ROUNDEDNESS OF PRICE AND RESOURCE DEPLETION:

(1)

ROUNDEDNESS OF PRICE AND

RESOURCE DEPLETION:

The impact of round prices at the supermarket checkout

Master’s Thesis

Alina Erdmann

Student Number: S3167062

M.Sc. Marketing Management

University of Groningen

Waldring 1

76337 Waldbronn, Germany

a.m.erdmann@student.rug.nl

(2)

2

MANAGEMENT SUMMARY

Visiting supermarkets shows that pricing strategies are very inconsistent: they might be rounded, end on the digit 9 or even end on any other digit. This inconsistency can also be found at the checkouts of supermarkets. Checkouts are an important place of the supermarket: every customer has to pass it and might spend a substantial amount of time there. It would only be comprehensible to make the most of this spot by optimizing pricing to the buying situation it implies and to the products it offers.

This piece of research seeks to find the optimal pricing strategy for products at the checkout, depending on their nature. It builds up on three major fields of theory: (1) the theory on rounded vs. non-rounded prices and their perceptional characteristics, (2) the theory on information processing by the consumer and the limitedness of cognitive resources and (3) the theory of the classification of products according to their hedonic and utilitarian level. It is hypothesized that a high degree of resource depletion results in higher purchase intention/sales at the checkout. Given a high degree of resource depletion, it is expected that purchase intention/sales will be higher for rounded prices than for charm prices. Also given a high degree of resource depletion, it is assumed that hedonic products lead to higher purchase intention/sales than utilitarian products. Finally, it is tested whether round prices lead to higher purchase intention/sales for hedonic goods than charm prices in a highly resource depleted state.

This paper presents two studies which have been undertaken to test the predictions. The first study is a survey experiment, conducted online and resulting in a sample at the size of 327, and served mainly to test the hypotheses since it enabled an accurate variable manipulation. None of the hypotheses could be accepted in the original approach due to insignificant results, yet tendencies could be found. Based on these, the following conclusions can be drawn:

 A high degree of resource depletion results in higher purchase/intentions at the checkout.

 Round prices are found to be more favorable than charm prices in a situation characterized by a high degree of resource depletion and thus by a feeling-based decision.

 Surprisingly, the utilitarian product type led to higher purchase intentions than the hedonic product type. However, limitations to this finding should be considered.

(3)

3

The underlying mechanism to make round prices more favorable in the described scenario is assumed due to the fit created by the affective dimensions of both round prices and hedonic goods. Also, round prices are easier to process than charm prices and enable a feeling-based decision which is favorable at a high degree of resource depletion. This might allow generalizing the results to other buying situations that evoke feeling-based decisions, either due to the product nature or due to low availability of resources. Round prices might be advantageous for hedonic products in general or in situations which require fast decisions. While parts of this have been suggested in literature, further research is needed to make reliable conclusions.

The second study was performed in the field, more precisely at a store of the Dutch retailer Albert Heijn. During a period of four weeks, prices at the checkouts were changed to round prices or charm prices. Even though prices overall increased after rounding them or setting them close to the next round number, unit sales of chocolate bars rose during the manipulation. Also, unit sales of chocolate bars were higher in the rounded price condition than in the charm price condition. Unit sales of fruit were very low; thereby contradicting what would be expected from the first study. While the first study was used for the hypotheses testing since accurate manipulation could only be achieved there, the findings of Study 2 might be more realistic with regards to the findings on fruit. Although the shelf in Study 2 was not evenly split in fruit and chocolate items, it displays the actual buying behavior of customers. As elaborated in the limitations of Study 1, survey participants might have been biased by the felt urge to present themselves in a socially acceptable way.

Both studies come with some limitations which need to be considered and might leave a gap for further research. The first study mainly did not serve to resource deplete participants as effectively as intended. Therefore, an alternative analysis has been introduced, which serves to bypass this limitation, but cannot replace the original analysis in confirming the hypotheses. The second study gave an interesting insight into practice, yet had drawbacks concerning the precise manipulation of variables other than the price.

(4)

4

(5)

5

PREFACE

During the past months I have been working on this piece of research as the final step towards graduating from the Marketing Master program of the University of Groningen. Working on my Master’s Thesis has certainly been intense and challenging, but also enriching and exciting. I am glad that I had the opportunity to do research on this interesting topic.

First of all, I would like to thank my supervisor, Prof. Dr. Laurens Sloot, for his guidance and support during this research. I could benefit very much from the given advice and the discussions on my topic. Secondly, I am grateful to the owner of the Albert Heijn store and her team for giving me the opportunity to conduct a field study and for supporting me in the setup. Thirdly, I would like to thank Dr. Wander Jager for the evaluation of my thesis. I would also like to thank my family and friends for their support, especially in spreading my survey and thereby helping me to reach the needed number of respondents. This leads me to my last point: thank you to everyone who participated in my survey!

Finally, I hope that you will enjoy reading my work and that it will provide you with interesting insights.

