Multichannel Price Discrimination in the Grocery Sector:
An Empirical Study Investigating the Effect of Higher Online Prices on
Online Purchase Intention under the Influence of E-Service Quality.
By Christina Brinkmann.1
Master Thesis Marketing Intelligence: EBM867B20.2018-2019.2 Supervisor: dr. A.E. (Arnd) Vomberg
Co-assessor: prof. dr. T.H.A. (Tammo) Bijmolt University of Groningen
Faculty of Economics and Business Nettelbosje 2
9747 AC, Groningen, The Netherlands June 17, 2019
Abstract
Online grocery shopping in Europe is still in early stages, but anticipated to gain more and more popularity within the upcoming five years. Area of concern is the unprofitability of current operating online grocery retailers. This paper aims to fill the research gap for price differentiations in the grocery sector. It addresses whether higher online grocery prices are accepted under the provision of a higher E-Service Quality, measured by means of online purchase intention, and therefore, can function as a resolution to the underlying problem of profitability. Data was gathered through an experimental survey, mainly conducted in Germany and the Netherlands. Primary, the effect of higher online prices on online purchase intention is investigated with the aid of (moderated) mediation analysis. Additionally, further potential influencing factors are taken into account. Findings reveal, that online purchase intention for groceries is low, regardless of a price discrimination and that E-Service Quality does not affect the online purchase intention in any way.
Keywords: Online Grocery Shopping, Multichannel Pricing, Price Discrimination, E-Service Quality, Perceived Value
1 Master Marketing Intelligence, student number: S3658112, email: c.brinkmann@student.rug.nl
Table of Contents
1. Introduction ... 1
2. Literature review ... 3
2.1 Multichannel Retailing ... 3
2.2 Pricing in Multichannel Retailing ... 4
2.3 The Effects of Price Differentiation on Customer Behavior ... 5
2.4 Online Grocery Retailing and Customer Channel Behavior ... 8
3. Research Model & Hypothesis Development ... 12
3.1 Hypothesis Development ... 12
3.1.1 The Effect of Multichannel Pricing on Purchase Intention ... 12
3.1.2 The Role of Perceived Value in the Relationship of Multichannel Pricing on Purchase Intention ... 13
3.1.3 The Role of E-Service Quality in the Relationship of Multichannel Pricing on Perceived Value ... 15
3.2 Research Model ... 15
4. Methodology... 16
4.1 Data collection ... 16
4.2. Study Design & Procedure ... 16
4.3 Measurement of Variables... 18 4.3.1 Independent Variable ... 18 4.3.2 Dependent Variable ... 18 4.3.3 Mediating Variable ... 19 4.3.4 Moderating Variable ... 19 4.3.5 Additional Variables ... 19 5. Data ... 23 5.1 Item Coding ... 23 5.2 Reliability Analysis ... 23 5.3 Data Cleaning ... 25 5.3.1 Outlier Detection ... 25 5.3.2 Missing Values ... 26 5.4 Data Quality ... 26 5.5 Sample Description ... 27 6. Experimental Analysis ... 28 6.1 Pre-Analysis ... 28 6.2 Mediation Analysis... 29
7. Main Analysis Results ... 33
7.1 Mediation Analysis Results ... 33
7.2 Moderated Mediation Analysis Results ... 37
7.3 Results summary ... 38
8. Discussion & Conclusion... 42
8.1. Discussion ... 42
8.2 Managerial Implications ... 44
8.3 Limitations & Recommendations for Future Research ... 46
8.4 Conclusion ... 48
References ... 49
1
1. Introduction
Online shopping enjoys, in times of Amazon and Alibaba, great popularity and is, especially in the fashion and electronic sector, a commonly used method to obtain a product. While online shopping is indispensable for consumers in these industries, it is to date less used to purchase groceries. Despite more and more traditional supermarkets expanding to the online channel in order to provide the possibility of ordering groceries online and get the products delivered, purchasing groceries online is only regularly used by 5% of the consumer in Europe (Nielsen Holdings plc, 2017). Hence, online grocery shopping is not yet adapted by European consumers. Regardless of the low demand, online grocery retailers already struggle with the profitability, as the higher costs of handling perishable products for the online channel cannot only be compensated by delivery fees (PwC, 2016). However, looking ahead in the future, research predicts that within the upcoming five years, 70% of the consumers will purchase groceries online and around 20% of grocery sales will be generated in the online channel (Nielsen & Institute Food Marketing, 2018).
In response to the prognosis of a tremendous increase in demand and the current situation of the unprofitable online grocery business, research needs to address approaches to improve profitability. Prior studies already focused on the operational part of online grocery retailers, such as possible optimization of the supply chain or improvement of the different delivery methods in order to lower costs (Boyer & Hult, 2005; Hays, Keskinocak, & de López, 2005; Kämäräinen & Punakivi, 2002). Especially the delivery within the last mile accounts for up to 50% of the logistical costs (Vanelslander, Deketele, & Van Hove, 2013) and therefore calls for improvement.
been decreased (Tigert & Ring, 2001). Consequently, a possible solution, that needs to be considered, might be an increase of product prices, as it would increase product margins and therewith the profit of the online grocery retailers (Wöhe & Döring, 2010). Research already investigated the feasibility of price discrimination by accessing the consumers’ perception and acceptance of price differences between the online and offline channel. However, main subject of investigation were non-FMCG industries with focus on the acceptance of higher offline prices (Bailey, 1998; Dodds, Monroe, & Grewal, 2006; Fassnacht & Unterhuber, 2016; Homburg, Lauer, & Vomberg, 2019; H. G. Lee, Westland, & Hong, 1999). Corresponding results agreed that higher prices are often perceived as unfair and are only accepted up to a slight percentage increase (Fassnacht & Unterhuber, 2016; Homburg et al., 2019). Nevertheless, these findings cannot simply be transferred to the grocery sector, due to industry and accompanying customer behavior differences (Hoyer, W., MacInnis, D., & Pieters, 2013). Especially, as a recent study found that in the German online grocery channel prices are about 16% higher compared to the offline channel (Fedoseeva, Herrmann, & Nickolaus, 2017), upcoming research needs to focus on it, and investigate how higher online prices are feasible and accepted by consumers in the grocery sector.
Based on the aforementioned and the request of the Marketing Science Institute (2018) for further research on the setting of prices in different channels, the underlying research aims at filling the research gap for price differentiations in the online grocery sector. This is addressed by questioning whether higher online grocery prices are accepted under the provision of a higher service quality, in terms of faster delivery, and measured by means of the intention to purchase groceries online.
3
2. Literature review
The subsequent subsections introduce related research trends in multichannel retailing in general and focus in particular on research relating to the price differentiation between channels. Furthermore, the concept of online retailing, especially in the grocery sector, will be outlined with particular focus on evidence regarding the influence of e-service quality. Therewith, the consumer and his/her behavioral differences between the channels will be discussed, and behavior influencing factors investigated.
