Master Thesis
The impact of goals on the customer decision process: A discrete
choice experiment
Final Version
June 22, 2018
Faculty of Economics and Business
MSc Business Administration Track: Marketing
Lisanne Reizevoort
Student number: 10581847 Supervisor: Dr. J.Y. Guyt
Statement of Originality
This document is written by Lisanne Reizevoort who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Table of Contents
Abstract 5
Chapter 1 Introduction 6
Chapter 2 Literature Review 9
The evolution of the customer decision process 9
Goal-directed customer behavior 12
Goal-directed choice models 14
Hypotheses and Research Framework 15
Chapter 3 Methodology 17 Pre-test 17 Research Design 18 Survey Design 19 Data Collection 20 Analytical Strategy 20 Chapter 4 Results 23 Preliminary Analysis 23 Manipulation Checks 24 Descriptive Statistics 25 Hypotheses Testing 26 Chapter 5 Discussion 41 Managerial implications 42 Theoretical implications 43
Limitations and Future research 44
References 46
List of Tables and Figures
Tables
Table 1 – Amount of participants per condition 20
Table 2 – Participants per category 23
Table 3 – Descriptive Statistics 25
Table 4 – Logistic regressions for Milk 35
Table 5 – Logistic regressions for Tomatoes 36
Table 6 – Logistic regressions for Granola 37
Table 7 – Logistic regressions for Wine 38
Table 8 – Logistic regressions for Shampoo 39
Table 9 – Willingness to pay 40
Figures
Figure 1 – The Funnel Metaphor 10
Figure 2 – Customer Decision Journey 11
Figure 3 – Goal-Based Customer Decision Model 15
Figure 4 – Conceptual model 16
Figure 5 – Interaction effect high price (milk) 31
Figure 6 – Interaction effect organic (milk) 31
Figure 7 – Interaction effect plastic package (tomatoes) 31
Figure 8 – Interaction effect medium price (tomatoes) 31
Figure 9 – Interaction effect organic (granola) 32
Figure 10 – Interaction effect medium price (wine) 33
Figure 11 – Interaction effect South Africa (wine) 33
Figure 12 – Interaction effect organic (wine) 33
Abstract
Like almost all behavior, customer behavior is goal-directed. Customers evaluate products according to the goals they want to attain. However, their customer decision process is influenced by the emergence of new digital channels. Customers are able to easily consume and produce all kind of information about brands and products, not only during their consumption choice, but also when they create their consideration set and even after purchase. Therefore, this study examines the influence of goals on the evaluation of products in multiple stages of the decision process. Contributing to existing literature that primarily focuses on the multiple-goal pursuit to examine differences in multiple-goal evaluation, this research incorporates the single-goal pursuit with the aim to find differences in product evaluation among different stages of the customer decision process. The study shows that stimuli from the environment can activate customer goals to elicit goal-directed information processing. A choice-based conjoint analysis is used to find patterns in the product evaluation between the conditions. Results show that customer decisions are influenced by goals, especially the evaluation of product attributes that differs from goal to goal. Moreover, the impact that goals have on the product attribute evaluation differs from the consideration set formation to the final purchase choice. The variation in the effect among product categories suggests that products differ in importance as means to the goal pursuit. This research offers insights for practical matters in the area of marketing, and for future research on goals in customer behavior.
Keywords: customer goals, single-goal pursuit, consideration set formation, customer decision process, choice-based conjoint.
Introduction
For a few years now, new technology has been disrupting the grocery industry dramatically. Customers are more empowered than ever before. How they were ten years ago just able to evaluate different product options in the brick-and-mortar stores, are they now connected anytime anywhere and do they use their mobile device while shopping (Demeritt, 2016). The Internet has increased their choices enormously, and they have the power to discover new brands, products or services whenever they want. The rise in available information about products and services has enabled them to evaluate all options thoroughly (Labrecque et al., 2013). They can create information about products, brands and companies by writing reviews, and share information through online social networks (Labrecque et al., 2013). Consequently, retailers are able to track customer’s buying behavior in-store and online, and use these data for direct marketing such as personalized ads and personal discounts (Savaglia, 2017). Overall, the presence of new technologies has changed the way customers and companies interact with each other, and the amount of touchpoints in the customer’s decision process has increased enormously. Accordingly, the customer decision process is now an ongoing process in which products and brands are evaluated continuously, even after the purchase (Court et al., 2009).
Besides, recently a change in the drivers of consumption has occurred in the food industry (Ringquist et al., 2016). This change is caused by a shift to a more sustainable lifestyle: people care more about their personal wellbeing and that of others, animals and the environment (Bender, 2017). Therefore, next to price, taste and convenience, customers now also take into account health and wellness, safety, social impact, experience, and transparency. These drivers can be seen as a translation of customers’ personal goals. Goals are the desirable end-states that customers deliberately try to attain, because of the perceived benefits (Gutman, 1997; Baumgartner and Pieters, 2008). People try to attain their goals through action, and therefor they incorporate them in their consumption choices. When a customer is in the supermarket and
has to decide which product to buy in a specific product category, he chooses the product that contributes to the attainment of a goal, and thus best meets idiosyncratic needs (Weber and Johnsson, 2009). To make this choice, customers evaluate the product attributes and consequences of a product consumption based on deliberate and automatic goals they aim to pursue (Gutman, 1997; Verplanken and Aarts, 1999). In fact, customers are constantly striving to complete multiple goals at a time. As a result, they have become more and more demanding and more difficult to please (Labrecque et al., 2013). For companies it is now the time to stand out, and to anticipate customers specific and changing want and needs. They have to find answer to question like: Which goals do customer incorporate in their decision process? Where are they looking for in a product to achieve their goals? How do they influence their decision process? And does their goals have a different impact on different stages of the decision process?
Former research has already studied the effect of goals on the customer decision process (Van Osselaer & Janiszewski, 2011; Swait et al., 2018; Dellaert et al., 2017; Souza, 2015). However, it does not provide insights into differences in the impact of goals between different stages of the decision process. Hence this study addresses the gap in literature by examining the following research question:
RQ: “How do customer goals affect the consideration set formation and the final purchase choice?”
In terms of theoretical relevance, this research will contribute to literature about goals in customer behavior. The aim is to study the impact of goals, which customers are pursuing and trying to attain, on the purchase choice in more detail. Since the customer decision process is seen as a multiple stage process in which customer evaluate different criteria, it interesting to examine the impact of goals on this evaluation. Recent research has already studied the effect
on several stages of the decision process separately (Swait et al., 2018; Souza, 2015). But, the differences in product attribute evaluation and product choice between the different stages of the decision process have yet not been examined. And, at a time where customers are more empowered, it is critical to get a better understanding about the differences in product evaluation between different stages of the decision process. So, this study will search for differences in product evaluation between multiple stages of the customer decision process affected by the goal pursuit, and will make use of a choice-based conjoint analysis.
