The effects of discount framings on consumer behaviour

The effects of discount framings on

consumer behaviour

1

The effects of discount framings on

consumer behaviour

By

Alexandra S. Djukanovic

University of Groningen

Faculty of Economics and Business

MSc Marketing – Intelligence

Master thesis

Supervisor: dr. J.E.M. van Nierop

Second supervisor: Prof. dr. L.M. Sloot

Author’s address: Menadostraat 10-A, 9715KX, Groningen

Phone number: +31630109119

E-mail: a.s.djukanovic@gmail.com

Student number: 1703439

2

Abstract

This research aims to identify differences between the various discount framings that retailers make use of when offering a price promotion to consumers. Five discount framing have been taken into consideration, namely ‘price off (-€)’, ‘percentage off (-%)’, a ‘from-for’ price, ‘buy one get one free’, and ‘get more pay less’. Furthermore, the interaction effect of product price level, brand preference and consumer demographics between discounts and consumer purchase decisions has been researched. Based on consumer characteristics and preferences several segments are identified. Results show significant as well as insignificant relationships, depending on product category and discount framing.

Preface

3

Contents

2.1.4 Buy one, get one free ... 8

2.1.5 Purchase more, pay less ... 9

2.1.6 No discount... 9

2.2 Quantity of sales ... 9

2.3 Regular price level ... 10

4 4.4 Main effects ... 24 4.4.1 Discount framings ... 24 4.4.2 Price level ... 25 4.4.3 Brand preference ... 26 4.5 Segments ... 31

5 Discussion & conclusion ... 38

6 Limitations & further research ... 40

References ... 41

Appendix A – Interview questions ... 44

Appendix B – Overview of choice sets ... 45

Appendix C – General results from survey ... 46

Appendix D – Overview of the part-worth utilities for ‘brand’ ... 47

Appendix E – Parameter estimates brand preference ... 48

Appendix F – Model fit ... 49

Appendix G – Class sizes ... 50

Appendix H – Part-worth utility levels per latent class ... 51

5

1 Introduction

6 for cognition are less likely to calculate the percentage terms to a monetary value and are thus more susceptible to the framing than consumers with a high need for cognition (Chatterjee et. al., 2000) Hence, discounts can take many forms and can have different influences on consumers.

This paper aims to find these differences between framings and will elaborate on the discussion about the most effective way of framing. As such, the problem statement will be as follows:

To what extent do different forms of framing price discounts influence the purchase decision of consumers?

Five different methods of framing will be analysed, which include price-off (-€), percentage-off (-%), a from-for price, buy one get one free, and get more pay less. All these different framings provide practically the same amount of price gain when an equal level of discount has been chosen. However, they might influence consumers in a different way.

7

2 Theoretical Framework

In their way towards finding the best buy, consumers are being faced with many offerings and discounts. When you walk through a supermarket, or any other store for that matter, you find different methods with which discounts are being framed. Why do stores make use of these various methods? Do consumers actually pay attention to these different framings or is a product being on discount simply enough? Chen, Monroe and Lou (1998) show, that the different methods of discount framing can affect consumer behaviour differently. It is now the question in what way they influence the consumers. Will consumers buy more when a discount in being framed in a certain way? Does it matter, whether consumers perceive a discount in percentages or in euros? If a discount is being framed with a notion to receive ‘one for free’, will more consumers purchase that product as it is in their nature to want as much as possible for the least amount of money? Surely will these discount framings have different effects on consumers. But are there any real consequences by reducing the price or providing an additional unit for free? According to Hardesty & Bearden (2003) price reduction will influence the consumer’s future reference price, while bonus packs (e.g. one plus one free) will avoid these direct price competition effects. Additionally, price discounts can take quality in doubt and alter brand equity (Palazon & Delgado-Ballester, 2009), whereas, a free gift will maintain the perception of product quality and increases the value of the purchase (Darke & Chung, 2005). In this regard both discount framings will be taken into consideration in this research. The framings to be considered will be discussed in detail in the following section.

2.1 Discount framings

Not surprising, the pricing of products is one of the most important factors that determines sales (Krishna et al., 2002). Retailers have many ways to alter their prices; from simple weekly discounts to continuous price reductions, which are leading to price wars (Van Heerde et al., 2008). As mentioned before, retailers provide consumers with many forms of discounts. The most common framings are perhaps the price discounts in percentages and in euros, which you can find in almost any store on a daily basis. Nevertheless, one can find many more variations on these. The framings to be considered within this research will be discussed below.

2.1.1 Percentage off

8 percentage discounts remain a common instrument. It has been argued that percentage discounts work best when product price levels are low (Chen, Monroe & Lou, 1998) and promotional benefit levels are high (Hardesty & Bearden, 2003), which indicates a high value for the consumer.

2.1.2 Euro off

Together with percentage discounts, discounts provided in euros are one of the most common forms of discount in practice (Yin & Dubinsky, 2004). A discount in euros immediately provides an overview of the amount that the consumer will save by purchasing the product. Euro discounts have been studied in comparison with percentage discounts by several researchers (DelVecchio et. al., 2007; Hardesty & Bearden, 2003). It has been argued that euro discounts have a larger effect when product price levels are high (Chen, Monroe & Lou, 1998) as consumer will value a larger euro gain higher compared to a smaller percentage gain (Chatterjee et. al., 2000).

2.1.3 From – For

The ‘from – for’ framing shows two prices, of which the discounted price is being compared to the reference price, i.e. the higher regular price. Lichtenstein et. al. (1991) refer to this framing as a semantic cue. The discounted price should attract more consumers since they encounter the purchase to provide superior value as a consequence of the reduced price (McKechnie et al, 2012), which is seen to be comparatively lower (Della Bitta, Monroe & McGinnis, 1981). The size of the discount should be taken into consideration carefully, as a small price reduction may be perceived as having little to no gain, while a large price reduction might question the quality of the product (Della Bitta, Monroe & McGinnis, 1981).

2.1.4 Buy one, get one free

9 2.1.5 Purchase more, pay less

Similar to the previous discount framing, the ‘purchase more, pay less’ discount is a variation on mixed price bundling. An example of these discounts is ‘get three, pay for only two’. In essence, the discount is the same and the units to be purchased are the same. The only difference here is the perception that one doesn’t receive an extra product, but one has the opportunity to save money. Whereas with a ‘get one free’ discount, one might perceive the free unit as a gain, in this case one might perceive the effects as a reduction in a loss (Chen et. al., 2012). For both the ‘purchase more, pay less’ and the ‘buy one, get one free’ discount framings applies that the consumable nature of the product determines the unit purchase (Li, Sun & Wang, 2007). Hence, retailers should carefully choose products which can be accompanied by these discounts.

