covers a broad selection of organizations, including supermarkets, banks, airlines, drug stores, and digital subscription services among others. The mechanisms these organizations use for communication, product or service delivery, and handling transactions with their customers are called channels (Berry et al., 2010).

As will be explained in the following sections, the retail market has changed dramatically in the last two decades (Verhoef et al., 2015). Strang (2013) described the transformation of the retail market as the evolution of retail toward omni-channel. This evolution occurred as a continuum of trends or stages: from single channel to multi-channel to cross-channel to omni-channel (see Figure 1). In the following paragraphs, each stage will be discussed.

Figure 1. The retail evolution toward omni-channel (Strang, 2013).

2.1.1 Single channel

In the 1990s, the availability and use of the Internet rapidly diffused among households and firms (Greenstein & Prince, 2006). This triggered the emergence of a new single channel business model, electronic commerce (e-commerce), which uses the Internet as its primary channel (Hortaçsu & Syverson, 2015).

The introduction of online shopping has had a disruptive impact on the retail market. Rigby (2011) described the developments in the retail market in this period as follows. A wave of new internet-based retailers entered the market – Amazon.com, Pets.com, etc. – with their e-commerce business models. After the burst of the dot-com bubble, which reduced the number of online retailers by half, the Internet was widely used for retail purchases and the e-commerce business model established as an economic reality (Rigby, 2011). In the U.S., e-commerce sales as a share of total retail sales have grown from 0.9 to 6.4 percent between 2000 and 2014. This equals a growth in nominal e-commerce sales of 1100 percent, while the total retail sales grew 55 percent in the same period (Hortaçsu & Syverson, 2015). Moreover, online retailing proved to be very profitable. Whereas traditional retailers averaged a return on investment (ROI) of 6.5%, Amazon, for example, had an ROI of 17% (Rigby, 2011).

These developments were recognized by traditional retailers as well and triggered the first transition in the evolution of retail toward omni-channel (Strang, 2013).

2.1.2 Multi-channel

The first transformation was from single channel into multi-channel. Retailers with a multi-channel strategy operate through two or more channels (Zhang et al., 2010). The concept of multi-channel retailing was not new. Sears, for example, started a mail order business in 1886 and added a physical channel in 1925 and an online channel, sears.com, in 1998 (Cao, 2014).

Although the concept of multi-channel was not novel, the amount of retailers pursuing it has rapidly expanded since the 1990s (Stone, Hobbs & Khaleeli, 2002). This period coincided with the rapid diffusion of the Internet (Greenstein & Prince, 2006) and the growth of e-commerce (Hortaçsu & Syverson, 2015).

Figure 2. The benefits of clicks-and-bricks (Prasarnphanich & Gillenson, 2003).

Avery et al. (2012) describe how online retailers started to value the advantages of offline channels as well. Some multi-channel retailers even outperformed their online competitors. Consequently, online retailers considered adopting this hybrid business model also by adding an offline channel. For example, Amazon complemented its online channel by installing lockers in shopping centers and stores, where consumers can collect their online-ordered products (Lewis et al., 2014).

Typical characteristics of the multi-channel model are: two or more channels are employed; channels are operated independently of each other, and often managed by separate departments (i.e. silo structure); channels offer multiple independent customer contact points; orders placed in different channels are processed by different distribution centers; retail-mix policies across channels have little in common – e.g. different price, brand, service, and product assortment policies between channels (Cao & Li, 2015; Strang, 2013).

Some retailers operate through non-integrated channels to prevent a sales reduction in existing channels caused by the introduction of an additional channel (Biyalogorsky & Naik, 2003; Falk et al., 2007). This phenomenon is referred to as cannibalization (Avery et al., 2012) and occurs if a new channel overly reflects an existing channel's capabilities (Deleersnyder et al., 2002; Moriarty & Moran, 1990) or overshadows the existing channel with superior capabilities (Alba et al., 1997). The key is, thus, to find ways in which the online and offline channels complement instead of compete with each other (Rigby, 2011). Even though cannibalization may initially occur, empirical evidence shows that adding channels eventually leads to an increase in new customers (Avery et al., 2012).

2.1.3 Cross-channel

The next step in the evolution was that multi-channel retailers introduced cross-channel integration activities. A silo structure potentially has some negative effects – e.g. poor customer satisfaction, poor inventory performance, and redundant resources – that may cause a decline in sales or margins (Cao, 2014). According to the literature, these negative effects can be countered by

Hybrid

clicks-and-bricks

Advantages of e-commerce over bricks-and-mortars

- Greater geographic exposure

- Greater product information and better product comparison

- One-to-one experience

- A broader range of product selections - Convenience and no waiting in line

Advantages of bricks-and-mortars over e-commerce

- Brand recognition and large customer base - Better maintenance and after-sales services - Established infrastructure

- Feeling, touching, and testing products before purchasing - In-store experience and human interaction

obtaining some degree of cross-channel integration (Berry et al. 2010; Neslin et al. 2006; Neslin & Shankar, 2009; Zhang et al. 2010). Cross-channel integration is defined as "the degree to which a firm coordinates the objectives, design, and deployment of its channels to create synergies for the firm and offer particular benefits to its consumers" (Cao & Li, 2015; p. 200).

This trend started when retailers focused on integrating and aligning the marketing mix across channels by offering consistent branding, pricing, promotion, and product assortment (Berman & Thelen, 2004; Cao & Li, 2015; Strang, 2013). Further cross-channel integration was achieved when retailers focused on integrating activities related to the transaction process, for example facilitating in-store pick-up and return of goods that are bought online (Berman & Thelen, 2004; Cao & Li, 2015), and managing customers across channels (Verhoef et al., 2015). Moreover, inventory, pricing, and customer data were shared across channels through the support of integrated information systems (Berman & Thelen, 2004; Cao & Li, 2015).

Several researchers studied the relation between cross-channel retailing and firm performance. Herhausen et al. (2015), for example, found evidence that online-offline channel integration resulted in channel synergies and competitive advantage instead of cannibalization. Cao and Li (2015) found that retailers with higher levels of cross-channel integration gained higher sales growth. Oh et al. (2012) examined the effect of channel integration through the use of IT. The authors found evidence that IT-enabled channel integration through improved exploitative competences (i.e. ability to improve current operations' efficiency) and explorative competences (i.e. ability to offer new services) indirectly enhanced firm performance.