In terms of managerial relevance, a better understanding of the effect of customer’ goals on the decision process can have an impact on the marketing communications and the development of new products in the industry for fast-moving consumer goods. If it turns out that goals affect the consumption choices differently from stage to stage in the customer decision process, brands can focus more on the customization of the marketing mix per stage of the decision process. For existing products, for instance, they can build a marketing communication strategy around the attributes which contribute to the attain of the goal in a particular stage. And for the development of products they can use customer data and trends to find out which goals are of great importance, so new product lines and a marketing strategy can be built around these goal(s). Since the initial consideration set is quite small in the ongoing customer decision making process, it is crucial to stand out (Court et al., 2009). With the new strategies companies can target more specifically through the entire process.
The remainder of this research is structured as follows. The first chapter provides a review of relevant literature and the proposed hypotheses. Subsequently, information about the research design and the methodology is discussed. Then, the hypotheses are empirically tested and the results are summarized. Lastly, the discussion provides the conclusion, theoretical and managerial implications, limitations, and directions for future research.
Literature Review
In this section the most relevant findings and core concepts from previous literature are discussed. It starts with the emergence of the customer decision process, and goes deeper into two types of customer decision processes. Then, it explains how customers use goals in their decision processes, and next how they are used in grocery consumption. Additionally, some relevant goal-direct decision models are presented. And finally, the hypothesis are introduced and visually presented.
The evolution of the customer decision process
In customer behavior a lot of research from psychological science is used. Especially, cognitive information processing that provides insights on the thought processes an individual goes through when making an decision. It is defined as “active, effortful planning, and goal-direct customer behavior that involves mediated intellectual activity” (Erasmus et al., 2001). In this manner, customers are viewed as a rational thinking individual who uses information and experience retrieved from memory to make reasoned purchase decisions. There are two major types of cognitive models: analytical models and prescriptive models (Bray, 2008). Analytical models are frameworks that help to explain, predict and design patterns in behavior of customers. Most analytical models follow a five-stage decision-making process, because it best fits the rational approach (Erasmus et al., 2001). The model contains the following stages: problem recognition, information search, evaluation of alternatives, purchase and post-purchase evaluation. From stage to stage customers carefully evaluate alternatives on their functional, emotional and economic value (Erasmus et al., 2001).
These insights of cognitive information processing are the base for multiple marketing theories. The initial marketing theories about customer decision processes are rooted in the 1960s (Lemon and Verhoef, 2016). Howard and Sheth (1969) were one of the first who created
such a theory, including a model of customers’ buying behavior. The theory divides the decision-making process in three levels: extensive problem solving, limited problem solving and habitual response behavior. It interprets customers as rational thinkers whose purchase behavior is stimulated by cues from marketing activities and their social environment. The process is repeatable so that customers collect and process information and learn from former choices, and there are five observable responses to the stimuli: attention, brand comprehension, attitude, intention and purchase (Howard and Sheth, 1969). Various marketing theories, such as the path-to-purchase models, are built upon this model (Bettman, 1979; Neslin et al., 2006). These models have in common that they provide insights in the interaction moments between the customer and a company, also known as touchpoints (Rawson et al., 2013). For years, the ‘funnel metaphor’ (Figure 1) has served as an important model in this field (Court et al., 2009; Edelman, 2010). At the wide end of the funnel, customers start their decision-making process with a bunch of brands, the consideration set, and while moving through the funnel they are exposed to many marketing activities, narrowing them down to a final choice.
Figure 1 - The Funnel Metaphor
However, this funnel is now seen as outdated. It is not able to capture the increase in touchpoints that is caused by the rise of digital channels that has empowered customers and increased the product choices (Court et al., 2009; Labrecque et al., 2013). Two-third of touchpoints are customer-driven, and only one-third is still company-driven (Court et al., 2009). Customers are in control, they filter helpful information from irrelevant information, and share content about
brands and product in their social network, offline and online. As a results, marketing departments are changing their perception to stay influential in the customer-driven touchpoints. They derive customer data from the digital technologies, and use it to track, map and optimize individual customer journeys. Court et al. (2009) developed a new approach, the customer decision journey (Figure 2). The decision-making process is seen as a circular journey, in which products and services are more actively evaluated, choices are added and removed from the consideration on an ongoing basis, and evaluation continues after a purchase choice has been made.
Figure 2 - Customer Decision Journey
To capture the post-purchase evaluation and the loyalty loop of the Customer Decision Journey, the customer behavior should be observed over a longer period of time. However, in this research it is not possible to perform a longitudinal study, because of a lack of time. Therefore, it will focus on the two main stages of the customer decision process: consideration set and the final purchase. The terminology of Souza (2015) is slightly adapted to formulate the different stages of the decision process. The universal set is the pre-determined set of options from which a consideration set is formed. This are “all alternative considered at the moment of choice” (Souza, 2015). From the consideration set follows a final purchase choice, a single alternative from the consideration set is selected (Souza, 2015).
Goal-directed customer behavior
It is well-recognized that individual behavior, including customer behavior, is goal-directed (Gutman, 1997; Baumgartner and Pieters, 2008). Customers formulate goals in such a way that they are desired or undesired states customer wants to attain or avoid because of its perceived benefits, resulting in approach goals or avoidance goals (Carver, 1996 cited in Souza, 2015). Customer goals can be categorized in three different categories with each their own motivation to be applied in a purchase decision process (Van Osselaer et al., 2005). Consumption goals are end-states obtained by the consumption. Because these goals are context-dependent, the value of a product is variable. Criterion goals are the goals that need to be achieved to meet the criteria of a good choice outcome. Process goals are end-states that a customer wants to obtain related to choice process, and therefore context-independent (Van Osselaer et al., 2005). Once a goal is set, customers start making consumption choices that best meet the end-states related to these goals.
Research has shown that choices can be better understood when a hierarchy of goals is used (Austin and Vancouver, 1996; Bagozzi and Dholakia, 1999). The hierarchical framework divides goals into groups: higher-level and lower-level goals, called the sub-goals. The theory behind this framework is the Means-end chain theory (Gutman, 1982). Gutman (1997) states that goals can be placed at each of three stages of the means-end chain model – attribute level, consequence level or end-value level, and that first lower level goals need to be attained to achieve a higher-level goal. For example, when the final goal is at the consequence level, sub-goals at the attribute level help to attain the final goal. The different types of sub-goals are spread to people’s behavioral plans, such that they are induced to take action to attain the goals (Ferguson & Bargh, 2004). As a result, people select means which best fit with the goal attainment in such a way that customers make several consumption choices that contribute to
the achievement of goals. Frequently, multiple consumption choices are made to attain a single goal.