2.1 Multichannel Retailing
Multichannel retailing is defined as the sum of actions of a retailer in order to sell its products to the consumer via at least two channels (Levy & Weitz, 2009). Therefore, it is the retailer’s choice and responsibility to determine the appropriate channel according to his business and include it in their channel mix (Konuş, Neslin, & Verhoef, 2014). The most common are store, internet, and catalog retailing, with each having different assets and drawbacks that influence the customer shopping experience (Levy & Weitz, 2009). Li, Lobschat, & Verhoef (2018) broadened the perspective of Levi & Weitz (2019) on multichannel retailing by incorporating the shopping behavior of the consumer into the approach of multichannel retailing. Their approach does not just take into account that the consumer uses different channels for information gathering and product purchase, but also considers that consumers tend to engage in showrooming and change between competing retailers, especially under the condition of a bigger online price dispersion (Gensler, Neslin, & Verhoef, 2017). Henceforth, it is essential for a retailer to identify the best operating strategy for the different channels in order to not loose turnover.
Cross-Channel Integration
benefits to its consumers.” (Cao & Li, 2015, p. 200). In consideration of the above mentioned, cross-channel integration is comprised, inter alia, by the extent to which a retailer provides a seamless and consistent shopping experience for the consumer over all channels. Cross-channel integration can be accessed in various ways, for example, by offering the consumer to purchase products online and pick them up in a retail store, providing the opportunity to check online the availability of a product in the preferred retail store or ordering products online and allow to return them in the offline store (Gallino, Moreno, & Stamatopoulos, 2014). Furthermore, the alignment of the assortment between the channels, as well as consistency in pricing and the offering of discounts determine the level of integration (Cao & Li, 2015).
2.2 Pricing in Multichannel Retailing
The setting of prices, as a factor of the marketing mix, is one of a retailer’s fundamental decisions. Especially in multichannel retailing, an increasing importance is attributed to the pricing strategy of a retailer as it is not only used to measure the extent of cross-channel integration2 (Ancarani & Shankar, 2004; Pan, Ratchford, & Shankar, 2002), but is crucial in regard to customer behavior. Empirical studies in pricing strategies lead to heterogeneous findings. However generally, it can be differentiated between either setting equal prices over all channels in order to radiate consistency to the consumer (Pan et al., 2002), or discriminating not just prices but also discounts between the channels (Homburg et al., 2019; Wolk & Ebling, 2010), with the latter being highly influential on the consumers’ channel behavior. Focus of most of the research on multichannel pricing lies in the two most important retail channels, which are the online and offline channel (Levy, M. & Weitz, 2009).
The theory of economics of information proposes that online prices need to be lower as not just search costs are lower, but the consumer has full access to nearly all price information (Biswas, 2002). Also the differences in the cost structure between the channels lead to the assumption of cost advantages in the online channel, due to less personal and rental costs for the companies, amongst others (Fassnacht & Unterhuber, 2016; Ratchford, 2009). Consequently, it is often expected by consumers that multichannel retailers tend to set prices online lower than offline (Jensen, Kees, Burton, & Turnipseed, 2003).
Despite theoretical argumentation, real shopping settings appear to be controversial with prices online also being found to be higher as compared to the offline channel. These findings are
5
often with respect to individual product groups (Brynjolfsson & Smith, 2003; H.-G. Lee, 1998). An interview with Tristan Blössel3, Strategic Business Manager at REWE International AG, a grocery retailer operating in Europe, validates the discrepancy of pricing found in research. While BILLA AG, the Austrian grocery chain of REWE International AG charges the same prices online and offline and customers receive the same bonusses across the channels, the German grocery chain REWE of the REWE International AG applies a different strategy. REWE retail stores are divided in three regionally different groups, with each having a different price level for their stores. For the REWE online store, prices are set countrywide equal to the highest regional price charged offline and do not allow for any bonuses offered offline. Hence, a price discrimination between the online and offline channel exists for REWE customers in two out of three operating regions.
With this evidence from practice and the understanding that current research, especially in the grocery sector, is missing comprehensive insights of the price differentiation online and offline, it is especially essential to examine how differences in price affect the customer behavior in these channels (Chen & Dubinsky, 2003).
2.3 The Effects of Price Differentiation on Customer Behavior The Effect of Price Differentiation on Consumers’ Purchase Intention
In order understand customer behavior and predict subsequent product purchases, purchase intention is a widely used indicator in the relevant literature (Grewal, Krishnan, Baker, & Borin, 1998). The consumer’s intention to purchase a product depends, amongst other factors, such as perceived quality and value, on the perceived product price (Zeithaml, 1988), which highlights the overall importance of price setting. Generally, the effect of price on the willingness to purchase has been found to be negative (Dodds et al., 2006). This effect is caused by the intuition of individuals. In every purchase situation a consumer is exposed to a price cue. In order to evaluate the trade-off between the price and the product consumers set the product price in relation to the obtained product and, if applicable, a reference price. A higher price in this trade-off intuitively causes a negative mental association and leads to a lower probability of purchase (Lichtenstein, Ridgway, & Netemeyer, 2006).
This tendency in the purchase intention can be found especially in the multichannel retailing sector. With prices being different, not just between retailers but also between the channels of
the same retailer, consumers adapt their purchasing behavior to the price difference. This behavior is explained by the theory of utility which states that people generally try to increase their personal profit from a purchase and thereby expand their personal utility (Horton, 1984). For a multichannel retailer this means that the consumer shifts from one to the other channel to exploit price differences between them. Nonetheless, it needs to be taken into account that the switching of channels is not solely due to price differentiation itself and the accompanying trade-off decision, but influenced (in)directly by a variety of additional factors, for example, attributes of the channel and price sensitivity of the consumer (Breugelmans & Campo, 2016). Especially price sensitivity, as a measure of the importance of price, has been found to influence the consumer perception of the price difference, and therefore affects the purchase intention (Chu, Chintagunta, & Cebollada, 2008; Shankar, Rangaswamy, & Pusateri, 2001). The study by Shankar, Rangaswamy & Pusateri (2001) also reveals that the online medium itself leads to an increased price sensitivity over all channels, due to not just an easier access to price information, but also an intensification of price search which results in a change in the corresponding value perception. Besides this general finding, research has not led to consistent insights on consumers price sensitivity for the different channels. Degeratu, Rangaswamy & Wu (2002) found that the price sensitivity is higher online, due to the signaling effect of promotions for price discount. Contrarily, the combination of the effect of price and promotion has led to different findings with the price sensitivity being lower online than offline. Chu, Chintagunta & Cebollada (2008) confirmed the findings of the combined effect by discovering that the price sensitivity of households for groceries is higher offline than online.