2.1.6 No discount

Evidently, products are not always being sold on discount. Most products, in fact, are not on discount. The regular price is an important indicator of quality, cost and perception regarding the product (Monroe, 1973; Gabor & Granger, 1966). Reasons for consumers to purchase these products could vary from brand preference to sticking to the usual list. Hence, a ‘no discount’ option needs to be taken into consideration as well. With regard to this research one might argue that a discounted product will cause a higher benefit to consumers than a regular priced product, as the obtained value will be equal, though, for a lower cost.

Based on the findings above the following hypothesis has been set up.

H1: The presence of a discount (framed as: euro off, percentage off, from-for, buy one get one,

purchase more pay less) will induce more sales than the presence of no discount.

2.2 Quantity of sales

10

2.3 Regular price level

Additionally Gendall et. all. (2006) argue that different ways of framing discounts work differently for different product category price levels. This means that the framing of a discount for low-price products has a different effect on consumers than the same discount framing will have for high-price products. The question remains to what extent the influence differs between the price levels of the products. As mentioned is section 2.1.1. percentage discounts work best for low-priced products. In their study, Chen, Monroe, and Lou (1998) also provide evidence that price discounts work best for high-price products. In the case of discounts versus free items, it has been argued that if the benefit to be received from the promotion is high, it is better to offer it as a price reduction (Palazon & Delgado-Ballester, 2009). In addition, this analysis will include multiple different framings aside price off and percentage off, for which the effect of price level on consumer behaviour is not known. In this regard, the research will include these differences between price levels. However, it is rather unlikely that high-price products will include discounts where multiple units need to be bought, given that they are not daily purchases. Therefore, this research will include differences between the price levels of products on a lower range: approximately 1 euro, 5 euro, and 10-15 euro. The hypothesis based on this issue is as follows:

H2: The price level of a product affects the influence of a discount (framed as: euro off, percentage

off, from-for, buy one get one, purchase more pay less) on consumer’s purchase decisions.

2.4 Brand preference

11 provided. Kim and Chintagunta (2012) find in their research that consumers’ preferences vary across social groups and consumption context. As a consequence, retailers should keep these differences in mind when thinking about promotional activities. Hence, the following hypothesis has been set up.

H3: An increasing brand preference for a product brand affects the influence of a discount (framed

as: euro off, percentage off, from-for, buy one get one, purchase more pay less) on consumer’s purchase decisions of that product brand positively.

2.5 Demographics

Finally, the relationship between discounts and sales can be affected by demographic aspects. As demographics change over time, so does consumer behaviour (Zeithaml, 1985). Based on their behaviour, consumers can possibly be divided into segments. Monroe & Della Bitta (1978) distinguish between quality conscious and price conscious consumers. These differences in characteristics and values can lead to an entirely different purchase decision. As consumer characteristics are partly grounded in their background, it is interesting to see how consumer demographics influence their purchasing behaviour with regard to various forms of discount framings. The following demographics will be discussed; gender, age, income level and household size. With regard to income level, one could argue that the higher the income, the less a consumer will pay attention to the provided discount, and vice versa. It would be interesting to see what the effect will be on the different discount framings. With regard to household size, one could argue that the larger the household size, the more a consumer will pay attention to multiple unit discounts, e.g. buy two get one free. Therefore, the following sub-question will be included:

H4: Demographics variables (i.e. gender, age, income level, and household size) affect the influence

of a discount (framed as: euro off, percentage off, from-for, buy one get one, purchase more pay less) on consumer’s purchase decisions differently.

Based on the outcomes of the above mentioned issues, with regard to the overall effect of discount framings on consumer’s purchase decisions, the main research question will be answered.

2.6 Conceptual model

12 effect of the discount on the purchase decision is dependent on the price level of the product, meaning that the way a low-price product is framed has a different effect than a high-price product will have. Additionally, a consumer’s brand preference is expected to influence its purchase decisions. Hence, the variable ‘brand preference’ will be included within the conceptual model. Moreover, demographic variables are expected to influence the purchases of consumers. To account for these influences the following demographic variables will be included: ‘gender’, ‘age’, ‘income level’, and ‘household size’.

All together, the conceptual model will be as follows;

13

3 Research Design

The following section will include an explanation of the research methods that will be taken and what the plan of analysis will be. In addition the limitations of this research will be taken into consideration.

3.1 Research Method

The research will contain two stages subsequent to one another. The first stage includes a qualitative research, where interviews will be conducted. The results from these interviews will initiate the basis for the succeeding quantitative stage, which will consist out of a conjoint analysis. Elaboration on both methods will follow.

3.1.1 Qualitative interviews

In-depth interviews, based on the direct questioning method, have been conducted with a six respondents. The respondents have been assigned on a random basis, with the goal to provide a deeper understanding on certain shopping behaviour of consumers and their responses towards certain price discounts. This method has been used to identify the underlying aspects which are important for consumers regarding price discounts when making purchasing decisions.

The final sample includes a total of 6 persons, 3 males and 3 females, ranging from the age of 23 to 51. Additionally, the sample included a variety of income levels and household sizes. The settings constituted a face-to-face scenery, where individuals have been interviewed one-on-one. The interviews are semi-structured and of a qualitative nature, where respondents have been asked open questions about their shopping habits and beliefs regarding product attributes in general and what they pay attention to when making their daily grocery shopping. Additionally, they have been asked about discounts in general, their view on discounts and their responses to discounts. Furthermore, respondents have been asked about their preference and behavioural changes regarding discounts. To elaborate on product categories, follow up questions have been posed with regard to shopping habits concerning certain products respondents purchase on sale. Here, the respondents will be asked about their last purchase that was on discounts, which product it was and how many units have been bought. A concept of the interview questions can be found in Appendix A.

14 3.1.2 Results interviews

The main findings from the conducted interviews are:

- When respondents shop on a daily basis, they usually purchase single units of a product. - When respondents shop on a weekly basis, they usually purchase in bulk.

- Respondents make use of multiple unit discounts when it comes to preservables or household items.

- When respondents shop on a daily basis, percentage discounts are preferred.

- Respondents are willing to switch supermarkets in case of particular discounts. This mostly concerns the preferred brand.

3.1.3 Quantitative analysis

Within this stage a questionnaire has been sent out to a panel, provided by Q&A, which is a research and consultancy firm based in the Netherlands. Use of the panel has resulted in an adequate representation of the Dutch population. The number of respondents in the final sample amounts to 763 individuals. Within the questionnaire respondent have first been asked about their demographics, after which they have been asked about their purchasing behaviour regarding grocery shopping in general as well as the selected product categories beer, milk and laundry detergent. Based on these shopping habits, respondent have been selected for subsequent question which are product specific. The remainder of the questionnaire consists out of a Choice-Based Conjoint Analysis. Respondents have been provided with several different choice sets of three alternative product brands. They have been expected to indicate which of the alternatives they would prefer. Additional to the three alternatives, a “none of these” option was provided. After being provided with the choice sets, respondents will be asked a follow up question, based on their previous choices. Here they have been asked to indicate the number of units they are willing to purchase. In the end, respondents have been asked to indicate their preference for the brands, on a scale from 1 to 7, that have been considered in the research.