2.1.4 Omni-channel

Most recently, attention has shifted towards omni-channel retailing (Verhoef et al., 2015).

Omni-channel management is defined as: "the synergetic management of the numerous available channels and customer touch points, in such a way that the customer experience across channels and the performance over channels is optimized" (Verhoef et al., 2015, p. 176).

The definition of omni-channel retailing underlines the difference between channels and customer touch points. In the multi-channel and cross-channel stages, channels are considered as customer contact points – e.g. mechanisms through which interaction between the firm and the customer takes place (Verhoef et al., 2015). In omni-channel retailing, the channel scope is broadened by including customer-brand interaction (Neslin et al, 2014; Verhoef et al., 2015). As such, mass-communication channels are included in the channel scope, whereas they are not considered channels in multi-channel and cross-channel retailing.

Verhoef et al. (2015) believe that future research will shift towards omni-channel retailing, and specifically invite researchers to do so.

2.2 Key drivers for multi-channel retailing

In this paragraph, the motivations of retailers to adopt a multi-channel strategy will be discussed. Even though some motivations have already been mentioned shortly in the previous paragraph, they will be further explained in the following sections.

Lewis et al. (2014) argue that three key drivers have motivated retailers to adopt a multi-channel strategy: (1) the Internet; (2) business benefits; and (3) consumers' changing shopping behavior. According to Strang (2013), the first two drivers motivated the adoption of multi-channel and cross-channel strategies; moving towards omni-cross-channel retailing was consumer-driven. Or put more specifically, the consumer's demand for a seamless experience across all channels drives omni-channel adoption. The three drivers will be discussed in the following sections.

2.2.1 The Internet

Overall, existing literature seems to agree that the Internet has been an important reason for the expansive adoption of multi-channel strategies (Hortaçsu & Syverson, 2015; Lewis et al., 2014; Strang, 2013; Verhoef et al., 2015). The emergence of e-commerce, which uses the Internet as its primary channel, and the troubles it caused traditional retailers, fueled speculations of the demise of brick-and-mortar retailers (Hortaçsu & Syverson, 2015). Instead, it led to the advent of hybrid clicks-and-bricks retailers (Lewis et al., 2014). Hence, these firms' motivation to adopt a multi-channel retail strategy was driven by the Internet.

2.2.2 Business benefits

Multi-channel retailing literature also seems to agree that the associated business benefits have been an important driver for retailers to adopt a multi-channel strategy (Lewis et al., 2014). Cao and Li (2015) argue that sales growth is the main economic argument. First, it gives them access to the multi-channel shoppers who have a higher sales volume – up to four times as much – and higher purchase frequency than single-channel shoppers (Ansari et al., 2008; Clifford, 2010; Konus et al., 2008), although others found conflicting evidence for these claims (Kushwaha & Shankar, 2008). Also, through the elimination of barriers (e.g. geographical barriers), multi-channel retailers can reach new customers and enter new markets (Lewis et al., 2014).

2.2.3 Consumers' changing shopping behavior

According to Lewis et al. (2014), the point has been reached that the multi-channel shopper can be considered as standard. Customers capitalize on the pros of the different channels and minimize the cons (Lewis et al., 2014). Figure 2 presents the pros of the online and physical channel. For example, through the online channel, customers have a broader product assortment available to them. The huge amount of product and price information reduces search costs and makes prices transparent and easy to compare (Lewis et al., 2014). Furthermore, the online channel removes the geographical barrier (Verhoef et al., 2015).

The physical store, on the other hand, has no delivery time and makes the products instantly available to the customer. Also, the customer is able to touch, see, hear, try or taste the product and have face-to-face interaction with the store's employees (Lewis et al., 2014).

Consumers increasingly use the different channels available to them interchangeably in the different stages of the purchase process (Verhoef et al., 2015). For example, a customer shops for a product in a store, where he can touch it and try it, while at the same time he consults his mobile phone to compare prices (Rapp et al., 2015). A contrasting example is research shopping, where a customer gathers information at home or at work on his laptop or tablet, locates it in a store nearby and goes there to buy it (Verhoef et al., 2007).

As said, the multi-channel consumer is now the norm. In addition to just having access to the retailer through different channels, consumers also increasingly desire a seamless shopping experience across channels, Strang (2013) argues. He illustrates this as follows:

They want to buy from anywhere – in a store, on a laptop or PC, or from their phones and tablets; they want to pick it up from anywhere – in a store, at their place of work, at their home, or sent to a friend; and they want to return it anywhere – to a store or back to a distribution point (Strang, 2013, p. 36).

Moreover, consumers increasingly use the social media channel. One form of usage is the exchange of experiences through customer-to-customer interaction, or another is to interact with the firm, for example with a customer service employee (Verhoef et al., 2015).

2.3 Theoretical framework

2.3.1 Multi-channel literature

To position this thesis within the broad literature field, this section presents an overview of the current state of the multi-channel literature.

Within the retailing literature, multi-channel retailing has gained considerable attention from academic researchers (e.g. Neslin et al., 2006; Neslin & Shankar, 2009; Verhoef et al., 2015). In the progression of the multi-channel retailing research, the article by Neslin et al. (2006) has played a very important part (Beck & Rygl, 2015; Verhoef et al., 2015). They define multi-channel management as: "the design, deployment, coordination, and evaluation of channels to enhance customer value through effective customer acquisition, retention, and development" (Neslin et al., 2006, p. 96). Although it is defined as multi-channel management, this definition is also referred to in this thesis when the term multi-channel is used in combination with other terms like strategy, retailing, model, etc.

Current multi-channel literature provides conceptual frameworks (e.g. Neslin et al., 2006; Zhang et al., 2010), analytical modeling (e.g. Neslin & Shankar, 2009; Yan et al., 2010), consumer utility maximization modeling (Bernstein et al., 2008), and evidence-based studies employing consumer surveys (e.g. van Baal, 2014; Verhoef et al, 2007), retailer surveys (e.g. Oh et al., 2012), and secondary data (e.g. Cao & Li, 2015; Xia & Zhang, 2010).

Multiple subtopics have been considered in the multi-channel retailing literature. Verhoef et al. (2015) distinguish three major research topics: (1) shopper behavior across channels; (2) retail mix across channels; and (3) the impact of channels on performance. Research within these themes can be done from different perspectives. Many multi-channel retailing researchers, especially in the marketing literature, take the customer's perspective (e.g. Cassab & MacLachlan, 2009; Verhoef et al., 2009). Others have taken the organization's perspective (e.g. Homburg et al., 2014; Pauwels & Neslin, 2015).