In addition, the evaluation of the numerical price difference depends, besides the numeracy4 of the individual, on the individual perception of the difference, which varies with the personal income. A higher income has been found to lower the perceived costs of consumption in general. Therefore, higher income decreases the awareness for a price difference which results in a lowered price sensitivity. Additionally, income is not a stand-alone variable, but is mostly correlated with the level of education (Roddy, Cowan, & Hutchinson, 1996). Consequently, the income, as well as the level of education, need to be considered as influencing factors of the individual price sensitivity, if the effect of price differentiation on purchase intention is explored.
7
The Effect of Price Differentiation on Consumers’ Value Perception
Different price levels have not just been found to influence consumers’ purchase intention but also their perception of a product’s value (Dodds et al., 2006; Zeithaml, 1988). Perceived value is an abstract concept that is based on personal evaluation of a product’s utility as a trade-off between price and non-price information. Due to its subjective character, the perception depends, besides the above mentioned factors, also on the situation and product category, and is also often confounded with quality or price (Zeithaml, 1988). Monroe & Krishnan (1985) explained the concept of perceived value by the conceptual model of Monroe (1979), who defined it as a construct of perceived quality and perceived sacrifice with both depending on the perception of price. In this context, a higher perceived price has been anticipated with a higher sacrifice of money, and resulted in a lower perceived value. Zeithaml (1988) agreed with this partially and argued, that a low price might results in a high perception of a products value, but also a higher price can be justified by higher quality, and consequently leads to a higher perceived value. Dodds, Monroe and Grewal (2006) confirmed the findings of Zeithaml (1988), showing that price has a negative effect on perceived value, and outlined that the perceived quality is positively associated with the perceived value. On the other hand, Sheth, Newman, & Gross (1991) argued that price and the perception of quality, as well as sacrifice, are too limited to explain perceived value. Their construct is based on emotional rather than utilitarian evaluation. Thus, they grounded their theory of perceived value on the functional, social, emotional, epistemic5 and conditional value obtained through a product. The extension of the former understanding of perceived value was also agreed on by Sweeney & Soutar (2001). They especially highlighted the social aspect during the formation of value perceptions. Summarizing the relevant literature in the area of formation of value perceptions, most researchers agree on the influence of price on this formation process, but many do not limit it to just this factor but argue that other factors need to be considered, such as product and service quality, social context and demographic conditions like income-level.
The Effect of Consumers’ Value Perception on Corresponding Purchase Intention
Following up on the above-mentioned, the construct of perceived value is not just affected, inter alia, by the price of a product, but is also a widely used predictor of purchasing behavior. Research has shown that the concept of perceived value has a major role in the formation of the consumers’ intention to purchase (Chang & Wildt, 1994), and positively influences their willingness to buy (Monroe & Krishnan, 1985). Agreeing on this, Grewal et al. (1998)
discovered that about 40% of the variance in the intention to purchase is explained by the perceived value of a product. Chang and Wildt (1994) additionally found a connecting role of perceived value between price and purchase intention. They did not just confirm the prior research in the influence of price and quality on perceived value, but also discovered the mediating role of perceived value on the effect of perceived price on purchase intention. Their study agreed on prior findings that the purchase intention is positively affected by perceived value. As perceived value has been found to have a positive influence on purchasing behavior, it does not just need to be considered to understand the customer behavior in between online and offline channels. Additionally, it should be aimed at determining further factors that lead to an enlargement of perceived value in order to increase the customers purchase intention.
2.4 Online Grocery Retailing and Customer Channel Behavior Industry Induced Differences of Retailing
Findings from research mentioned in subsections 2.1, 2.2 and 2.3 are mainly conducted for non-food industries. Focus of this was primarily on online and offline price dispersion and customer behavior within product categories such as books and CDs (Bailey, 1998), cars (Lee, Westland, & Hong, 1999), or electronic products (Dodds et al., 2006; Homburg et al., 2019). However, these results are not transferrable to fast moving consumer goods (FMCG), especially the grocery sector (Hoyer, MacInnis, & Pieters, 2013), as not just the industry but also the consumer behavior differs in this sector. One main distinction with focus on industry differences is that the competition for usual retail products, especially online, is worldwide, while in the grocery sector it is more countrywide, or even local (Chu et al., 2008). This more localized competition is mainly due to the perishable nature of many groceries that does not allow worldwide shipping (Mortimer, Fazal e Hasan, Andrews, & Martin, 2016). Another distinction is that purchasing groceries is done more often, and therefore is a repeated action, while shopping other products is done irregularly within small time intervals (Chu et al., 2008). Thus, grocery shopping needs to be considered as more automated, and is therefore substantially different from shopping products in other categories.
Channel Induced Differences of Retailing
9
Additionally, the online channel in general provides the advantages of no necessity to travel to a store, and therefore leads to time saving and higher flexibility. More precisely, customers do not depend on closing hours, do not have to carry the products from the store and profit from a bigger assortment. On the other hand, also disadvantages arise with the online channel, such as no prior quality inspection of sensory products6, delivery fees, and waiting time, as well as missing personal purchasing assistance (Degeratu et al., 2002; Ghazali, Mutum, & Mahbob, 2006; Grewal, Iyer, & Levy, 2004).
The Effect of Channel Difference in the Grocery Sector
Focusing more in-depth on channel distinctions in the grocery industry, research confirmed prior findings that were based on overall channel distinctions, and also their effect on customer behavior. Online grocery shopping7 deviates from offline grocery shopping by several factors. One main differentiation between the channels in the grocery sector is the size of the assortment. While in general the assortment is bigger online than offline for non FMCG products, in the grocery sector, it has been found to be the opposite (Campo & Breugelmans, 2015). As cross-channel integration is a fundamental factor in the success of multicross-channel retailers, as elaborated in subsection 2.1, it is crucial for grocery retailers to improve the alignment of the assortment online and offline. Especially with regard to the aspect that customers are used for shopping groceries offline, not being able to purchase their commonly used products and brands online will negatively influence their tendency to purchase in the online channel.
Another factor that fundamentally influences the different perception of the channels is the perishability and sensory-character of grocery products. Accompanying with the online channel the option of self-selection of these products is missing and therefore no inspection of the product quality or characteristics, such as scent, prior to the purchase, is possible. (Chu, Arce-Urriza, Cebollada-Calvo, & Chintagunta, 2010; Degeratu et al., 2002; Mortimer et al., 2016). Additionally, Hand, Dall’Olmo Riley, Harris, Singh, & Rettie (2009) highlighted that customers need to change their shopping habits when purchasing groceries online due to not having the products arranged in aisles like in a retail store, but having to pick them from a list on the website. Henceforth, customers are required to change thinking prior to the online
6Sensory products are defined as products that product characteristics’ are usually identified by human
senses such as smell and touch (Degeratu et al., 2002)
purchase in order to be efficient and fast in picking necessary products. Besides that, the advantages that have been found in regard to online shopping in general and have been mentioned before, do hold for the category of groceries as well and might even be more important. Per example the weight and timing of a grocery purchase are elaborated. Grocery purchases are not just heavy in weight but also mainly done during peak hours. The restriction in opening hours and usual working hours lead to a relatively small slot of free time that most people use for grocery purchases. Henceforth, stores are often crowded which leads to longer waiting times and an inconvenient shopping experience (Morganosky & Cude, 2002).
