Price sensitivity
on the web and
in the stores
Researching the price sensitivity and the influence of switching costs on
consumers in online & offline sales channels.
Rinze Postma
Price sensitivity on the web and in the stores.
Researching the price sensitivity and the influence of switching costs on
consumers in online & offline sales channels.
Master Thesis
MSc Marketing; Management Track University of Groningen Faculty of Economics and Business
9-11-2014
Supervisor: dr. prof. Laurens Sloot 2nd supervisor: mr. Sebastian Sadowski
Management Summary
The main objective of this research is aimed at researching the differences in price sensitivity of consumers in online and offline sales channels. In a rapid changing world were online shopping becomes more important, and price competition amongst retailers is high, having knowledge about these topics may be important for retailers.
This research tries to assess the differences in price sensitivity between online and offline sales channels. Furthermore the roles of perceived switching costs and the type of products is examined. A distinction is made between three different types of switching costs, namely procedural, financial and relational switching costs. Furthermore this research divides products in two groups, non-durable and durable. It is expected that perceived switching cost will have a reducing influence on the price sensitivity of consumers. Moreover it is expected that consumers perceive price sensitivity differently amongst both types of products. To be more precise, it is expected that the perceived price sensitivity of consumers in the offline and online sales channel is mediated by the perceived switching costs. Furthermore it is expected that the type of product is moderating the relationship between the sales channel and the perceived switching costs. Reasoning for this is that consumers are expected to perceive higher switching costs when they shop for durable goods in the offline channel. The previously described is summarized into the research question of this paper:
To what extent does price sensitivity differ between the offline and online channel, taking into account the retail sector, and how can the several types of switching costs as perceived by consumers explain these differences.
The research was conducted by spreading an online questionnaire. A total number of twenty possible scenarios were developed, and each respondent was presented with only one scenario. The final database consisted out of 209 respondents, which were used for the analysis.
perceived financial switching cost in the offline and online channel. However when it comes to durable products, perceived financial switching costs are much higher in the online sales channel than in the offline sales channel. The exact role of the perceived switching cost with regard to the perceived price sensitivity however, stays unanswered in this research paper. It was expected that the differences in perceived price sensitivity in both sales channels would be mediated through perceived switching costs. However no evidence of such a mediating role for perceived switching costs has been found. Furthermore it was found that when consumers perceive high procedural switching costs they become less sensitive towards pricing.
Preface
This paper is my master thesis for the MSc Marketing at the University of Groningen. It is the final product of my 5 years of study at the University of Groningen. In the year 2009 I started with the bachelor Business Economics at this University. After four years I’ve completed this bachelor and had to make up my mind about which Master I wanted to try to obtain. I’ve doubted along time between a master Finance and a master Marketing. Since both fields of study had my interest, it was hard to make a final decision. However in the end, after a lot of doubting, a Master marketing became the study of my choice.
As I preferred some courses more over others, I’ve knew from the beginning that I made the right choice, when choosing for a master Marketing. During this master I focused on the management track. Especially the Retail Management course stood out from the others in my experience. I really enjoyed participating in this course, moreover because of the fact that I already obtained a lot of practical experience in the field of retail. It soon became clear that I also wanted to write my master thesis for this subject, and the final product lies here in front of you as a reader. Although I enjoyed my time as a student at the University of Groningen, the time for me has come to take the next step.
I would like to finish this preface by thanking some people. First I would like to thank prof. dr. Laurens Sloot. He has proven to be of great help and support during the process of writing this master thesis. He was always willing to help and to provide feedback. Furthermore I would like to thank my fellow thesis group members. They’ve also provided me with help, feedback and, sometimes, moral support when needed. Finally I want to thank my parents, my closest family and my friends for their unremitting moral support during my years of study.
TABLE OF CONTENT
MANAGEMENT SUMMARY ... 3
PREFACE ... 5
1. INTRODUCTION ... 7
2. PROBLEM STATEMENT AND RESEARCH QUESTION ... 8
3. LITERATURE REVIEW ... 9
3.1 ONLINE ... 9
3.2 PRICE SENSITIVITY & PRICE ELASTICITY ... 12
3.3 SWITCHING COSTS ... 14
4. CONCEPTUAL MODEL & HYPOTHESES ... 17
5. METHODOLOGY ... 20
5.1 THE RESEARCH DESIGN ... 21
5.1.1 General Elements ... 21
5.1.2 Pre-‐set variables ... 21
5.2 MEASURING THE CONCEPTS ... 22
5.2.1 Measuring Price Sensitivity ... 22
5.2.2 THE PERCEIVED SWITCHING COSTS ... 23
5.2.3 Control Variables ... 23
5.3 QUESTIONNAIRE AND SCENARIOS ... 24
5.4 PLAN OF ANALYSIS ... 25
5.4.1 General tests ... 25
5.4.2 Testing the direct effects ... 26
5.4.3 Testing Moderation ... 26
5.4.4 Testing Mediation ... 26
6. THE RESULTS ... 27
6.1 DESCRIPTIVE DATA ABOUT THE SAMPLE ... 28
6.2 THE RELIABILITY OF THE CONSTRUCTS ... 29
6.3 INVOLVEMENT WITH THE PRODUCTS ... 30
6.4 TEST OF NORMALITY ... 31
6.6 TESTING OF THE HYPOTHESES ... 32
6.5.1 Main Effect Sales Channel & Type of Product ... 32
6.5.2 Moderating Effects Type of Good ... 34
6.5.3 The mediating effect of Switching Costs ... 37
7. DISCUSSION ... 38
7.1 DISCUSSION OF THE DIRECT EFFECTS ... 39
7.2 DISCUSSION OF THE MODERATING EFFECT ... 42
7.3 DISCUSSION OF THE MEDIATING EFFECT ... 43
8. MANAGERIAL IMPLICATIONS ... 43
8. LIMITATIONS & DIRECTIONS FOR FUTURE RESEARCH ... 47
REFERENCES ... 49
APPENDIX ... 52
APPENDIX 1: RESULTS PRE-‐TEST ... 52
APPENDIX 2: OVERVIEW OF PERCEIVED SWITCHING COSTS ITEM SCALES ... 53
APPENDIX 3: QUESTIONNAIRE ... 54
APPENDIX 4: THE SCENARIO’S ... 58
APPENDIX 5: RELIABILITY ANALYSIS CONSTRUCTS ... 62
APPENDIX 6: INVOLVEMENT CHECK PRODUCT TYPES ... 64
APPENDIX 7: SKEWNESS AND KURTOSIS ... 65
APPENDIX 8: ANOVA RESULTS DIRECT EFFECTS ... 66
APPENDIX 9: ANOVA RESULTS MODERATING EFFECT ... 67
1. Introduction
In the past decennium the world has changed significantly. Internet, smartphones and tablets have become part of our daily routines. In a couple years of time we have become more and more connected to each other, and we have instant access to information of any kind. According to figures from the Dutch Central Agency for Statistics (CBS), 97% of the Dutch people have access to an Internet connection and 72% of the Dutch own a smartphone with an Internet connection. Furthermore approximately 50% of the Dutch households own a tablet.