Besides, the influence of goals on product choices can be explained by the goal compatibility framework (Markman and Brendl, 2000). In general, customers are evaluating alternatives based on product attributes and available information about products and brands (Court et al., 2009). In this way a value is created for each single product and the product with the highest values will be selected to purchase. In addition to this, the goal compatibility framework suggests that the value of a product is depending on the goal a customer wants to achieve, such that product attributes and product information are evaluated against the active goal. Sometimes this evaluation occurs effortless and sometimes a lot of deliberation is required (Kopetz et al., 2012). Because customer’s goals are variable over time, also product choices change over time (Markman & Brendl, 2000). This means that products are valued higher when it contributes to the goal pursuit, and that when the goal is achieved, the value decreases if the product is not a means for the attainment of another goal (Ferguson & Bargh, 2004).
Additionally, research has shown that customers incorporate multiple goals in one consumption choice (Kernan and Lord, 1990; Schmidt and Deshon, 2007; Swait and Argo, 2012). Kruglanski (2002) states that when the goals are equally important one consumption choice can lead to the accomplishment of each of these goals. However, oftentimes the importance of a goal is weighed against the other goals, depending on circumstances and personal cognitive strategies. Hence, one goals is then prioritized over another and some goals will be neglected, what could lead to inaccurate depictions of evaluations (Kernan and Lord, 1990).
Goal-directed choice models
Last decade, several researchers constructed a customer decision processes with the incorporation of the multiple goal pursuit incorporated to generate new insights in customer behavior (Swait et al., 2018; Dellaert et al., 2017; Van Osselaer & Janiszewski, 2011). They primarily focused on the multiple-goal pursuit and created goal-directed choice models to create new insights in the evaluation of goal in customer decisions. Van Osselaer & Janiszewski (2011) developed a goal-based choice model that consists of multiple tasks. In the first task, customers select products for evaluation that are presented or retrieved from memory, or a mix of both. Secondly, they evaluate each alternative based on its level of hinder or help to achieve a goal, and they asses a weight of importance to each goal based on the decision to be made. And lastly, they sum the weights of goal attainment per alternative, evaluate this and make a choice (Van Osselaer & Janiszewski, 2011).
Besides, Swait et al. (2018) composed the multiple-goal-based choice model (MGBCM). This model consists of three dimensions: prior goal weighting, goal specific attribute allocation and goal weight adaption. First, customers set some goals and allocate a weight to each goal prior the actual purchase decision process. Then they decide how well each alternative performs on each goal based on the attributes of the alternatives. Lastly, they modify the goal weights based on the actual product assortment, thus more weight to more attainable goals based on the assortment, and the final choice is made. As a future research direction, Swait et al. (2018) suggested to extend the MGBCM by putting the model into a multiple-stage decision-making process, and to incorporate the consideration set formation.
Yet relatively little research has been done to the effect of goals on consideration set formation. In this field Souza (2015) establishing the Goal-Based Choice Set Formation framework. In this framework goal thresholds mediate the formation of the choice set such that a screening/recruiting process occurs. Customers form a consideration set by screening the
universal set based upon the attainment of the goals which are set as a threshold. Only when both goals can be attained the alternative will be part of the consideration set. By testing the model, he found that goals act as drivers for variations in attribute importance in the consideration set formation stage (Souza, 2015).
Hypotheses and Research Framework
To extend prior research, this study incorporates the consideration set formation (CSF) and the final purchase decision (FPD) in one model, called the Goal-Based Customer Decision Model.
Figure 3 - Goal-Based Customer Decision Model
Since previous research has shown that goals affect product attribute evaluation in the consideration set formation and in the final purchase decision (Swait et al., 2018; Souza, 2015), there is perceived that people value product attributes differently from one goal to another in each stage of the customer decision process. This leads to the following hypothesis:
Hypothesis 1: Customer goals affect the impact of product evaluation on the consumption choice.
- H1a: In CSF, the effect of product attributes evaluation on product choice is moderated by consumption goals.
- H1b: In FPD, the effect of product attributes evaluation on product choice is moderated by consumption goals.
This study also looks into the differential effect between the moderation in the CSF and the FPD. There supposed to be an effect, but given the lack of literature pointing in either direction the directionality is left as an empirical question. This leads to the second hypothesis:
Hypothesis 2: There will be a differential effect between the goal moderation in the CSF stage and the goal moderation in the FPD stage.
The differential effect can be explained as follows. There is a difference in the attribute evaluation between the goal in each stage separately, and this difference in evaluation is also different between the two stages of the decision process. This could lead to the two following situations:
1. In CSF some product attributes are more important for goal 1 compared to goal 2, while in FPD these attributes are more important for goal 2 than for goal 1.
2. In both stages the same product attributes are more important for goal 1 than for goal 2 (or vice versa), but the difference in importance is different from one stage to another.
The visualization of the hypotheses is shown in the following conceptual model:
Methodology
To understand the importance of goals in the customer decision process, a choice-based conjoint analysis is conducted. This provides a way to estimate a real-time decision-making process (Aguinis and Bradley, 2014). First, a pre-test is done to determine the goals. Then, the main study is conducted of which the sample, the research design and the procedure will be discussed in this section. Lastly, the analytical strategy will be discussed.
Pre-test
A pre-test is done to determine the goals that are used in the main study. Two surveys are conducted, one for product categories to find goals in a broad sense, and one for specific products to stay close to the setting of the main survey. Vegetables and fruit, breakfast and snack were used as product categories, and milk, coffee and wine as specific products. The pre-test was done among 15 respondents who were randomly assigned to one of the surveys.
Based on the laddering interview technique survey questions are formulated with the aim to generate a set of goals that covers all levels of the means-end chain (Gutman, 1982; Pai and Arnot, 2013). This technique generates insights in customer perceptions of attributes, consequences and values in relation to a product (category), what can be used to generate high-level and low-high-level goals for the main study. The surveys were extended with some questions in which participants had to indicate to what extend they take into account a particular goal when making a consumption choice.
The open questions resulted in a huge list of goals. Some goals were only mentioned once, other several times, and multiple goals were similar to each other. Finally, the following two goal categories were constructed from the results:
1. A moment of full enjoyment
Research Design
The main research has a conjoint design. In marketing, it is the most used technique to analyze trade-offs in customer choices (Cattin and Wittink, 1982). Most important for this research is that it can estimate patterns in customer preferences based on product choices, and reflects on choices customers make from day-to-day (Green et al., 2001). There are several types of conjoint analysis. Although the two-factor-at-a-time method seems to be more representative of the real-life behavior of a customer (Green and Srinivasan, 1978), for this research the full-profile method is used. Compared to the two-factor-at-a-time method, in which only two product attributes are exposed, the full-profile approach presents multiple product attributes with each multiple levels. This could lead to an information overload, but per category only four product attributes with a maximum of three attribute levels are used (Appendix A). The conjoint analysis is done among six product categories: milk, tomatoes, granola, wine, shampoo and coffee. Coffee consists of four subcategories: coffee beans, ground coffee, coffee pads and coffee cups.