Following up on these findings it is important to recognize the deviations between the channels, and that corresponding (dis-)advantages are valued differently based on person’s preferences. In line with the subjective evaluation, research has led to the consensus that the channel differences are the key influencing factor for the consumer’s preference for or against online grocery shopping. Morganosky & Cude (2002) found that time saving, and convenience are the main drivers of purchasing groceries online. This was agreed upon by Huang & Oppewal (2006), who found that paying a delivery fee is preferred over travelling 15 minutes to a retail store. Apart from that, Hand et al. (2009) discovered that a person’s situational circumstances, such as having physical or health constraints, or an infant, are the main reasons for online shopping preference. Counter arguments for the online channel are in conjunction with the channel properties. Research by Degeratu et al. (2002) and Dodds et al. (2006) has shown that the missing product tangibility in the online channel, especially for sensory products, leads to a preference for the traditional retail channel. Summarizing these findings of past research, it can be concluded that the decision to purchase groceries online is based on both a subliminal and subjective evaluation of the channel (dis-) advantages as well as influenced by personal preferences and situational context.
The Influence of Quality in the Grocery Sector
11
a result of the channel properties of the online channel, it does not provide information related to intrinsic product properties, and therefore enhances the evaluation of a products quality on extrinsic attributes (Teas & Agarwal, 2000).
The quality of the service, in turn, is determined as the outcome of an elaboration of the received service. Thereby, the contrast between expectations and reality plays a crucial role in the assessment of the level of service quality (Zeithaml, 1988). Grönroos (1984) already made a distinction between the technical and functional service quality, with the latter being more important in order to generate a high level of service quality for the customer. The way how service is delivered to the customer identifies the level of functional service, for example in the context of online grocery shopping, with regard to which delivery methods are offered to the consumer. Verhoef & Langerak (2001) found that the findings from Grönroos (1984) also hold for the concept of online grocery shopping. Henceforth, the intention for purchase groceries online is indeed dependent on the delivery options for the ordered products.
3. Research Model & Hypothesis Development
3.1 Hypothesis DevelopmentResearch in the last years put a lot of focus not just on the grocery sector in general, but also the difference in customer behavior for FMCG between the online and offline channels (Chu et al., 2008; Degeratu et al., 2002; Marimon, Vidgen, Barnes, & Cristóbal, 2010). Especially the perception of price (Ariely & Lynch, 2001), the variation in price sensitivity due to channel differences (Chu et al., 2008; Kalyanaram & Little, 2002), as well as the difference in customer behavior, based on product characteristics and situational factors (Chu et al., 2010; Hand et al., 2009), have been addressed. Despite the fact that these different aspects have been examined more extensively, as elaborated on above, little was researched on possible price dispersions between the online and offline channels for grocery purchases. Building upon the past theoretical findings, and extending these for the FMCG industry, hypotheses for online grocery shopping under the effect of channel price discrimination and different service levels will be developed in the subsequent subsections.
3.1.1 The Effect of Multichannel Pricing on Purchase Intention
Price differentiation between channels in the grocery sector needs to be evaluated from two perspectives. On the one hand, the sole perception of the differentiation between the channel prices needs to be considered. On the other hand, the asymmetry in this perception and its effect on the price sensitivity, due to product category differences, also calls for deliberation.
13
Besides the consumer expectations for prices in different channels, the different channel characteristics also influence the consumer’s price sensitivity. Current research generally agrees on the theory of Alba et al. (1997) and Bakos (1997) that the online channel lowers the costs of search for price information and therefore leads to a higher price sensitivity in the online channel. However, the electronic channel does not only provide easy access to price information, but especially a multitude of quality information which leads not only to a decrease in the price sensitivity of the consumer, but also allows for higher prices in the online channel (Ancarani, 2002; Ariely & Lynch, 2001). The research of Ariely and Lynch (2001) is based on differentiated products like wine. For these kinds of products, the decrease in the price sensitivity, due to the availability of quality information, exceeds the increase in the price sensitivity in the online channel. Much to the contrary, the act of grocery shopping is understood as a basic process, which is characterized by the low-involvement of the consumer. Due to its repetitive nature and the low costs associated with grocery shopping, consumers do not require many detailed quality information per product (Raijas, 2002). Therefore, it can be assumed that the positive effect of non-price information on the price sensitivity is not applicable in the grocery sector. Following this assumption, a higher price sensitivity exists for the grocery shopping online compared to offline, which does not allow for higher online prices in grocery retailing. The customers’ expectation of lower prices online in combination with a higher price sensitivity for the online grocery sector leads to the following hypothesis:
H1: A higher online price has a direct negative effect on the online purchasing
intention.
3.1.2 The Role of Perceived Value in the Relationship of Multichannel Pricing on Purchase Intention
be perceived as unusual or frightening, which causes the impression of an unjustified price discrimination and therefore, leads to a decrease in the perceived value. Overall it is expected that the positive and negative perception of the situational context are balanced, and the effect of price is stronger than the effect caused through the situational context. As research in general agrees on the negative effect of price on perceived value (i.a. Dodds et al., 2006; Zeithaml, 1988), the following is hypothesized:
H2a: A higher online price is negatively related to the perceived value.
Investigating the effect of perceived value on purchase intention, empirical research agrees that perceived value is a strong predictor for the purchase intention (Zeithaml, 1988), especially after it has been found to explain about 40% of the variance of the intention to purchase (Grewal et al., 1998). Referring to the grocery sector, perceived value is expected to have a similar positive effect. As mentioned previously, the process of value perception is also influenced by the situational context. In the situation of grocery shopping, the channel itself also represents a situational context, and therefore influences the purchase intention. Accounting on this and considering that the process of value perception for groceries is also channel dependent and leads to a channel dependent perceived value, the perceived value of a product may be divergent for the different channels. In turn, this can also lead to different purchase intentions per channel. However, the overall positive effect of the construct of perceived value on purchase intention is expected to hold for the online channel as well. Henceforth the following hypothesis is assumed for the online channel:
H2b: A higher perceived value is positively related to online purchasing intention.
15
H2: The effect of a higher online price on the online purchasing intention is mediated
by the perceived value.
