Master Thesis
Msc Business Administration – Strategy & Innovation
“The effect of mass-customization strategies on the
Willingness To Pay for lingerie”
Table of Contents
1 Introduction ... 3
1.1 Background of the study ... 3
1.1.1. Theoretical background ... 3
1.1.2. Practical background ... 5
1.2. Purpose of the study ... 7
1.3. Problem identification ... 7
1.4. Problem statement ... 8
1.5. Relevance of the study ... 9
1.5.1. Theoretical relevance ... 9
1.5.2. Practical relevance ... 9
1.6. Problem domain and scope ... 10
1.7. Structure of the study ... 11
2. Theoretical framework ... 12
2.1. Mass-customization Strategies ... 12
2.1.1. Aesthetic Mass-customization ... 13
2.1.2. Functional Mass-customization ... 14
2.1.3. Combination of different customization types ... 15
2.1.4. Individualized design ... 15
2.2. Motivations to use online Mass-customization ... 16
2.3. Toolkits ... 17
2.4 Strategies to use for the lingerie industry ... 18
3. Conceptual model ... 19 3.1. Direct effects ... 19 3.2. Moderation effects ... 23 3.2.1. Age ... 23 3.2.2. Cupsize ... 24 3.3. Conceptual model ... 25 4. Research Methodology ... 26 4.1. Research strategy ... 26 4.2. Sample selection ... 27 4.3. Sample description ... 27 4.4. Data collection ... 28 4.5. Measurement variables ... 30 4.6. Data analysis ... 31 4.6.1. Regression models ... 31
4.6.2. Linear Mixed Model ... 32
4.7 Model validity ... 33
5. Results and Analysis ... 33
5.1 Assumptions for regression ... 33
5.2. Correlation-matrices ... 34
5.2.1. Pearson correlation for Bra ... 34
5.2.2. Pearson correlation for Set ... 35
5.3. Fixed Effects using least squares dummy variable model (LSDV) ... 35
5.4. Hausman test ... 36
5.5. Linear Mixed model ... 37
5.6. Summary of empirical results ... 40
6. Conclusion and Discussion ... 41
6.1. Conclusion ... 41
6.2 Discussion ... 42
6.3 Limitations and Further Research ... 45
References ... 47
Appendices ... 50
Appendix 1 Interface Double Dutch Design customization tool ... 50
Appendix 2 Cupsize variable made categorical ... 51
Appendix 3 Scatterplots for Bra and Set ... 52
Appendix 4 Correlation matrix for Set ... 53
Appendix 5A Simple Fixed Effects Linear Model ... 54
Appendix 5B Comparing the Fixed effects regression methods ... 55
Appendix 5C Panel regression with random effects ... 56
Appendix 5D Linear Mixed Model with random effects ... 57
Appendix 5E Linear Mixed Model with random effects and Age interaction ... 58
Appendix 5F Linear Mixed Model with random effects and Cupsize interaction ... 59
1
Introduction
1.1 Background of the study
1.1.1. Theoretical background
As fabrication processes became more efficient in the early 20th century, the tailored approach to production diminished. The production revolution, accelerated by Henry Ford, improved the production volume of products, but also took the customers voice away. As Henry Ford said in 1922 (Ford and Crowther, 1922): ‘Any customer can have a car painted any color that he wants so long as it is black’.
However, nowadays product life cycles are becoming shorter, while customer’ demands for innovative products in a variety of sizes, styles and fabrications are increasing (Research-Assistance, 2010). Part of this trend is the upcoming of mass-customization. Mass-customization is defined by Tseng & Jiao (2001) as a strategy concerned with offering products that meet individual's needs and preferences while these offerings are produced with near-mass production efficiency. Forrester (Gownder, 2013) defines mass-customization as a product strategy in which:
“Customers can tailor a product’s appearance, features, or content to their own specifications, resulting in an individualised product experience for the customer”.
only those products that meet their individual designer taste and “fit requirement” (Hyun-Hwa, 2011).
James Gilmore and Joseph Pine II already said it in 2000: “Roll over, Henry Ford. Today you can have any color you want, as long as it’s the one you want”. This was the beginning of the mass-customization era, however it took more than a decade before it finally hit an inflection point (Isern, 2010).
There are increasingly more examples of the fact that the time for mass-customization has finally come. Due to several authors, mass-customization can offer a solid and innovative value proposition in a time where customers are asking for individuality and authenticity (Pine, 1995). Berger & Piller already stated in 2003 that there is a tendency towards an experience economy, a design orientation and, most importantly, a new awareness of quality and functionality that demands durable and reliable products corresponding exactly to the needs of the buyer. However, it was not until recently when companies embraced the phenomenon and the industry was ready to produce mass-customized products instead of mass-produced products.
Where mass-production largely removed individuals from its equation by producing unidirectional products to a mass market of customers or target segments, mass-customization restores the individuality of customers by letting the customer participate in design. Consumers can make choices around particular features (co-design) and unique built-to-order products are made for direct delivery to the customer (individualized design) (Berger & Piller, 2013). Customers gain control again and receive a product that is tailored to their individual preferences and needs. Also, companies that offer customization are able to use consumers as their merchants - continuously gaining insights from customized designs and fine-tuning products in a feedback loop that helps companies stay one step ahead of the competition (Spaulding & Perry, 2013).
1.1.2. Practical background
The theories of mass-customization being the ‘next big thing in product strategy’ (Pine, 1999) have outstripped practice for years, with mass-customization still remaining the underdog in production. Unsuccessful real-world examples have made many people doubt on the efficacy of mass-customization. However, according to several authors, these failures have not been failures of the concept itself but have been failures of execution (Piller, F., Moeslein, K., and Stotko, C., 2004; Gowdner, 2013). Dell already started with mass-customization in 2001. This was probably more by chance, because the small company could only afford to produce built-to-order custom orders (Gardner, 2010). However, nine years later, Dell stated in 2010 that mass-customization had become too complex and costly for the main customer (Gardner, 2010). They could not pursue mass-customization as a strategy anymore, due to the price pressures the company faced. This was mainly because the computer market was shifting. Dell faced a consumer market where the price of a laptop dropped from $1650 in 2001 to $976 in 2009, as PC competitors (and many companies in other sectors) shifted production to inexpensive Asian factories.