Although multi-channel retailing has been broadly studied, empirical evidence on how and if channel integration leads to better firm performance is thin (Cao & Li, 2015).

This thesis adopts the organization's perspective as it examines the impact on firm performance. It provides empirical evidence based on secondary data and is positioned in the third major research stream.

2.3.2 Multi-channel taxonomy

Beck and Rygl (2015) performed a systematic review of the multi-channel retailing and found that there is no common meaning among the various terms operationalized for channel interaction and integration. They discovered that the term multi-channel is used for integrated channels, interacting channels, non-integrated channels, and non-interacting channels. Similarly, cross-channel is used for the interaction of both channels that are integrated, as well as channels that are not integrated.

Multiple categorizations for multi-channel modes have been proposed. Beck and Rygl (2015), for example, propose a classification of multi-channel retailing across two dimensions: (1) who controls or triggers the interaction; the customer or the firm; and (2) how many channels are considered; one, some, or all. Cao and Li (2015) provide a different categorization based on the channels' integration level, which results in four modes of multi-channel retailing. Relying on grounded theory (Strauss & Corbin, 1994) they conducted an exploratory, inductive study and distinguished four evolutionary stages of multi-channel integration:

1. Silo mode (level 1 integration): retailers sell through two or more channels, but independently operate them

2. Minimal integration (level 2 integration): retailers jointly optimize established channels, focusing on marketing communication-related activities

3. Moderate integration (level 3 integration): retailers jointly optimize established channels, focusing on activities related to consumer transactions

4. Full integration (level 4 integration): retailers jointly optimize established channels, focusing on the seamless shopping experience of the consumer

Additionally, the study resulted in a total of 27 indicators that cover front office and back office activities and elements of organizational structure related to multi-channel integration. Each indicator is ranked 1-4 according to the level of integration it represents (see Appendix A for an overview of the indicators and their corresponding integration level).

For several reasons, this categorization is considered useful and appropriate for this thesis. First, as demonstrated in Table 1, the categories or stages correspond well with the evolutionary stages discussed in paragraph 2.1. Second, it covers elements related to front-end and back-end operations and the organization's structure. Employing an integrated multi-channel strategy requires changes in all these elements (Cao & Li, 2015; Sousa & Voss, 2006; Zhang et al., 2010). Third, it provides a ranking of multi-channel integration, which can be used as a practical measurement tool for the quantitative research part of this study, as will be illustrated in paragraph 3.3.2.

Table 1. Comparison of multi-channel stages.

Multi-channel integration stage (Cao & Li, 2015)

Level of integration

Evolutionary stage toward omni-channel retail

(Strang, 2013)

Silo mode 1 Multi-channel

Minimal integration 2 Cross-channel; with low to moderate integration

Moderate integration 3 Cross-channel; with moderate to high integration

Full integration 4 Omni-channel

2.3.3 Dynamic capabilities

In strategic management literature, the resource-based view is the leading theoretical stream Bharadwaj, 2000). In multi-channel retailing literature, it is gaining importance as well, as several multi-channel researchers have adopted the resource-based view (e.g. Oh et al., 2012; Zhu & Kraemer, 2005; Zhu et al., 2015).

According to the resource-based view, organizational resources and capabilities that are costly to copy are fundamental drivers of superior financial performance (Bharadwaj, 2000). More specifically, firms create value by obtaining a sustainable competitive advantage through the process of combining and exploiting heterogeneous resources that are valuable, rare, and difficult to imitate (Barney, 1991).

The dynamic capability theory builds on the resource-based view by arguing that instead of through the resource-picking process, the capability-building process drives a competitive advantage (Makodok, 2001). It is argued that in highly competitive, rapidly changing environments the ownership of resources is not enough to sustain a competitive advantage (Teece, 2007). The changing external requirements may cause valuable resources to become less valuable and, consequently, neutralize a firm's superior position. The dynamic capability theory explores the firm's ability to reconfigure its resources and adapt to rapid market changes (Eisenhardt & Martin, 2000; Teece et al., 1997). To maximize their fit with today's highly competitive and rapidly changing environments, firms need to be able to adapt and diversify their activities and resources or even reinvent themselves (Eisenhardt & Martin, 2000).

Especially important is the capability of building competences out of combined resources (Teece, 2007; Teece et al., 1997). The creativity involved in this process and the resulted interconnectedness of the integrated resources makes these competences hard to imitate (Pavlou & Sawy, 2006; Oh et al., 2012). This makes it difficult for competitors to identify the key success factors (King, 2007; Lavie, 2006; Pil & Cohen; 2006), which, in theory, provides the focal firm with excellent performance (Coates & McDermott, 2002).

As described earlier in this chapter, technological developments and consumers' changing shopping preferences make today's highly competitive retail market a rapidly changing environment (Lewis et al., 2014; Verhoef et al., 2015). Many firms adopt a multi-channel strategy to adapt to these environmental changes (Verhoef et al., 2015). This makes the dynamic capability perspective particularly fitting to study how retail firms gain a competitive advantage.

2.3.4 Multi-channel capability

Building on the dynamic capability theory, this thesis adopts the idea that in order to successfully operate a channel strategy retailers need to possess a particular type of dynamic capability: multi-channel capability. This concept will be developed in the remainder of this section.

metrics tailored for multi-channel integration, and data integration and consumer analytics competences and resources.

Neslin & Shankar (2009) combined these elements and identified thirteen important issues distributed over five key decision areas: analyze customers; develop multichannel strategy; design channels; implement; and evaluate. The decision areas represent five successive stages in the process of developing and implementing a multi-channel strategy. To structure this process, the authors developed a multi-channel customer management decision (MCMD) framework. Without going into too much detail of explaining the framework, but to illustrate the complexity of the process and the required resources and competences to execute it, the MCMD framework is presented in Figure 3 (for a full discussion of the framework, see Neslin & Shankar, 2009).

Figure 3. Multi-channel customer management decision (MCMD) framework (Neslin & Shankar, 2009).

Hence, firms need capabilities to develop and reconfigure their existing resources, competences, and data in order to design, deploy, coordinate and evaluate the optimal mix of channels, integrate the retail mix across channels, and implement customer management across channels (Neslin et al., 2006). These capabilities can be referred to as multi-channel capabilities. Then, multi-channel capability is a type of dynamic capability, since it refers to the ability of a retail organization to integrate, build, and reconfigure its competences to adapt to the changing requirements in the external environment to create or sustain value (Luo, 2001; Teece et al., 1997).