With the emergence of Internet, smartphones and tablets, online shopping has become more and more of an important player. It can be said that Internet and online shops are a disruptive phenomenon for the traditional retailers, the so-called brick and mortar stores. Times, for instance, named online shops the ultimate replacement for the traditional shopping malls in an article named “Kiss your mall goodbye: Online shopping is cheaper, quicker and better.” Furthermore Mintel (2013) is expecting that the revenue from online sales in Europe will double till 323 billion euro in 2018. Figures of the CBS show, that in 2013 in the Netherlands, 82% of the people with an Internet connection made at least one purchase through an online store. So online retailing is “booming” already, and it is expected that this only will continue. Zalando, a large player in the European market, increased its revenue with 52% in 2013 (retaildetail.eu). A larger player on the Dutch market, Wehkamp.nl, is investing for over 100 million euro’s in what has to become the worlds largest fully automated distribution centre. The centre will have 480.000 so called pick locations, and allows Wehkamp.nl to deliver placed orders within a day.
degree of information transparency is favourable for consumers, the opposite is true from an internet retailers perspective. As Grewal et al. (2004) discuss in their paper there have come along quite some market consequences with internet retailing. One of these consequences is an increase in price competition. Due to high price transparency consumers can easily make use of so called site-bots like, beslist.nl or the tweakers.net pricewatch, to compare prices and to pick the retailer that is offering the cheapest price. Furthermore, retailers are tracking each other in order to keep ahead on each other when it comes to price. When online retailer x is lowering its price, online retailer y will notice this almost instantly. So online shopping has even more strengthened the price competition for both online and offline retailers. Online shopping has led to more intertype competition where margins are even under more pressure, both for offline and online shopping channels. However despite high transparency and price competition Clemmons et al. (2002) found that even on the online market there exists price dispersion. This price dispersion exists because consumers can for instance experience switching costs and therefore are willing to pay a price premium for a product at a certain online retailer. Switching costs can be for instance the time that consumers have to spend in searching across different online stores. Also creating a new shopping account at another online retailer, or “learning” how processes work at another online retailer are examples of switching costs.
So pricing is an important factor in a retailer’s strategy, if not one of the most important issues in marketing (Bijmolt et al., 2005). However as argued before despite the high transparency there are still dispersion in prices in both the offline and online channel. So it may be that consumers are less price sensitive than thought, regardless the transparency of price and that they are willing to pay a certain price premium because of perceived switching costs. If this is true this may have implications for both online and offline retailers when it comes to the “race-to-the-bottom” with regard to prices. It would imply that online retailing can be less price competitive and thus online retailers can have the opportunity to create higher margins in such a way that online retail channels actually can be profitable.
2. Problem Statement and research question
The above leads to the following research question:
To what extent does price sensitivity differ between the offline and online channel, taking into account the retail sector, and how can the several types of switching costs as perceived by consumers explain these differences.
To be able to answer the research question above, the following questions have to be answered:
How price sensitive are consumers in offline channels? How price sensitive are consumers in online channels? Does price sensitivity differ between different product types?
To what extend do consumers experience switching costs in online channels? To what extend do consumers experience switching costs in offline channels?
Do consumers experience switching costs differently across different types of products? In order to be able to answer these questions, first theoretical insight will be provided in the literature review. Based on the literature review a conceptual model and hypotheses will be provided. After the conceptual model the research methodology will be discussed, followed by the results. The results of the research will be discussed in the Next chapter. After the discussion, the conclusions and managerial implications will be provided. This paper will conclude with limitations and directions for future research.
3. Literature Review
3.1 OnlineThe pioneer in the field of internet shopping is Amazon.com. This online shop, with its roots in the United States, started business in 1995. From that moment on Amazon, and the industry as a whole, developed itself very quickly. Having a look at the figures shows that in the U.S. online spending in 2013 was $268 billions, which implies a market share of 11.6% in the total U.S. retail sector for online shopping. The market share of online shopping as a part of the total retail revenue is the largest in the United Kingdom. In the U.K. online shops are accounting for 13.5% of the total retail sales. Having a look at the average number in the European Union (E.U.) shows that online retail is generating 7.2% of the revenue in the total E.U. retail market. Having a closer look at the home market where this research is focusing on, the Netherlands, we see that online shopping accounts for 7.1% of total retail sales and that there are 45.000 different Dutch web shops. (Retailresearch.org; thuiswinkel.org).