To create the conjoint design, the LMA design method is used in Rstudio. This method creates the conjoint design from the orthogonal main-effect array (Aizaki, 2012), in which there is no correlation between the attributes of the alternatives within one choice set (Johnson et al., 2006). The ‘questionnaire-function’ is used to create survey questions related to the consideration set formation (CSF), the first stage of the customer decision process. The system has provided over a hundred of questions per product category, each questions reflected one universal set consisting of 10 alternatives. For each product category, five questions were selected to include in the survey.
To create the consideration sets for the questions related to the final purchase choice (FPD), the second stage of the customer decision process, data from the first data collection is used (see survey design below). The alternatives that were mostly chosen and most in
combination with each other in the CSF, are used to create the consideration sets for the second data collection.
Survey Design
The research is conducted through an experiment because it provides a way to discover a causal relationship between the variables (Field and Hole, 2003). It was split into two online surveys (Appendix B), administered via Qualtrics, to collect data for the different stages of the customer decision process. In both surveys, participants were first randomly assigned to one of the following goal manipulations. Goal manipulation was used, because literature has shown that when a goal is activated people are enabled to process goal-related knowledge and information in such a way that it contributes to the attainment of the goal (Kopetz et al., 2012).
1. Imagine that you are shopping in the supermarket and are deciding about what to buy. The recent weeks have been challenging for you. Therefore, you want to treat yourself to products for which you think it to be a moment full of enjoyment when you consume this. Hence, you decide to buy only products that will contribute to this.
2. Imagine that you are shopping in the supermarket and are deciding about what to buy. Recently, you have become more aware of environmental problems. Therefore, you want to contribute to less environmental damage and a better welfare of animals, and at the same time you want to attain a healthier lifestyle.
Then, in the first survey participants were randomly assigned to CSF or both stages. In CSF, they had to form their consideration set from the provided universals set. In both stages, first they had to form their consideration set, and from this self-constructed consideration set they had to make a final purchase choice. The second survey captured the FPD, in which participants had to make final purchase choices from the provided consideration sets. In CSF, participants
were randomly assigned to three product categories, in both stages to 2 product categories, and in FPD to 5 product categories.
Lastly, in both surveys participants were asked to fill out some questions regarding goal incorporation and their demographics.
Data Collection
For the data collection, no restrictions were made regarding the population, because everyone is used to make grocery consumption choices. This makes the population large and diverse in their goals. Therefore, non-probability sampling is used, in particular volunteer-convenience sampling and snowball sampling. Respondents are reached through Facebook and WhatsApp (Goodman, 1961). In total 245 respondents completed the survey. Table 1 shows how many participants were assigned to each condition.
Table 1 – Amount of participants per condition
Manipulation stage of customer decision process
CSF FPD Both stages
Manipulation type of goal
Environment Condition 1: N=32 Condition 2: N=43 Condition 3: N=45
Enjoyment Condition 4: N=40 Condition 5: N=43 Condition 6: N=42
Analytical Strategy
To conduct a statistical analysis of the collected conjoint data, a logit model should be implemented. A part-worth can be determined for each product category by performing dummy variable regressions. Subsequently, the value or utility of each product in the universal sets and the consideration sets can be calculated by implementing the attributes in the following utility function.
Utility = α + *++ *,-, + *.-. + … + *0-0 + 1
where k represents the product attributes, -0 the dummy variables for the presence of
the attribute, and *0 the regression coefficient that indicates the value of each attribute. By
comparing the part-worth range of one attribute against the sum of part-worth ranges of all attributes, the importance of one attribute can be determined.
However, the aim of this study is to examine differences in product evaluation between customer goals in different stages of the customer decision process. Therefore, a logistic regression will be used, because it is an appropriate way to analyze data in which the dependent variable, the response of the participant, only takes two values (Rodríguez, 2007). There is assumed that the dependent variable follows a binominal distribution, which is often used to model the number of successes, in this context the number of product choices. Besides, the logit supposed to be a linear function of the predictors, and is denoted as:
23456 (89) = ;9<*
where, in the model for this research, ; represents a dummy variable for each product attribute and * represents the coefficient of the product attribute. This coefficient can be interpreted as follows: the presence of a product attribute increases or decrease the probability that a product is chosen compared to the reference level of the attribute. The reference level is the absence of that attribute or another level of that product attribute.
In the analysis, first a model will be created for each product category to examine the moderation effect of goals in each stage of the decision process. The model is as follows:
23456 (89) = *+ + *,-, + ⋯ + *0-0 + *,-, ->?@. + ⋯ + *0-0 ->?@.
This model provides insights in the difference in importance of attributes for product choice among customer goals. The terms *,-, and *,-, ->?@ will be compared to test for a significant
Additionally, the willingness to pay (WTP) will be captured by the analysis because the CBC design is viewed as a strong method to measure the WTP. The WTP generates important insights for marketers about customer demand and the composition of existing and new products (Jedidi and Jagpal, 2009). Specifically, it indicates how much a customer is willing to pay for a product. Because the CBC study provides a function that describes how customers evaluate different product attributes, the customer reservation price for a product can be computed (Jedidi and Zhang, 2002). Following Jedidi and Zhang (2002) terminology, the reservation price is the price for which a customer is indifferent between buying and not buying a product. The reservation price R for product profile j can be denoted as:
B(C) =xEaGH αH
where xE describes the attribute levels of product profile j, aGH is the vector of the
associated coefficient for customer i, and αH the price effect. For each product category the price
effect will be computed separately. The so-called ‘exchange rate’ between utility and money will be used to define the value of each unit in utility (Jedidi and Zhang, 2002). As a result, the WTP for each product attribute can be computed (Appendix F).
Exchange rate = aGPQR STHUV
(high price in € − low price in €)
Lastly, a second logit model will be composed to examine the differential effect for each product category. The model is as follows:
23456 (89) = *+ + *,-, + ⋯ *0-0 + *,-, ->?@ + ⋯ *0-0 ->?@ + *,-, -\]^ + ⋯ *0-0 -\]^ + *,-, ->?@-\]^ + ⋯ *0-0 ->?@-\]^
In this model are three-way interaction variables included, because they can provide insights into the difference in importance for the customer goals in the CSF compared to the FPD.
Results
In this section, first a preliminary analysis of the data is provided. Then, a manipulation check is performed and descriptive statistics are discussed. Lastly, the hypotheses are tested.
Preliminary analysis
A check of frequencies was conducted to examine missing values. Cases that were not completely finished or were missing any values are excluded listwise from the data. Unfortunately, a mistake was uncovered in the universal sets which were composed in Rstudio. Some universal sets included two of the same product options, so each double choice one had to be excluded from the universal set in the analysis. As a result, the universal sets were different in the amount of alternatives. Besides, data for the category coffee are excluded from the analysis. Because the subgroups include a low number of respondents (Table 2), the statistical power will miss out and this can results in wrong implications. Preparation of the dataset is discussed in more detail in Appendix C.