Alina Erdmann

(6)

6

1. INTRODUCTION

It is Friday, 5:30 pm. On your way home from work you stop by at a supermarket to get groceries for the weekend. With a full shopping basket you head to the checkouts. There you see some tasty chocolate bars – why should you not take one with you? Or are you rather feeling like buying a healthy snack like an apple or a banana? They only cost one euro. That is a good deal, right?

Choosing a pricing strategy is an important step for a retailer. The price might be part of the retailer’s positioning and contribute to the image created in consumers’ minds. It might as well set the frame for competitive analyses and influence the financial performance. Commonly, consumers might judge lower priced items to be of lower quality. However, the effect of the pricing strategy is not as easy as that. It is not only the price as a whole that creates different perceptions, but even its composition and digits.

In retail, different types of prices are used. Some are rounded and end on the digit 0, while others end on the digit 5, 9 or sometimes on a seemingly random digit. Generally, non-rounded, charm or odd prices are set close to a round number (e.g. €3.95 or €3.99 instead of €4.00) (Gendall et al., 1997). Different pricing strategies can even be observed within one store. It is however remarkable that especially prices ending with the figure 9 are used very frequently (Holdershaw et al., 1997). Prices ending with the digits 9, which are in fact only €0.01 below the rounded equivalent, are thought to be perceived as much cheaper and to create demand (Gendall et al., 1997).

Until recently, charm prices were mostly argued to be the better option since they are believed to stimulate sales. However, there are different findings on the effectiveness of rounded versus charm prices and there is literature arguing for the superiority of either option. Concerning charm prices, it can also be argued that they lead to negative image effects e.g. when they are perceived as a sign of inferior quality. This is why more and more arguments for round prices have been brought up. A recent paper from Wieseke et al. (2016) provides empirical support that round prices can indeed be favorable in some settings.

(8)

8

situation, price perception and processing could substantially differ from other situations and thus require different strategies.

Two per cent of grocery sales are realized at the checkout (Miranda, 2008). This might not be a very high number. Still, the checkout area of retail stores is a very valuable spot since every customer needs to pass it and occasionally spends a substantial amount of time there. It is a common practice in retail to dispose some products at the checkout to encourage impulse purchases (Almy & Wootan, 2015). Single items at the checkout are normally priced higher than bigger packs of the same product. Visiting a store of the Dutch retailer Albert Heijn has shown that unit sales per day are quite restricted: only a few units per single product are sold per day, seldom reaching 10 units. Prices were partly rounded but price setting did not seem to follow a consistent system.

At the checkout, the dominating product groups are sweets like chocolate bars. However, there are also products that are rather utilitarian than hedonic such as apples or bananas. It has already been suggested in literature that the effectiveness of pricing strategies can depend on the product category. However, authors disagree in their findings: while Choi et al. (2014) argue that charm prices match hedonic products, Wadhwa & Zhang (2015) point out that rounded prices and hedonic goods create a fit. It remains arguable which tactic suits best to realize the highest sales at the supermarket checkout.

The research questions arising from the previous reflections are: How does the roundedness of

price influence purchase intention/sales when the consumer is more or less resource depleted? Further, what is the role of the product type and how does it influence the effectiveness of the pricing strategy? What is the outcome of putting all three factors, degree of resource depletion, roundedness of price and product type, together?

A relatively clear managerial implication of this research is to provide retail stores with the information if one pricing strategy is superior to the other in the checkout setting. If this is found and implemented, this could lead to higher sales at the checkout. The study could also reveal differences concerning the product type and give indications on which product type(s) should be offered at the checkout or which are not favorable to be presented there.

(9)

9

challenge the current status of literature on impulse purchasing and add up to the findings on the favorability of rounded prices in specific situations. Also, it provides another view on the effectiveness of price forms per product type and thereby might give a response to the differing findings in current literature.

In the following, the theoretical background to this thesis will be presented. The three fields of topic will be brought together and set into relation in chapter 3. This will lead to the formulation of the hypotheses. The variable measurement as well as the study setup of Study 1, the survey experiment, will be presented in chapter 4. It also comprises the analysis of results and their discussion. Chapter 5 deals with the second study, conducted in the field. Parallel to the previous chapter, it contains the study design, the analysis of results as well as their discussion. Afterwards, the results of both studies are jointly discussed. Finally, managerial as well as theoretical implications are formulated.

2. THEORETICAL FRAMEWORK

This chapter serves to provide the theoretical background for this piece of research. The current status in literature is presented and delivers the basis for the conceptual development. The chapter includes theories on pricing strategies and compares and confronts differing findings. Further, it will be elaborated on the limited resource model and the theory of impulse purchasing. Lastly, the theoretical background to hedonic vs. utilitarian goods will be given.