2.3.5 IT capabilities

In this section the concept of IT capabilities will be explained and why it is crucial for multi-channel retailers to possess these capabilities.

Drawing upon the dynamic capability theory, a firm's IT capability refers to its ability to effectively develop, integrate and utilize IT resources in combination with other resources to create a unique competitive advantage (Bharadwaj, 2000).

logistical back-end activities with front-end marketing activities, retail organizations need to integrate and align functional areas, such as marketing, inventory, order fulfillment, and product returns (Mollenkopf et al., 2007). To achieve such operational and strategic coordination, IT capabilities – i.e. to mobilize and deploy IT-based resources conjointly with other resources – are required (Sanders, 2008). Otherwise, it will be very challenging to support a consistent multi-channel offer (Lewis et al., 2014).

IT-based resources can be classified into three categories (Bharadwaj, 2000). First of all, retailers need to have a supporting IT infrastructure in place to offer multi-channel services (Pentina & Hasty, 2009). To sell online, for example, a website needs to be designed and managed. But also, to integrate inventory and logistics across channels, it is necessary to have IT systems in place to integrate data across all channels (Lewis et al., 2014). Next, IT human resources are needed to provide technical and managerial skills to use and develop the infrastructure (Zhu & Kraemer, 2005). An infrastructure without the ability to use it effectively does not enhance firm performance (Oh et al., 2012; Pavlou & Sawy, 2006). Hence, IT personnel are essential to apply these technologies successfully. The third type of IT-based resources is intangible IT assets. These include knowledge assets, which are embedded in the skills and experience of the employees; customer orientation through the sharing and integration of customer data across channels; and synergies due to the sharing of resources and capabilities across departments (Bharadwaj, 2000).

Oh et al. (2012) provided empirical evidence to support the positive relation between IT capabilities and firm performance. They found that IT-enabled retail channel integration increased the ability of retail firms to improve the efficiency of their current operations as well as the ability to offer new services, both of which in turn enhanced the performance of the firm.

2.4 Hypotheses development and conceptual model

Building on the theoretical concepts that have been introduced and discussed in the previous paragraph, this paragraph will explain the relationships between the main concepts. Based on this, hypotheses will be formulated so the relationships can be tested empirically.

Multi-channel integration and firm performance

A successful multi-channel strategy requires firms to possess multi-channel capabilities. These capabilities refer to the ability to develop and reconfigure their existing competences, resources, and data in order to design, deploy, coordinate and evaluate the optimal mix of channels, integrate the retail mix across channels, and implement customer management across channels (Neslin et al., 2006). In order to achieve a higher level of integration across channels, a higher degree of multi-channel capabilities is required. According to the dynamic capabilities theory, higher levels of dynamic capabilities provide firms with superior performance (Oh et al., 2012).

Thus, it can be expected that firms with a higher level of multi-channel integration have a higher level of multi-channel capabilities, which, based on the dynamics capabilities theory, should enhance the firm's performance. Hence, it is hypothesized:

Hypothesis 1: A higher level of firm multi-channel integration has a positive effect on firm performance.

IT capabilities and multi-channel integration

A firm's ability to effectively develop, integrate and utilize IT-based resources in combination with other resources to create a unique competitive advantage is referred to as IT capability (Bharadwaj, 2000). For multi-channel retailers IT-based resources are crucial in order to effectively integrate operations and information flows across functional departments and channels (Cappiello et al., 2003; Markus, 2000; Mollenkopf et al., 2007; Oh et al., 2012; Pentina & Hasty, 2009; Vickery et al., 2003).

As explained in the previous section, a higher level of multi-channel integration requires a higher level of coordination and integration (Neslin et al., 2006), and hence, a higher level of IT capabilities to support it. Following this logic, it can be expected that firms with a higher level of IT capabilities are better able to effectively develop and deploy an integrated multi-channel strategy. As hypothesized in Hypothesis 1, a higher degree of multi-channel integration leads to better performance. Therefore, it is hypothesized:

Conceptual model

The theoretical concepts and their hypothesized relationships are visualized in the conceptual model, as presented in Figure 4. In addition to the main concepts, several control variables are included in the model: industry sector, year, firm size, and competitor relative. The relationship between these variables and firm performance has been established in prior research and, hence, can be expected to influence firm performance. Therefore, they are included in the model. The control variables will be discussed in section 3.3.4. Firm multi-channel integration Firm performance Firm IT capabilities + (H1) + (H2)

Figure 4. Conceptual model.

Control variables

Industry Sector

Year

Firm size

3 Methodology

From the discussion of the current state of the multi-channel literature in section 2.3.1, it follows that the theory base is quite established. Nevertheless, empirical evidence on the subject of interest – i.e. the effect of multi-channel integration on firm performance – is still inconclusive (Cao & Li, 2015). In this context, theory-testing research can be conducted to (partly) close the literature gap (van Aken et al., 2012). Van Aken et al. (2012) describe theory-testing research as a four-step process: (1) identification of business phenomenon and literature gap; (2) identification of concepts, and development of hypotheses and conceptual model; (3) large scale data collection and statistical analysis; and (4) discussion of results, theoretical and practical implications, and direction for future research.

The first two steps are covered in the first two chapters. The fourth step of the theory-testing process will be described in chapters four (results) and five (discussion and conclusion).

In the remainder of this chapter, the third step will be described. First, a description of the research setting will be given. Second, the employed data collection method will be explained, followed by the measurements used in this study. Next, a description of the data analysis method will be given. Finally, the reliability and validity will be addressed.

3.1 Research setting

To test the hypotheses, data are collected from publicly listed U.S. retailers. These firms are considered to provide suitable data for this research, because they provide an appropriate mixture of multi-channel integration development, an appropriate amount of competitiveness, and widely available public information (Cao & Li, 2015).

3.2 Data collection

Data are collected for the years 2011 to 2014. A timeframe of four years is chosen to balance on the one hand the need for sufficient data, and on the other hand the effort needed to collect these data. Furthermore, this is the latest timeframe possible, since the performance data for the fiscal year 2015 were not yet available for all companies when the data were collected. Most of these data become available between January 2016 and July 2016, and for some companies even later.

annual report through a Form 10-K. These forms are accessible through the EDGAR database on the website of the SEC.