Other unsuccessful examples are the ones that were limited by technical factors, as for Levi’s. Levi’s offered customized jeans but did not offer color options because the production process was limited. However, color was a key desire of buyers and customers did not get the value they were searching for with this offering of Levi’s (Gownder, 2013). Furthermore nonlocal manufacturing, where speed of delivery to the buyer was not provided, has been one of the reasons that mass-customization was not executed properly (Gownder, 2013). Last, the immaturity in digital experiences has often let mass-customization fail, as was for example the case for Hallmark and American Greetings. Both companies offered kiosks for consumers to design and print their own greeting cards. However, these kiosks did not remember customers. They could not create an account that saved their previous orders. Therefore, there was no learning curve and every new customization experience by the same customer had to start from scratch. Also, the customization process was very slow. This immaturity in the kiosk strategy prevented the development of a learning relationship with customers (LA Times, 1996).
manufacturing by increasing the delivery time to three weeks, which is the ‘sweet spot’ for delivering customized products (Spaulding & Perry, 2013). This way, there is time to produce the customized offering but the customer does not have to wait too long. To increase the delivery time and product replenishment for mass-customized offerings in Western countries even more, a shift can be seen from producing in the Middle East to producing in Europe and the U.S.A. (Business of Fashion, 2016).
In recent years, mass-customization has succesfully spread to many different sectors, from drinking water by Kraft that offers MiO: a personalized flavored water offering (http://www.makeitmio.com), to candy by M&M’s that has launched a product line of personalized M&M’s candies (http://www.mymms.com). Buyers can upload digital photos or write their own text, which are printed on the candies with edible ink and then shipped to buyers. Thus, to date, mass-customization has been offered in many sectors. Retail and shoes are the industries where mass-customization works best, foremost for female consumers (Piller, Moeslein and Stotko, 2004). Standardized products are mass-produced and sold on the shelf, meeting only the mean preferences of an average customer in a market segment. This implies that a major group of customers stays somewhat dissatisfied with standard offerings, even when it comes to what seem to be mature markets. This general finding is confirmed by the Outsize study (1998), analyzing consumer needs when buying clothes and shoes. A shortfall in matching fit and style was identified, especially in the up-market segment. The variety of clothes and shoes provided today is not sufficient to fulfill the heterogeneous needs of customers (Piller & Muller, 2004). Consumers are curious about the customization concept and would pay a premium for the benefits of mass-customization. Especially female consumers seem to be willing to invest in customization, so that they do not have to compromise between fit and style any longer (Piller, Moeslein and Stotko, 2004) .
Lingerie industry
comfortably. We can distinguish two main customer groups in the lingerie market and a combination of the two:
(1 Functionally driven: a woman who sees the bra purely as a comfortable and functional piece of garment. It is important the bra should fit for the lowest price possible.
(2 Emotionally driven: a woman who sees a bra as a fashion item, wants to look sexy and is willing to pay for value. (Research-Assistance, 2010)
Despite the success stories in other sectors, mass-customization is not known as a strategy for the main lingerie brands yet. Reason for this could be that the large companies do not have the supply chain that mass-customization needs, as the speed of delivery and pallet of choice options that is needed to let mass-customization work. However, start-ups are not embracing mass-customization yet either. With the trends above in mind, and mass-customization working for a lot of companies in other sectors of the retail industry, it would be very interesting to see if mass-customization would also work for the lingerie industry.
1.2. Purpose of the study
The first purpose of this study is to empirically validate the claim of several authors as Pine (1999) and Gownder (2013), that the mass-customization era has arrived and could work for the lingerie industry as well. Also, this study aims to research which mass-customization strategy would apply best to this sector and if women would pay a premium for different customized offerings. Previous studies have shown that there might be a promising market for customized offerings but not all kinds of customization processes work for every industry. For mass-customization, a distinction can be made between functional customization and design (aesthetic) customization (Piller e.a., 2006). The two main target groups in the lingerie industry can be divided in functionally driven and emotionally driven (Research-Assistance, 2010), which is a good fit with functional and design mass-customization. There is no prior research about using functional and / or design mass-customization in the lingerie industry in relation to Willingness to Pay, so this research will focus on both types of mass-customization.
1.3. Problem identification
consumers seem to be willing to participate in customization, so that they do not have to compromise between fit and style any longer. This study aims to research if and which mass-customization strategies would work in an industry entirely focused on women. Identifying the influence of using mass customization on a lingerie website, can give insight into the market conditions in which mass customized programs may be successful and those in which they may fail.
This research aims to connect the subject of mass-customization with the influence on the willingness to pay (WTP) in the lingerie industry. WTP is defined here as the price at which a consumer is indifferent between purchasing and not purchasing (Breidert, Hahsler and Reutterer, 2006; Gensler et al. 2012).
This research investigates the relationship between four independent variables (design customization, functional customization and after that Age and Cupsize) and their influence on the WTP when using mass-customization.
1.4. Problem statement
The central theme of this study is the effect of online mass-customization on the WTP for lingerie. The problem statement is stated as follows: “ To what extent does the use of mass-customization strategies influence a customers’ willingness to pay for lingerie?”
The following section will outline the subquestions that accompany the problem statement. Elaborating on the lack of evidence regarding the effect of mass-customization on the WTP in the lingerie industry, this section presents several subquestions. There are two main strategies to explore for mass-customization in lingerie. First, it should be tested whether the use of aesthetic (design) mass-customization leads to a higher WTP in the lingerie industry. Therefore, the first subquestion is stated as follows:
1. To what extent does a design (aesthetic) mass-customization strategy influence the WTP for lingerie?
Second, it will be tested if the use of functional mass-customization influences the WTP for a consumer in the lingerie industry. Thus, the second subquestion is stated as follows:
2. To what extent does a functional mass-customization strategy influence the WTP for lingerie?
Third, the potential effect of multiplying the two different customization-strategies will be researched.
Also, we are interested in the effect of age. Would younger women, of the millenium generation (also known as generation Y), perceive more value by self-design than older women, and thus pay more for customization than women of a higher age?
4. To what extent does a higher age lead to a lower WTP for customization in the lingerie industry?
Finally, we are interested if women with a larger cupsize, who need more support from bras and have less options of bras to buy in lingerie-stores, would pay more for customized bras than women with a smaller cupsize. Thus our last subquestion will be stated as follows:
5. To what extent are women with a larger cupsize willing to pay more for customized lingerie than women with a smaller cupsize?
1.5. Relevance of the study
1.5.1. Theoretical relevanceThere is no clear framework available about which customization strategy works for the various sectors. Previous studies have shown that there might be a promising market for customized offerings (Outsize study, 1998; Piller, Moeslein and Stotko, 2004; Piller & Muller, 2004). However, customization has to be customized as well. Most authors state that the effect of mass-customization is highly dependent on the specific premiums and their respective retail prices, providing a challenge for both scholars and product strategists to draw conclusions for a specific industry where mass-customization has not extensively been researched yet. There is no research about mass-customization available for the lingerie industry. It would be a good addition to the mass-customization research in the various other sectors as for shoes, t-shirts and watches (Franke and Piller, 2004). Also, there we are using one of the first toolkits that are standardized for different companies in this study (and not a company-owned toolkit), this study could get more accurate results than previous studies and results that are also applicable to other companies and sectors.