Thuiswinkel.org is a Dutch organization, which is actively monitoring shopping behaviour of consumers online. In Table 1 there can be found an overview of how large the share of online revenue of the different categories are in the Netherlands. It can be seen that especially the travel sector is of large significance in the total revenue share of online sales. However if we have a look at the growth of the revenue in each sector over the year 2013 it can be concluded that with a revenue growth of 3% the sales of holidays online seems to have stabilized. Categories that show growth potential are: insurance (16%), telecom (16%), computer supplies (16%), clothing (16%), Food & Care (15%), music (19%) and toys (27%). The overall average growth of online revenue over the year 2013 was 8%. In contrast to that there is the growth number of the traditional retail. Over the year 2013 Dutch offline retailers did not realize a revenue growth, but had to take a loss of -2% in which revenue declined. (CBS.nl)
Category Online revenue share
Travel, Hotel & Flighttickets 36%
Hard- en software 8%
Telecom 14%
Clothing 8%
Books, CD’s, DVD’s, Games 6%
Electronics 7%
Food, health, home & gardening 5%
Others 15%
Table 1: Categories and revenue shares
market. This view of online retailing, as being an exclusive domain for new players, a market where traditional retailers had nothing to find for, has changed over the years. Nowadays being active with an online shop is seen as an essential part of a multichannel strategy (Grewal et Al., 2004). A lot of traditional companies, which started as brick and mortar retailers, expanded their presence in the marketplace with the introduction of an online store. Those retailers are combining the benefits of online shopping with the benefits of physical stores, which leads to synergy for those retailers. In a reaction to this response of traditional retailers, e-tailers are reacting in several ways. One strategy a lot of e-tailers do follow is optimizing the whole online shopping process. They are optimizing their sites and the ordering and shipping processes. There are also e-tailers who open a physical “front end” store. These are stores used for promotional issues and to let customers get in touch with the e-tailer. Furthermore there are also e-tailers who expand from online to offline stores (open physical stores by themselves or team up with an existing brick and mortar retailer). In general their can be defined two strategies which companies can pursue when it comes to online shopping. The first one is being an online-only retailer. Examples for online-only companies are, for instance, Bol.com and Wehkamp.nl, their only sales channel is the online channel. They do not have any physical stores where the consumers can come to. The second strategy is the so-called click-and-mortar strategy. This strategy means that a retailer is selling through both physical stores and the online channel.
missing the experience that comes with actually visiting a real store and they are not able to actual see and touch the product before they buy it.
So it can be said that the rise of online retailing has had a significant effect on the retail market, and can be classified as a disruptive phenomenon. One of the most important consequences of online retailing is that it made the retail market as a whole much more transparent when it comes to price. Furthermore the power of consumers rises substantially because they have access to much more information of any kind. As a consequence they are no longer condemned to the favour of the retailer. Consumers are getting more powerful and with that less loyal (Schoenbachler & Gordon, 2002).
3.2 Price Sensitivity & Price Elasticity
As stated before price is one of the most important attributes in marketing and in consumer’s decision-making processes (Gijsbrechts, 1993). A lot of decisions that consumers make in their daily life are strongly influenced by the prices they have to pay. Each consumer has different preferences and for that every consumer is differing in their willingness-to-pay for different products and different product categories. In practice this is implying that for instance consumer A, who is a so called “gadget freak”, is willing to pay more for a laptop than consumer B who is not technology minded at all. This price sensitivity is a fundamental aspect of the many elements of a retailers policy when it comes to for instance design of promotions, private label pricing decisions and the spot where certain products should be placed on a shelve. (Hoch et al., 1995). The price sensitivity of consumers toward different products and product categories is influenced by several factors. Such factors, which influence consumer’s price sensitivity, are amongst others the product category, consumer’s income, educations and household size (Hoch et Al., 1995). For instance, a low income household that consists of five people is likely to be more price sensitive than a single person household.
can also have a positive sign, although it is very unlikely. A positive sign would imply that if prices increase, sales also increase.
A lot of literature can be found about the price elasticity of demand. However a lot of the research is somewhat outdated and therefore only takes the price elasticity in traditional retail settings into account. In general, based on the available literature, it can be said that the price elasticity for demand differs amongst product categories (Tellis, 1988). Gordon et al. (2013) did a study for an American supermarket chain. They indeed found that there is a difference in price elasticity between different products and product categories. For instance, they found an elasticity of -0.37 for toilet paper, which suggest that the demand for toilet paper is price inelastic. Other products like carbonated soft drink or mayonnaise show an elasticity of respectively -2.81 and -3.84 which implies that these products appear to be price elastic. Furthermore Gordon et. Al. (2013) found that the price elasticity of demand is not a static measure, but that it is influenced and changing overtime. They found that the price sensitivity, and with that the price elasticity of demand, varies countercyclical with economic growth or decline. The Most interesting finding is that the share-of-wallet of a product category seems to be the most important driver for price sensitivity and price elasticity. So the larger the share-of-wallet for a certain product category, the more price sensitive consumers are, and the more price elastic the category is.
strategy, so online and offline stores like HEMA, de Bijenkorf and H&M, or if the retailer is a pure online retailer like BOL.COM and Wehkamp.nl.
So as the dominant opinion may be that online shopping has made the retail market more competitive and prices more elastic, there are also signals that are objecting to this opinion. It can thus be said that it is not per definition the case that online shopping has made consumers more sensitive to prices. Therefore the question arises if consumers are really more price sensitive on the internet than in traditional retail stores.
3.3 Switching Costs
Due to higher price transparency and decreasing customer loyalty in a highly competitive market, customer retention and customer relationship management (CRM) have become rapidly of more importance (Burnham et Al., 2003). The most important measure, most firm use in case of CRM, is customer satisfaction. The majority of firms believe that when they excel in customer satisfaction, by providing the best services and quality products, they will gain the best customer retention rates as possible. Reichfeld (1996) defines this way of reasoning as the so-called “satisfaction trap”. Companies are so focused, and with that investing a lot of money on achieving customer satisfaction, that they outrun other drivers. Customer satisfaction is not the only way that leads to customer retention. There are also other elements, like switching costs, that play a role in the likelihood of a customer to stay loyal to a certain company.
In literature there can be found many definitions of switching costs. Some definitions are very broad where others are very narrow. This research uses a slightly adjusted definition from the one that Burnham et Al (2003) provide in their paper. Switching costs are defined as follows: Switching costs are the onetime costs customers associate with the process of switching from one retailer to another retailer. So if a customer decides to switch to another retailer, those customers will perceive switching costs. The word switching cost is covering a broad spectrum of different kinds of costs that customer can perceive when they are switching.
Switching Cost Description
Economic Risk Cost These are costs that come with the uncertainty of switching to a new provider. These costs occur because of insufficient information that a consumer has about the new retailer.
Evaluation Cost Costs in terms of time and effort a consumer has to invest in researching a retailer they see as a potential alternative.
Learning Cost Cost in terms of time and effort a consumer has to invest in getting familiar with the process, skills and know-how that is needed at the new potential retailer.
Setup Cost Cost in terms of time and effort a consumer has to invest in setting up the new relationship or setting up/installing the new product. Benefit loss Cost Cost that are linked to created economic benefits at the current
retailer a consumer is shopping. Examples of these costs are losing accumulated points in a loyalty program or obtained benefits like discounts, which the consumer will not get at a new retailer and so loses them at his or her old retailer.