If no multi-channel related data were available in the Form 10-K, additional information was obtained through the firm's annual reports on their website. In some cases these annual report contained additional information to the Form 10-K. However, in most cases they were similar. If still no data were found, the firm's website was explored, and when this process did not yield any result, the firm was labeled as no data available and the firm was removed from the sample.

The initial sample contained 138 companies, of which 21 were labeled no data available. Excluding the years in which no data were available (e.g. firm was not listed yet or anymore), 453 annual reports in total were analyzed of the remaining 117 companies. 13 of them were labeled as single channel retailers over the entire time frame, and consequently removed from the sample. For the remaining 104 firms, cross-channel integration data were collected.

Next, data for the other variables (e.g. financial information) were collected from the COMPUSTAT database.

For some firms no financial data were available over the entire time frame, for others partially. First, firms with no financial data were removed from the sample. Finally, due to unavailable financial data for some firm-year observations, some firms were left with a single observation. These were also excluded from the sample. The final sample comprised 98 firms and 372 firm-year observations.

The data were documented and saved in an Excel spreadsheet (which is available on request) containing nearly 143 sheets; one 'multi-channel integration scorecard' for each company (see Appendix A for an example scorecard), and several summary sheets containing financial data.

There is some variation in how different companies define their fiscal year. Some companies' fiscal year runs from December 31 till December 31, while others from January, February, March or even June or September. Even though they cover different periods between firms, for this research the fiscal years are compared. For example, if a company's fiscal year ended in march 2013, the data are registered for the year 2012. Since the other variables are measured per fiscal year as well, inconsistencies between firm-year observations are not likely.

3.3 Measurement

In this paragraph a description will follow of the different measures used in this research, why they were used, and how they were determined.

3.3.1 Dependent variable

Sales growth

The first performance indicator is sales growth ("Firm sales growth"), because the desire to increase sales is one of the main drivers for retailers to adopt a multi-channel strategy (Cao & Li, 2015; Lewis et al., 2014). Sales growth at time t is calculated as:

𝐹𝑖𝑟𝑚 𝑠𝑎𝑙𝑒𝑠 𝑔𝑟𝑜𝑤𝑡ℎ𝑡 = 𝐹𝑖𝑟𝑚 𝑠𝑎𝑙𝑒𝑠𝑡 − 𝐹𝑖𝑟𝑚 𝑠𝑎𝑙𝑒𝑠𝑡−1 𝐹𝑖𝑟𝑚 𝑠𝑎𝑙𝑒𝑠𝑡−1

Here, 𝐹𝑖𝑟𝑚 𝑠𝑎𝑙𝑒𝑠𝑡 is the reported sales – i.e. 'Revenue - Total' in COMPUSTAT – at time t and 𝐹𝑖𝑟𝑚 𝑠𝑎𝑙𝑒𝑠𝑡−1 is the reported sales at time (t – 1). For example, to calculate the sales growth at time t = 2011, the reported total revenue of fiscal years 2011 (t) and 2010 (t – 1) were used.

Abnormal sales growth

Potential industry or economy-wide effects may influence firm performance. To control for these effects, a firm's abnormal performance can be measured and is, for this reason, commonly used in academic research (Xia & Zhang, 2010). Abnormal sales growth ("Abnormal sales growth"), at time t is calculated as:

𝐴𝑏𝑛𝑜𝑟𝑚𝑎𝑙 𝑠𝑎𝑙𝑒𝑠 𝑔𝑟𝑜𝑤𝑡ℎ𝑡 = 𝐹𝑖𝑟𝑚 𝑠𝑎𝑙𝑒𝑠 𝑔𝑟𝑜𝑤𝑡ℎ𝑡− 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 𝑠𝑎𝑙𝑒𝑠 𝑔𝑟𝑜𝑤𝑡ℎ𝑡

Here, 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 𝑠𝑎𝑙𝑒𝑠 𝑔𝑟𝑜𝑤𝑡ℎ𝑡 is calculated as the average sales growth of all firms in the initial sample (N=138) at time t, provided the availability of the data.

ROA

An integrated multi-channel strategy requires significant investments in, for example, IT infrastructure such as an e-commerce supporting website and IT systems that support the integration and sharing of data across channels and departments (Lewis et al., 2014). ROA can be used as an indicator of the ability of the firm to earn profits from these investments. This measure has been used in many previous studies as a performance metric (Xia & Zhang, 2010).

Because the effect on ROA is of interest, the change in ROA ("ROA Change") is used to measure this effect. Change in ROA at time t is calculated as:

𝑅𝑂𝐴 𝐶ℎ𝑎𝑛𝑔𝑒𝑡 = 𝑅𝑂𝐴𝑡 − 𝑅𝑂𝐴𝑡−1 Here, 𝑅𝑂𝐴𝑡 is calculated as:

𝑅𝑂𝐴𝑡 = 𝑁𝑒𝑡 𝐼𝑛𝑐𝑜𝑚𝑒𝑡 𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠𝑡

Here, 𝑁𝑒𝑡 𝐼𝑛𝑐𝑜𝑚𝑒𝑡 and 𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠𝑡 are COMPUSTAT metrics 'Net Income (Loss)' and 'Assets - Total', respectively, at time t.

3.3.2 Independent variable

activities and elements of organizational structure, and indicate the level of integration. Figure 5 gives an overview of the indicators and their corresponding integration level. The second column lists the descriptions of the indicators and the third column lists the corresponding integration level. For example, indicator 9 is the activity of 'allowing online consumers to browse the inventory in-store', which is considered to be a multi-channel integration activity at the third level.

Figure 5. Example scorecard Best Buy.

In order to use this measurement tool, insight in firms' multi-channel related activities is needed. Through the analysis of corporate annual reports this insight can be gained. Publicly listed organizations issue these documents to their shareholders and publish them online to communicate their annual progress (Rodrigues & Minshall, 2015). Annual reports have been used extensively in academic research (Rodrigues & Minshall, 2015). For this study, the annual reports of the sample firms were analyzed to identify multi-channel related activities. These activities were linked to one of the 27 key indicators. As such, the qualitative data from the annual reports were made quantitative.

Conform the research of Cao and Li (2015) all activities reported in previous years were assumed to be existent in the following years, unless otherwise was reported.