3.1.3 The Role of E-Service Quality in the Relationship of Multichannel Pricing on Perceived Value
Chang and Wildt (1994) did not just reveal the mediating role of perceived value in the relationship of price on purchase intention, but also found that perceived value is positively influenced by perceived quality. Marimon et al. (2010) already confirmed that this effect can be found for online supermarkets as well. They assessed the effect on perceived value by using the E-S-QUAL scale developed by Parasuraman et al. (2005). Their research lead to the result that the dimensions ‘efficiency’ and ‘privacy’ have no significant impact on perceived value, while the dimensions ‘system availability’ and ‘fulfillment’ have positive significant effects, with the latter being the strongest predictor of perceived value. Henceforth, the following relationship is hypothesized:
H3: The effect of a higher online price on the perceived value is moderated by the
level of e-service quality.
3.2 Research Model
Figure 1: Research Model
H3
H2b H2a
H1 Multichannel Pricing
• equal price between offline & online shopping basket
• 20% higher price for the shopping basket online
Purchase Intention Online
E-Service Quality
• High level of fulfillment (delivery within 2 hours) • Low level of fulfillment
(delivery within 7 hours)
Perceived Value Of The Online Shopping Basket
Covariates
Expenses on groceries/week Planned purchases (yes/no) Amount of people/household Gender
Perceived Price Fairness Perceived Time Pressure Enjoyment of shopping Need for touch
4. Methodology
The experimental design consists of a decision-making situation between the online and offline channel based on a shopping scenario for groceries. Participants were asked to evaluate the presented scenario, which is manipulated in form of different price and service levels. After being introduced to the scenario, participants rate the value of the offer, their personal intention to choose the online channel as well as answer further personal and demographical questions.
4.1 Data collection
In order to test the hypothesized relationships between the different variables, an experimental study was conducted. The survey took place between April 12th and May 6th, 2019. It was spread over several social media channels (e.g. Facebook, Instagram) in order to randomize the sample. Participants were offered the choice to answer the survey either in English or in Germany.
4.2. Study Design & Procedure
The conducted study used a 2x2 between-participants experimental design, resulting in four different experimental conditions. It comprises the price difference between the online and offline channel (same price/price surplus of 20% online) and the level of fulfillment (delivery within a time frame of 2 hours/7 hours) as a measure of E-Service quality (Table 1). Participants were randomly assigned to one of the conditions.
IV
Fulfillment
delivery within 2 hours (high service quality)
delivery within 7 hours (low service quality) Equal price online and
offline
14.89€
delivery within 2 hours
14.89€
delivery within 7 hours
20% higher online price 17.87€
delivery within 2 hours
17.87€
delivery within 7 hours
Table 1: Experimental Conditions and Manipulations
17
weekly grocery purchase is imminent, and the presented exemplary shopping basket conforms their household needs of a week. For the purpose of not having biases due to differences in the in the perceived urgency of getting the groceries, participants were also told that they do not need the exemplary groceries as soon as possible, but that it is sufficient to get them on the next day. Afterwards, the exemplary shopping basket was introduced, which consists of four commonly used products. The shopping basket is included to make the setting more tangible and vivid. Therefore, the selection of items includes only generally used products. Moreover, the number of items is deliberately kept small to not distract the participants with the shopping cart content. Additionally, the items in the shopping basket do not depict any brands. This was done on purpose since research by Degeratu et al. (2002) and Dodds et al. (2006) has shown that brand names are more valued online and therefore, could lead to biased results in the present study. Furthermore, their research found differences in the effect of sensory and non-sensory attributes of products on consumer choice behavior in the online channel. According to these findings, the exemplary shopping basket includes both types of products with an equal share in order to prevent an inclination in the preference for one channel due to sensory properties of the single products in the shopping basket.
After being familiarized with the general setting, the participants were exposed to the price of the shopping basket to be paid at their offline supermarket. Additionally, they were provided with the price for the option to buy the same shopping basket at an online grocery shop and get the groceries delivered on the next day. To avoid reluctance for the online purchase, due to doubts about the condition of the fresh products after delivery, it was mentioned that the online shop guarantees that all products will be delivered in a flawless condition and are ready to consume. By stating both prices, online and offline, it is assured that all participants are aware of a price difference between the channels and their purchase intention is not biased by a missing reference price.8 The introduction part ends with indicating that choosing to not shop groceries is not an option and participants can solely choose between shopping them online or in a traditional supermarket. This information was included to ensure that participants do not have inner restraints when it comes to the measurement of the purchase intention.
The first manipulation of the experiment was implemented by setting the online price equal to the price of the shopping basket in the supermarket or 20% higher compared to the offline
price.9 The second manipulation comprises the level of fulfillment. Therefore, the delivery of the groceries could either take place within a time frame of two or seven hours during the next day (see Appendix A for a full scenario description).
After being introduced to one of the four conditions, the participants were asked several questions on the perceived value, perceived price fairness and purchase intention. Following the advice of Hauser & Schwarz (2015), attention checks were included after the fundamental measures of the study, to ensure that participants were not influenced in their interpretation of the crucial questions. Furthermore, the participants were asked about other factors that might influence their channel choice for grocery shopping such as price consciousness, convenience, time pressure, the need for touch and cognition and the enjoyment of grocery shopping. Finally, participants were asked to answer some demographic questions concerning age, gender, income and education (for the complete questionnaire see Appendix B).
4.3 Measurement of Variables 4.3.1 Independent Variable
The independent variable is categorical and represents the two levels of price that the shopping basket can take. The first level is represented by the equal price in both channels while the second level is represented by a 20 percent higher price for the same shopping basket in the online channel.
4.3.2 Dependent Variable
The online purchase intention is included as a dependent variable and is accessed by a four-item measure on a 7-point Likert scale (1=’strongly disagree’, 7=’strongly agree’). It was adapted from Dodds, Monroe, & Grewal (2006) and adjusted for the purpose of the survey. The intention to do something results not always in the corresponding behavior, because several other factors, such as subjective norms amongst others influence a person’s behavior (Ajzen, 1991; Ajzen & Fishbein, 1970). Even though people attempt to reflect their intentions and preferences by their choices (Drolet, Luce, & Simonson, 2008), both constructs are not always in line. In order to be able check for this bipolarity, it is necessary to test if the purchase intention for the shopping basket online is consistent with the consumers behavior under the condition of choice between the online and offline channel. Consequently, it was measured on a
19
point Likert Scale whether the respondent would rather buy the shopping basket offline or online (0=’preferably offline’ to 7=’preferably online’).
4.3.3 Mediating Variable
In order to test the hypothesis of the mediating effect of perceived value on the purchase intention, consumers were asked to evaluate their perceived value of the shopping basket. Corresponding to the construct of perceived value from Zeithaml (1988), who defined perceived value as a tradeoff between the costs and the benefits that a customer has to face, the participants were asked to rate five item statements that account for this comprise in the customers decision, on a 7-point Likert scale. The scale resembles previous approaches to measure the so defined concept of perceived value (Dodds et al., 2006; Marimon et al., 2010; Sirdeshmukh, Singh, & Sabol, 2002).