2.1 Pricing Strategies

(10)

10

The frequent use of 9-ending prices gives occasion to question their actual ability to create greater demand. Gendall et al. (1997) deliver empirical support to the economic theory which says that if price goes down, in general demand goes up. Tested with a small set of products, the authors found that charm prices overall indeed foster demand. Further, they identified that charm prices are met with higher sensitivity especially for the low-priced grocery items covered in the study. They explain that this might be due to higher price awareness for these products since they are purchased regularly. They argue that it might also be due to the fact that a price decrease of e.g. €0.01 is relatively stronger in the context of small prices than high prices. However, their work lacks an explanation on how this effect occurs. Additionally, the study setup is not fully consistent with realistic circumstances.

Gedenk & Sattler (1999) argue that the best option for retailers is to use 9-ending prices. However, consumers’ perceptions are influenced by the mere roundedness of prices both positively and negatively. Stiving & Winer (1997) describe two image related effects of the last price digit. Consumers being exposed to a 9-ending price might interpret this as a price discount which leads to higher sales; this is called the price-image effect. The quality-image effect however says that consumers might perceive the charm priced item’s quality to be relatively lower. This consequently is a negative effective evoking from the non-roundedness of the price.

In addition to producing positive quality image effects, some authors even argue for the superiority of round prices to stimulate demand (Bray & Harris, 2006) thereby contradicting economic theory. In a pay-what-you-want context, consumers were also found to prefer round prices (Lynn et al., 2013), possibly partly due to the ease of processing round numbers.

(11)

11

2.2 Limited Resource Model and Impulse Purchasing

In this paragraph, the way how consumers process information and react to information in general will be explained. Depending on the elaboration continuum, the amount of thought spent on the incoming information, it is spoken of System 1 or System 2 thinking. The first is a fast system characterized by low effort and low amount of thought. For example, consumers do not pay a lot of attention to messages of promotions (as in Petty & Cacioppo, op. 1986), but process the information on the so-called peripheral route. In this case, cues rather than arguments can be persuasive. This means that message elements trying to persuade on the peripheral route are designed to create e.g. mere liking, thus trying to evoke a rather feeling-based decision. The latter system on the contrary requires a higher amount of thought spent on the incoming information. In so-called central route processing consumer involvement is strong and high attentive. Persuasion depends on the quality of arguments presented. For this reason, advertising planned to evoke central route processing, rather encompasses product-related attributes, thus requiring cognitive effort (Petty & Cacioppo, op. 1986).

(12)

12

appealing in this respect (e.g. chocolate cake over fruit salad). On the contrary, consumers choose cognitively appealing products given resource availability. Study layouts differed to the extent that Shiv & Fedorikhin (1999) made their participants decide between two options, whereas Vohs & Faber (2007) presented an array of products that could be purchased or not.

Wadhwa & Zhang (2015) also take into account the limitedness of processing resources in one of their studies. When processing resources are depleted or limited, they argue that people usually rely more on feelings when it comes to the purchase decision. On the contrary, given the availability of processing resources, the decision is built upon cognitive efforts. Wadhwa & Zhang (2015) linked this to the processing of prices and found out that under the condition of resource depletion (and thus for a decision based on feelings) the rounded prices produced better purchase intentions. The theory of regulatory fit is a critical underlying factor to Wadhwa and Zhang’s (2015) findings. Cesario & Higgins (2008) studied the theory in the context of providing consumers with advertising messages fitting their promotion- or prevention-focus. They found that experiencing a fit led more to feeling right and produced more positive attitudes. These findings are transferred to the topic of the nature of the purchase decision in this piece of research.

2.3 Hedonic vs. Utilitarian Goods

Products can generally be classified into two groups: hedonic vs. utilitarian goods. Hirschman & Holbrook (1982) define hedonic consumption as linked to “multisensory, fantasy and emotive aspects of product usage experience”. Bazerman et al. (1998) add that hedonic products are what the consumer affectively “wants”, whereas utilitarian consumption is characterized by cognition and the thought of “shoulds”. Further, utilitarian consumption can be ascribed the goal of fulfilling functional needs, thereby having a rather practical than emotional character (Strahlevitz & Myers, 1998).

(13)

13

goods that are highly affective and fruit salad for products that are strong on the cognitive dimension in one of their studies.

Choi et al. (2014) combine the topic of different product types with the issue of the roundedness of prices. They found that consumers choice for hedonic instead of utilitarian goods is positively influenced by charm prices. Charm prices thereby seem to be more effective for hedonic goods. This effect is mediated by guilt reduction: consumers perceive the charm priced item as less expensive and thereby justify their buying decision – but only for hedonic goods. Wadhwa & Zhang (2015) state the opposite in one of their studies. Despite previous literature having argued for the low price image of non-rounded prices and thus their demand-fostering character when price is an important decision criteria as well as for perceived inferior quality, they found that non-rounded prices lead to higher perceived quality and higher anticipated purchase satisfaction when the consumption goal is utilitarian. The other way around, they found more positive results for rounded prices when the consumption goal was hedonic.

3. CONCEPTUAL DEVELOPMENT

This chapter sets the before mentioned theories into relation and shows how they are expected to influence each other in the conceptual model. Based on this, the hypotheses are developed.