1.5.2. Practical relevance
their customized products? Which strategy will work for lingerie? These are all questions that are relevant for product strategists in the lingerie industry. This research aims to answer those questions. Also, the real dynamism in mass-customization can be found among small and medium-size businesses (SMBs) that built their product strategies from the ground up (Gownder, 2013). Therefore, it would be very interesting to look from a SMB perspective in the lingerie industry. Product strategists will succeed when offering the perfect fit for apparel together with customer-chosen colors and styles (Gownder, 2013). For a strategy that requires a relatively long learning curve, there is no time to waste in at least exploring a mass-customized offering. Also, mass-customization leads to building a relationship with customers and to cultivate a whole new dimension of customer loyalty (Pine, 1999). Additionally, differentiation plus loyalty equals margin. As mass-produced competitors are increasingly seen as nonsubsitutes, mass-customized products can ask other prices than the prices for the mass produced products. This margin grows, due to a higher volume of repeat purchases. Last, mass-customization holds the promise of helping bring manufacturing back to the US and EU due to the need for local production and highly skilled labor (Pine, 1999; Gownder, 2013).
1.6. Problem domain and scope
The domain of this research is the use of mass-customization strategies. Two of the independent variables of this research, functional mass-customization and aesthetic (design) mass-customization, are considered two of the mass-customization strategies that can be used. With this research we will focus on the lingerie industry and take the variables age and cupsize into account in order to come up with the right strategies for this industry. The Premises below form the basis upon which this research rests. Delimitations define the scope of the research.
Premises
• There is little formal research to substantiate the value of mass-customization in the lingerie industry.
• Continued rapid growth in industries that use mass-customization strategies and products will challenge management to develop effective ways identifying the appropriate mass-customization strategies for their industry and apply this to the production processes.
• The production of products is shifting back slowly from the Middle East to Europe and the U.S.A. .
• Improper execution of mass-customization by companies may be manifested in poorly choosing the mass-customization strategies to use for their particular industry, resulting in lower price standards and not being able to implement mass-customization as a product strategy.
• This research assumes that the customization tool used herein is representative of customization tools in the lingerie-industry.
• This research assumes that the use of students and the Double Dutch test-panel as subjects is generalizable to the lingerie community in Europe.
Delimitations
• This research will not consider contextual variables related to the composition of a group with respect to any psychological variables like: level of familiarity with mass-customization as a concept, experience level with the mass-customization tool, abilities, motives, personal preferences and needs.
• This research will not study the usefulness or utility of a particular mass-customization tool. In other words, this research will not consider specific hardware and/or software alternatives for the tool as an independent variable.
• This research assumes that the mass-customization design process at the level researched herein is generalizable to more complex products and systems in the retail industry.
In sum, this research will aim to provide more insight into the importance of different mass-customization strategies and their effects on the lingerie industry. Additionally, we aim to shed light on the possible moderating effect of age and cupsize of women on this relationship.
1.7. Structure of the study
2. Theoretical framework
2.1. Mass-customization Strategies
In this framework we will elaborate on the different strategies for mass-customization to use. Noble & Kumar (2008) state that a functional and emotional design value can be offered by customization with the following design strategies: utilitarian, kinesthetic, and visual design. Noble & Kumars model wants to bridge the gap between the often creatively-driven world of design, and the concerns with tangible returns on investment that drive the rest of a business. The study states that functional benefits (utilitarian design) are the core of a product and are needed to achieve kinaesthetic or visual design.
Berger & Piller (2003) offer a similar definition of the types of customization into functionality, fit (ergonomic) and style (aesthetic). Berger and Piller (2003) state that the three types of customization can be reached in different ways and combined as well.
One year later, Franke and Piller (2004) elaborate on this distinction and add ‘Comfort’ to the ‘Fit’ section with a focus on measurements. Style contains modifications aiming at sensual or optical senses as selecting colors, styles, applications or flavours. Fit and Comfort is based on the fit of a product with the dimensions of the recipient, for example tailoring a product according to a body measurement. Functionality is an option in regard to functionality or interfaces of the product, i.e. selecting power, precision or cushioning. Franke and Piller (2004) state that functionality (e.g. cushioning or selecting the insole of shoes) is often overseen when mass-customization is addressed. Hermans (2012) agrees with the above and states that a customization toolkit gives the user the opportunity to engage in three types of customization, based on the three functions a product can possess. This is the same as Noble & Kumar (2008) stated: utilitarian, kinaesthetic or visual.
In the next section we will elaborate on the three mass-customization strategies, based on the studies of the mentioned authors and the framework of Gownder.
2.1.1. Aesthetic Mass-customization
Franke, Schreier and Kaiser, (2010)
According to several authors (Noble & Kumar, 2008; Hermans, 2012; Pine, 1999), aesthetic customization includes changing the decorative design of a product. Noble & Kumar (2008) call this visual design but their definition is almost similar: Visual design is primarily focused on the generation of emotional value. (Noble & Kumar, 2008, page 274). There are numerous tactics that can be applied under the strategy of aesthetic (visual) design. Trends have for example a big influence on aesthetic design. In this, a high design approach often comes in play (Noble & Kumar, 2008). For some people, the emotional satisfaction of using and being seen with this products could even outway any performance deficiencies they may have. High-end womens’ shoes by makers such as Jimmy Choo and Manolo Blahnik can cost over
‘I feel a greater sense of psychological ownership because I designed it myself.’ ‘When I design the product myself, it has more personal value to me.’
frequent wearers (Noble & Kumar, 2008). However, design or style is something some consumers, and women in particular when it comes to fashion, are willing to pay much more money for to possess. For many, the emotional satisfaction of using these products and, perhaps even more, being seen with these products, outweighs the performance deficiencies that may occur. Franke and Piller (2004) state, as explained in the previous section, that style contains modifications aiming at sensual or optical senses as selecting colors, styles, applications or flavours.
When looking at comparitable studies in the retail industry, there are several interesting findings about aesthetic mass-customization. In a study of Piller et al. (2002) the authors found that while the optimal price for style customization for women is clearly above the average price for a standard pair of shoes, men’s WTP for style customization is lower than that for standard shoes. In an exploratory study in the watch market (Swatch alike fashion watches), Franke and Piller (2004) performed a set of four experiments with a total of 717 participants, in which users created their own customized watches. The self-designed watches are highly heterogeneous and diverse in style, confirming the trend reported in the literature, that today’s users have very distinct preferences for aesthetic customization. Other examples of aesthetic customization are the Keds sneakers, which allow customers to design aesthetic customized shoes, with the option of adding Disney images. Placing Mickey Mouse on a red sneaker with blue laces – choices made by the buyer – represents aesthetic customization (Franke and Piller, 2004; Gownder, 2013).