4.3.4 Moderating Variable
In order to test hypothesis 3, which assumes a mediating effect of e-service quality on the relationship of price on perceived value, a condition that reflects the provided e-service quality was included in the survey. With regard to the limitations for this research and the findings of Marimon et al. (2010) that the dimension ‘fulfillment’ is the strongest predictor of perceived value, e-service quality will be simulated only using the dimension of ‘fulfillment’. ‘Fulfillment’ is referred to as the “extent to which the site’s promises about order delivery and product availability are fulfilled” (Marimon et al., 2010, p. 4). Henceforth, two levels of delivery time have been included as a categorical variable that either represents a low level of fulfilment (delivery within a seven-hour time frame) or a high level of fulfillment (delivery within a two-hour time frame).
4.3.5 Additional Variables
As consumer characteristics might have an influence on the purchase intention, they can influence the experiment in the following three ways: they could either function as competing mediator, as competing moderator in the experimental setting or serve as information provider about the respondents’ shopping behavior. Thus, the following variables were additionally measured during the experiment.
Perceived Price Fairness
test whether this effect can also be found under the manipulation of the service quality of the e-grocery retailer. Therefore, participants were asked to rate three items on a previously used 7-point Likert Scale that evaluate fairness, acceptability and justifiability of the online price (Homburg et al., 2019).
Demographics
Former research in the field of grocery shopping has shown that differences in demographics can have an impact on the shopping behavior and influence the personal price importance (Degeratu et al., 2002; Verhoef & Langerak, 2001). Consequently, the insertion of gender, age, income level and education is important in order to test and control for possible moderating effects of demographic variables on the influence of price on the mediating variable. An extensive register of the demographic variables can be found in Appendix B.
Price Consciousness
Price consciousness is generally defined as the extent to which a customer’s focus lies mainly on price awareness and aiming to pay the lowest price (Lichtenstein et al., 2006). In order to compare if the moderating effect of price consciousness may be stronger than the effect of quality of the e-service, a five-item scale, based on previous research by Lichtenstein et al., (2006) and Wakefield & Inman (2003), was included in the survey. Participants were asked about their effort in regard to price comparison and the resulting personal value gain arising from that.
Time Pressure
21 Convenience
Research has shown that one of the most relevant reasons to use online-grocery shopping is convenience (Chu et al., 2008; Rohm & Swaminathan, 2004). Convenience in this context refers to situational factors such as no waiting in line, no need to pick and carry groceries (Fedoseeva et al., 2017) and also to additional offerings, such as personalized and saved shopping lists (Degeratu et al., 2002). The scale is adapted from Rohm & Swaminathan (2004) and measures five items in regard to the convenience associated with the internet.
Enjoyment of shopping
While online shopping for groceries may save time and is very convenient for some consumers, for others the enjoyment of shopping is lower online as they prefer going through the store, being exposed to smells of food and get inspired by the assortment (Bellenger & Korgaonkar, 1980). Thus, enjoyment of shopping was not found as an factor that triggers online grocery shopping (Rohm & Swaminathan, 2004). To test whether this also affects the influence of a higher online price on the perceived value of the online shopping basket, a five-item scale from Gehrt & Shim (1998), focusing on the several factors of pleasure of offline shopping, was included in the survey.
Online Shopping Categories
In order to acquire insights on the participants’ online shopping behavior in general, they were asked which product categories they usually buy online. The categories offered for selection are based on the eight most popular online shopping categories worldwide in 2018 (Statista, 2018). Additionally, participants were offered the opportunity to state further categories. Under the condition that participants have chosen or not chosen ‘Groceries’ as a usually purchased category, they were asked for their main reasons for engaging or not engaging in online grocery shopping. Besides further understanding of the participants behavior in general, this aspect was also included to validate the results from the underlying survey. Especially further influencing factors, such as ‘No availability of online grocery shopping in my area’ or ‘Physical disability to shop groceries in a supermarket’, which are not covered in the conceptual model of this thesis but might influence the participants’ attitude towards online grocery shopping, can be distinguished.
Need for touch
account for potential influences on the participants’ individual attitude towards the online shopping basket and the decision for or against it. Therefore, participants were asked to rate their level of agreement to the four of eight statements developed by Peck & Childers (2003) on a 7-point Likert-Scale.
Need for cognition
23
5. Data
5.1 Item Coding
Some variables in the survey have been reversed coded. Reverse coding means that the survey questions is asked in such a way that the higher scores becomes the lower one (7 = 1 or 5 = 1) and vice versa. In order to account for the type of questioning, the reverse questioned items were recoded. Additionally, the level of employment was inquired in a descending order (1 = ”working full time” to 5 = ”High-School student” and 6 = ”prefer not to answer”). In order to align this variable with the remaining variables, the coding was changed into ascending order (1 = ”prefer not to answer” to 6 = ”working full time”) .
5.2 Reliability Analysis
Following up on the recoding of the items, one needs to test whether the associated items of the different independent variables can be combined into factor variables. The combination of the items depends on their internal consistency which has been accessed through a reliability analysis. Internal consistent variables are characterized by a Cronbach’s Alpha that is higher than 0.6 (Bagozzi & Yi, 1988). Table 2 shows that only the items for the variables ‘Convenience’ and ‘Need for Cognition’ do not reach the threshold of 0.6 and therefore, cannot be combined into factor variables. Henceforth, only the internally consistent variables have been combined by averaging the values of the items per factor variable.
Variable Number of Items Cronbach's Alpha Internally Consistent
Purchase Intention Online 4 .96 yes
Perceived Value 5 .92 yes
Perceived Price Fairness 3 .93 yes
Price Consciousness 5 .71 yes
Perceived Time Pressure 4 .85 yes
Convenience 4 .59 no
Enjoyment of Shopping 5 .82 yes
Need for Touch 4 .88 yes
Need for Cognition 3 .50 no
Further investigation of the internally inconsistent variables led to the insight that Cronbach’s Alpha would exceed the threshold of 0.6 if the prior reverse coded item for the variable ‘Need for cognition’ and the single item that was not reverse coded for the variable ‘Convenience’, would be deleted. This indicates a low inter-item correlation and allows for the assumption that the difference in the formulation of the questions might led astray the respondents. Nevertheless, the two variables need to be excluded from further analysis, as the construct of questions was successfully used in prior research (see Rohm & Swaminathan (2004) and Ailawadi, Neslin, & Gedenk (2001) and Wood & Swait (2002)), but does not lead to internally consistent variables for the underlying sample.
The effect of two the manipulated variables ‘Price’ and ‘E-Service Quality’ was measured in two dimensions. The first dimension is the objective measurement on the intention to purchase the shopping basket online which bases on the four items questioned in the survey, and that merge to the variable ‘Purchase Intention Online’. Dimension two is the choice preference between the offline and online channel, namely the variable ‘Offline-Online Choice Preference’, which was accessed by one question. It is therefore subjective and based on self-assessment of the respondents. In order to determine whether the respondents reviewed both variables similarly and the subjective and objective measurement are in line, firstly the frequency distribution was investigated. Figure 2 shows that the distribution is not completely equal, but both variables have the same tendency.