3.1.4 Choice-based conjoint

The method of analysis includes a choice-based conjoint, which askes respondents to choose a product according to their preferences within a set of given alternatives. The input for the choice-based conjoint analysis will be discussed below.

3.1.5 Choice sets

15 the variety in price level. The category milk includes prices within the lowest price level, the category beer includes prices within the highest price level, and the category laundry detergent includes prices in between those two levels. As can be inferred from the conducted interviews, respondents usually purchase preservable products when it comes to items that are on sale. One respondent answered: “It’s mostly drinks and preservable items I buy multiple units of.” Hence, the choice has been made to include long-lasting milk within the analysis, retaining the characteristic of a daily produce. All three products provide the opportunity to buy bulk, which enables a more equal comparison of the effects of discounts across all product categories.

Within the product category, the choice has been made to include brands that are considered to be equal in price. This means that all brands are A-brands, with higher quality. The prices of the chosen brands are so close to each other, compared to other brands within that same product category, that price should not affect purchase decisions between these brands. The inclusion of the Euroshopper brand, for example, might affect purchase decisions due to its very low price as such, that nothing can be said about the brand preference for that product. Therefore, the included brands have been chosen to resemble one another.

The following products with according brands and prices have been included in the research. Prices were based on real-time prices from Appie.nl.

Table 1 – Overview regular prices per brand

Milk Beer Laundry detergent

Campina €0,97 Heineken €13,39 Omo €6,29 Friesche Vlag €0,93 Grolsch €13,39 Robijn €6,19 Melkunie €0,95 Warsteiner €13,39 Ariel €6,05

For the category milk, the product includes 1 litre long-lasting whole milk. For the category laundry detergents, the product includes a 1,3l or 1,5l flask. For the category beer, the quantity will constitute a 24x 30cl or 24x 33cl crate.

16 free’. This means that three units need to be purchased, which is considered a sound quantity. A decrease to two units would mean a 50% discount, which is a rather large percentage. Also, an increase to four units, as to accommodate a 25% discount, can be seen as a rather large number of units. Hence, a 33% discount, with a 33% discount rate seems most appropriate. Furthermore, a no-discount option is also included.

3.1.6 Conjoint design

The conjoint design will include the attributes ‘brand’ and ‘discount’. ‘Brand’ will have three levels, whereas ‘discount’ will consist out of six levels. This might lead to a ‘number of levels effect’ as the attribute ‘discount’ has twice as much levels compared to ‘brand’, which can lead to consumers perceiving the estimated relative importance of ‘discount’ higher (Hair et. al., 2010).

The design will include the following alternatives for the product category beer:

Table 2 – Overview alternatives within the beer category

Heineken, 24x0,3l, €13,39 Grolsch, 24x0,33l, €13,39 Warsteiner, 24x0,3l, €13,39

From €13,39 for €8,93 From €13,39 for €8,93 From €13,39 for €8,93

33% discount 33% discount 33% discount

€4,46 discount €4,46 discount €4,46 discount

2+1 free 2+1 free 2+1 free

3=2 3=2 3=2

No discount No discount No discount

17 Literature suggest that conjoint analysis should limit itself to approximately 20 choice sets per respondent (Sawtooth Software, 2013) in order to avoid repetition and monotony with respondents. Additionally, the minimum number of choice sets should be one and a half times the number of parameters to be estimated (Sawtooth Software, 2002). Therefore, it has been decided that each respondent received a maximum of 15 choice sets, based on their previous purchases for each product. This means that if a respondent indicated to have purchased all three product categories, beer, milk and laundry detergent, he/she has received 5 choice sets per product category. If the respondent has indicated to only purchase beer, he/she has received 15 choice sets within the beer category. The choice sets for the product categories have been shown in rotating order to the respondents. Meaning that, in the case of multiple products, respondents have been shown a choice set of one product category followed by a choice set of another product category. This will decrease the loss of respondents’ attention as choice sets will be rated. Furthermore, the sequence in which choice sets have been shown has also been rotated among respondents. As such, consecutive respondents will receive a different version of choice sets.

The choice sets provided to respondents will be shown with pictures of the products. See an example of the pictures for laundry detergent in Figure 2. This has been taken place as a simple representation of a product shelf, where the different brands in one product category have been portrayed next to each other in a 2 by 2 grid. Respondents were asked to visualize a visit to the supermarket, with the intention to buy the products in question and base their current decisions on their daily shopping behaviors, in order to keep a realistic mind-set as possible.

18 3.1.7 Pre-test

Before the final survey has been sent out, a small pre-test has taken place. This pre-test has been conducted among 41 respondents from the panel of Q&A. The test has been used to check for irregularities within the survey. Based on these results some amendments have taken place. It turned out that the questions on product categories led to some confusion. Hence, the product categories have been specified with more detail. This means, instead of listing the categories as beer, milk and laundry detergent, it was changed to a crate of beer, long-lasting milk and liquid laundry detergent. Furthermore, the indicated time for the survey was approximately ten minutes. Nevertheless, the results showed that an average of six minutes was needed for respondents to fill out the questionnaire. This has been altered in the final survey. Noteworthy is also the few respondents who indicated that the method of asking the questions, the choice based conjoint, was bothersome and too long. Nevertheless, the length of the questionnaire remained the same, at 15 choice sets per respondent, whilst being aware of its monotonous nature.

3.2 Data collection

The quantitative data was collected though an online survey. This survey has been sent out to a panel provided by Q&A, Research and Consultancy. The data will include an adequate representation of the Dutch population, hence improving the reliability and the quality of this research. The following section will discuss the tests to be used in analysis.

3.2.1 Plan of Analysis

The results from the conjoint analysis have been analysed in two ways. IBM SPSS 19 will be used to analyse general results from the data, which will include cross tables. Furthermore a multinomial regression analysis will take place, from which basic relationship between variables can be identified. This will be explained later on. Additionally, Latent Gold 4.5, software, will be used to analyse the data received from the choice based conjoint analysis. The data will be edited according to the existence of extreme responses. Extreme responses refer to the respondents who have always indicated the same option within each choice set i.e. always the ‘none of these’ option. The fit of the model will be investigated based on the measures of the CAIC (Consistent Akaike Information Criterion) and the R². The CAIC indicates the way in which the observed covariance matrix differs from the expected covariance matrix, therewith, adjusts the model fit depending on sample size (Nylund et. al, 2007).