2.1.2. Functional Mass-customization
Franke, Schreier and Kaiser, (2010)
Noble and Kumer (2008) state that functional benefits (utilitarian design) are the core of a product and are needed to achieve the other features as design and fit. Utilitarian design focuses on the practical benefits that a product may provide.
Franke and Piller (2004) agree with this by stating that functionality is a mass-customization strategy regarding functionality or interfaces of the product, i.e. selecting power, precision or cushioning.
‘If someone else had made the same object I wouldn’t care as much. But I value it because I made it myself.’
According to Gownder (2013), a functional customization strategy is changing actual utility – the features or combination of features. For example, building-to-order a pair of jeans with four pockets instead of two changes the feature of the product. The taste of food, like a design-your-own chocolate bar, is also functional, there it deals with a fundamental feature of the product: how the food tastes.
2.1.3. Combination of different customization types
Berger and Piller (2003) state that the three types of customization can be reached in different ways and combined as well. When the color of a product changes, this would only be an aesthetic change. When the shape of a product changes, this could be both influencing the ergonomics and aesthetics of a product. So by changing one product attribute, one or more of the three mass-customization types can be realized. Furthermore, Franke and Piller (2004) found that design (style, colour and heel) and custom fit were equally important for a customers’ decision to purchase a customized pair of shoes. However, customers indicated explicitly the possibility to combine custom design with fit as the most important purchase factor.
2.1.4. Individualized design
2.2. Motivations to use online Mass-customization
According to Pine II, Peppers and Rogers (1995) and Pine & Gilmore (2000), three key trends will accelerate the implementation of mass-customization in different sectors and its use for many different products. These three trends are: (1 Higher customer expectations because of the adoption of digital technologies by customers. Customers’ adoption of digital technologies has changed the way they interact with companies and with one another. With online digital experiences (from Amazon.com to Facebook) customers can increasingly customize what they see and do. The result of this customized digital environment is that customers nowadays want everything tailored to what they want – even fysical goods they buy (Pine, 1999; Hermans, 2012; Gownder, 2013).
(2 The customer-facing technologies are becoming cheaper. Ten years ago, developing a configurator cost 1 million dollars and mass-customization has often failed due to these high costs. Nowadays, start-ups in the configuration toolkit area are popping up everywhere, what makes it easier for makers to present their customized products online (we will elaborate on this in section 2.3). This way, barriers to use mass-customization, such as the high costs of configurators and other web-tools, are overcome.
3) The customer-facing technologies of tomorrow will be revolutionary. Technologies are not only becoming cheaper but also more rich and plentiful. Technologies for customization will not only become increasingly available for smartphones and tablets but also more radical innovations will emerge. The rise of 3D scanning and printing, as well as technologies as Microsoft’s Kinect (http://www.microsoftstore.com), will offer tools that can totally change the mass customization playground.
When looking at motivations for females in particular, a study on women’s shoes of Franke and Piller (2004) found that female respondents use customization ‘to make the fashionable shoe more comfortable’ and to improve the price–quality ratio of customization. They are kind of satisfied with the footwear designs offered today, but do not want to compromise anymore when it comes to style and fit. This conclusion is confirmed by other studies on consumers’ demands for individualization of apparel and footwear (Kieserling 1999, Zitex 1999). These studies state that the most important benefit of customization for these goods is to minimize today’s compromise between fit or comfort and design. (Franke & Piller, 2004).
2.3. Toolkits
In conventional product design the designer completely defines a product. This in contrast to mass-customization where a certain amount of control is given to the consumer. This amount of control is often given by using a customization toolkit. (Hermans, 2012).
A toolkit is a solution space where customers can design their favorite product online (Berger & Piller, 2003). The designer of the toolkit already determined what the consumer will be able to customize. In other words: a solution space is constrained. This in contrary to a design space, which is seen as an infinite space. (Hermans, 2012).
According to Hermans (2012), using the right customization tools is very important for mass-customization to become successful. The better and richer the experience with a configurator (toolkit) is, the more appealing it will be for buyers to build mass-customized products. In the past, customization toolkits were often made by the manufacturer of the product. This often created complicated tools with few or no connection to what customers would want in a configurator (Berger & Piller, 2003). Also, because every company was designing a toolkit themselves, there was no learning experience from each other in order to let the tools operate smoothly. This changed in the last couple of years by start-ups which are providing simple online customization tools for customers and makers. These companies only focus on the technology in order to let the toolkit operate smoothly and make the toolkit available for all companies. These technologies can be plugged in into an existing e-commerce store. Some of the pioneers in this are Citizen Made (http://www.citizenmade.co) and Makers Row (http://www.makersrow.com).
In addition to this, Von Hippel (2006) gives five conditions to which high-quality toolkits must fulfill: (1) enable users to carry out complete trial-and-error cycles, (2) offer a wide solution space (many options or variables), (3) easy to operate even without formal training, (4) contain commonly used modules to provide a starting point, and (5) custom designed products can be made by the manufacturer properly without delays or having to change the production process drastically. In this study, a customization toolkit is used that corresponds with all the needed features stated by the authors in this section.
2.4
Strategies to use for the lingerie industry
Women can increasingly choose from prints and lingerie models. A few examples are Victoria’s Secret (www.victoriassecret.com) and Soma Intimates (www.soma.com) who offer many different bra models and brief (panty) designs. However, almost no lingerie-brands offer (online) customization. Fruit of the Loom started a customized brand in 2010 (Lingerietalk, 2010) but stopped in 2013 with the offering of these bras. There are no resources why this brand ended so quickly, but one of the reasons could be that only three colors and two cupsizes could be chosen. The wide solution space, Von Hippel (2006) is talking about, is not given. The pioneer in creating a wide solution space and offering customization in lingerie is Double Dutch Design, a new brand for customized lingerie (http://www.doubledut.ch). Double Dutch Design focuses on customizing prints and colors (aesthetic design) but also offers several functional features, as the length of the straps and a brapocket to store your money in. Furthermore, Double Dutch Design uses prints made by Dutch Designers, to create a brand image of quality design. This fits with the ‘high-end design’ approach of Noble & Kumar (2008). In relation to this, it would be interesting to research if aesthetic customization, functional customization, or a combination, would be the best customization strategy to pursue for lingerie companies.