Monetary Loss Cost Those are the financial monetary costs that a consumer has to make for switching to the new retailer, except the cost of the product the consumer buys. Examples of these costs are for instance deposits or registration fees.
Personal Relationship Loss Costs
Cost that consumers experience in the form of breaking a bond of identification with people who the consumers have interacted with for a long time. For instance a consumer can be very familiar with the employees at a retailer. Switching to another retailer will translate into losing a familiarity that will not be directly available at the new retailer.
Brand Relationship Loss Cost
Affective cost that consumers experience when they break with a brand they have identified themselves with for a long time. This is experienced as a cost because consumers form associations with a brand and it becomes part of their identity. So switching to a new retailer is in sort of way experienced as breaking with his or her own identity.
All the defined switching costs summed up in Table 2 can be divided into three main types of switching costs:
• Procedural switching cost
These are switching cost that broadly can be defined as time and effort a consumer has to invest when switching to a new retailer. Economic risk, evaluation, setup and learning cost cover the procedural switching cost.
• Financial switching cost
As the name implies these switching cost can be measured in a monetary value. Benefit and Monetary loss cost cover these financial switching costs.
• Relational switching cost
These switching cost can be defined as psychological and emotional discomfort that a consumer experiences when switching to another retailer. Personal relationship and brand relationship loss costs cover the relational switching costs.
Burnham et al (2013) found several factors that are of influence on the perceived switching costs. Two of these factors are with regard to the market characteristics of markets where switching costs do exist, these are provider heterogeneity and product complexity. With regard to provider heterogeneity it was found that if consumers experience high provider heterogeneity in a market (so large differences between providers), these consumers perceive high switching costs. The same accounts for product complexity; if a consumers has to make a buying decision with regard to complex products (products with a lot of options), then consumers also experience high perceived switching costs.
As mentioned before the retail market has become increasingly more competitive. In his research Klemperer (1987) found that switching costs make markets less competitive. When switching costs exist in a market, a market becomes less transparent. If consumers experience switching cost it will be harder for them to judge if prices are really lower at an other retailer and if a potential switch will be really favourable for them. It is even so that Weiss and Heide (1993) found in their research that perceived switching cost have positive influence on consumer’s retention behaviour.
4. Conceptual Model & Hypotheses
The theoretical discussion in the previous chapter provides the bases for the conceptual framework and hypotheses of this research. Figure 1 provides the conceptual model that depicts the suggested relationships. Subsequent the assumed hypotheses are described.
Figure 1: Conceptual Model
more sensitive toward pricing in the online channel. Therefore the first hypotheses, which is related to the main effect of channel choice on price sensitivity is as follow:
H1: Shoppers show higher price sensitivity when they shop online instead of offline.
The second hypothesis is stated with regard to different types of products and the perceived price sensitivity of consumers. As described several researches are conducted regarding this subject. Tellis (1988) and Gordon et al (2013) both found that consumers differ in price sensitivity across different type of products/product groups. In addition to this, Gordon et al (2013) found that the share-of-wallet of a product is an important factor in the price sensitivity of consumers towards different products. How higher the share-of-wallet of a product, how higher the price sensitivity of consumers is. In the conceptual model of this paper the distinction is made between non-durable and durable products. It is expected that consumers will show higher price sensitivity for durable goods, as the share of wallet of those products is in general larger, than non-durable goods. Therefore the second hypotheses which is related to the main effect of type of product on price sensitivity is as follows:
H2: Shoppers show higher price sensitivity when they shop for durable products than for non-durable products.
The third and fourth hypotheses are set with respect to the assumed relationship between channel choice and perceived switching costs, and the moderating role of type of product on the relationship between channel choice and perceived switching cost.
H3: Shoppers perceive lower switching cost when they shop online instead of offline.
H3a: Shoppers perceive lower procedural switching cost when they shop online instead of offline. H3b: Shoppers perceive lower financial switching cost when they shop online instead of offline. H3c: Shoppers perceive lower relational switching cost when they shop online instead of offline
Subsequently it is expected that type of product will have a moderating effect on the relation between sales channel and perceived switching costs. As described before it is expected that consumers will perceive lower switching costs in the online channel. In addition to this it is expected that this effect is weaker for non-durable products then durable products. Reasoning behind this is that Burnham et al (2013) also found that product complexity is of influence on the perceived switching costs. Since high information transparency exists in the online channel (Granados et al, 2012) it is expected that because of this people perceive durable products less complex in the online channel than in the offline channel. This because they can easily find information about those products, and if they can read more information about the product it is likely to assume that consumers experience the product as less complex. Since non-durable products are assumed to be less complex to understand it is expected that the effect of the higher information transparency will be less strong for these products. So since it is expected that because of high information transparency consumers will especially perceive durable products as less complex in the online channel, and with that are expected to perceive lower switching costs, the following hypothesis is stated:
H4: The effect of the sales channel on perceived switching cost is weaker for non- durable products then for durable products.
H4a: The effect of the sales channel on perceived procedural switching cost is weaker for non- durable products than for durable products.
H4b: The effect of the sales channel on perceived financial switching cost is weaker for non- durable products than for durable products.
H4c: The effect of the sales channel on perceived relational switching cost is weaker for non- durable products than for durable products.
switching costs in a market. So if consumers perceive higher switching costs it is assumed that they have less access to transparent information and show higher retention behaviour. From this it is derived that consumers will be less price sensitive when high switching costs exist in a market. It is assumed that perceived switching cost have a mediating role. The channel choice has an effect on the price sensitivity of consumers and this effect is mediated through the perceived switching costs that appear in both channels. To be able to test for a mediating effect, the direct effect of switching costs on price sensitivity has to be tested. Therefore the following hypothesis is defined:
H5: Perceived switching cost will have a significant negative effect on consumers’ price sensitivity.
H5a: Procedural switching cost will have a significant negative effect on consumers’ price sensitivity. H5b: Financial switching cost will have a significant negative effect on consumers’ price sensitivity. H5c: Relational switching cost will have a significant negative effect on consumers’ price sensitivity.
The expectation is that there will be found a positive relationship between online channel choice and price sensitivity. So if people shop online they are expected to be more price-sensitive. The same accounts for the type of product. It is expected that there will be a positive relation between durable products and price sensitivity. So if people shop for durable products they are expected to be more price sensitive than when they shop for non-durable products. Furthermore it is expected that perceived switching cost play a mediating role between channel choice and price sensitivity. It is expected that higher perceived switching cost are assumed to lower the price sensitivity of consumers. Furthermore it is expected that the type of product will have a moderating effect on the relationship between the channel choice and the perceived switching cost. It is expected that when consumers shop for non-durable goods the relationship between channel-choice and perceived switching cost will be weaker than in case of durable goods.