Figure 1: Conceptual Model

H3 H1

Degree of resource depletion - low - high Purchase intention Sales Product type - Hedonic - Utilitarian

- Combined (hedonic & utilitarian) Roundedness of price

- Rounded price - Charm price

Control variables - Mood

- Buying Impulsiveness Scale - Usual purchase frequency

(14)

14

The underlying mechanism to this study is the effect of the degree of resource depletion on purchase intentions or sales. Consumers who are resource-depleted due to prior activities subsequently have less energy to exert self-control. Without self-control, they are prone to purchase items that they were not planning to buy (Vohs & Faber, 2007) and might buy items that they do not need (Beatty & Ferrell, 1998). Retailers can benefit from this effect by placing items at the checkout, a place where consumers are likely to be resource-depleted after having finished their shopping trip. This effect has been studied frequently and it has been shown that resource depletion has led to higher purchase intentions/sales. In this study’s setting, it is also expected that resource depleted consumers will utter higher intention to buy an item. Based on this, the first hypothesis is formulated:

H1: A high degree of resource depletion results in a significantly higher purchase intention at the checkout setting of a supermarket than a low degree of resource depletion.

In addition to the direct effect of resource depletion on purchase intention/sales, the type of price strategy could have a moderating effect. In a resource-depleted situation, consumers are expected to rather take feeling-based decisions since cognitive resources are scarce. As shown by Wadhwa & Zhang (2015), round prices are also processed affectively rather than cognitively. This matches the behavior of a resource-depleted consumer. Through the induced fit, higher purchase intention/sales are expected. This fit is expected to overrule literature findings that argue for a higher general effectiveness of charm prices (e.g. Gedenk & Sattler, 1999).

(15)

15

Figure 2: Expected moderating effect of round prices

H2: Given a high degree of resource depletion, implementing rounded prices for products at the checkout setting of a supermarket results in a significantly higher purchase intention than implementing charm prices.

Additionally, the product type could have a moderating effect on the relationship between the degree of resource depletion and purchase intention/sales. In literature, there are different findings regarding the effect of the product type. Vohs & Faber (2007) and Shiv & Fedorikhin (1999) do not agree on the effect of the product type: the first authors found increased impulse purchases for both hedonic and utilitarian products, whereas the latter authors argue that hedonic purchases are more likely in the state of resource depletion. Although they do not explicitly test it, it can be inferred from Wadhwa & Zhang (2015) that there should be a fit created from being resource depleted and the nature of the purchase decision which is feeling-based for hedonic products. Therefore, it is expected that there will be increased purchase intentions for hedonic goods, which translates into the third hypothesis. Figure 3 shows the expected moderating effect of the product type.

low high P u rch ase in ten tio n

Degree of resource depletion

Expected moderating effect of round prices

(16)

16

H3: Given a high degree of resource depletion, the hedonic product type results in a significantly higher purchase intention at the checkout setting of a supermarket than utilitarian products.

Further, there might as well be a combined effect of all three factors: degree of resource depletion, roundedness of price and product type. Choi et al. (2014) found that consumers choose hedonic over utilitarian goods when the products are charm priced and that this effect was moderated by guilt reduction. Wadhwa & Zhang (2015) yet identified rounded prices to produce better outcomes for hedonic goods. For the development of this hypothesis, the affective and cognitive dimensions of the variables are taken into account. In the state of resource depletion, information is rather processed feeling-based which requires less cognitive effort. Rounded prices as well as hedonic goods also stand for affective decision making. Combining these three levels of the variables, the highest purchase intention/sales might be produced due to an underlying induced fit as described earlier. This fit is expected to overrule the findings of Choi et al. (2014) and leads to the fourth hypothesis, as illustrated in Figure 4:

low high P u rch as e in ten ti o n

Degree of resource depletion

Expected moderating effect of the product type

(17)

17

H4: Given a high degree of resource depletion, implementing rounded prices for the hedonic product type at the checkout setting of a supermarket results in a significantly higher purchase intention than implementing charm prices.

4. STUDY 1: SURVEY EXPERIMENT

The hypotheses developed in chapter 3 were tested in a survey experiment. The variable measurement as well as the study settings will be explained in this chapter. Further, the collected data will be analyzed and discussed. This chapter concludes with an elaboration on the limitations of the survey experiment and suggestions for further research.

The research is built upon a 2x2x3 factorial design. The factor levels are low vs. high degree of resource depletion, round vs. charm price and hedonic vs. utilitarian vs. a combination of hedonic and utilitarian goods. This produces twelve different factor level combinations.