Using this process, a scorecard for each sample firm is documented in the Excel spreadsheet. Included on the sheets are citations of the concerning activities as reported in the annual report, and a link to the online location of the annual report. A filled out example of a scorecard can be found in Figure 5 (citations have been replaced with an X for readability reasons). A copy of an example sheet is attached in Appendix A, from where it can be seen that the cited excerpts from the annual report are documented as well.

Figure 5 represents the scorecard of Best Buy Co. Inc. It shows that the company reports level 1 and level 3 activities in their annual report of 2011. The multi-channel integration level is the level of the highest ranked indicator, which in 2011 is level 3. Similarly for 2012, 2013 and 2014, the integration levels are 3, 3 and 4, respectively.

3.3.3 Moderating variable

The moderating variable is IT capabilities ("IT Capabilities"). Previous research has been done under the assumption that firms with higher IT capabilities, on average, spend more on IT. For this reason, IT expenditure is often used to measure IT capabilities (Bharadwaj et al., 1999; Oz, 2005; Shin, 2004). There are some difficulties obtaining secondary IT spending data: IT investments are sometimes otherwise accounted for than other capital investments; the definition of IT expenditure varies across firms, which leads to the fact that some of the IT costs are charged to other, non-IT accounts; finally, many companies do not want to publish their IT investment budget because they regard it as confidential (Oz, 2005) and strategically important and sensitive information (Bharadwaj et al., 1999). Oz (2005) even regards obtaining IT expenditure figures as "extremely difficult, if not impossible" (p. 793).

There are only two sources with secondary, publicly available IT spending data: Informationweek (IW) magazine and Computerworld magazine (Bharadwaj et al., 1999). Especially IW is often used in IT capabilities research (e.g. Shin, 2004). IW collects its data from 500 manufacturing and service firms. This dataset contains, however, only a small portion of retailers – e.g. 7 out of 500 in 2011. Moreover, some of these 7 firms were private firms. Hence, the IW dataset was unsuitable because of the marginal overlap with this study's sample. For similar reasons, Computerworld was an unsuitable data source for this research.

Considering that IT expenditures have often been used in research as an indicator for IT capabilities, and Capex and SGA could be considered proxy measures for IT expenditures, Capex and SGA will be used as proxy measures for IT capabilities.

Table 2. Capex specifications excerpt, mentioning multi-channel related investments (emphasis added).

Company Citation from annual report

Hancock Fabrics, Inc. (annual report fiscal year 2014):

During 2014, expenditures for investing activities consisted of store fixtures and leasehold improvements relating to seven new stores, six relocated units, six remodels and development costs related to the re-launch of the Company website. Staples, Inc.

(annual report fiscal year 2012):

We expect a modest increase in capital spending in 2013 resulting from

investments in our online businesses and our other strategic growth initiatives.

The Bon-Ton Stores, Inc. (annual report fiscal year 2011):

We believe capital investments for information technology are necessary to support our business strategies. We are continually upgrading our information systems to improve efficiency and productivity. Included in the 2012 capital budget are expenditures for numerous information technology projects, most

notably efforts to enhance our online presence and selling tools.

Bed Bath & Beyond Inc. (annual report fiscal year 2013):

Capital expenditures for fiscal 2014, principally for information technology

enhancements, including omnichannel capabilities, new stores, existing store

improvements, and other projects are planned to be approximately $350 million.

Table 3. SGA specifications excerpt, mentioning multi-channel related expenses (emphasis added).

Company Citation from annual report

Kohl’s Corporation (annual report fiscal year 2011):

SG&A increased primarily due to store growth, increased advertising, and

investments in technology and infrastructure related to our E-Commerce business.

The Finish Line, Inc. (annual report fiscal year 2012):

The $22.3 million increase in selling, general and administrative expenses [...] was primarily due to the following: [...] (2) an increase in marketing expense to drive

traffic to our website and our stores; (3) investments to support the Company’s technology upgrades, digital platform and omni-channel strategy

Stage Stores, Inc. (annual report fiscal year 2011):

The increase in SG&A expenses in 2011 was primary due to increases in expenses

related to eCommerce

Aéropostale, Inc (annual report fiscal year 2011):

The increase in SG&A as a percentage of net sales was largely due to a $5.6 million

increase in volume related e-commerce transaction expenses, $4.2 million more

store payroll expense from new store growth, and $2.6 million of higher marketing expenses.

Capex

An integrated multi-channel strategy requires significant investments in IT resources. For example, an e-commerce website needs a supporting IT infrastructure to process its online sales, by disseminating the purchase information across channels and functional departments, such as inventory, order processing, and the physical store in case of 'buy online/pick-up in-store' (Lewis et al., 2014).

investments are needed. It is calculated as the difference between the current and previous Capex, divided by the previous Capex:

𝐶𝑎𝑝𝑥𝑐ℎ𝑎𝑛𝑔𝑒𝑡 = 𝐶𝑎𝑝𝑒𝑥𝑡 − 𝐶𝑎𝑝𝑒𝑥𝑡−1 𝐶𝑎𝑝𝑒𝑥 𝑡−1

Here, 𝐶𝑎𝑝𝑒𝑥 𝑡 and 𝐶𝑎𝑝𝑒𝑥𝑡−1 are denoted by the reported 'Capital Expenditures' in COMPUSTAT at times t and (t – 1), respectively.

SGA

The traditional assumption is that an increase in the ratio of SGA is a negative indicator for future firm performance (Anderson et al., 2007). However, Anderson et al. (2007) claim that a deliberate increase in SGA can be a signal of managers' positive beliefs in future performance. This claim is backed by empirical evidence (Anderson et al., 2007; Baumgarten et al., 2010). Increases in SGA potentially create intangible assets (Banker et al., 2006; Banker et al., 2011), such as improvements in operational processes and enhancements of IT systems (Baumgarten et al., 2010). To examine whether intangible assets – e.g. IT capabilities – are created, the relative change in SGA ("xsgachange") will be used as a measure. It is calculated as the difference between the current and previous SGA, relatively to the previous SGA:

𝑥𝑠𝑔𝑎𝑐ℎ𝑎𝑛𝑔𝑒𝑡 =

𝑆𝐺&𝐴𝑡 − 𝑆𝐺&𝐴𝑡−1 𝑆𝐺&𝐴𝑡−1

Here, 𝑆𝐺&𝐴𝑡 and 𝑆𝐺&𝐴𝑡−1 are denoted by the reported 'Selling, General and Administrative Expenses' in COMPUSTAT at times t and (t – 1), respectively.