Second, to further analyze the compliance of objective and subjective evaluation of preference, a reliability analysis was conducted on the two variables. As the Cronbach’s Alpha of purchase intention online and offline-online choice preference is 0.91, reliability is given. Henceforth,
25
the measurement of a higher intention to purchase the shopping basket online is in line with the respondents self-reported preference to choose the online channel for this purchase. Thereupon, the two variables have been combined into the variable ‘Online Purchase Intention’, which will be further referred to as ONLINEPI.
5.3 Data Cleaning
In order to improve the data quality and ensure that the results of the analysis are not biased, the data needs to be cleaned. Cleaning includes the detection of errors and unusual values as well as the recognition of inconsistencies, so-called outliers, and dealing with these observations (Rahm & Do, 2000).
5.3.1 Outlier Detection
For the underlying sample outliers can only occur for questions for which responding consist of a free text entry. Only four questions in the survey used this type of questioning. For all the other variables, a scale between 1 to 7 or a multiple-choice option was prescribed. Henceforth, outliers could only be identified for the variables, listed in the following. For the sake of identification of outliers, boxplots have been created in SPSS.
Grocery Shopping per Week
For the times of grocery shopping per week three outliers have been detected. SPSS detected the frequencies of 150, 50 and 7 as outliers. As 7 equals the amount of days per week, this observation might be unusual, but still possible for people who do not plan their meals and go to the supermarket every day. Therefore, just the two observations have been excluded from the analysis.
Average Spending per Week on Groceries
as outliers as well. As these observations stated that they purchase groceries for two or more people, these values also seem to be realistic and are therefore kept in the analysis, too.
Amount of People
SPSS detected four outliers for the variable amount of people groceries are purchased for. With respect to the prior mentioned correlation, two observations can be kept in the analysis as these are the families of 8 people that have been mentioned before. Henceforth, only two observations can be detected as outliers. One stated that groceries are purchased for 100 people, which is unrealistic, while the other indicated to purchase for 10 people by only spending 4 Euros per week.
Age
For the variable age 12 outliers have been detected. As the age of these respondents varies between 42 and 67, which is a normal age, no observation has been excluded from the analysis.
5.3.2 Missing Values
As respondents have not been forced to answer, they had the ability to skip questions. While values for reasons of (not) purchasing groceries are not missing at random, because they depend on the respondents answer in regard to what he or she bought online in the last six months, the other values are missing at random. To assure that the missing values do not affect the results of the analysis, the cases with at random missing values have been deleted pairwise in further analysis. Hence, the observations have still been used for every part of the analysis for which they have a valid value.
5.4 Data Quality
In order to determine whether the respondents have paid attention to the described scenarios in the survey, two attention checks regarding the price and the manipulation of e-service quality have been included (Oppenheimer, Meyvis, & Davidenko, 2009)10. Considering both checks, only 71% of the sample would have paid full attention. Contrary to this approach, Vannette (2017) revealed that judging on the basis of failed attention tests in order to exclude these participants from the sample would harm the data quality. Following his recommendation, only respondents that failed both attention checks will be excluded for further analysis. In the sample this applies to 5% of the sample which equals 8 respondents that have to be excluded.
27
The remaining sample consists of 139 respondents, after excluding the detected outliers and the respondents that failed both attention checks. Since the participants have been randomly assigned to one of the four possible scenarios, it needs to be ensured an equal distribution of scenarios over the remaining sample has to be ensured. The analysis led to the conclusion that, with a variation of maximum 5 observations per condition (3%), a nearly equal distribution exists.
5.5 Sample Description11
Demographics
After data adjustments the sample contains of 139 randomly selected people who completed the survey. 60% (N = 84) of the sample are female, while the remaining part is male. A gender equality could not be reached. Additionally, the average age of the respondents is relatively low, amounting to 29 years (SD = 10.17, min. = 18, max. = 67). 63% (N = 87) of the respondents answered the survey in English, the rest preferred German. The sample mainly consist of university students (52%, N = 72) and full-time workers (32%, N = 45) of whom 70% (N = 97) have a bachelor’s or master’s degree. In line with the amount of students in the sample, the personally estimated monthly income is below €1.200 in 51% of the cases (N = 71), while the rest is nearly split equally over the three remaining income categories: 17% (N = 24) for €1,200 to €2,000; 10.1% (N = 14) for €2,000 to €3,000; 12% (N = 17) for > €3,000).
Grocery Purchase Behavior Variables
The respondents in the sample go to the supermarket nearly three times per week (M = 2.71, SD = 4.38) and spend around 64 Euros per week on groceries (M = 63.93, SD = 54.28), ordinarily. These expenses are on average based on purchases for two persons (SD = 1.29). Even though 84% (N = 116) of the sample plan their grocery purchase in advance, which could be a main driver for online grocery shopping due a not direct availability of products, only 16% (N = 22) of the sample have purchased groceries online in the last six months. The possibility of shopping other categories online such as clothes (70%, N = 97), shoes (45%, N = 62) or books/movies/music (39%, N = 54) has been used more often.
6. Experimental Analysis
The following section constitutes the examination of the existence of the hypothesized relationships between the dependent variable and the chosen predictors. Furthermore, it aims to validate the effect and direction of the hypotheses. Henceforth, a moderated mediation analysis will be conducted to measure the conditional indirect effect of ‘Price’ on ‘Online Preference’ following the model 7 of Andrew F. Hayes (Hayes, 2017). In order to do so, pre-analysis was conducted before that to assure that the moderated mediation is not biased due to correlations of the predictors
6.1 Pre-Analysis Dependent Variable
A plot of the dependent variable ‘Online Purchase Intention’ shows that the variable does not follow a normal distribution as it is skewed to the right (figure 3). This is confirmed by the Shapiro-Wilk test which assumes normally distributed data in its null hypothesis. As this test is highly significant (b =.87, p < .01), the null hypothesis needs to be rejected, leading to the conclusion that the dependent variable is not normally distributed, and thus, more observations exist for a lower online purchase intention (figure 3). As a normal distribution is not necessary to conduct mediation analysis, which uses bootstrapping in an OLS regression, non-normality in the data does not affect further analysis (MacKinnon, Lockwood, & Williams, 2004).
Figure 3: Histogram & Normal Q-Q Plot of Online Purchase Intention
29
prefer to purchase the shopping basket offline than online since a value of 3.5 for the ‘Online Purchase Intention’ reflects indifference between the offline and online channel.