4.1 Variable Measurement

Dependent variable: Purchase intention. The purchase intention was measured with the

eleven point purchase probability scale by Juster (1966) (see Appendix I). The participant was asked about the probability that he or she would purchase the item in question. This scale has proven to be superior in predicting sales over other types of scales such as binary scales only including yes or no answers. There is a similar scale reduced to 5 points, however Juster (1966) has found the 11 point scale to be more appropriate. Kalwani & Silk (1982) argue for increased scale liability with an increased number of points as well; nevertheless, Morwitz

utilitarian hedonic P u rch as e in ten ti o n Product type

Expected moderating effect of round prices at high degree of resource depletion

(18)

18

(2012) points out that this is not valid for infinite expansion. Despite the large support of the Juster scale, there are some limitations to it. Brennan & Esslemont (1994) elaborated on the ability of the Juster scale to predict purchases of fast-moving consumer goods. They put a special focus on brand share rates, which were predicted accurately; however, the purchase rates as such were overestimated by 5 to 6 percent.

Control variable: Mood. Following the example of previous studies on impulse buying (Shiv

& Fedorikhin, 2002, Vohs & Faber, 2007), a scale was introduced to control for possible mood effects. For this study, the mood “at the moment” will be asked for since the degree of resource depletion also refers to the very moment of taking part in the study. Three five-point bipolar scales were used to assess the respondents’ mood: happy-unhappy, relaxed-stressed and interested-bored, thus each ranging from a positive to a negative denoted state. They were combined into one scale afterwards.

Control variable: Usual purchase frequency of products. Naturally, the purchase probability

might also depend on if the participant is used to buying the specific product. Some participants might just never buy any of the displayed items. This can somewhat be tackled by giving a variety of products to choose from, but there are still certain space limits. Also, products at checkouts in reality were used to be able to draw relevant conclusions. To control for usual purchase frequencies, participants were asked how often they usually buy the displayed items and answered on a five-point Likert scale from “never” to “very often” for each item displayed. Based on this information, a score was developed to enable controlling for how frequently participants buy what they were exposed to in the experiment.

Control variable: Impulse buying tendency. Participants might differ in their general

(19)

19

Consumers with a high BIS score are assumed to generally be more prone to impulse purchases than those with low scores. The scale has been proven to be valid and internally consistent (Rook & Fisher, 1995, Weun et al., 1998). Still, the scale has its limitations: in Vohs’ & Faber’s (2007) study the BIS score did not serve to explain differences in spending when participants were not resource depleted.

Additional analysis: Hedonic and utilitarian level. As described before, products are

classified by the consumer on both dimensions (Batra & Ahtola, 1991). Despite the listed products which are generally rather hedonic or utilitarian, the participants’ perceptions might differ. Their opinion had thus to be captured for the products included in the study in order to be able to assess participants’ perceptions. Batra & Ahtola (1991) tested this dimensionality using the semantic differential scales developed by Osgood et al. (1978), which consist of pairs of words that are opposite in meaning. The hedonic level factor was found to load the most on pairs like pleasant-unpleasant or nice-awful, whereas the utilitarian factor was found to load on the pairs such as useful-useless or beneficial-harmful. For the survey experiment, a four item five-point scale was used for simplicity and consistency reasons. It consisted of the above mentioned word pairings (pleasant-unpleasant, nice-awful, useful-useless, beneficial-harmful), since they are expected to be able to measure the underlying hedonic factor as well as the utilitarian factor.

Variable Type Measurement Based on

Purchase probability Dependent variable Eleven point scale Juster (1966)

Mood Control variable Three item five-point

bipolar scale

Shiv & Fedorikhin, 2002, Vohs & Faber, 2007 Usual purchase

frequency

Control variable One item five-point scale -

Impulse buying tendency

Control variable Four item five-point

Likert scale

Rook & Fisher (1995)

Hedonic/utilitarian level Additional analysis Four item five-point

bipolar scale

Batra & Ahtola (1991), Osgood et al. (1978) Table 1: Overview on variable constructs

4.2 Design and Procedure

(20)

20

sample size of 327 respondents from which 110 were male and 217 were female. The majority of respondents stated to be students (70.3%). The respondents’ age ranged from 18 to 65 and produced an average of 27.

Respondents were either presented with the task to do mental arithmetic (high degree of resource depletion condition) or with general questions about their grocery shopping habits (low degree of resource depletion condition). Mental arithmetic as a complex task has been used before to achieve resource depletion (Hamilton et al., 2007; Ryu & Myung, 2005). The arithmetic tasks included pictures of products that can be found in supermarkets. The presented task was to add two prices including cent details imagining that these two items would be bought by the respondent. The task thus served as a resource depleting element as well as a cover story and as a mean to let respondents imagine a shopping situation. The same product pictures were shown in a different context to the other group to balance any possible effects of the product pictures.

(21)

21 Figure 5: Shelf design

(22)

22

The main effect as well as the moderating effects are planned to be tested with regression analyses and analyses of covariance. Multi-item scales, as for example used for the control variable buying impulsiveness scale, are tested for reliability and reduced to fewer items with factor analyses.