5. Methodology
In the previous chapter the hypotheses and with that the corresponding conceptual model of this research are discussed. To test the conceptual model an empirical research will be
5.1 The research design 5.1.1 General Elements
In order to be able to answer the research question that is stated in the introduction of this paper a quantitative research will be performed. This quantitative research will be conducted by means of an experiment. A 2x2 factorial design, which corresponds with subsequently channel choice (2 levels) and type of retailer (2 levels) will be used. In order to be able to make the experiment operational a questionnaire will be developed and used. The questionnaire will be spread among Dutch residents in the age group 16-75. This because people in this age category are assumed to have enough knowledge to make independent choices and because it is assumed that they are likely to have enough knowledge to be able to shop online. It has to be mentioned that there is awareness about the fact that people who are 16 and 17 years old are not legally allowed to make purchase decisions without permission of their parents/legal guardian. However in current social life it is not uncommon that people in this age group make buying decisions with their own knowledge, so because of this they are also targeted with the questionnaire. The questionnaire will be conducted online and will be spread through snowball sampling. The questionnaire will be posted on social media websites such as facebook and twitter, and will be spread through e-mail.
To actually test the hypotheses the respondent will first be presented with a text scenario. In this scenario the respondent is asked to identify himself/herself with a certain situation. For example the respondent can be asked to imagine a situation where the respondent is going to a physical store and he/she has to buy a book. After being introduced to the situation the respondent is asked to fill in the questionnaire. The questionnaire will contain questions, which are used to measure the different constructs. Furthermore questions will be included which will be used as a control variable, for instance involvement with the product. At the end of the questionnaire some last questions with regard to demographic variables like age, gender and income will be asked.
5.1.2 Pre-set variables
Sales Channel: With regard to the channel choice the definition of this variable can be straightforward. Offline sales channel implies a retail setting in a physical store. Online sales channel implies a web-based retail setting, i.e. online web shops.
Type of product: With regard to type of product, durable vs. non-durable, this research sticks with the classic definition economic definition of durable goods. This states that durable goods are goods that are assumed to have a lifetime of longer than 3 years (Sullivan & Sheffrin, 2003). So non-durable goods are assumed to have a lifetime no longer than 3 years. In order to be able to test if there is a possible moderating effect of type of product the experiment should at least contain a product that is perceived as non-durable and a product that is perceived as durable by consumers. To make sure that the result will be generalizable for both types of products, five products for each category are selected. For non-durable products these are detergent, deodorant, an ink cartridge, a book and a pair of shoes. For durable products a refrigerator, a vacuum cleaner, a laptop, a watch and an office chair are selected. A pre-test was conducted to check if, based on involvement, this division of the products over both groups was right. The results of the pre-test, which can be found in Appendix 1 show that the products are assigned to the right group and people indeed make the distinction between durable and non-durable products.
5.2 Measuring the concepts 5.2.1 Measuring Price Sensitivity
To be able to measure the sensitivity of respondents towards prices, a single-item scale is used. The scale is used to apply an experimental treatment on consumers in such away that their perceived price sensitivity should be influenced. The single-item-scale consists of a 7-point likert scale, which runs from (1) absolutely not to (7) absolutely sure. For the questionnaire the scale will be translated into Dutch. The scale can be found below:
You have just read about the situation you are in when you want to buy a [insert product]. However, just suppose that at that moment you receive a message, for instance on your iPhone, that you can buy the same [insert product] 10% cheaper at (scenario 1) another store in the neighbourhood / (scenario 2) another web-shop. Would you switch from store/web-shop?
5.2.2 The perceived switching costs
Burnham et al. (2003) did an extensive research on the different types of switching cost consumers can perceive. They developed 8 scales that measure if respondents perceived a specific switching cost. All the items in the scales used are measured on a five point likert scale, where 1 represented strongly disagree and 5 represented strongly agree. In this research the different scales that were used for the 8 types of switching cost will be pooled into three main scales. The three scales represent the three main types of switching cost as described in the literature review. Furthermore the scales will be translated into the Dutch language. Subsequently the scales will be slightly adjusted from a 5 point based likert scale to a 7 point based likert scale. This is done to give the respondents a better feeling to indicate their feelings/perceptions. As this is quit an extensive scale, for reading purposes, it can be found in Appendix 2.
5.2.3 Control Variables
Several control variables will be used in this research. This is done to be able to measure a purer effect of the sales channel and type of product on the perceived price sensitivity. The same accounts for the effects of sales channel of perceived switching costs, and the effect of perceived switching costs on perceived price sensitivity. Several demographic variables will be used along with a variable that controls for the general price sensitivity of a consumer. The questions for the demographic variables, and the item-scale for general price sensitivity can be found below.
Questions for demographic variables: 1. What is your gender?
o Male o Female
3. What is your highest level of education? o Basisonderwijs o VMBO o HAVO o VWO o MBO o HBO o WO
Furthermore an five-item-scale is used to measure the general price sensitivity of consumers. This is used since it might be that consumers who are less sensitive towards price in general are also less sensitive for the influences of the price discount. The same applies in reverse; consumers who are more sensitive towards price in general, might be more sensitive for the influences of a price discount. To be able to measure the general price sensitivity of the respondents towards prices, several items that can be used to measure this concept are derived from the Handbook of marketing scales by William O. Bearden. For this research a 5 item scale is composed. The scales each consist of a 7-point likert scale, which runs from (1) strongly disagree to (7) strongly agree. For the questionnaire the scale will be translated into Dutch. The scale can be found below:
Price Sensitivity Scale
1. I like to be aware of all possible options before buying an expensive item. 2. I keep up with sales being offered by (department) stores in my area. 3. I feel that I can save a lot of money by using coupons.
4. I often shop at more than one store in order to find the best price for products 5. When (grocery) shopping, I always look at the price per ounce or price per unit
information.