Table 2
Participants per category
Category CSF FPD Total 1. Milk 65 117 151 2. Tomatoes 73 120 159 3. Granola 60 108 146 4. Wine 64 117 150 5. Shampoo 62 113 148 5. Coffee 68 32 68 5.1 Coffee beans 19 6 19 5.2 Ground coffee 28 16 28 5.3 Coffee pads 8 3 8 5.4 Coffee cups 13 7 13 Total 72 86 245
Manipulation Checks
In the surveys some questions were added related to the incorporation of enjoyment and environment in grocery decisions, in general and in this experiment. On a scale from 0% to 100%, results show that on average participants incorporate enjoyment in their grocery consumption in general to an extent of 63%, and environment 60,11%.
In order to find out if the enjoyment goal manipulation was successful, a one-sample and an independent sample t-tests are conducted. The one-sample t-test to check for differences in enjoyment incorporation between in general and in this experiment, in the enjoyment condition. And the independent sample t-test to check the difference in enjoyment incorporation in the experiment between the two goal conditions. In the enjoyment condition, enjoyment incorporation in the experiment (M=67.32, SD=22.27) is significantly higher than incorporation in general grocery shopping (M=60.08, SD=19.60), (t(113) = 3.47, p<.01). In addition, enjoyment incorporation in the experiment is significantly higher in the enjoyment condition (M=67.32, SD= 22.27), than in the environment condition (M=58.18, SD=23.17), (t(243) = -3.137, p<.01). Therefore, the manipulation for environment was successful.
Besides, it was checked if the environment goal manipulation was successful. A one-sample t-test shows that in the environment condition, environment incorporation in the experiment (M=71.51, SD= 22.23) is significantly higher than incorporation in general grocery shopping (M=59.16, SD= 24.16), (t(130) = 6.359, p<.01). Moreover, an independent sample t-test shows that environment incorporation in the experiment is significantly higher in the environment condition (M=71.15, SD=22.23) than in the enjoyment condition (M=60.95, SD=26.02), (t(243) = 3.427, p<.01). So, there can be concluded that also the manipulation of the enjoyment goal was successful.
Descriptive Statistics
The total sample of this research counts 245 participants, 66.5% female and 33.5% male. Their households consist at least of one person, with a maximum of nine persons. But, most participants, 84.1%, were part of a household consisting of less than four persons. Their monthly grocery expenditure varied from less than €100,- to more than €500,-. Although, 40% spends between €100,- and €200,- and 26,1% between €200,- and €300,-.
Table 3 - Descriptive Statistics
N Mean Std. Dev. Min Max Scale
Gender 245 0.67 0.473 0 1 0-1
Grocery Expenditure 245 2.6 1.242 1 6 1-6
Household 245 2.13 1.250 1 9 Persons
Appendix D shows the correlation matrices. The effects are tested for each product category separately, therefore a correlation matrix is conducted for each product category. They provide correlation coefficients which show a statistical relationship between the variables. The product attributes for milk, except dairy-free, are correlated with the product choice (p<.01). Low price is the strongest predictor with a positive relationship of r=0.256 to product choice. Apart from the type of wine and the medium price level, also all product attributes of wine are correlated the product choice. Only the high price level and Germany are negatively related to product choice. For tomatoes, all product attributes are correlated with product choice (p<.01). Plastic package (r=-.186), medium price (r=-.065) and high price (r=-.197) are negative correlated, what means that when a product has a plastic package, medium price and/or high price, the product is less often chosen. Some of the attributes of granola are correlated with product choice: muesli&nuts, seeds&nuts, all price levels and organic (p<.01). Also for shampoo only a few of the attributes are correlated with product choice: smooth&shine, all price levels and prevents dandruff (p<.01). For both product categories, low price and high
price are the strongest predictors. Low price has a positive relationship (rgranola=.273, rshampoo=.220) and high price a negative correlation (rgranola=-.248, rshampoo=-.160) with product choice.
Hypotheses Testing
First of all, it is tested if goals moderate the effect of attribute evaluation on product choice. Therefore, logistic regressions are carried out for each product category in both the CSF and the FPD. Results are shown in table 4 to 8. Besides, for each attribute that has a significant effect on the product choice, the willingness to pay (WTP) is computed (table 9).
Regression 1 in table 4 for the category Milk shows that in CSF low price, medium price, and high price are significant less important (* = -1.097, * = -.993, * = -1.006, p <.01) and no fat and organic are more important (* = 1.008, p <.01; * =.424, p <.05) for product inclusion in consideration set in the environment condition than in the enjoyment condition. Similarly, regression 2 shows that in FPD low price, medium price, and high price are significant less important (* = -1.661, * = -1.620, * = -2.310, p <.01) and no fat, dairy free and organic are more important (* = .862, * =.756, * =.424, p <.01) for the final product choice in the environment condition than in the enjoyment condition. These findings are quite logical, because people who want to contribute to better environment make consumption choices that contribute to this, while people who want to attain enjoyment not or to a lesser extent. The willingness to pay contributes to this with the finding that people are willing to pay between €1.15 and €1.82 more for organic milk in the environment condition than in the enjoyment condition, when making a final purchase decision.
For Tomatoes, regression 1 in table 5 shows that the product attributes local farmer and paper package are respectively significant more (* =.439, p =.026) and less important (* = -.760, p =.018) for product inclusion in consideration set in the environment condition than in
the enjoyment condition. People in the enjoyment condition are willing to pay more for a kilo of tomatoes from a local farmer in both stages of their decision process (€0.38 in CSF, €0.57 in FPD), and people in the environment condition are willing to include tomatoes from the local farmer in their consideration set if the price is even €0.14/kilo higher. However, in FPD no package, plastic package, paper package, price low, price medium and organic are significantly different from enjoyment to environment. The price levels and organic are more important (* = 1.057, * = 1.031, * = .753 p <.01) for the final product choice. People in the enjoyment condition are willing to pay €0.47 more for a kilo of tomatoes that is organic, and people in the environment condition are willing to pay another €0.24 on top. Besides, all types of package are less important (* = -1.347, p <.05; * = -1.989, * = -1.873, p <.01) and people’s willingness to pay in the environment condition is less for each type of package than in the enjoyment condition, respectively €-0.43, €-0.63 and €-0.60. This implies that customers who try to attain the environment goal care less about product package than customers who try attain enjoyment. On the one hand this is surprising, because environmental aware customers are perceived as people who care about the package and would prefer no package and paper package over plastic package. On the other hand, customer who want to attain enjoyment may infer from the packaging whether or not the product has the ability to make them enjoy it, and therefor perceive it as more important.