3.3.4 Control variables

In existing literature, the impact of several variables on firm performance is known. The effect of these variables will be controlled for. First, firm size has been known to positively influence firm performance (Oh et al., 2012). Here, firm size ("Size") is calculated as the natural log of 'Assets - Total' as reported in COMPUSTAT.

Second, the variation in performance due to competitors' relative advantage in adopting a multi-channel integration strategy is controlled for (King et al., 2004). First, for each firm the median level of integration (MCI) of its within-sector competitors was calculated. Then, the variable ("Competitive advantage ") was coded as follows. If the competitors' median was higher than the focal firm's MCI, it was coded 1. It was coded –1 if the competitors' median was lower, and 0 if they were equal (Cao & Li, 2015).

sectors, dummy variables were created for the different retail sectors ("Sector") (Cao & Li, 2015; Oh et al., 2012).

3.4 Data analysis

The dataset for this research has a panel structure and is referred to as panel data, as the structure of the dataset is based on the pooling of observations on a cross-section of a number of N individuals (e.g. firms) over a T number of time periods (Baltagi, 2013). In this particular research it refers to the pooling of 372 yearly observations of N = 98 U.S. retailers over a time span of T = 4 years between 2011 and 2014.

The basic model for panel data regression is:

𝑦𝑖𝑡 = 𝛼 + 𝛽 Χ𝑖𝑡′ _{+ 𝑢𝑖𝑡 (i = 1, ..., N; t = 1, ...,T ),}

where i denotes firms, countries, etc., and t denotes time (Baltagi, 2013). Thus, the subscript it represents the firm-year observation. Here 𝑦𝑖𝑡 is the outcome of firm i at time t, Χ𝑖𝑡 is the it-th observation on a K amount of explanatory variables, and 𝑢𝑖𝑡 is the error component to capture disturbances (Baltagi, 2013). The intersect 𝛼 and coefficient 𝛽 are not firm or time specific; they are estimated from the regression of all observations.

This basic regression model does not have a moderating effect yet. Typically, a moderating effect is modeled by adding a product of the main terms (e.g. 𝑋1 and 𝑋2) that need to interact (Balli & Sørensen, 2013):

𝑦𝑖𝑡 = 𝛼 + 𝛽1𝑋1𝑖𝑡+ 𝛽2𝑋2𝑖𝑡 + 𝛽3𝑋1𝑖𝑡𝑋2𝑖𝑡 + 𝑢𝑖𝑡

The independent terms 𝑋1 and 𝑋2 are referred to as main terms and the product 𝑋1𝑋2 is referred to as the moderating (i.e. interacting) term (Balli & Sørensen, 2013). The moderating effect is captured in 𝛽3.

Furthermore, a moderating regression should always contain the main terms – i.e. 𝑋1 and 𝑋2, as it is likely that the moderating term – i.e. 𝑋1𝑋2 – is correlated with the main terms, and leaving them out may result in a significant moderating effect that is biased (Balli & Sørensen, 2013).

In econometric studies, it is common to perform a plain regression without the moderating effect prior to a moderated regression (Balli & Sørensen, 2013). By comparing the plain (i.e. non-interacted) main effect with the moderated main effect, the impact of the moderator on the main effect can be determined.

the moderating variable IT capabilities. Models 3 and 4 are similar to models 1 and 2, respectively, but replace capxchange by xsgachange as measurement for IT capabilities.

Furthermore, a random effects model will be used, instead of a fixed effects model. In a fixed effects model, time-invariant differences between firms are controlled for (Kohler & Kreuter, 2005). However, because variation within sectors is marginal over time, and sector dummy variables are included in the model, the random effects model is a better fit (Cao & Li, 2015).

It follows from the previous paragraph that there are K = 6 explanatory variables 𝛸: MCI, IT Capabilities, Competitive advantage, Sector, Year, and Size. This results in the following regression models.

Model 1 represents the plain regression model, with capxchange as measurement for IT capabilities:

𝑦𝑖𝑡 = 𝛼 + 𝛽1𝑀𝐶𝐼𝑖𝑡+ 𝛽2𝑆𝑖𝑧𝑒𝑖𝑡 + 𝛽3𝑌𝑒𝑎𝑟𝑖𝑡 + 𝛽4𝑆𝑒𝑐𝑡𝑜𝑟𝑖𝑡 + 𝛽5𝐶𝑜𝑚𝑝𝑒𝑡𝑖𝑡𝑖𝑣𝑒 𝑎𝑑𝑣𝑎𝑛𝑡𝑎𝑔𝑒𝑖𝑡 + 𝛽6𝑐𝑎𝑝𝑥𝑐ℎ𝑎𝑛𝑔𝑒𝑖𝑡 + 𝑢𝑖𝑡

In model 2, the moderating effect 'MCI × capxchange' is added:

𝑦𝑖𝑡 = 𝛼 + 𝛽1𝑀𝐶𝐼𝑖𝑡+ 𝛽2𝑆𝑖𝑧𝑒𝑖𝑡 + 𝛽3𝑌𝑒𝑎𝑟𝑖𝑡 + 𝛽4𝑆𝑒𝑐𝑡𝑜𝑟𝑖𝑡 + 𝛽5𝐶𝑜𝑚𝑝𝑒𝑡𝑖𝑡𝑖𝑣𝑒 𝑎𝑑𝑣𝑎𝑛𝑡𝑎𝑔𝑒𝑖𝑡 + 𝛽6𝑐𝑎𝑝𝑥𝑐ℎ𝑎𝑛𝑔𝑒𝑖𝑡 + 𝛽7𝑀𝐶𝐼𝑖𝑡 × 𝑐𝑎𝑝𝑥𝑐ℎ𝑎𝑛𝑔𝑒𝑖𝑡 + 𝑢𝑖𝑡

Model 3 is similar to model 1, but substitutes xsgachange for capxchange:

𝑦𝑖𝑡 = 𝛼 + 𝛽1𝑀𝐶𝐼𝑖𝑡+ 𝛽2𝑆𝑖𝑧𝑒𝑖𝑡 + 𝛽3𝑌𝑒𝑎𝑟𝑖𝑡 + 𝛽4𝑆𝑒𝑐𝑡𝑜𝑟𝑖𝑡 + 𝛽5𝐶𝑜𝑚𝑝𝑒𝑡𝑖𝑡𝑖𝑣𝑒 𝑎𝑑𝑣𝑎𝑛𝑡𝑎𝑔𝑒𝑖𝑡 + 𝛽6𝑥𝑠𝑔𝑎𝑐ℎ𝑎𝑛𝑔𝑒𝑖𝑡 + 𝑢𝑖𝑡