3. Conceptual model
3.1. Direct effects
For customers, the decision to buy customized products is basically the result of a simple economic equation (Franke and Piller, 2003): if the (expected) returns exceed the (expected) costs, the likelihood that customers choose mass customization will increase. Returns work in two ways: first, possible rewards from a special shopping experience such as flow experience or satisfaction with the fulfilment of a co-design task, and second the value of product customization - i.e. the increment of utility a customer gains from a product that fits better to his needs than the best standard product attainable) (Dellaert and Stremersch 2003; Franke and Piller 2003). Costs of mass customization for consumers are: (1) the premium a customer has to pay for the individualized product compared to a standard offering; and (2) the drawbacks of the customers’ active participation at value creation during the configuration process (i.e. purchasing complexity, uncertainty, co-design risk, etc.; see Huffman and Kahn 1998, Dellaert and Stremersch 2003, Piller et al. 2004). In this study, we will focus on the premium a customer is willing to pay for the customized product compared to a standard offering.
Pine already stated in 1999 that there is a tendency towards an experience economy, a design orientation and, most importantly, a new awareness of quality and functionality that demands durable and reliable products corresponding exactly to the needs of the buyer. There are so many different customization features that it is hard where to focus on in an industry that does not use mass-customization yet. Previous customization studies in the retail industry will be used to base this research on.
Franke and Piller (2004) asked participants, who reported an interest in footwear customization, which customization options they would prefer (see Figure 2 below, showing the mean values). For women, design is the most important feature, followed by fit and functional features. One of the main insights from the survey is that fit, comfort and style (design) customization are considered almost equally important for customization. Deeper analysis of customer needs in focus group discussions, however, indicated that fit and comfort are the most important criteria in the consumers’ buying decision while colour, material and the heel length are considered as interesting but not vital parameters for customization.
Figure 2. Source: Franke & Piller, 2004
In a subsequent study of Franke and Piller (2004), the subjects stated that design (style, colour and heel) and the custom fit were equally important for their decision to purchase a customized pair of shoes. Many customers indicated explicitly the possibility to combine custom design with fit as the most important purchase factor.
As is showed at Figure 3,Spaulding, E. & Perry, C. (2013) found that the sweet spot for a price premium is 25 per cent, based on athletic shoes, dress shirts and handbags. The focus should, according to these authors, be on partly customized (so modular) customization.
Franke & Piller (2003) state in their study on watches, that customizing leads to even higher premiums: the WTP for a self-designed watch exceeds the WTP for standard watches by far, even for the bestselling standards (Swatch models) of the same technical quality. As can be seen in Figure 4 below, the WTP for a self-designed watch was 50 percent higher.
When looking at these studies, we can state that the average WTP for customization is a Figure 3. Source: ‘Sweet spot for customized products on key dimensions, based on athletic shoes, handbags and dress shirts’. Spaulding & Perry (2013
)
(individualized measurements) are the three most important parameters.
Furthermore, the women’s lingerie industry is growing, yet its consumer base requires both fashion-forward merchandise as well as products that fit women comfortably. We can distinguish two main customer groups in the lingerie market:
(1 Functionally driven: a woman who sees the bra purely as a comfortable and functional piece of garment. It is important the bra should fit for the lowest price possible.
(2 Emotionally driven: a woman who sees a bra as a fashion item, wants to look sexy and is willing to pay for value. (Research-Assistance, 2010)
An interesting point is that these two groups correspond with the two mass-customization groups that several authors made: design and functional mass-customization (Berger and Piller, 2003; Franke and Piller, 2004; Gownder, 2013).
Thus, looking at the needed and available options for women in the lingerie industry, and the mass-customization strategies that are needed most by females, we will focus on modular customization and make a distinction between aesthetic customization and functional customization.
The aesthetic customization in this study will include chosing the colors and print for the cups and straps. On this particular subject, Piller er al. (2002) found that the optimal price for style customization for women is clearly above the average price for a standard pair of shoes. Combined with the studies mentioned above, as the study of Franke and Piller (2003) (where for self-designed watches, with the same technical quality as Swatch watches (so no functional customization) customers still paid a premium) we expect that the use of design customization will have a positive effect on the WTP for lingerie. Therefore, the first hypothesis will be as follows:
H1: the use of design (aesthetic) mass-customization will have a positive effect on the WTP for lingerie.
The bra pocket and the length of the straps are functional features that can be modified and will therefore be a part of functional customization. As mentioned above, this customization strategy is seen as a vital component of customization by several authors (Noble and Kumar, 2008; Franke and Piller, 2004). However, women stated in several studies that design is their main reason for customization (Piller et al. 2002; Franke and Piller, 2003; Franke and Piller, 2004). The studies are not entirely homogeneous about the most important customization strategy for women. However, all studies state that functional customization has a positive effect on the WTP for lingerie. Therefore our second hypothesis will be stated as follows:
That combining several customization strategies works is seen at the customized shoe market as well. Adidas, for example, offers the whole set of customization options: comfort, fit, graphic design and functionalities. They can charge premiums up to 50 per cent on the suggested retail price. Nike ID on the other hand, only offers style customization (design) and can only ask premiums of 10 per cent (Franke and Piller, 2004). This is a difference of 40%! In a subsequent study Franke and Piller (2004) found that design (style, colour and heel) and custom fit were equally important for a customers’ decision to purchase a customized pair of shoes. As stated before, women used customization ‘to make the fashionable shoe more comfortable’ and to improve the price-quality ratio. Women do not want to comprise between style and fit anymore and many customers indicated explicitly the possibility to combine custom design with fit as the most important purchase factor (Kieserling, 1999; Zitex, 1999; Franke & Piller, 2004). Therefore, we assume that the combination of functional and design customization will lead to higher premiums than offering the premiums apart. The third hypothesis is therefore stated as follows:
H3: the combination of design and functional customization strategies will lead to a higher premium than using the two strategies separately as premiums.
3.2. Moderation effects
3.2.1. Age
subject such that the positive effect of mass-customization on sales is stronger when the subject is younger. Our hypothesis is constructed as follows:
H4: a younger age will have a positive effect on the WTP when using online mass-customization.
3.2.2. Cupsize
Our last hypothesis focuses on the bra-cupsize of women. In Europe the average bra size increases fast. 15 years ago, the average bra size in Europe was 75B, in 2010 this increased up to 80C. 30 per cent of the European women have cup D or larger (Nu.nl, 2013). Regarding bra size, Dutch women count for a third place in Europe. However, average lingerie brands are often not offering bras for plus size women or women with relatively large breasts according to their chest circumference. This targetgroup has to visit lingerie specialist-stores to find a fitting bra, which leaves them less choice than the women with regular cupsizes. Besides, those bras that are comfortable are mostly not that fashionable and when they are, the price is often very high when compared to what most lingerie brands offer. Furthermore, lingerie for cup D+ is often only limited available and quickly out of stock. Therefore, we expect that women with a larger cupsize would be willing to pay more for a customized bra that fits their style and comfort needs. Therefore, our last hypothesis will be as follows: H5: a larger cupsize will have a positive effect on the WTP when using online mass-customization.