19 more sales compared to no discounts, will be analysed based on the part-worths of the discount framings. These will indicate the utility preference, the obtained value when a particular condition occurs (Hair et. al., 2010), of respondents regarding the discount methods of which we can infer that a higher preference will lead to more sales. In this way, the attribute importance can be established. The second hypothesis, regarding price level, will be analysed by comparing the part-worth utility levels for the discount framings between the three products, beer, milk and laundry detergent, which all represent a different range of price levels. In order to make any conclusion about the significance of their differences a T-test needs to be conducted. The third hypothesis, regarding an increase in brand preference, will be researched by conducting a multinomial logistic regression analysis in IBM SPSS 19, as mentioned before. The dependent variable will be purchase choice, and the independent variables will include the survey questions regarding brand preference, which were reported on a 1-7 scale. Finally, the last hypothesis regarding demographic characteristics will be analysed by means of the latent class function in Latent Gold 4.5, which divides the sample into separate classes based on choice behaviour and characteristics. After these classes have been established another look will be taken at the first hypothesis regarding the influence of discounts, to see whether the results obtained from the analysis also hold in different classes.

3.2.2 Limitations

20

4 Results

The following section will provide an overview of the data and an analysis on the results that have been obtained from the research.

4.1 The sample

The total number of responses received amounts to 852, of which 89 (10%) have indicated to never purchase the respective products. The remainder of the sample (90%) does purchase at least one of the respective products, and has been qualified for further research. The number of respondents in the final sample amounts to 763 individuals, 361 males (47%) and 402 females (53%). The average age is approximately 48 years. The average household size is approximately 2,5 persons. The gross annual income level lies between €30.000 and €40.000. More detailed results of the sample can be found in Appendix C.

4.2 Extreme responses

The data has been checked for extreme responses. With extreme responses are meant respondent who have always selected one singular option. In total a number of 150 respondents (20%) have always selected the same option, of which 89 respondents have selected the ‘none of these’ option continuously. These respondents have been deleted from the dataset. The other respondents, which have continuously chosen one brand, have also indicated a higher preference for that chosen brand in many cases. As such, these responses have been retained in the analysis based on brand preference.

21 purchase the product, without actually buying the quantity needed to receive the discount. This is being referred to as the ‘discount communication effect’.

4.3 General findings

The following section will provide an overview of the general finding obtained from the data. These include frequencies and crosstabs with regard to consumers’ shopping habits and brand categories. 4.3.1 Shopping habits

Table 3 presents the results of consumers’ visits to supermarkets regarding grocery shopping. More than half of the respondents (53%) go shopping once or twice a week. Only a small number of respondents (5%) go shopping for groceries every day.

Table 3 – Shopping habits 1-2 times a week 53%

3-4 times a week 34% 5-6 times a week 8%

Every day 5%

Of these respondents, 47% has purchased beer, 75% has purchased milk and 10% has purchased laundry detergent before. Table 4 shows these differences per gender. Noticeable is the large difference between males and females when it comes to the purchase of milk and laundry detergent. More females purchase milk, whereas twice as many males purchase laundry detergent.

Table 4 – Category purchase

Total Male Female

Beer 47% 49% 46%

Milk 75% 68% 81%

Laundry detergent 10% 14% 7%

22 Table 5 – Frequency of category purchase

Beer Milk Laundry detergent

Daily 0% 3% 0%

Weekly 16% 61% 4%

Monthly 35% 24% 51%

Less than monthly 44% 11% 42%

Other 5% 1% 2%

4.3.2 Brand preference

Consumers can have preference for a product or a brand. In the following section a closer look will be taken at the brand preferences of respondents per product category. The preference can be analysed in two ways. First, based on the Likert-scale questions from the summary an average level of brand preference can be determined for each brand. Second, the function of attribute importance in Latent Gold will show part-worth utilities per brand. The results from both methods will be compared to each other to determine whether results are constant.

Beer

Table 6 (Appendix D) shows the part-worth utility for each brand within the aggregate level, which includes overall results represented by only one class. The results are significant (P=1,00E-14) and show that Grolsch has the highest utility level (0,1429) compared to the other brands.

Respondents have indicated their preference for the brand on a scale from 1-7. The graph below (Figure 3) shows the results from these preferences. Grolsch has the highest preference with an average of 4,2, followed by Heineken and Warsteiner with an average rating of 3,86 and 3,34 respectively. These results correspond with the results in table 6 (Appendix D) obtained from Latent Gold. 3,86 4,20 3,34 0 1 2 3 4 5

Heineken

Grolsch

Warsteiner

23

Milk

Table 7 (Appendix D) shows the part-worth utility for each brand within the aggregate level, which includes overall results represented by only one class. The results are significant (P=3,00E-55) and show that Friesche Vlag has the highest utility level (0,00) compared to the other brands.

Respondents have indicated their preference for the brand on a scale from 1-7. The graph below (Figure 4) shows the results from these preferences.

Friesche Vlag has the highest preference with an average of 4,35, followed by Campina and Melkunie with an average rating of 4,26 and 3,73 respectively. These results are in line with the results in table 7 (Appendix D).

Laundry Detergent

Table 8 (Appendix D) shows the part-worth utility for each brand within the aggregate level, which includes overall results represented by only one class. The results are significant (P=8,60E-71) and show that Ariel has the highest utility level (0,8902) compared to the other brands. This difference is small compared to the utility level of Robijn (0,8889).

Respondents have indicated their preference for the brand on a scale from 1-7. The graph below (Figure 5) shows the results from these preferences.

4,35 3,73 4,26 1 2 3 4 5

Friesche Vlag

Melkunie

Campina

Figure 4 – brand preference ‘milk’

3,52 4,47 _4,37 1 2 3 4 5

Omo

Ariel

Robijn

24 Ariel has the highest preference with an average of 4,47, followed by Robijn with an average rating of 4,37. Omo has the least preference with an average of 3,52. These results are in line with the results in table 8 (Appendix D).

4.4 Main effects

The following section will include an analysis of the results obtained regarding the influence of discount framings on consumers’ purchase decisions. Furthermore, results will be provided on the influence of price level and brand preference on the purchase decisions of consumers. In addition, the differences between consumer characteristics, i.e. demographics, and purchase decisions will be discussed.

4.4.1 Discount framings

It has been argued that the presence of a discount (framed as: euro off, percentage off, from-for, buy one get one, purchase more pay less) will induce more sales than the presence of no discount. The results regarding this issue will be discussed below.

Beer

Table 9 shows the part-worth utility for each discount within the aggregate level, which includes overall results represented by only one class. The part-worth utility that is highest in value has the largest preference. The results are significant (p=0,0095) and show that the ‘2+1 free’ discount has the highest utility level (0,2531), whereas, the ‘euro off’ discount has the lowest utility level (-0,1143). Remarkably, the presence of no discount has a higher utility level than the ‘euro off’ and ‘percentage off’ discounts.