Moreover, the regression for product inclusion in consideration set for Granola shows that price levels are perceived as less important product attributes (* = .566, * = .965, * = -.897, p <.05), and muesli&nuts and gluten free (* = .601, p = .021, * = 1.036, p = .006) as more important attributes for the environment goal than for the enjoyment goal (Table 6). In FPD, also the price levels are less important, but next to gluten free also organic (* = .850, * = 1.036, p <.01) is perceived as a more important attribute for final choice for the environment goal than for the enjoyment goal. People in the environment condition are willing to pay
between €0.32 and €0.55 for organic granola on top of the €0.41 that people in the enjoyment are willing to pay compared to non-organic granola.
The regressions for Wine in table 7 show that the goal moderation has an effect on different product attributes in CSF than in FPD. In CSF, high price is less important and organic is more important (* = -.673, p =.017; * =.842, p =.000). However, the willingness to pay for a bottle of organic wine is only €0.07 higher in the environment condition. In FPD, South Africa and France are less important and medium price is more important for environment than for enjoyment (* = -1.078, p =.001; * = -.1063, p =.017; * = 1.146, p =.16). People in the environment condition are willing to pay between €0.14 and €0.26 less for a bottle of South African wine compared to the WTP of €0.28 of people in the enjoyment condition, compared to a German wine. The same holds for French wine, but people in the enjoyment condition are willing to pay €0.52 more for a French wine than for a German wine.
Lastly, the regressions for Shampoo in table 8 show that in CSF price low, price high and strong aroma are significantly different from enjoyment to environment (* =.590, p =.024; * = -1.374, p =.000; * = -.468, p =.016). These differences imply that customers who want to attain enjoyment care more about the strength of the aroma and about a high price, and less about a low price. So, customers who want to attain enjoyment are willing to pay more for a shampoo with a strong aroma than environmental aware customers. However, the results for the WTP show that the difference is only €0.02. In FPD only strong aroma is significantly different from enjoyment to environment (* = -.612, p <.01). In this stage, the difference in WTP is €0.09. So, environmental aware customers are willing to pay €0.09 less for a shampoo with a strong aroma than customers who want to attain enjoyment who are willing to pay €0.05 more for strong aroma compared to soft aroma.
Altogether, there can be stated that there is a moderation effect of goals in both the CSF and the FPD. A large number of attributes are evaluated differently from the enjoyment goal to
the environment goal in both CSF and FPD. However, the difference in WTP between the goal conditions are quite low for most of the product attributes. This implies that the moderation effect is not extensive. Nevertheless, hypothesis 1 is confirmed.
Additionally, there is tested for differences in consideration set size. Former research shows that the consideration set size is affected by information processing, for instance involvement in the consumption, the consumption context, and familiarity with the consumption context (Howard and Sheth, 1969; Divine, 1995; Aurier and Zaichkowsky, 2000). However, the independent sample t-tests show no significant differences in consideration set size between the goal conditions and between people who participated only in the CSF and people who participated in the CSF and the FPD, neither for each individual product category, nor for all product categories together (Appendix E). It was also checked if people who were more involved in the goal situation (more than 60% incorporation of goal in general grocery shopping), created larger consideration sets than people who were less involved in the goal situation. Again, no significance difference was found.
Also, an additional test is done for the FPD to check for differences in attribute evaluation between participants in the two-stage survey and in the one-stage survey. For each product category the overall logistic regression in FPD was split into two logistic regression, one-stage and two-stage. Results show differences in the signs of the product attribute parameters between the type of stages. This implies that there are differences in product importance. However, most of the parameters are not significant. Therefore, we can infer that there is no statistical difference in goal moderation in FPD between the two-stage and the one-stage.
To test the differential effect between the goal moderation in both stages, for each product category a logistic regression is used, which includes three-way interaction variables CSF x Environment x ‘product attribute’ (table 4-8). When this variable is significant, it implies that the importance of a specific product attribute in the environment conditions varies across CSF and FPD compared to enjoyment in FPD, and also that that the importance of a specific product attribute in CSF varies across environment and enjoyment compared to enjoyment in FPD.
First of all, the overall regression model for milk is statistically significant (_.(30) = 1002.68, p =.000), and explains 23.9% of the variance in product choice (R2 = .239). The three-way interaction terms for high price and organic are significant predictors of product choice (* =1.288, p =.048; * =-.801, p =.018). This means that there is a differential effect for the product attributes high price and organic in the category milk. The interaction effect for high price shows that high price is more important in CSF than in FPD (Figure 5). In CSF high price is more important for environment than for enjoyment, and in FPD high price is more important for enjoyment than for environment. In each stage, medium price and low price are more important for product choice than high price. Besides, the interaction effect for organic shows that organic is more important in FPD than in CSF (Figure 6). In both stages, organic is more important for environment than for enjoyment, but in FPD high price in environment increases the probability that a product is chosen compared to enjoyment to a greater extent than in CSF. Secondly, the overall model for tomatoes is also statistically significant (_.(30) = 1509.26, p =.000), and explains 31.8% of the variance in product choice (R2 =.318). Results in table 4 show that there is a differential effect of the product attributes plastic package and price medium (* = 1.482, * = -1.173, p <.05). The interaction effect for plastic package shows that each type of package is more important in CSF than in FPD, and in each stage more important for enjoyment than for environment (Figure 7). However, plastic package is for enjoyment in FPD slightly more important than no package and in each of the other conditions less important,
while plastic package is in each condition less important than paper package. The interaction effect for medium price shows that in each condition a low price is more important than a medium price compared to a high price (Figure 8). A medium price is most important for the environment goal in FPD, and the least for the enjoyment goal in FPD. CSF is in between them, and medium price is slightly more important for enjoyment than for environment.
Figure 5 – Interaction effect high price Figure 6 – Interaction effect organic
Note: The y-axis represents the probability that a product is chosen with the reference alternative as benchmark.
(See Appendix G for calculations)
Figure 7 – Interaction effect plastic package Figure 8 – Interaction effect medium price
Note: The y-axis represents the probability that a product is chosen with the reference alternative as benchmark.
(See Appendix G for calculations)
0 10 20 30 40 50
Low Price Medium Price High Price
Pr ob . p ro du ct c ho ic e (% ) Enjoyment in FPD Evironment in FPD Enjoyment in CSF Environment in CSF 0 2 4 6 8 10 12 14 16
Not organic Organic
Pr ob. pr oduc t choi ce ( % ) Enjoyment in FPD Evironment in FPD Enjoyment in CSF Environment in CSF 0 1 2 3 4 5 6 7
No package Plastic package Paper package
Pr ob. pr oduc t choi ce ( % ) Enjoyment in FPD Evironment in FPD Enjoyment in CSF Environment in CSF 0 5 10 15 20 25
Price low Price medium Price high
Pr ob. pr oduc t choi ce ( % ) Enjoyment in FPD Evironment in FPD Enjoyment in CSF Environment in CSF
The overall model for granola is also statistically significant (_.(30) = 1019.77, p
=.000), and it explains 27.8% of the variance in product choice (R2 =.278). Besides, there is a differential effect on the moderation of goals on the importance of organic for product choice (* = -.680, p =.05). The interaction effect shows that organic is overall more important in the environment goal than in enjoyment goal (Figure 9). For environment organic is more important in FPD than in CSF, and for enjoyment organic is slightly more important in CSF than in FPD.