5.3 Questionnaire and Scenarios
respondent needs to buy a detergent. The respondent is asked to imagine as he or she is before the shelves with detergents and the respondent picks a bottle of detergent he or she wants to buy. Furthermore it is mentioned that, of course, the respondent is very much aware of the fact that the detergent is also available at other shops in the neighbourhood. After being introduced into this short scenario the respondent will be presented with the questions regarding the constructs in the conceptual model. The first question points back to the scenario the respondent is introduced to. It states that the respondent has to imagine that it receives a message, which states that the respondent can buy the product 10% cheaper at another store/webshop. This first question is intended to apply an experimental treatment to the respondents. The second part of the questionnaire consists of some final questions with regard to the respondent’s demographics like gender, age and income. The questionnaire and an overview of all the possible scenarios a respondent can be presented with can be found in Appendix 3 and Appendix 4.
As this is an 2x2 factorial experimental-design with an total of 10 different product there will be a total of 20 different possible scenarios. As 20 scenarios would take a lot of time to answer, the likelihood of a respondent that will finish a questionnaire will be negligible in case of a full factorial design. Therefore it is chosen that a respondent only has to fill in one possible scenario. The scenarios are based on the sales channel and type of product and will be randomly and evenly distributed among consumers. Furthermore a minimum of ten respondents for each scenario will be considered as needed in order to be able to make the product groups generalizable. With twenty different scenarios this implies that a minimum of two hundred respondents is needed.
5.4 Plan of analysis
5.4.1 General tests
constructs. To be able to create these constructs a reliability analysis will be performed on the different item scales. This is done to check if the constructs are internally consistent enough. To be able to create the constructs a Cronbach Alpha measure of at least 0.6 is needed. If this is the case the constructs are internally consistent at an acceptable level. Besides this a test of normality will be conducted. This is done to check if the data follows a normal distribution or not. For this a skewness and kurtosis test will be performed. If the values of both of the test lie between -1 and 1 it can be assumed that the data follows a normal distribution.
5.4.2 Testing the direct effects
To test the two direct effects of sales channel and type of product on perceived price sensitivity a factorial ANOVA, also called a Two-Way ANOVA will be used. This test is used as both independent variables, sales channel and type of product, are nominal variables (consisting both out of two groups) and the dependent variable, perceived price sensitivity, is a scale variable. An ANOVA will show if there exists a significant difference between the groups. The control variables used in this research are added into the ANOVA as covariates. 5.4.3 Testing Moderation
Since the same conditions apply for the moderation hypotheses; the effect of type of product on the relation between channel choice and perceived switching cost, there will also be made use of ANOVA’s. The three sub-hypotheses, which are set with regard to the moderation effect, will be tested with separate ANOVA tests. The control variables will be added again as covariates.
5.4.4 Testing Mediation
To test if there is mediation effect there can be made use of several options. In this research the model that was developed by Baron & Kenny (1986) is used. The model consists of four steps that have to be taken to test if there is mediation. Those steps are described below.
1. Test if the independent variables sales channel and type of product have a direct effect on the dependent variable price sensitivity. (the c-path)
2. Test if the independent variables sales channel and type of product have a direct effect on the proposed mediating variables of perceived switching costs (the a-path)
4. Test if the independent variables and the proposed mediating variables together have an effect on the dependent variable price sensitivity. (the b/c path).
In Figure 2 below the described paths are visualized.
The first and second step of the model will be tested as described in previous paragraphs, about the testing of the direct and moderating effects. If these conditions hold there is continued with the third step, testing the direct effect of perceived switching costs on perceived price sensitivity. Since these are all scale variables these effects will be tested with regression analysis.
If the first three steps of the Baron & Kenny model will hold, the fourth step will be conducted to test if there exists mediation. Conducting a regression analysis, which tests the effect of the independent and mediating variables on the dependent variable, will do this. Baron & Kenny make the distinction between 3 levels of mediation, namely no mediation, partial mediation and full mediation. There is no mediation when the significance level of the model does not change after the b-path is added into the model. There is partial mediation if the significance level of the model is reduced but stays significant after adding the b-path into the model. Full mediation exists if the model becomes insignificant after the b-path is added into the model.
6. The Results
In this section the results of the quantitative research will be summarized. First there will be a descriptive analysis in which the demographics of the sample will be discussed. Furthermore a reliability analysis is described, in which the internal consistency of the defined constructs will be tested. Subsequently the hypotheses and the conceptual model, which are discussed in
Sales Channel &Type of product Price Sensitivity Switching Costs C’ A B
6.1 Descriptive Data about the sample
A total of 331 respondents have started the questionnaire. However, a total number of 83 respondents did not complete the questionnaire, which is a dropout-rate of 25,1 per cent. Another 39 entries were found with extremely aberrant answers. These entries were also not included in the final data set.
Therefore the final dataset, which is used for the analysis, consists of 209 respondents. From these 209 respondents, 39.2% are male (82) and 60.8% are female (127). By far most of the respondents belong to the age-group 20-40 years old (70.8%). In comparison 8.1% of the respondents belong to the age-group younger than 20, and 21.1% of the respondents belong to the age-group 41-65. Especially respondents with an age of 20-25 are overrepresented. Furthermore 9.1% respondents have a secondary school education level, 26.3% of the respondents has a “MBO” education level, 33% has a “HBO” education level and 31.6% of the respondents has an academic education level. When it comes to net income of the respondents 53.6% earns less than €10.000 per year, 12.4% earns €10.000-€20.000, 22.5% earns between €20.000 and €30.000, 8.6% between €30.000 and €40.000 and 2.9% of the respondents earn more than €40.000 per year.
What is most notably, when it comes to the demographics of the respondents, is that there are a lot of respondent’s aged 20-25, with a low income and a high education level. This can be explained by the fact that a respectively large group of students have filled in the questionnaire. To get an insight in how representative this sample is for the Dutch population a comparison of the sample data and data from the Dutch agency for statistics (CBS) can be found in Table 3.
Demographic Variable Sample Statistics
WO 31.6 n/a 17.3 Yearly Net. Income*
<€10.000 53.6 45,4 34.7 €10.000-€20.000 12.4 9,2 12.6 €20.000-€30.000 22.5 9,1 31.8 €30.000-€40.000 8.6 8,7 15.5 €40.000-€50.000 1.9 8,3 4.0 >€50.000 1.0 19,2 1.5
* Are CBS statistics over 2012, as data over the year 2013 is not available yet.