Table 9 – Attribute levels ‘beer’ Table 10 – Attribute levels ‘milk’

Attributes Class1 Wald p-value

No discount 0 15,2 0,0095 3=2 0,09 Euro off -0,11 2+1 0,25 Percentage off -0,06 From/for 0,01 Milk

Table 10 shows the part-worth utility for each discount within the aggregate level. The results are significant (p=0,018) and show that the ‘percentage off’ discount has the highest utility level (0,1537), whereas the ‘2+1 free’ discount has the lowest utility level (-0,2051). As with the beer

Attributes Class1 Wald p-value

25 category, it is remarkable to see that the presence of no discount attains a higher utility level than certain other discounts.

Laundry Detergent

Table 11 shows the part-worth utility for each discount within the aggregate level. The results are highly significant (p=2,70E-37) and show that the ‘from/for’ discount has the highest utility level (1,123), whereas the presence of no discount has the lowest utility level (0,00). Furthermore, all discount framings create higher utility levels than the presence of no discount.

Table 11 – Attribute levels ‘laundry detergent’

Attributes Class1 Wald p-value

No discount 0 181,401 2,70E-37 3=2 0,9 Euro off 0,9 2+1 0,95 Percentage off 0,96 From/for 1,12

From the results we can conclude that the presence of a discount will not always induce more sales than the presence of no discount. This is only the case in the laundry detergent category. Hence, H1 is partially supported.

4.4.2 Price level

It has been argued that the price level of products affects the influence of discount framing on consumers’ purchasing decisions. The results regarding this issue will be discussed below.

In this research milk represents the lower price level, laundry detergent represents the middle price level, and beer represents the higher price level. The part-worth utility levels for each particular discount per product category needs to be compared in pairs. This holds that for example the effect of a ‘ThreeIsTwo’ discount will be compared between the categories milk and laundry detergent. Subsequently the comparison will be made between milk and beer, and beer and laundry detergent. In order to find significant differences between these groups a T-test needs to be conducted based on the individual z-values. The values for the T-test will be derived from the following formula:

26 variations between milk and laundry detergent are significantly different from one another. This hold for all five discount formats. Additionally, the variations between laundry detergent and beer are also significantly different from each other, which again holds for all five discount formats. On the other hand, the only significant difference found between milk and beer concerns the ‘TwoPlusOne’ discount format. Other framings between these two product categories are not found to have significant differences. The positive (negative) signs attached to the values indicate that the influence of the discount on the first product category mentioned in the header is higher (lower) than the second product category. In the case of the comparison between milk and laundry detergent it can be said that the influence of all discount framings have a larger effect on laundry detergent than on milk. This is in line with the expectations that the higher the price level, the higher the perceived value for consumer is. Nevertheless, when comparing laundry detergent with beer, it can be concluded that the influence of all discount framings have a larger effect on laundry detergent. This is not in line with the expectations.

Table 12 – Product category difference

Milk-Laundry Detergent Milk-Beer Laundry Detergent-Beer

3=2 -6,84** -1,18 5,83** Euro off -6,78** 0,52 7,26** 2+1 -7,90** -2,92** 5,07** Percentage off -5,74** 1,42 7,31** From/For -7,26** 0,59 7,94** Df 937 711 920

** Significant at the 99% confidence level

Based on these results, it can be concluded that in some instances the price level of products does influence the effects of discount framing on consumers’ purchase decisions. Hence, H2 is partially supported.

4.4.3 Brand preference

This research has argued that brand preference affects the influence of discount framing on consumers’ purchasing decisions. In order to analyse this effect a multinomial logistic regression has been conducted with the discount per brand as the dependent variable. The dependent variable is categorical of nature and will contain 19 levels (6 discounts x 3 brands + 1 ‘none’ option), as such parameters will be estimated for each stage, i.e. the choice for Heineken when no discount is present or the choice for Heineken when a ‘ThreeIsTwo’ discount is present. This relationship can be represented by the following formula:

27 P = the probability that brand i will be chosen when discount j is provided

C = the constant

= the parameter of preference for brand i when discount j is provided The results regarding this issue will be discussed below.

Beer

The variance is not very well explained in this model; Cox and Snell 0,03 and Nagelkerke 0,031. Yet the model seems to be significant (p=,014). Table 13 (Appendix E) shows the parameter estimates regarding the effect of brand preference on purchase choice. The results show that, apart from the effect of PreferenceGrolsch on GrolschTwoPlusOne and PrefrenceWarsteiner on WarsteinerPercentageOff, none of the effects are significant. Hence, no statements can be made regarding their influence. On the other hand, significant results were found for the variables PreferenceGrolsch and PreferenceWarsteiner (p=0,01 and p=0,00 respectively). The effects of brand preference for Grolsch show a positive relationship with GrolschTwoPlusOne (B=0,11, p=0,01) and the effects of brand preference for Warsteiner show a positive relationship with WarsteinerPercentageOff (B=0,25, p=0,00). Hence, an increase in the preference for Grolsch will result in an increase of purchase choice for Grolsch when a ‘two plus one free’ discount is provided. The same holds for Warsteiner, an increase in the preference for Warsteiner will result in an increased purchase choice for Warsteiner when a ‘percentage off’ discount is provided.

Milk

The model shows significant results (p=,000) with the following pseudo R squares: Cox and Snell 0,446 and Nagelkerke 0,449. Table 14 shows the parameter estimates regarding the effect of brand preference on purchase choice for the milk category. The results show that the effects of brand preference for FriescheVlag are significant (p=0,00) and positively related to purchase choice of FriescheVlag milk regarding all discount formats (ThreeIsTwo (B=0,66), EuroOff (B=0,55), TwoPlusOne (B=0,56), PercentageOff (B=0,54), FromFor (B=0,56)) including the ‘no discount’ option (B=0,53). Hence, an increase in the brand preference for FriescheVlag will result in an increase of purchase choice for the brand FriescheVlag, regarding of whether a discount is being provided or not.

28 increase in the brand preference for Melkunie will result in an increase of purchase choice for the brand Melkunie, regarding of whether a discount is being provided or not.

Moreover, the results show similar effects for the brand Campina. The effects of brand preference for Campina show a significant (p=0,00) and positive relationship to the purchase choice of Campina regarding all discount formats (ThreeIsTwo (B=0,70), EuroOff (B=0,81), TwoPlusOne (B=0,86), PercentageOff (B=0,86), FromFor (B=0,72)) including the ‘no discount’ option (B=0,69). Hence, an increase in the brand preference for Campina will result in an increase of purchase choice for the brand Campina, regarding of whether a discount is being provided or not.