Figure 9 – Interaction effect organic
Note: The y-axis represents the probability that a product
is chosen with the reference alternative as benchmark. (See Appendix G for calculations)
Moreover, the model for wine includes three three-way interactions which are statistically significant. The model itself is also statistically significant (_.(30) = 684.39, p=.000) and explains 16.6% of the variance in product choice (R2 =.166). Medium price, South Africa and organic are the product attributes for which a differential effect is found (* = -1.414, p =.01; * = 1.161, p =.004; * =.602, p =.029). The interaction effect for medium price shows that it is more important in CSF than in FPD (Figure 10). In CSF medium price is more important in the enjoyment goal than in the environment goal, while in FPD it is more important in the environment goal than in the enjoyment goal. Besides, the interaction effect for South Africa shows that the importance is only slightly different from the reference alternative
0 5 10 15 20 25
Not organic Organic
Pr ob. pr oduc t choi ce ( % ) Enjoyment in FPD Evironment in FPD Enjoyment in CSF Environment in CSF
Germany, between 0 and 2 percent (Figure 11). The importance is almost the same for enjoyment and environment in CSF and environment in FPD, while the importance is higher for enjoyment in FPD compared to the other conditions. Lastly, the interaction effect for organic shows that organic is more important in FPD than in CSF, although there is only a little difference between enjoyment in FPD and environment in CSF (Figure 12). In both stages, organic is more important for environment than for enjoyment.
Figure 10 – Interaction effect medium price Figure 11 – Interaction effect South Africa
Note: The y-axis represents the probability that a product is chosen with the reference alternative as benchmark.
(See Appendix G for calculations) Figure 12 – Interaction effect organic
Note: The y-axis represents the probability that a product
is chosen with the reference alternative as benchmark. (See Appendix G for calculations)
0 5 10 15 20 25 30 35
Price low Price medium Price high
Pr ob. pr oduc t choi ce ( % ) Enjoyment in FPD Evironment in FPD Enjoyment in CSF Environment in CSF 0 2 4 6 8 10 12
South Africa France Germany
Pr ob. pr oduc t choi ce ( % ) Enjoyment in FPD Evironment in FPD Enjoyment in CSF Environment in CSF 0 0,5 1 1,5 2 2,5
Not organic Organic
Pr ob. pr oduc t choi ce ( % ) Enjoyment in FPD Evironment in FPD Enjoyment in CSF Environment in CSF
However, no differential effect is found for the product category shampoo. The model explains 18.3% of the variance in product choice (R2 =.183) and it is statistically significant
(_.(30) = 354.79, p=.000) but none of the three-way interactions is statistically significant. So,
the differences in attribute importance from the enjoyment goal to the environment goal are not different from the CSF to the FPD.
Altogether, results show that there is a differential effect, but it depends on the product category and the included attributes. For example, the product attribute organic has a differential effect in multiple product categories, but the effect is different for each product category. It can be inferred that the importance of a product attribute per stage of the decision process in relation to the goal depends on the product category. This can be explained by the fact that products are selected as means to attain a consumption goal. Thus not all product categories has the same function and the value of an attributes depends on the product of which it is an attribute. Concludingly, hypothesis 2 is partially confirmed.
Table 4 - Logistic regressions for Milk
Independent Variables (1) Choice in CSF (2) Choice FPD (3) Choice Overall
B SE B SE B SE Low price 1.777** .195 1.829** .268 1.838** .268 Medium price .841** .198 .501 .301 .506 .301 No fat -.175 .171 .200 .199 .201 .199 Low fat .614** .196 1.673** .233 1.677** .233 Dairy free -.438** .144 .002 .149 .002 .149 Organic .422** .149 1.082** .183 1.084** .183
Env. x Low price -1.097** .284 -1.661** .369 -1.651** .369 Env. x Medium price -.993** .255 -1.620** .391 -1.613** .391 Env. x High price -1.006** .324 -2.310** .566 -2.307** .567
Env. x No fat 1.008** .223 .862** .280 .864** .280
Env. x Low fat .332 .263 .140 .348 .141 .348
Env. x Dairy free .317 .187 .756** .213 .757** .214
Env. x Organic .424* .197 1.219** .275 1.221** .275 CSF x Low price 2.002** .326 CSF x Medium price 2.403** .318 CSF x High price 2.071** .433 CSF x No fat -.375 .262 CSF x Low fat -1.066** .304 CSF x Dairy free -.437* .206 CSF x Organic -.664** .236
CSF x Env. x Low price .537 .464
CSF x Env. x Medium price .604 .465
CSF x Env. x High price 1.288* .651
CSF x Env. x No fat .137 .357
CSF x Env. x Low fat .188 .436
CSF x Env. x Dairy free -.441 .283
CSF x Env. x Organic -.801* .337 Gender -.263** .099 0.070 0.116 -.117 .075 Expenditure .142** .041 0.005 0.049 .083** .031 Household -.012 .036 -0.015 0.045 -.021 .028 Constant -2.076** .270 -3.987** 0.402 -4.048** .381 _. R2 358.86** .175 580.19** .278 1002.68** 239
Note: High price and High fat are reference categories
Table 5 - Logistic regressions for Tomatoes
Independent Variables (1) Choice in CSF (2) Choice FPD (3) Choice Overall
B SE B SE B SE Local farmer 1.204** .138 1.798** .198 1.795** .199 No package -.262 .154 -.383* .169 -.382* .169 Plastic package -.760** .175 -.244 .235 -.239 .235 Price Low 1.884** .164 1.909** .252 1.900** .252 Price Medium .736** .185 .191 .280 .185 .280 Organic 1.165** .150 1.466** .206 1.466** .206
Env. x Local Farmer .439* .197 .291 .277 .313 .277