Conditional differences
Aside from testing the hypothesized relationships from subsection 3.1, it is interesting to examine if there is a significant difference in the online purchase intention for the shopping basket across the four experimental conditions in the survey. A univariate analysis with post-hoc comparison (Turkey HSD) was conducted. Results show a highly significant effect of the experimental condition on online purchase intention between groups (F(3, 134) = 3.67, p < .05). The pairwise comparison of conditions shows that only the conditions same price/high fulfillment (M = 3.96, SE = .34) and higher price/low fulfillment (M = 2.53, SE = .32) significantly differ from each other (∆𝑀 = 1.43, 𝑝 < .05). However, for the remaining conditions no significant difference between the means could be detected.
Multicollinearity Analysis of Independent Variables
For the following mediation analysis, multicollinearity can be an issue. Due to the fact that PROCESS v3.3 add-in from Andres F. Hayes (Hayes, 2017) will be used to perform a mediation analysis and this add-in produces an error if multicollinearity is a problem, a separate test for multicollinearity would not be necessary. Nevertheless, for the sake of completeness, a further test for multicollinearity was performed in order to obtain the variance inflation factor (VIF) scores. As all VIF scores are below the threshold of 5 (Craney & Surles, 2002), which would reflect moderate multicollinearity, it can be concluded that multicollinearity is not a problem for the following analysis.
6.2 Mediation Analysis
Total Effect Model
Figure 4: Statistical Model for Mediation
Summed up, it can be derived that the direct effect of PRICE on ONLINEPI equals c’ and the indirect effect of PRICE on ONLINEPI through PVALA equals ai * bi with the total effect C=
ai * bi + c’. Following up on the approach from Zhao, Lynch, & Chen (2010), neither the indirect
effect c’, nor the total effect C need to be significantly different from zero. In order to have a mediating effect, only the indirect effect of X on Y over M, which is ai * bi, needs to be
significant.
The model is estimated in SPSS using the model 4 in the PROCESS v3.3 add-in from Andres F. Hayes (Hayes, 2017). This add-in includes bootstrapping. Bootstrapping is a resampling procedure that draws various sub-samples from the original sample with replacement and thereby assures that the size of the sub-sample is the same as the original. The add-in PROCESS provides the possibility to estimate bootstrapping intervals that allow to judge about the significance of the indirect effects of the mediator (Hayes, 2017; Zhao et al., 2010). Henceforth, a 10000 bootstrap sample and a 95% confidence interval will be used for the estimation.
31
may arise due to respondents who answered the survey either in English or German. For interpretation purposes (Echambadi & Hess, 2007), the relevant variables in the hereby conducted analyses were centered to the mean (i.e. Age) or centered to the average category of the variable (i.e. Employment status, category: “out-of-work”). Centering was either applied by subtracting the variable’s overall mean from the value of each observation, with ten decimal places for the purpose of accuracy, or by subtracting the value of the average category (i.e. 3 for Employment status), from the value of each observation.
6.3 Moderated Mediation Analysis
If the hypothesized mediation will be confirmed, a moderated mediation analysis will be conducted in order to test whether hypothesis 4 from subsection 3.1 holds. In addition to mediation, moderated mediation not only measures the indirect effect of independent variable X on the dependent variable Y under the mediator M, but indicates also the effect of X on M which is moderated by the moderator W and reflects the conditional indirect effect of X on Y. Statistically, the model differs from the mediation model as follows, including the additions which are marked in green.
Figure 5: Statistical Model for Moderated Mediation
Based on this, the direct effect of PRICE on ONLINEPI equals c’ and the conditional indirect
effect of PRICE on ONLINEPI through PVALA, which is moderated by SERVQ, equals
(a1i + a3i * W) * bi. As mentioned above, the significant difference from zero of the conditional
indirect effect of X on Y through M, under the different levels of W, needs to be significant in order to distinguish a moderated mediation (Zhao et al., 2010)
33
7. Main Analysis Results
The following subsections will provide the results of the analysis of the different statistical models from sections 6.2 and 6.3. Based on this, a summary of the results will be provided offering an overview of the acceptability of the developed hypothesis from section 3.1. This section ends with the final estimated equation and the interpretation of the different parameter estimates in the context of the underlying experiment.
7.1 Mediation Analysis Results
Following the method of Baron & Kenny (1986) to report the analysis outcomes, the total effect C, as well as the individual effects a and b and the direct effect of X and Y, namely c’ will be investigate. Most important, the indirect effect a*b will be in distinguished and reported. Based on these effects an overall conclusion of the existence and type of mediation will be inferred based on Hayes (2017) and Zhao et al. (2010).
Without Covariates
The first conducted analysis examined whether a simple mediation of the effect of PRICE on ONLINEPI through the mediator PVALA exists. For this analysis, which focuses on the existence of mediation in the underlying dataset, no covariates have been added to the model. As a consequence of not including any covariates, the sample size for this model is 138. The results of this mediation analysis indicate that, even though the explained variance of the total model is with 39% not very high, the overall model is highly significant (R2 = 0.39, F(2, 135)
= 8.91, p < .01).
Second, in order to test for mediation, the different paths a and b, taken from the statistical model in figure 4, need to be investigated separately. The constant for the effect of X on M, is positive and highly significant (b = 4.83, t(136) = 29.08, p < .01). Thus, it reflects that the perceived value of the exemplary shopping basket equals 4.83 when the price for the shopping basket is the same online and offline (i.e. PRICE = 0). The ai-path from the statistical model, which is the effect of X on M when the price for the exemplary shopping basket is higher online than offline (i.e. PRICE = 1), is negative and highly significant (b = -1.31, t(136) = -5.72, p < .01). This outcome implies, that the perceived value of the shopping basket is lower by 1.31 than under the condition of the same price online and offline and therefore equals 3.52 (i.e. 4.83 – 1.31). Also the bi-path, which is the effect of M on Y, is highly significant (b = .85, t(135) = 8.45, p < .01), meaning that an increase of the perceived value by 1 leads to an increase in the online purchase intention for the shopping basket by 0.85. Since both the constant and the direct effect of X on Y, which equals the c’-path in the statistical model, are not significant (p > .05), it can be assumed that mediation exists.
Following upon these findings, it needs to be verified that the resulting total effect is based on the mediating effect through PVALA. The indirect effect of X on Y through M, which equals ai * bi in the statistical model, is highly significant (b = -1.11, SE = 0.22, 95% CI [-1.56, -0.71]. As zero is not included in the confidence interval, the mediating effect of PVALA is confirmed.
The combination of the insignificance of c’ and the high significance of the indirect effect ai * bi as well as of the total effect C leads to the conclusion that the effect of PRICE on ONLINEPI is only given through the mediator PVALA. Hence, an indirect-only mediation exists (Zhao et al., 2010). The effect of X equals therefore ai * bi, which equals -1.11, meaning that a higher price for the shopping basket online decreases the online purchase intention by 1.11. Equation 1.3 summarizes the total full model.