Additionally, the results show some significant effects for the preference of one brand and the purchase choice of a second brand. For example, the preference of Melkunie also shows positive correlation with the purchase choice of FriescheVlag regarding all discount formats. Though, this effect is smaller compared to the preference of FriescheVlag. The preference of Campina on the other hand shows a negative correlation with four of the discount formats regarding the purchase choice of Friesche Vlag. It could be argued that consumers with a preference for Melkunie are more likely to purchase FriescheVlag product as well, whereas consumers with a preference for Campina are less likely to purchase FriescheVlag products.

Furthermore, significant (p=0,03 and p=0,00) results have been found for the preference of FriescheVlag with two discount formats. The preference for FriescheVlag shows a negative relationship with the purchase choice of Melkunie when the discount formats ‘ThreeIsTwo’ and ‘EuroOff’ are provided (B=-0,23 and B=-0,39 respectively). It could be argued that consumers with a preference for FriescheVlag milk are less likely to purchase the Melkunie brand.

Finally, the preference for FriescheVlag milk and the preference for Melkunie shows significant effects concerning the purchase choice of Campina with regard to all discount formats. FriescheVlag shows a negative correlation with all discount formats including the ‘no discount’ option, whereas Melkunie shows positive correlations with all of these. It could be argued that consumers with a preference for FriescheVlag are less likely to purchase Campina, while consumers with a preference for Melkunie are also likely to purchase Campina milk.

29 purchase Melkunie. In addition, it could be argued that Melkunie has a lower brand equity, as consumers with a preference for Melkunie are also likely to purchase Campina and FriescheVlag.

Table 14 – parameter estimates for brand preference ‘milk’

Label Intercept Preference FriescheVlag Preference Melkunie Preference Campina FriescheVlag NoDiscount -3,52** 0,53** 0,23** -0,16* FriescheVlag ThreeIsTwo -4,33** 0,66** 0,22* -0,23** FriescheVlag EuroOff -3,58** 0,55** 0,25** -0,17* FriescheVlag TwoPlusOne -4,29** 0,56** 0,24** -0,09 FriescheVlag PercentageOff -3,33** 0,54** 0,24** -0,23** FriescheVlag FromFor -3,89** 0,56** 0,23** -0,13 Melkunie NoDiscounnt -4,53** -0,15 0,8** 0 Melkunie ThreeIsTwo -4,86** -0,23* 0,88** 0,06 Melkunie EuroOff -4,37** -0,39** 0,86** 0,06 Melkunie TwoPlusOne -5,56** 0,09 0,8** -0,13 Melkunie PercentageOff -4,51** -0,13 0,71** 0,05 Melkunie FromFor -4,72** -0,05 0,8** -0,08 Campina NoDiscount -5,38** -0,32** 0,41** 0,69** Campina ThreeIsTwo -5,61** -0,34** 0,51** 0,7** Campina EuroOff -5,75** -0,22** 0,37** 0,81** Campina TwoPlusOne -6,53** -0,2* 0,38** 0,86** Campina PercentageOff -6,13** -0,28** 0,41** 0,86** Campina FromFor -5,49** -0,19* 0,39** 0,72**

* Significant on the 95% confidence level ** Significant on the 99% confidence level

Laundry Detergent

30 Furthermore, similar results can be found for the brand Ariel. Results show significant positive effects (p=0,00) of brand preference for Ariel on purchase choice regarding all forms of discount provided for the brand Ariel (ThreeIsTwo (B=0,70), EuroOff (B=0,73), TwoPlusOne (B=0,79), PercentageOff (B=0,68), FromFor (B=0,75)) including the ‘no discount’ option (B=0,84). Hence, an increase in the brand preference for Ariel will result in an increase of purchase choice for the brand Ariel, regarding of whether a discount is being provided or not.

Moreover, results for Robijn follow a similar pattern, showing significant (p=0,00) positive effect for brand preference of Robijn on the purchase choice for Robijn. This holds for all discount formats (ThreeIsTwo (B=0,81), EuroOff (B=0,76), TwoPlusOne (B=0,84), PercentageOff (B=0,78), FromFor (B=0,72)) including the ‘no discount’ option (B=0,85). Hence, an increase in the brand preference for Robijn will result in an increase of purchase choice for the brand Robijn, regarding of whether a discount is being provided or not.

In addition, the results show some significant effect for the preference of one brand and the purchase choice of a second brand. When it comes to the purchase choice of Omo, significant results have been found regarding the preference of Robijn for all discount formats, except for the ‘no discount’ option. All results show a positive correlation between the preference of Robijn and the purchase choice of Omo, indicating that a consumer with a higher preference for Robijn is also likely to purchase Omo when a discount is being provided. The preference for Ariel shows only a significant negative relationship with the purchase choice for Omo when the discount format ‘ThreeIsTwo’ or ‘PercentageOff’ is provided, indicating that consumers with an increasing preference for Ariel are less likely to purchase Omo when the abovementioned discount are provided.

Also, significant results have been found for the preference of Omo with the purchase choice of Ariel regarding four discount formats (ThreeIsTwo, EuroOff, PercentageOff, and FromFor). All these results show a positive correlation, indicating that consumers with an increasing preference for Omo are also likely to purchase Ariel when the abovementioned discounts are being provided. Significant results have been found for the preference of Robijn with the purchase choice of Ariel regarding the discount formats ‘TwoPlusOne’ and ‘PercentageOff’, which indicate a positive correlation. It can be argued that consumers with an increasing preference for Robijn are also likely to purchase Ariel when the aforementioned discounts are provided.

31 between the preference for Ariel and the discount format ‘FromFor’, which shows a non-significant result. Omo shows a positive correlation with all discount formats including the ‘no discount’ option, whereas Ariel shows negative correlations with all remaining discount formats. It could be argued that consumers with a preference for Omo are also likely to purchase Robijn, while consumers with a preference for Ariel are less likely to purchase Robijn.

To summarize, it could be argued that Ariel has a stronger brand equity than Omo and Robijn, as consumers with an increasing preference for Ariel are less likely to purchase the other brands, whereas consumers with an increasing preference for Omo and Robijn are also likely to purchase Ariel. In addition, consumers with a preference for Robijn are also likely to purchase Omo and vice versa.

Based on the above results it can be concluded that in some instances brand preference does indeed influence the effect of discount framings on consumers’ purchase decisions. Hence, H3 is partially supported.

4.5 Segments

Several researchers have argued that consumer differ in their choices, see section 2.5. As such, this research argues that their characteristics, based on demographics, will influence the effect of discount framings on consumers’ purchase decisions. The results regarding this issue will be discussed below.

First, segmentation will take place based on the latent class function in Latent Gold 4.5. For each product category the results will be extracted for one to eight classes. Based on the model fit, one solution will be chosen regarding the optimal number of classes per product category. Following, the classes will be compared and differences will be identified.