Table 3: Descriptive statistics
As can be seen in Table 3 the respondents in the sample are not representative for the Dutch population. To correct for this skewness the data that is collected will be weighted. The data will we weighted by means of age. The groups 16-19 and 20-40 are overrepresented in the sample and they will be assigned a lower weighting in the sample. The assigned weighting for these two groups are respectively 0.90 (7.3/8.1) and 0.53 (37.9/70.8). The group 41-65 is underrepresented so they will be assigned a higher weighting in the sample. The weighting for this group is 2.59 (54.8/21.1). Table 3 shows that after a weight is placed on the sample it becomes slightly more representative for the Dutch population in terms of gender and age. 6.2 The reliability of the constructs
In order to be able to measure the constructs, which are displayed in the conceptual model of this paper, different item scales were constructed. To be able to combine these items into the defined constructs, a reliability analysis has to be performed for the different item scales. In order to determine if different items can be combined into one construct, each construct needs a Cronbach Alpha (CA) measure of at least 0.6. When an item scale has a CA of below 0.6 the scale is not internally consistent and the items cannot be combined into a construct. Before conducting the reliability analysis, one of the items of the scale that is measuring perceived procedural switching cost is being recoded, as this item is measured on a reversed likert-scale. In Table 4 the different constructs with their corresponding CA measure can be found.
Construct Cronbach Alpha Items
Procedural Switching Cost 0.739 8
Financial Switching Cost 0.611 2
Relational Switching Cost 0.922 4
General Price Sensitivity (control variable) 0.725 5 Table 4: Cronbach Alpha measures of the constructs
Initially the construct of financial switching cost consisted out of 3 different items. However the reliability analysis for this construct showed that with those 3 items the construct was not internally consistent (CA: 0.575). If one item is removed from the case the CA of the construct is 0.611, which can be considered as internally consistent. The four other constructs are all found to be internally consistent. In Appendix 5 the output of the reliability analyses can be found.
6.3 Involvement with the products
In this paper, amongst others, type of products is a construct. With regard to type of product the distinction between non-durable and durable products has been made. For the questionnaire 5 non-durable products and 5 durable products were selected. To be able to determine which products consumers marked as durable and non-durable, a pre-test was conducted (Appendix 1). To check if the respondents also perceived this distinction during the filling of the questionnaire a one-way ANOVA test is performed. To test if this distinction exists a control variable has been created. This control variable measures the involvement per product assuming that low-involvement products can be perceived as non-durable and high-involvement products as durable. In Table 5 the high-involvement per product can be found, the output of this ANOVA test can be found in Appendix 6.
DF 9
F 3.976
Significance .000 Table 5: Involvement per product
The analysis shows that all products that are labeled as non-durable have a mean below 4, which indicates low involvement. In contrast all products labelled as durable have a mean higher than four, which indicates high involvement. The ANOVA results show that with an F-value of 3.989 and a P-F-value of .000 the difference between the products is statistically significant. Furthermore also a statistical significant difference is found when the products are pooled into the two different groups. With an F-Value of 31.025 and a P-Value of .000 there is a statistical significant difference between the groups durable and non-durable. So it can be concluded that the respondents indeed perceived the ten different products in the way they were supposed to. Detergent, deodorant, ink cartridge, books and shoes were perceived as non-durable. A refrigerator, vacuum cleaner, laptop, watch and office chair were perceived as durable products.
6.4 Test of Normality
To gain a better insight in the data the skewness and kurtosis test statistics are conducted. The skewness test indicates if the data follow a normal distribution or not, and if the data set is asymmetric. The kurtosis test indicates how peaked or flat the data is in comparison to a normal distribution of data. The key test results can be found in Table 6, the output of the test can be found in Appendix 7.
Table 6: Test of normality
As can be seen al the values of the skewness and kurtosis test lie between -1 and 1, so the data is distributed normally. Price sensitivity, procedural switching costs and financial switching costs are somewhat skewed to the left. This indicates that the extreme values are located at the
left of the mean and most other values at the right of the mean. Relational switching costs are skewed to the right. This indicates that for this variable the extreme values lie right from the mean, were most other values are located at the left from the mean. The kurtosis statistics shows that price sensitivity, procedural switching costs and financial switching costs are platykurtic, meaning that the distributions for these variables are wider peaked and flatter, than in case of a normal distribution. The kurtosis value for relational switching costs indicates that the distribution is leptopkurtic, meaning that this variable has a more peaked and steeper distribution than a normal distribution. Since all the test values lie between the range -1 and 1, a normal distribution of the data can be assumed.
6.6 Testing of the hypotheses
In this section the testing results of the hypotheses are described. First the hypotheses regarding the direct effects of channel-choice and type of product on the perceived price sensitivity are tested. After the direct effects the moderating effect of type of product on the relation between shopping channel and perceived switching cost is tested. The section concludes with the testing of the assumed mediating effect of perceived switching costs. 6.5.1 Main Effect Sales Channel & Type of Product
In Chapter 4 of this research the hypotheses for the assumed direct effects regarding sales channel and product type on perceived price sensitivity are described. The following two hypotheses regarding the two assumed main effects were stated:
H1: Shoppers show higher price sensitivity when they shop online instead of offline.
H2: Shoppers show higher price sensitivity when they shop for durable products than for non-durable products.
To test the hypotheses a Two-Way-ANOVA test was performed of which the main result can be found in table 7 and the output in appendix 8.
Df F Significance Estimated Beta
Independent Variables
Sales Channel 1 .521 .472 .870
Type of Good 1 12.066 .001* -.262
Control Variables
Income 1 1.606 .208 .150
General Price Sensitivity 1 13.207 .000* .354
Levene’s Test 3 1.945 0,127 -
* Significant at 1% level
Table 7: Results of the two-way ANOVA of the direct effects
As can be seen in the table above there is only a statistical significant effect of type of good on the perceived price sensitivity of consumers at a 1% level. There is no statistical significant effect of sales channel on perceived price sensitivity. Furthermore the Levene’s test shows that it can be assumed that the variance of the dependent variable is equal across the groups. Figure 3 is visualizing the direct effect of product type on price sensitivity. It can be seen that consumers perceive a higher price sensitivity for durable products than they do for non-durable products. In figure 4 the effect of channel choice on the price sensitivity is visualized. Although the price sensitivity is slightly higher in the online channel (as was expected), there is no significant difference between both channels. Figure 5 is visualizing the differences in price sensitivity for non-durable and non-durable products in both sales channels. It can be seen that, as also displayed in figure 3, that overall consumers are less price sensitive for non-durable products then for non-durable products. However these differences are smaller between both product types in the online channel then it is the case in the offline channel. Also non-durable products show a higher perceived price sensitivity in the online channel in Figure 3: Visualizing the Direct Effect of Type Of
Product. 3 4 5 6
Non Durable Durable
3 4 5 6
Ofaline Online Figure 4: Visualizing the Direct Effect of channel choice. 3 3,5 4 4,5 5 5,5 6 Ofaline Online Non-‐Durable Durable
comparison to the offline channel. For durables it is the other way arround, they show a slightly lower price sensitivity in the online channel then in the offline channel.