In sum, our hypotheses are constructed as in the table above. The dependent variable is the WTP for lingerie. As independent variables, we will use functional and design mass-customization strategies as well as the combination of these two strategies. As moderating
Hypothesis Expected influence
Hypothesis
H1: the use of design (aesthetic) mass-customization will have a positive effect on the WTP for lingerie
+
H2: the use of online functional mass-customization will have a positive effect on the WTP for lingerie
+
H3: the combination of design and functional customization strategies will lead to a higher premium than using the two strategies separately as premiums
+
H4: a younger age will have a positive effect on the WTP when using online mass-customization
+ H5: a larger cupsize will have a positive effect on
the WTP when using online mass-customization
3.3. Conceptual model
MC = Mass-customization
The use of a functional MCstrategy
The use of a design MC strategy
Willingness to Pay for lingerie
Age
The combination of functionaland design MC strategies
4. Research Methodology
This section of the report explains the research strategy, the sample and the data collection process as well as the independent and dependent variables and the plan of analysis.
4.1. Research strategy
Because there is no prior research for mass-customization in the lingerie industry available and it is also interesting to look at a female only industry, this study will focus on the lingerie sector.
The study uses a quantitative research method (linear mixed model) in order to investigate to what extent the use of a mass-customization strategy influences the Willingness to Pay for lingerie. In order to measure the relationships between the different constructs, a customization toolkit with open questionnaire was designed and used. This is in line with the study of Dellaert & Stremersch (2005), the study of Franke & Piller (2004) and the terms a customization toolkit should meet according to Hermans (2012). The interface of the customization tool is showed at Appendix 1. Compared to a possible research design of studying the interactions with multiple toolkits, this research strategy has the advantage of providing deeper insight. This will for example lead to a higher internal validity. Furthermore, because this toolkit has been tested by several other retail companies before and adjusted to their feedback, it can be assumed that the external validity is not as limited as would be the case for a new customization toolkit. Next to customizing a bra, the focus of the study will be on the fact if customizing a set (a bra and a brief) instead of a single bra will multiply the effect of a higher WTP. This because women most of the time buy lingerie in sets (Research-Assistance, 2010).
This study aims to research what the average WTP is for the different customization strategies and what amount of variance can be explained by the different variables. In order to do this properly, a linear mixed model will be used. A ‘normal’ hierarchical regression or ANOVA Repeated Measures would not suffice, there in the current model, each subject has several measurements and the errors for those measurements are correlated within-subjects (Stata, 2013).
isfeasible for this study and hence only a respondent’s hypothetical WTP will be measured. This is in line with the study of Piller and Muller (2004). In a recent study Miller et al. (2011) found that the bias in hypothetical WTP, as compared to real WTP, is limited. The hypothetical WTP estimates derived from the questionnaire are thus expected to lead to valid outcomes.
4.2. Sample selection
Double Dutch Design is a company that offers customized bras for women with cup D+. It aims to revolutionize the bra industry with offering customized products that satisfy the individual needs and preferences of their customers. Therefore, it is important for Double Dutch Design, to know which customization strategy would work and set their pricing strategy accordingly. At the time of writing, Double Dutch Design does not have sales yet and just closed their first fundinground of EUR 82.500. There the author of this paper is the founder of Double Dutch Design, she has access to the Double Dutch testpanel and client database. However, to avoid biased results, three main groups of participants were constructed: women from the Double Dutch testpanel, students interested in mass-customization processes and random (potential) customers that were asked through social media to take part in the experiment.
4.3. Sample description
Table 2: Descriptive Statistics sample Double Dutch Design, 2015
The average age of the respondents, all females, is 30,7 years. The mean of the clothingbudget is 119.97 euro with a maximum of 500 euros. According to CBS (2011), the average clothingbudget in the Netherlands is 151.75 EUR. This means our sample has a budget that is a little below the average budget. What has to be taken into account is that also a few women outside the Netherlands have filled in the questionnaire, what could have shifted the outcome a little.
The average cupsize in this sample is a C. The current average cupsize in the Netherlands is a D (Research-Assistance, 2010), but is has to be taken into account that more than 70% of the women do not know their correct cupsize (Dagelijks.nl, 2015). Women often think that their cupsize is one size smaller than it is in real life. Therefore, this group seems to be a representative sample for cupsize.
4.4. Data collection
First of all, to gain reliable data, a comprehensive and thorough literature study was conducted in order to provide a theoretical foundation and to design a conceptual model. Furthermore, the focus of this research, and the different factors to measure the WTP for customization in the lingerie industry appropriately, could be derived from this research. Secondary literature as journals, statistics of the Central Bureau of Statistics and other related databases, Google scholar and other archival data are used in this literature study.
In order to test whether self-designing with a customization toolkit influences the Willingness to Pay for lingerie, data collection took place via the universal mass-customization toolkit of Double Dutch Design, made by the Citizen Made toolmaker (citizenmade.co) and a questionnaire to measure the WTP for design customization, functional customization and a combination of these two strategies. This will be done for a single bra and for a set.
Descriptive Statistics
N Minimum Maximum Mean Std. Deviation
There the author of this paper is the founder of Double Dutch Design (www.doubledut.ch), access to the company was already established. The incentive for respondents to take part in the study is their designed Double Dutch bra design. All women had to customize and fill in all the specifics of their favorite bra to win it. This way, it was ensured that women really used the customization tool. All questions were asked in Dutch and in English there most women that took part speak Dutch. Overall, the procedure ensured that the WTP only referred to the graphic design and functional features and not to the perception of the bras’ base technical quality. This because the bra model does not change but only the design and functional options can be chosen.
After answering three control questions (clothing budget per month, age and cupsize), subjects are asked what they currently would pay for a plain black bra and for a plain black set. This way, their current willingness to pay for a basic, non-customized bra and a basic, non-customized set is measured. After this, participants engage in self-designing behavior. They see the Double Dutch Design customization toolkit (see Appendix 1) and are instructed online to create a bra design according to their own preferences. Each participant has to use the same tool, which creates a controlled environment and quarantined reliable answers.
There are six main features that can be customized: three design options and two functional options. The functional options are: straps length where normal length and shorter length are available and a bra pocket or no bra pocket. The design options are: straps color where six color options are available, cups color where participants can choose between four base colors, designer art print by two Dutch Designers, and decorations including four different colored bows. Subjects are not informed about the price range of a normal Double Dutch Design bra to avoid anchoring distortion (Wertenbroch and Skiera 2002).