Beer

Table 16 (Appendix F) shows the model fit for one to eight classes in the beer category. The CAIC value is lowest (3450,253) for a model with four classes. The corresponding R² is 0,5885.

32 Based on these results, a model with four classes will be retained.

Table 18 shows the importance of the attributes brand and discount. Interestingly, all classes place a higher importance toward brand, except for class 3, which slightly prefers discount over brand.

Table 18 – Attribute importance

Maximum Class1 Class2 Class3 Class4

Brand 2,6128 2,7642 1,0871 2,9599 Discount 1,4224 0,8246 1,3818 0,8

None 0 0 0 0

Relative Class1 Class2 Class3 Class4

Brand 0,6475 0,7702 0,4403 0,7872 Discount 0,3525 0,2298 0,5597 0,2128

None 0 0 0 0

33 Table 20 (Appendix I) show the demographics of the latent classes. Class 1 exists mainly out of males (61,8%), and consumers within the range of 30-59 years (0,59%). Most households (40,5%) contain two persons and the income level (18,4%) lies within the €20.000 - €29.999 range. Class 2 exists out of slightly more males than females (51,4% and 48,6% respectively). The age of consumers has been nicely distributed over all categories. Most of the consumers (42,9%) make up a household of two persons, and 22,8% obtain an income level between the range of €20.000 - €29.999. Class 3 exists out of slightly more males (56,3%) than females (43,6%) as well. In addition, this class is increasingly represented by elderly people, of which 36% are above 60+ years. Most household sizes (45,5%) include two persons, and the most common income level lies between the range of €20.000 - €29.999. Finally, class 4 includes the most males (69,8%) and consumers within the range of 45-59 years (44,2%). The most common household size includes two persons (44,18%). Additionally, this class contains consumers with a slightly higher income level compared to other classes. 24% of the consumers within this class have an income level between the range of €30.000-€39.999. This last is in line with the earlier results from table 18, which indicates that the utility derived from the brand is much higher than the utility derived from a discount. This difference is also higher compared to other classes that prefer brand over discount.

Milk

Table 21 (Appendix F) shows the model fit for one to eight classes in the milk category. The CAIC value is lowest (3401,0725) for a model with four classes. A lower CAIC value indicates a better fit of the model. The corresponding R² is 0,5946.

Table 22 (Appendix G) shows the class sizes for each of the models from one to eight classes in the milk category. In case of four classes, the class sizes are not equal in size. There in one class representing over 40% of the total respondents, while the other classes represent about 27%, 19% and 12%. A change to a model with five classes evens out these differences by splitting the largest class. Nevertheless, the last class becomes one which represents about 7% of the total sample, which is very small. Furthermore, the table shows the prediction error for each of the models from one to eight classes in the milk category. Prediction error is lowest (0,180) in the case of four classes. An increase to five classes and a decrease to three classes will increase the prediction error.

34 Table 23 shows the importance of the attributes brand and discount for each latent class. Class 1, class 3 and class 4 derive a higher utility from the brand than from a discount, whereas class 2 derives a slightly higher utility from the discount.

Table 23 – Attribute importance

Maximum Class1 Class2 Class3 Class4

Brand 2,3897 0,9704 3,0393 1,323 Discount 0,5271 0,9916 1,0989 0,8224

None 0 0 0 0

Relative Class1 Class2 Class3 Class4

Brand 0,8193 0,4946 0,7344 0,6167 Discount 0,1807 0,5054 0,2656 0,3833

None 0 0 0 0

35 Table 25 (Appendix I) shows the demographics of the latent classes. Class 1 is equally distributed with males (50,8%) and females (49,2%), of which the largest segment (31%) is within the range of 45-59 years old. The most common household size (34,1%) consists out of two persons, and the most common income level (26,7%) lays between €30.000 - €39.999. Class 2 and Class 3 include slightly more males (54,8% for both) than females (45,2% for both). The largest group represents the age group between 45-59 years in both class 2 and class 3 (30,1% and 37,5% respectively). Additionally, the most common household size in both classes includes two persons (40,3% and 49,2% respectively) and the most common income level in both classes lays between €30.000 - €39.999 (20,7% and 22,6% respectively). What is also noteworthy, 16,1% of the consumers in class 3 have an income level of more than €70.000. Finally, class 4 consists out of the most males (68,9%). The largest group of consumers (31,7%) are within the ages of 45-59 years and the most common household size consists out of 2 persons. Also, the most common income level lays between €20.000 - €29.999.

Laundry Detergent

Table 26 (Appendix F) shows the model fit for one to eight classes in the laundry detergent category. The CAIC value is lowest (6223,3897) for a model with six classes. The corresponding R² is 0,5977. A model with six classes can be too much, therefore, another look will be taken at other possibilities with a lower number of classes. Reducing the model to five or even four classes will not lead to a large loss of model fit. However, reducing the model to three or even two classes reduces the model fit with a substantial amount. In addition, a reduction to five or four classes, will lead to an R² of 0,5652 and 0,5286 respectively.

Table 27 (Appendix G) shows the class sizes for each of the models from one to eight classes in the laundry detergent category. In the case of six or five classes, multiple classes come to existence which represent only a small percentage of the sample. A reduction to four classes evens out these differences, including two classes which are larger in size (33,3% and 31,9%) and two classes which are smaller in size (17,6% and 17,3%). Furthermore, the table shows the prediction error for each of the models from one to eight classes in the laundry detergent category. Prediction error reduces as the number of classes increase. Prediction error is lowest (0,186) in the case of seven classes. When reducing classes, the prediction error slightly increases for each step. Nevertheless, in the case of four classes the prediction error is still relatively small (0,228).

36 Table 28 shows the importance of the attributes brand and discount for each latent class. Class 1 derives a higher utility from brand compared to discount, whereas all three other classes derive a higher utility from discount than from brand. When it comes to class 2, this different is rather small though.