4.3 Descriptive Results

Representativeness. The survey sample does not fully represent distributions from the Dutch

society as shown in Table 2 (Centraal Bureau voor de Statistiek, 2016; Statistics Netherlands, 2016). More women than men took the survey, whereas the Dutch population is quite evenly split in the two genders. Due to the large amount of students who took the survey, the sample lacks to represent other groups in society and does not display regular age and occupation distributions of the Dutch society. The strong representation of students is due to the writers own occupation being a student and to mainly having respective contacts. Further, the survey was primarily distributed on Facebook which is mainly used by younger people.

Sample Dutch population

N 327 17 Mio. Gender Male 33.7% 49.6% Female 66.3% 50.4% Age 0-20 years 10.8% 22.5% 20-40 years 80.8% 24.5% 40-65 years 8.4% 34.8% 65-80 years 0.0% 13.8% 80 years+ 0.0% 4.4% Occupation Student 70.3% 4.1%

Table 2: Sample representativeness

Control variables. The control variables mood and impulse buying tendency were measured

on multi-item scales. Before including them in further analysis, they were reduced to their underlying factors using factor analysis. For the impulse buying tendency scale another preparing step was needed. One item had to be reverse-coded since it was opposite in meaning from the other items.

(23)

23

communalities. The KMO should be >0.5 and measures if data is likely to factor well. The Bartlett’s test should be significant and thus enable the rejection of the null hypotheses (H0: Variables are uncorrelated). Lastly, communalities, the percent of variance in a given variable explained by all the extracted factors, should be >0.4. The KMO and Bartlett’s test criteria have been met for both variables measured with multi-item scales. The last item of the mood scale had to be dropped, since its communality was <0.4.

All factor analyses were performed using the principal component analysis (PCA). The analyses also included rotation with the Varimax method, which serves to get a better interpretability of the factor loadings. For the factor analyses of the control variables mood and impulse buying tendency the number of factors was determined by the Eigenvalue criterion, which says that only components with an Eigenvalue >1 should be chosen. This resulted in one factor for each of the variables, thus all items loaded on the same factors with all of them found to have a strong loading. When only one factor is extracted, no rotated factor loadings output is given. In this case, it is referred to the unrotated factor loadings.

The internal consistency of the impulse buying tendency items was assessed based on Cronbach’s alpha. The scale is internally consistent when the value is >0.6. The mood variable was measured based on two items only. In this case, the internal consistency is tested by correlating the two items rather than by using Cronbach’s alpha. Eisinga et al. (2013) found that Cronbach’s alpha would only be appropriate under narrow conditions. While they argue for the Spearman-Brown test to deliver the most reliable results, the Kendall correlation was used here due to the ordinal scale of the items in questions. The correlation has to be significant for internal consistency. Table 3 shows the outcomes of the analyses.

Constructs Items Scale Factor

loadings Internal consistency Mood M1. Happy-unhappy M2. Relaxed-stressed Two item five-point bipolar scale 0.858 0.858 r=0.412 p<0.01 Impulse buying tendency

BIS1. I often buy things spontaneously.

BIS2. "Just do it" describes the way I buy things. BIS3. I carefully plan most of my purchases. BIS4. I often buy things without thinking.

Four item five-point Likert scale 0.811 0.829 0.753 0.735 α=0.789

Table 3: Overview on factor analysis results

(24)

24

buys the displayed items on average and allows making comparisons among conditions in which different shelves and thus different products were displayed. To ensure comparability with variables of other scales, the score was standardized. This had not do be done with the factor scores since they are standardized by default.

The hedonic and utilitarian level variable was also transferred into a mean score. High ratings on the scales of the hedonic items and low ratings on the scales of the utilitarian items were counted as a high hedonic level. Low ratings on the hedonic items’ scales and high ratings on the utilitarian items’ scales are counted as a low hedonic, or utilitarian, level. Two items had to be coded reversely to allow this treatment. The final score indicates a high hedonic level for low values and a low hedonic level or high utilitarian level for high values. Values could range from 1 to 5, thus a score of 2.5 means equally strong hedonic and utilitarian perception of the displayed items. For the participants in the hedonic condition the score should be as low as possible since this would mean that they actually perceived the displayed items as hedonic. The mean score was reported to be 2.36. While a score of 1 would have been the optimal value, the score still does not exceed 2.5. Thus, participants perceived the displayed items as rather hedonic on the average. On the contrary, the score should be above 2.5 for the utilitarian condition. The score is 3.02, thus this assumption is met. The recorded perceptions in the combined shelf condition have to be looked at separately. The findings are similar to the ones in the single product type conditions: the mean score on the perception of the assumingly hedonic products equals 2.36 and 3.07 for the assumingly utilitarian products. Thereby both values meet the requirement to be below and above 2.5. From this it can be concluded that participants roughly perceived the displayed items in the assumed way. Other implications of this finding will be discussed in the limitations.

Assumptions. To be able to conduct the planned tests, some assumptions have to be met.

Those include:

a) Independence of observations: this is given due to the method of data collection (between-subject design).

b) No significant outliers: assessing the distribution of the dependent variable per condition revealed four outliers in the hedonic condition. Those four cases were excluded from further analysis. This reduced the sample size to 323.