Env. x No package -.256 .334 -1.347* .582 -1.417* .582 Env. x Plastic package -.600 .314 -1.989** .573 -2.068** .572 Env. x Paper package -.760* .321 -1.873** .571 -1.935** .571 Env. x Price low .397 .231 1.057** .377 1.094** .377 Env. x Price medium -.126 .259 1.031** .397 1.048** .397
Env. x Organic .168 .214 .753** .294 .763** .294 CSF x Local Farmer -.598* .242 CSF x No package 1.865** .467 CSF x Plastic package 1.227** .442 CSF x Paper package 1.743** .444 CSF x Price low -.028 .301 CSF x Price medium .546 .336 CSF x Organic -.308 .255
CSF x Env. x Local Farmer .124 .340
CSF x Env. x No package 1.173 .670
CSF x Env. x Plastic package 1.482* .652
CSF x Env. x Paper package 1.190 .654
CSF x Env. x Price low -.698 .442
CSF x Env. x Price medium -1.173* .474
CSF x Env. x Organic -.595 .363 Gender -.426** .093 .046 .112 -.229** .070 Expenditure .061 .035 -.007 .046 .030 .028 Household -.057 .034 .007 .047 -.037 .027 Constant -2.605** .243 -4.618** .420 -4.419** .398 _. R2 884.46** .341 537.28** .259 1509.26** .381
Note: Paper package and High price are reference categories ** significant at .01 level * significant at .05 level
Table 6 – Logistic regressions for Granola
Independent Variables (1) Choice in CSF (2) Choice FPD (3) Choice Overall
B SE B SE B SE
Price low 2.397** .196 2.908** .429 2.914** .429
Price medium 1.515** .195 1.054* .458 1.055* .459
Muesli & Nuts -.498** .184 -.180 .195 -.180 .195
Muesli & Fruit -.092 .170 -.589** .193 -.590** .193
Gluten free .148 .146 -.571** .171 -.572** .172
Organic 1.035** .146 .970** .180 .972** .180
Env. x Price low -.566* .267 -1.174** .274 -1.183** .275 Env. x Price medium -.965** .314 -1.004* .428 -1.023* .428 Env. x Price high -.897* .352 -1.888* .781 -1.993* .780 Env. x Muesli & Nuts .601* .261 .044 .269 .042 .269 Env. x Muesli & Fruit -.407 .252 .007 .277 -.002 .277 Env. x Gluten free .591** .216 .85** .250 .853** .250
Env. x Organic .365 .216 1.036** .272 1.048** .272
CSF x Price low .160 .256
CSF x Price medium 1.147** .352
CSF x Price high .700 .511
CSF x Muesli & Nuts -.313 .268
CSF x Muesli & Fruit .500 .257
CSF x Gluten free .719** .225
CSF x Organic .053 .231
CSF x Env. x Price low .634 .382
CSF x Env. x Price medium .073 .530
CSF x Env. x Price high 1.104 .856
CSF x Env. x Muesli & Nuts .553 .374
CSF x Env. x Muesli & Fruit -.404 .374
CSF x Env. x Gluten free -.264 .330
CSF x Env. x Organic -.680* .347 Gender -.315** .108 .015 .118 -.153* .078 Expenditure .157** .045 -.013 .049 .077 .033 Household -.040 .048 .008 .052 -.021 .035 Constant -2.860** .207 -3.411* .486 -3.463** .467 _. R2 569.67** .298 439.48** .252 1019.77** .278
Note: Nuts & seeds and High price are reference categories ** significant at .01 level * significant at .05 level
Table 7 - Logistic regressions for Wine
Independent Variables (1) Choice in CSF (2) Choice FPD (3) Choice Overall
B SE B SE B SE Red .012 .132 -.185 .154 -.185 .154 Price low 1.071** .201 2.456** .343 2.465** .343 Price medium .544** .175 .706** .270 .716** .270 South Africa .289 .180 1.471** .239 1.472** .239 France 1.224** .198 2.751** .324 2.750** .323 Organic .057 .128 .945** .151 .947** .151 Env. x Red .148 .175 -.003 .222 -.004 .221
Env. x Price low -.192 .239 .695 .394 .677 .393
Env. x Price medium -.290 .284 1.146* .474 1.121* .474
Env. x Price high -.673* .283 -.104 .665 -.161 .664
Env. x South Africa .094 .237 -1.078** .321 -1.067** .320
Env. x France -.378 .265 -1.063* .444 -1.052* .444 Env. x Organic .842** .107 .232 .217 .240 .217 CSF x Red .197 .203 CSF x Price low 1.633** .339 CSF x Price medium 2.856** .402 CSF x Price high 3.028** .480 CSF x South Africa -1.183** .299 CSF x France -1.526** .379 CSF x Organic -.890** .198 CSF x Env. x Red .150 .282
CSF x Env. x Price low -.872 .460
CSF x Env. x Price medium -1.414** .552
CSF x Env. x Price high -.515 .722
CSF x Env. x South Africa 1.161** .398
CSF x Env. x France .674 .517 CSF x Env. x Organic .602* .275 Gender -.112 .089 .193 .106 .017 .068 Expenditure .016 .028 -.006 .044 .007 .024 Household -.015 .028 -.016 .044 -.015 .024 Constant -1.836** .228 -4.992** .461 -4.920** .443 _. R2 240.46** .111 342.45** .181 684.39** .166
Note: High price and Germany are reference categories
Table 8 - Logistic regressions for Shampoo
Independent Variables (1) Choice in CSF (2) Choice FPD (3) Choice Overall
B SE B SE B SE Price low .323* .154 1.432** .256 1.451** .256 Price medium .064 .159 .417 .306 .423 .306 Volume .079 .155 .328 .179 .327 .179 Repair .094 .149 .421* .179 .420* .179 Strong aroma -.188 .126 .376** .142 .372** .142 Dandruff free .796** .134 .584** .162 .589** .162
Env. x Price low .590* .261 .124 .293 .129 .293
Env. x Price medium -.247 .263 -.095 .351 -.080 .351
Env. x Price high -1.374** .297 -.589 .483 -.568 .482
Env. x Volume .227 .237 .262 .256 .262 .256
Env. x Repair -.081 .230 .180 .258 .180 .258
Env. x Strong aroma -.468* .195 -.612** .204 -.609 .204
Env. x Dandruff free .282 .206 .097 .231 .093 .231
CSF x Price low .719** .273 CSF x Price medium 1.488** .299 CSF x Price high 1.848** .348 CSF x Volume -.249 .236 CSF x Repair -.327 .233 CSF x Strong aroma -.560** .190 CSF x Dandruff free .205 .210
CSF x Env. x Price low .441 .392
CSF x Env. x Price medium -.186 .439
CSF x Env. x Price high -.823 .566
CSF x Env. x Volume -.036 .349
CSF x Env. x Repair -.261 .346
CSF x Env. x Strong aroma .142 .282
CSF x Env. x Dandruff free .189 .309
Gender -.057 .098 .065 .109 .019 .072 Expenditure .067 .042 -.017 .047 .025 .031 Household .067 .040 -.006 .047 .032 .030 Constant -1.692** .220 -3.223** .339 -3.389** .319 _. R2 354.79** .183 253.88** .141 643.34** .171
Note: Smooth & Shine and High price are reference categories