Table 28 – Attribute importance

Maximum Class1 Class2 Class3 Class4

Brand 3,3091 2,6385 0,7437 0,9605 Discount 2,0957 2,6949 2,7395 1,8462

None 0 0 0 0

Relative Class1 Class2 Class3 Class4

Brand 0,6123 0,4947 0,2135 0,3422 Discount 0,3877 0,5053 0,7865 0,6578

None 0 0 0 0

37 Table 30 (Appendix I) show the demographics of the latent classes. Class 1 includes slightly more males (53%) than females (47%). The majority within this class (65%) is over 45 years old. Also, the most common household size (38,5%) includes two persons and the most common income level (27,4%) ranges from €20.000 - €29.999. Class 2 is equally divided across males and females (50,3% and 49,7% respectively). The majority (56,4%) is older than 45 years and 43,9% of the households include two persons. Additionally, the two most common income levels range between €40.000 - €49.999 and €50.000 - €59.999 (19,1% and 18,7% respectively), which is fairly higher than in other classes. Class 3 consist out of slightly more females (57,1%) than males (42,9%), of which the largest group (42,1%) is over 60 years old. The majority (50,3%) is part of a household that consists out of two persons. Furthermore, the most common income level (36,2%) ranges from €30.000 - €39.999. Finally, class 4 consists out of slightly more males (52,3%) than females (47,7%). The ages within this class are equally distributed, with the largest group (29,4%) falling in the age level of 45-59 years old. Additionally, the most common household size (33,7%) includes 2 persons, and the most common income level (23,2%) ranges from €20.000 - €29.999. Also noteworthy, 18,1% of the consumers in class 4 have an income within the range of €60.000 - €69.999.

38

5 Discussion & conclusion

The goal of this paper is to identify significant results regarding the effect of discount framings on consumer purchase decisions within three product categories: i.e. milk, laundry detergent and beer. The discount framings that have been taken into consideration are: euro off, percentage off, from-for, buy one get one, and purchase more pay less. Along the way this paper has reviewed several aspects that can have an influence on this effect, which are product price level, brand preference and consumer demographics.

The purchase frequency of the products differs per category. Results show that milk is bought most frequently, with 61% of the respondents making weekly purchases. Beer and laundry detergent are bought on a monthly or less than monthly basis. This is as expected due to the nature of the products. As milk is a daily used product, this category is purchased most often. Beer and laundry detergent might also be used on a daily basis, however, one purchased unit includes multiple uses.

Within the beer category the brand Grolsch has the highest preference, followed by Heineken and Warsteiner. Within the milk category the brand FriescheVlag has the highest preference among respondents, closely followed by the brand Campina. Mulkunie is the least preferred brand. Within the laundry detergent category the brand Ariel is most preferred, with only a small difference with Robijn, which is number two. Omo is the least preferred brand among respondents. These results were found in both the regression analysis, with the likert-scale based questions from the survey, and the attribute importance, derived from the choice based conjoint analysis. The consistency found in both methods adds to the reliability of the analysis.

39 dependent? Or do the consumers not perceive the discount framings as substantial. One justification could perhaps be found within the manner of the research. Respondents were asked to visualize their visit to the store where only three brands were present. Nevertheless, they might still have another brand in mind when making a choice. This would then lead to numerous ‘no purchase’ options.

Additionally, it has been argued that a different price level will affect the influence of the discount framings on consumers’ purchasing decisions. Results show that there are significant differences between the milk and laundry detergent and between laundry detergent and beer. However, only one significant result has been found between milk and beer, which concerns the ‘TwoPlusOne’ discount. As such, H2 is partially supported. Nevertheless, one would assume an even larger difference between the milk and beer category due to its larger price difference. These mixed results indicate that the differences might not be due to price level differences in the first instance, but perhaps due to some other reasons. Other factors that can cause these differences could be the variation in product category. While milk and beer are beverages, laundry detergent is a completely different product from these. The different nature of these products might be a stronger influential, thus, overshadowing the effect of the different price levels.

Furthermore, this research has taken a focus on consumers’ brand preference and its possible effects on consumer choice. Within the beer category hardly any significant results have been found unfortunately. Conversely, within the milk and laundry detergent category results show that an increase in brand preference will have a positive influence on brand choice for that product regarding all discount formats. Based on these results it can be concluded that H3 is supported regarding the milk and laundry detergent category.

Following, based on consumer characteristics, segments have been identified. These segments have been analysed according to differences regarding purchasing behaviour. The results show that consumers are heterogeneous in their purchase decisions concerning discounts.

40 been identified between product price levels. Nevertheless, the exact source of the difference is still debatable.

6 Limitations & further research

41

References

Chatterjee, S., Heath, T. B., Milberg, S. J. & France, K. R., 2000. The Differential Processing Of Price In Gains And Losses: The Effects Of Frame And Need For Cognition. Journal Of Behavioral Decision

Making, 13(1), pp. 61-75.

Chen, C.-C., Chen, P.-K. & Huang, C.-E., 2012. Brands and Consumer Behavior. Social Behavior and

Personality: An International Journal, Issue 40(1), pp. 105-114.

Chen, Monroe & Lou, 1998. The Effects Of Framing Price Promotion Messages On Consumers' Perceptions And Purchase Intentions. Journal Of Retailing, 74(3), pp. 353-372.

Darke, P. R. & Chung, C. M., 2005. Effects Of Pricing And Promotion On Consumer Perceptions: It Depends On How You Frame It. Journal Of Retailing, 81(1), p. 35–47.

Drozdenko & Jensen, 2005. Risk And Maximum Acceptable Discount Levels. Journal Of Product &

Brand Management, pp. 264-270.

Foubert, B. & Gijsbrechts, E., 2007. Shopper Response To Bundle Promotions For Packaged Goods.

Journal Of Marketing Research, Volume XIIV, p. 647–662.

Friedman, L., 1967. Psychological Pricing In The Food Industry. In: In Almarin Phillips And Oliver

Williamson, Eds. Prices: Issues In Theory. Practice, And Public Policy. Philadelphia: University Of

Pennsylvania Press, pp. 187-201.

Gabor, A. & Granger, C., 1966. Price As An Indicator Of Quality: Report On An Enquiry. Economica. Gamliel, E. & Herstein, R., 2007. The effect of framing on willingnedd to buy private brands. Journal

of Consumer Marketing, 24(6), pp. 334-339.

Gamliel, E. & Herstein, R., 2011. To Save Or To Lose: Does Framing Price Promotion Affect Consumers' Purchase Intentions?. Journal of Consumer Marketing, 28(2), pp. 152-158. Gamliel, E. & Herstein, R., 2012. Effects of Message Framing and Involvement on Price Deal Effectiveness. European Journal of Marketing, 46(9), pp. 1215-1232.

Gendall, P., Hoek, J., Pope, T. & Young, K., 2006. Message Framing Effects On Price Discounting.

Journal Of Product & Brand Management, pp. 458-465.

Giuliana, I., Pozzani, A. I., Chen, V. A. & Perfi Gomes, . M. B., 2012. Influence Of Discount Price Announcements On Consumer’s Behavior. Rae: Revista De Administração De Empresas, 52(6), pp. 657-671.

Grewal, D., Marmorstein, H. & Sharma, A., 1996. Communicating Price Information Through Semantic Cues: The Moderating Effects Of Situation And Discount Size. Journal Of Consumer

Research, Issue 23, p. 148–155.

Gupta, S. & Cooper, L. G., 1992. The Discounting Of Discounts And Promotion Thresholds. Journal Of