(25)

25

However, data points generally rather lie closely to the expected line in the Normal Q-Q plot. Additionally, the sample size was quite large (N>300). Therefore, a normal distribution of the dependent variable is assumed.

d) Homogeneity of variances: this can be assessed using the Levene’s test for homogeneity of variances. If the null hypothesis does not have to be rejected (p>0.05), it can be assumed that the variance of the dependent variable is equal across groups. This will be checked and reported before the respective ANCOVAs.

e) Linear relationship of covariate and dependent variable at each level of independent variable: this was tested by creating scatterplots for each control variable with the dependent variable. The data points were grouped per level of the independent variable. The same procedure was followed with the moderators. All control variables (mood, impulse buying tendency, usual purchase frequency) have approximately met the criterion. f) Homogeneity of regression slopes: to test whether there are interactions between the

covariates and independent variable and moderators, interaction variables were introduced. There were some interactions with the control variable buying impulse tendency, but the respective regression coefficients were rather low. The independent variable and the moderators all interacted with the control variable usual purchase frequency and produced somewhat higher regression coefficients. The control variables will first be kept in the model as they might be essential. During the hypotheses testing, it will be argued why certain control variables might be excluded in further steps.

g) No multicollinearity: for the regression, multicollinearity needs to be excluded. The independent variable and moderators have been regressed on each other. Then, all respective variance inflation factors (VIF=1/(1-R2)) were calculated. All calculated values were low; multicollinearity can thus be excluded.

4.4 Hypotheses Testing

Independent variable: Degree of resource depletion. First of all, a dummy regression was run

(26)

26

of the variation of the dependent variable (adj. R²=0.237). The effect of the degree of resource depletion was not found to be significant (p=0.493).

Another possibility was to exclude the control variable usual purchase frequency from the model due to the before found significant interaction with the independent variable and the moderately strong coefficient. However, this does neither improve the model nor does it produce a significant effect of the degree of resource depletion.

To get an idea of how group means differ from each other, even if not significantly, an ANCOVA was run. Homogeneity of variances can be assumed (p=0.11). The mean of purchase probability is slightly higher in the high degree of resource depletion as presented in Table 4. Yet this cannot be judged to be more than a tendency.

Degree of resource depletion Mean purchase probability (estimated) Difference Sig.

low 3.901 -0.199 0.49

high 4.101 0.199 0.49

Included control variables: Impulse buying tendency, usual purchase frequency. Table 4: Differences in mean purchase probability

Concluding, H1 is not supported. A high degree of resource depletion did not result in a significantly higher purchase intention than a low degree of resource depletion.

Moderator analysis: Roundedness of price. To test whether the roundedness of price has an

impact on the effect of the degree of resource depletion on purchase probability, another dummy regression was run. As a starting point all control variables were included; however, mood was again not found to be significant and was thus excluded. The model fit was rather low (adj. R2=0.236). The interaction effect of the degree of resource depletion and the roundedness of price was not significant (p=0.446).

Since the interaction of the control variable usual purchase frequency with the price moderator was significant, it could be argued to exclude it from the analysis. However, neither the moderating effect becomes significant nor is there a significance difference in means within the high degree of resource depletion. The adjusted R2 value also becomes smaller, thus the model fit does not increase.

(27)

27

charm prices. This difference was not significant, but can as well be viewed as a tendency. The outcome is presented in Table 5.

Degree of

resource depletion

Roundedness of price Mean purchase probability (estimated) Difference Sig.

low charm 3.882 -0.042 0.92 round 3.924 0.042 0.92 high charm 3.843 -0.488 0.25 round 4.331 0.488 0.25

Included control variables: Impulse buying tendency, usual purchase frequency. Table 5: Differences in mean purchase probability

The group differences are illustrated in Figure 6. The gap between purchase probabilities within the high degree of resource depletion condition is remarkably larger than in the low degree of resource depletion. However, the difference is not as high in absolute terms.

(28)

28

Concluding, H2 is not supported. A high degree of resource depletion and implementing

rounded prices for products at the checkout did not result in a significantly higher purchase

intention than a high degree of resource depletion and implementing charm prices.

Moderator analysis: Product type. Next, it was tested whether the product type has an impact

on the effect of the degree of resource depletion on purchase probability. For that purpose, a dummy regression was performed. It included all control variables first, but was continued without the control variable mood since it did not have a significant effect. The model explained a minor part of the variation of the dependent variable (adj. R²=0.232). The interaction effects of the degree of resource depletion and the product type dummies were not significant (p=0.805; p=0.674).

Since the interaction between the product type and the control variable usual purchase frequency was found significant and produced moderately strong regression coefficients, the test was also conducted excluding the respective control variable. In the regression analysis, still no significant overall interaction effect was found. For the subsequent analysis the control has still been excluded to get a clearer view on group differences.