Only after a complete customized bra is created, the participants are granted access to a questionnaire. This experimental set up assures that all the participants have the opportunity to experience the same customization process. It is not possible to order the customized bras in real life because the web shop is not yet actively working.
4.5. Measurement variables
Name Type Definition Scale
WTP (Willingness to pay)
Dependent Variable WTP is defined here as the price at which a consumer is indifferent between purchasing and not purchasing (Breidert, Hahsler and Reutterer (2006); Gensler et al.2012 ).
Metric in EUR
Functional Independent Variable Investigates customers WTP for online functional customization (Dellaert & Dabholkar, 2009; Piller & Müller, 2004)
Metric in EUR
Design Independent Variable Investigates customers WTP for online aesthetic
customization (Dellaert & Dabholkar, 2009; Piller & Müller, 2004)
Metric in EUR
Current Bra Independent Variable Investigates customers current WTP for a plain, black bra (Dellaert & Dabholkar, 2009; Piller & Müller, 2004)
Metric in EUR
Current Set Independent Variable Investigates customers current WTP for a plain, black Set (Dellaert & Dabholkar, 2009; Piller & Müller, 2004)
Metric in EUR
Age Moderation / Control
Variable
Investigates if customers with a lower age (Millenium generation vs. non-millenium generation) would be willing to pay a higher premium for customized lingerie (Franke and Piller, 2004; Garnder, 2010; Spaulding, E. & Perry, 2013 )
Metric in Years
Cupsize Moderation / Control Variable
Investigates if women with a higher cupsize are willing to pay a higher premium for design and / or functional customization (Research-Assistent, 2010; Lingerie 1. Divided in low (A – C) and high (D – G) cupsize. 2. Numeric where A=1, B=2, C=3 etc. Clothing Budget Control Variable The Clothing Budget per month Metric in EUR
and the cupsizes D to G are the second category. To devide the groups, we looked at the median of this group. For grouping A,B and C together, this would contain 62% of all respondents and would be the best deviation. For the linear mixed model, we changed the cupsizes into numbers. This in order to see if the WTP for lingerie changes when the cupsize goes up (A=1, B=2, etc.).
4.6. Data analysis
4.6.1. Regression modelsA linear regression model is used and then altered for the different tests. The basic model is: 𝑃 = 𝛽!+ 𝛽!𝑚! + 𝛽! 𝑚!+ 𝛽!"𝑚!"+ 𝜀!
P = WTP
f = functional customization d = design customization s = Set
𝜀! = error term per individual
𝑚 : dummy variable for the 𝛽 parameters. 𝑚 = 1 or 𝑚 = 0.
𝛽! : reference category. (what would you currently pay for a plain, black bra?)
𝛽! : Design (Aesthetic customization) link to customization tool www.doubledut.chà what would
you pay for a bra with a graphic design and color of your choice?
𝛽! : Fit (Functional customization) (what would you pay for a black bra with better functional features
than your current bra?)
df = Design + Fit (what would you pay for a bra with functional and aesthetic customization) link to customization tool www.doubledut.ch
Control and moderator variables:
Ø budget (what do you spend on average on clothing per month?) Ø age (in years)
Ø cupsize (numeric where A=1, B=2 etc.)
This model will be altered for the different models (fixed effects and mixed model).
The mixed effect model deals with double indexed random variables ( 𝑦!"). The first index is for
𝑌!" = 𝛼!+ 𝑋!"𝛽!+ 𝜀!"
Where
𝛼! 𝑖 = 1 … . 𝑛 is the unknown intercept for each individual (n entity-specific intercepts)
𝑌!" is the dependent variable (DV) where i = entity and j = measurement 𝑋!" represents one independent variable (IV)
𝛽! is the coefficient for that IV
𝜀!" = vector of random errors
This is the individual regression for each group. The second level regression used tries to explain variation in regression coefficients:
𝛼!= 𝛾!+ 𝑍!!𝛾!+ 𝑢! 𝛽! = 𝛿!+ 𝑍!!𝛿!+ 𝜀!
When the second equation is substituted to the first one, it will become as follows: 𝑌!"= 𝛾!+ 𝑍!!𝛾!+ 𝑋!"𝛿!+ 𝑋!"𝑍!!𝛿!+ 𝑢!+ 𝜀!"
The fixed effects are: 𝛾!, 𝛾!, 𝛿!, 𝛿! . The random effects are 𝑢! (between-entity error) and 𝜀!" (within-entity error). 𝑍! = the random effects design matrix for group i (Stata, 2013).
4.6.2. Linear Mixed Model
In this study, WTP is treated as a categorical factor with four levels: the WTP for a plain black bra, the WTP for design customization, WTP for functional customization and WTP for the combination of those two. The same model has been fit for Set. The model for Set has not been further researched in this study, due to time and limited space of the study.
There were also additional random effects present: age, cupsize and clothing budget. In the design, multiple measures per subject were taken. Subjects were asked what they would pay for a plain black bra, and were asked what their WTP would be for a bra with the use of the several customization strategies and a combination of the strategies. When looking at the assumptions of the linear model, this would violate the independence assumption (Field, 2005; Winter and Grawunder, 2012): multiple responses from the same subject cannot be seen as independent from each other. Therefore, a random effect is added, a random intercept, for the variable ‘ID’. This way, the non-independence can be solved by assuming a different ‘baseline’ value for each subject. Thus, the by-subject variation in overall WTP has been accounted for.
account. It is predicted that the WTP for customization is modulated through Age and Cupsize. The Linear Mixed Models procedure allows specifying factorial (IV) interactions, which means that each combination of factor levels can have a different linear effect on the dependent variable (Baltagi, 2013).
4.7 Model validity
Scatterplots and a Pearson correlation matrix are used to test the linearity and further validity of the model. F-test and Chisquare values are used for the overall goodness of fit. Subsequently a Hausman test is computed and AIC and BIC tests are used to compare models. After this, a mixed model with fixed and random effects will be used to account for the correlated errors and dependency of errors in this study. Furthermore, we use a standardized customization toolkit, used by many companies in the past. Also, the questions and the customization process were tested by 50 women of the Double Dutch testpanel before the actual study was done. These two factors help explore and verify the model feasibility and its generalizability (Cameron & Trivedi. (2010).
5. Results and Analysis
This section of the report explains the main results of the research. First the Pearson correlation matrices are presented, followed by a Linear Fixed Effects model and a Linear Mixed model.