Model 4, builds on model 3 by adding the moderating effect 'MCI × xsgachange':

𝑦𝑖𝑡 = 𝛼 + 𝛽1𝑀𝐶𝐼𝑖𝑡+ 𝛽2𝑆𝑖𝑧𝑒𝑖𝑡 + 𝛽3𝑌𝑒𝑎𝑟𝑖𝑡 + 𝛽4𝑆𝑒𝑐𝑡𝑜𝑟𝑖𝑡 + 𝛽5𝐶𝑜𝑚𝑝𝑒𝑡𝑖𝑡𝑖𝑣𝑒 𝑎𝑑𝑣𝑎𝑛𝑡𝑎𝑔𝑒𝑖𝑡 + 𝛽6𝑥𝑠𝑔𝑎𝑐ℎ𝑎𝑛𝑔𝑒𝑖𝑡 + 𝛽7𝑀𝐶𝐼𝑖𝑡 × 𝑥𝑠𝑔𝑎𝑐ℎ𝑎𝑛𝑔𝑒𝑖𝑡 + 𝑢𝑖𝑡

Each model will be run three times. In the first run Firm sales growth will be the dependent variable 𝑦. In the second and third run Abnormal sales growth and ROA Change will denote the dependent variable 𝑦, respectively. This results in a total of twelve regressions, which are run using the software Stata version 14. A copy of the commands entered in Stata, are presented in Appendix B.

3.5 Reliability and validity

several summary sheets containing financial data. The scorecards contain a link to the website where the annual report can be retrieved, and paraphrases of the multi-channel related activities from the annual reports. This way the multi-channel activities are retraceable in the annual report. The formulas on the summary sheets are also intact, so it is traceable how the financial measures were calculated.

To perform the statistical analysis, a set of commands was entered into the software program Stata. A copy of the list of commands can be found in Appendix B. Also, a copy of the output from Stata can be found in Appendix C to Appendix G, which presents the regression results. This should further increase the traceability of the research process.

Reliability independent variable

To establish the reliability of the independent variable, ten firms were randomly selected after the collection of the multi-channel data from the annual reports was completed. For these firms, the data collection process was repeated independently of the first attempt. Comparing the results of the second attempt with the first attempt did not yield any differences. Within the limitation that the second attempt was done by the same researcher, the data was considered valid.

Reliability dependent, moderating and control variables

The process of collecting and calculating the financial data for the dependent, moderating and control variables was repeated three times independently from each other. The results were compared with previous attempts. After the second attempt, some discrepancies were found between the first and second version of the dataset. The causes for these differences – mostly caused by an error in a formula – were traced and solved. As such the first and second version of the dataset were equal. A third attempt was performed for validation. The third version of the dataset was equal to the first two and, therefore, the dataset was considered to be reliable.

Validity independent variable

The independent variable was measured using a measurement tool that was developed, explained and empirically tested in an article (Cao & Li, 2015) published in the Journal of Retailing. This journal is ranked Very good, the second highest ranking (Top and very good journals, 2014) by the University of Groningen.

The validity of the independent variable measurement was reflected on as follows. If the academics reviewers of the Journal of Retailing considered the research based on that particular measurement tool valid for publishing, it is valid to be used in this thesis.

Validity dependent, moderating and control variables

4 Results

This study's dataset has a panel structure, since there are time-series observations for each of the multiple firms (Baltagi, 2013). The dataset is unbalanced with 372 observations over four years for 98 firms. Table 4 presents the distribution of the firm-year observations. The 'Pattern' column shows the pattern of the subsets of observations. A '1111' pattern means that data was obtained for all four years, '_111' means that data for the last three years were obtained, '_11_' means that only data was obtained for 2012 and 2013 and not for 2011 and 2014, etc. Hence, for nearly 84% of the firms observations were obtained for all years within the timespan.

Table 4. Firm-year distribution; unbalanced dataset.

Frequency Percent Cumulative Pattern

82 83.67% 83.67% 1111 6 6.12% 89.80% _111 6 6.12% 95.92% 111_ 2 2.04% 97.96% __11 1 1.02% 98.98% _11_ 1 1.02% 100.00% 11__ 98 100.00%

4.1 Descriptive statistics

The basic descriptive statistics of the sample firms are reported in Table 5. The mean level of multi-channel integration is 2.51, the median is 3, and the lowest and highest levels were 1 and 4, respectively.

Table 5. Descriptive statistics.

Variable Mean Median Std. Dev. Min Max

Firm sales growth 0.0560 0.0424 0.1060 –0.3863 0.5158

Abnormal sales growth –0.0102 –0.0276 0.1191 –0.4569 0.4813

ROA Change –0.0075 –0.0001 0.0916 –0.6032 0.5667 MCI 2.5131 3.0000 1.2916 1.0000 4.0000 capxchange 0.2066 0.1308 0.5596 –0.7919 6.3764 xsgachange 0.0598 0.0451 0.1223 –0.5701 0.7861 Size 7.2193 7.2359 1.7170 3.3390 12.2295 Competitor Advantage 0.0188 0.0000 0.8975 –1.0000 1.0000

of level 4 over the four-year time span can be clearly seen. Levels two and three do not change considerably.

Figure 6. Distribution of multi-channel integration per year.

Figure 7. Distribution clustered by multi-channel integration level.

Table 6 presents the average performance metrics per year. Panel A presents the average sales growth for the sample firms. It shows that each year, on average, retailers with an integration level of 4 underperform in comparison to level 1 retailers. Level 2 and 3 provide mixed results. Panel B presents the average abnormal sales growth figures and shows similar results. The same can be said for the average change in ROA (see panel C), except for 2013 where level 4 firms outperform level 1 firms.

0% 10% 20% 30% 40% 50% 2011 2012 2013 2014

Silo mode (level 1)

Minimal integration (level 2) Moderate integration (level 3) Full integration (level 4) 0% 10% 20% 30% 40% 50%

Silo mode (level 1) Minimal integration (level 2)

Moderate integration (level 3)

Full integration (level 4)