Furthermore, several control variables were included in the analysis. Table 7 shows that there was found a statistical significant effect for gender and the general price sensitivity of consumers. Having a look at the corresponding estimated beta’s it can be seen that in general women are less price sensitive than man. Furthermore, if consumers are more price sensitive in general they are found to perceive higher price sensitivity in this experiment.
6.5.2 Moderating Effects Type of Good
To test if there is a direct effect of sales channel on perceived switching cost and a moderating effect of type of good on the relation between sales channel and the three different perceived switching costs separate ANOVA analysis will be performed. Also in this analyses gender, age, income and general price sensitivity will be added as control variables. Regarding to these assumed relations the following hypotheses were stated in Chapter 4 of this research: H3: Shoppers perceive lower switching cost when they shop online instead of offline.
H3a: Shoppers perceive lower procedural switching cost when they shop online instead of offline. H3b: Shoppers perceive lower financial switching cost when they shop online instead of offline. H3c: Shoppers perceive lower relational switching cost when they shop online instead of offline.
H4: The effect of the sales channel on perceived switching cost is weaker for non- durable products than for durable products.
H4a: The effect of the sales channel on perceived procedural switching cost is weaker for non- durable products than for durable products.
H4b: The effect of the sales channel on perceived financial switching cost is weaker for non- durable products than for durable products.
H4c: The effect of the sales channel on perceived relational switching cost is weaker for non- durable products than for durable products.
Df F Significance Est. Beta
Independent Variables Dependent Variable
Table 8: Results of the MANOVA test
The core results of the performed ANOVA analyses can be found in Table 8, the output of the test can be found in Appendix 9.
The table above shows that sales channel has a statistical significant effect on the financial and relational switching cost. Furthermore the type of product has a significant moderating effect on the relationship between sales channel and perceived financial switching costs. Having a closer look at the control variables which were added in the model, it can be seen that income has a significant effect on the perceived relational switching cost. Besides this the price sensitivity of consumers has a significant effect on the perceived financial switching costs. As all the other p-values are >.05 there is no statically significant effect on the 5% level for those relationships. So there is no proof of a direct effect of channel choice on perceived procedural switching cost. Neither is there a moderating effect of product type on the relationship between channel choice and procedural and relational switching costs. Figures 4 and 5 are visualizing the statistical significant relationships.
Channel*P_Type PSC 1 .168 .683 FSC 1 6.622 .012* 1.319 RSC 1 1.261 .264 Control Variables Gender PSC 1 .104 .748 FSC 1 .256 .614 RSC 1 1.910 .170 Age PSC 1 .008 .927 FSC 1 .926 .338 RSC 1 1.155 .285 Income PSC 1 .237 .628 FSC 1 .090 .765 RSC 1 8.362 .005* -.410
Gen. Price Sensitivity PSC 1 1.872 .174
FSC 1 8.151 .005* -.288
RSC 1 .079 .779
Figure 6: Visualizing channel choice and financial switching costs, moderated by type of good.
Figure 7: Visualizing channel choice and relational switching cost.
As can be seen in Figure 6 above, in the case of non-durable goods people perceive almost no differences in financial switching cost between offline and online shopping. However with regard to durable products there exists a large difference, people then perceive much higher financial switching cost in case of online shopping. Figure 7 shows that overall people perceive higher relational switching cost when they shop for durable goods then when they shop for non-durable goods. Furthermore by looking at the average it can be concluded that people perceive overall higher relational switching cost in case of offline shopping then in case of online shopping. Summarizing, hypothesis H3a is rejected as there is no significant relationship, hypotheses H3b is rejected as the relationship is not as expected. For non-durable products consumers do not perceive financial switching cost differently in the different channels. In case of durable products consumers perceive much higher financial switching costs online instead of lower. Hypothesis H3c is accepted, since people perceive
rejected, as they are not significant. Hypothesis H4b is accepted, as it is significant and the relation is as expected. Figure 4 shows that with regard to financial switching cost there is the case of disordinal interaction. Figure 4 shows that type of product almost has no effect on the relationship of sales channel on perceived financial switching cost in case of non-durable products. However the opposite is true for durable products. It can be seen that type of product has a strong effect on the relationship between channel choice and perceived financial switching cost. Where in the offline channel financial switching costs are perceived much lower than it is the case in the online channel.
6.5.3 The mediating effect of Switching Costs
In this last paragraph of this chapter the mediation model will be tested. This will be done by using the steps of Baron & Kenny (1986) that are described in the plan of analysis. Step 1 and Step 2 have already been tested in de previous paragraphs. As type of product has no significant effect on the perceived switching costs no mediation can exist for this variable. Furthermore sales channel has only a significant effect on the financial switching costs and the relational switching cost so it can be said that there is no mediation effect possible for sales channel on the price sensitivity through the procedural switching cost. To continue the testing of a possible mediation effect step 3 of the model of Baron & Kenny is performed, which is corresponding to hypotheses 5, the direct effect of switching cost on price sensitivity. This relation is tested by means of a regression analysis. The hypothesis is as follows:
H5: Perceived switching cost will have a significant negative effect on consumers’ price sensitivity.
H5a: Procedural switching cost will have a significant negative effect on consumers’ price sensitivity. H5b: Financial switching cost will have a significant negative effect on consumers’ price sensitivity. H5c: Relational switching cost will have a significant negative effect on consumers’ price sensitivity.
The results of this regression analysis can be found in Table 9. The output of the regression can be found in Appendix 10.
Df Beta Significance
Independent Variables
Procedural Switching Cost -.206 .022*
Financial Switching Cost -.002 .980