5.1 Assumptions for regression
The variable cupsize was made categorical to be able to use in the correlation matrix. See Appendix 2 for this process. From the scatterplots about the WTP for a plain, black Bra (see Appendix 3), it is showed that there is a positive relationship between the WTP for a current black bra and the use of the two mass-customization strategies. The same accounts for the relationship between the WTP for a current black bra and the combination of using functional and design customization. Therefore, it can be assumed that there does appear to be some linear relationship between those variables and the WTP of women for a current black bra. For the WTP for a set and the WTP for a plain, black Set (bra plus brief), the WTP for a customized designed set, a better fitting set, and a combination of design and functional customization of a set, there does appear to be some linear relationship as well (see Appendix 3). Furthermore a Skewness Curtosis test for normality was executed. A goodness of fit test for normality of the error term, 𝜀! will be conducted, as is unobserved, and the residuals 𝑒! are used.
test (1987), to be able to use the test for a small amount of data. Samples from a normal distribution have an expected skewness and expected excess kurtosis of 0 and an approximate chi squared with 2 degrees of freedom. The chi squared is adjusted for the fact that a small sample is used. The output of the command xtreg in Stata was changed into residuals and a sktest was executed, as can be seen in Figure 4 below. The p-value for Skewness and the p-value for Kurtosis are significant for the regression residuals on the 0.1 level for Skewness and 0.00 level for Kurtosis. Therefore, it can be stated that the errors are distributed normal.
In order to meet the assumption of regression that the error variance should be constant,
heteroscedasticity-consistent standard errors are used (with the option ",r"). Also, as is mostly used in
regression and time-series modelling, we make use of the assumption that the errors or disturbances ui have the same variance across all observation points.
5.2. Correlation-matrices
5.2.1. Pearson correlation for Bra
In Table 3, the Pearson Correlationmatrix for WTP for a Bra is showed. All three independent variables (BraDesignCust, BraFunctCust and BraDesignandFunct) have a statistical significant linear relationship on the p < 0.01 level with the WTP for a plain, black bra. Also, all three independent variables have a positive statistical significant linear relationship with the control variables (age, cupsize and clothingbudget) and with the other independent variables on a p < 0.05 level. The variable Cupsize has a positive statistical significant linear relationship on a a p < 0.05 level in relation to the three independent variables. Therefore, this variable can be used in the linear mixed model. Interesting is that Age also has a positive statistical significant linear relationship on the p < 0.01 level with all the WTP categories (current black bra, design, fit and a combination). It can be assumed that there is a linear relationship between a higher age and a higher WTP for lingerie.
residuals 452 0.0782 0.0000 35.13 0.0000 Variable Obs Pr(Skewness) Pr(Kurtosis) adj chi2(2) Prob>chi2 joint Skewness/Kurtosis tests for Normality
. sktest resid
Table 3
5.2.2. Pearson correlation for Set
In Appendix 4, a correlationmatrix for Set is showed.
All three independent variables (SetDesignCust, SetFunctCust and SetFunct+DesignCust) have a statistical significant linear relationship on the p < 0.01 level with the WTP for a plain, black bra. Also, all three independent variables have a positive statistical significant linear relationship with the other independent variables on a p < 0.05 level. Design and fit customization for a set and the WTP for a combination of the two customization-strategies for sets (SetFunct+DesignCust) have a statistical significant linear relationship on the p < 0.01 level with the WTP for a current set. The variable Cupsize has a positive statistical significant linear relationship on a a p < 0.05 level in relation to all three independent variables. Identical to the correlations for Bra, the variable Age has a positive statistical significant linear relationship on the p < 0.01 level with all the WTP categories. It can be assumed that there is a linear relationship between a higher age and a higher WTP for a lingerie-set.
Due to time and space constraints of this study, the data for Set will not be further explored in the linear mixed model. From now on, the study will only focus on the effects of buying a Bra.
5.3. Fixed Effects using least squares dummy variable model (LSDV)
Now that the linear relationships between the variables and other assumptions are tested, the different premiums for customization will be tested. It will be checked if the premiums for the different variables are significant and what the correlation of the premiums is with the control variables Cupsize, Age and Clothingbudget.
Stata (Stata/SE 12.0, 2011) with the command ‘xtreg fe’ is used (Baltagi, 2013; Greene, 2012; Cameron & Trivedi, 2010) to perform a linear fixed effects model analysis of the relationship
Pearson Correlationmatrix for Bra 1 2 3 4 5 6
1. Bra Design Customization strategy 1 ,910** ,935** .149 ,412** ,314**
2. Bra Functional Customization strategy ,910** 1 ,945** .182 ,408** ,364** 3. Bra Design + Functional Customization strategy ,935** ,945** 1 ,194* ,364** ,337**
4. Clothingbudget per month in Euros .149 .182 ,194* 1 .156 -.108
5. Age ,412** ,408** ,364** .156 1 -.044
6. Cupsize (small (A-C) or big (D-G) ) N
,314** ,364** ,337** -.108 -.044 1
113 113 113 113 113 113
**. Correlation is significant at the 0.01 level (2-tailed). *. Correlation is significant at the 0.05 level (2-tailed).
the data is declared as panel data. As fixed effects, the different customization strategies wtp_2, wtp_3 and wtp_4 are entered, with wtp_1 as reference category (black bra). For every individual there is a dummy included. The first row of Table 4 in section 5.4. (and in Appendix 5B in detail) shows this fixed effects linear model with individual fixed effects. Explanatory variables that do not vary for an individual could not be estimated (e.g. age, budget, cupsize) and were omitted from the analysis.
Based on the results of the F-test, with a significance of 0.00, we assume that this model is consistent for the true parameters. The R-squared within is 0.47, which means that 47% of the variance in the model can be explained by the measured variables. The overall R-squared is 0.06. This means that the fixed effects can explain only 6% of the variance. This is very low but panel datasets are a combination of cross-sectional and time-series datasets, and it should be expected that the reported explanatory power of panel datasets to lie in between (Baltagi, 2013; Stata, 2013). When linear regression is executed instead of the xtreg fe- command in Stata, the reported R squared will likely be much higher. This because in the xtreg, fe- calculation, the explanatory effects of the intercepts are washed oud and if just linear regression is run, those explanatory effects are still accounted for (Cameron and Trivedi, 2010). Therefore, a comparison with two other regression methods (OLS and areg) is made. It is showed (see Figure 5 below) that all three methods show the same results. Therefore it can be concluded that the Rsquared of the OLS method (areg method) can be used. Now, it can be seen that the amount of variance explained by the fixed effects is 91%.
Figure 5
Furthermore, the intraclass correlation, the rho-value is 0.911. It can therefore be assumed that 91% of the variance is due to differences across panels.
5.4. Hausman test
To be certain that a mixed model (so a model with random effects) is the best fit for this data, a Hausman test is conducted, which compares the Fixed effects analysis above with a Panel regression with random effects. It will be tested whether the unique errors (ui) are correlated with the regressors, the null hypothesis is that they are not (Greene, 2012). In other words, the null hypothesis is that the