Running an ANOVA delivered insights on how the groups differ from each other concerning the purchase probability. The assumption of equal variances was violated (Levene’s test: p=0.00). The overall direct effect of the product type was found significant (p=0.00), but is not part of the conceptual model being tested. However, this contributes to explaining the significant differences between some of the groups. The results are presented in Table 6.

(29)

29 Degree of

resource depletion

Product type Mean purchase probability

(estimated)

Compared to Difference Sig.

low hedonic 2.603 utilitarian -2.201 0.00 combined -1.464 0.02 utilitarian 4.804 hedonic 2.201 0.00 combined 0.737 0.50 combined 4.067 hedonic 1.464 0.02 utilitarian -0.737 0.50 high hedonic 3.003 utilitarian -2.244 0.00 combined -1.165 0.11 utilitarian 5.247 hedonic 2.244 0.00 combined 1.079 0.16 combined 4.168 hedonic 1.165 0.11 utilitarian -1.079 0.16

Included control variables: Impulse buying tendency. Table 6: Differences in mean purchase probability

The above found results are also shown in Figure 7. The significant difference in purchase probabilities between the hedonic and the utilitarian product type can clearly be seen. As described before, this holds for both conditions of the degree of resource depletion.

(30)

30

Participants exposed to a combined shelf were additionally asked how likely they would be to choose chocolate respectively fruit if they had to choose something. The results of these questions are somewhat in line with the findings above. The means reported for fruit are slightly higher and confirm the before indicated preference for fruit of the participants. However, the means of purchase probability do not as strongly differ from each other between chocolate and fruit within the same degree of resource depletion condition.

Due to these results, H3 is not supported. A high degree of resource depletion and the

hedonic product type did not result in significantly higher purchase intention at the checkout

setting of a supermarket than a high degree of resource depletion and the utilitarian product

type. In fact, across both levels of the degree of resource depletion, the utilitarian type resulted

in a significantly higher purchase probability than the hedonic product type.

Combined moderator analysis: roundedness of price and product type. Lastly, a full model

regression was performed. As a first step, all control variables were included. Further, three-way interaction variables were introduced for this analysis. The control variable mood did not have a significant effect. Therefore, it was excluded from the model. The model fit was rather low (adj. R2=0.228). The model did not show significant findings for neither direct effect, nor two way interactions or three-way interactions. Excluding the control variable usual purchase frequency from the regression due to the lack of independence also does not serve to find significant interaction effects and decreases the model fit. For subsequent analyses it was still excluded to get a clearer view on group differences.

(31)

31 Degree of resource depletion Roundedness of price

Product type Mean purchase

probability (estimated)

low

hedonic charm 2.899 round 0.628 0.42

round 2.272 charm -0.628 0.42

utilitarian charm 4.469 round -0.819 0.26

round 5.288 charm 0.819 0.26

combined charm 3.985 round -0.163 0.83

round 4.148 charm 0.163 0.83

high

hedonic charm 2.642 round -0.638 0.41

round 3.279 charm 0.638 0.41

round 5.456 charm 0.435 0.58

round 4.810 charm 1.283 0.12

Figure 8 illustrates the findings for the high degree of resource depletion condition which is in the focus of this piece of research. For all product types, purchase probabilities were higher in the round price condition, with the largest difference between charm and round prices found in the combined shelf condition.

(32)

32

Finally, H4 has to be rejected. A high degree of resource depletion, the hedonic product type and round prices did not result in a significantly higher purchase intention at the checkout setting of a supermarket than a high degree of resource depletion, the hedonic product type and charm prices.

4.5 Alternative Analysis

One important limitation is that mental arithmetic might effectively resource deplete some respondents but not all of them. To control for this, respondents were asked to state how difficult they found taking the sums on a scale from 1 (easy) to 5 (difficult). Based on this, an alternative variable opposing the following two groups was formed: those that were in the high degree of resource depletion condition and answered the difficulty question with 4 or 5 versus those that were in the same condition but answered the difficulty question with 1, 2 or 3 and those that were in the low degree of resource depletion condition. In the following this variable will be referred to “high difficulty”.

Independent variable: high difficulty. The same procedures as before were followed with this

variable. First, all control variables (mood, impulse buying tendency and usual purchase frequency) have been included. In the following mood was excluded since it did not show a significant effect. The model only served to explain a small part of the variation in the dependent variable (adj. R2=0.243). The effect of high difficulty on purchase probability was not found significant (p=0.09).

Interaction was also found between the high difficulty variable and the control variable. Excluding it led to a decrease in the model fit (adj. R2=0.044), but the effect of high difficulty on purchase probability could be found significant (p=0.04).

(33)

33

High difficulty Mean purchase probability (estimated) Difference Sig.

No 3.834 -0.848 0.04

Yes 4.683 0.848 0.04

Moderator analysis: Roundedness of price. In the following, the moderating effect of the

roundedness of price was tested again; now concerning its impact on the effect of high difficulty on purchase probability. First, a regression was run. The control variables impulse buying tendency and usual purchase frequency had a significant effect. The model fit was rather low (adj. R2=0.247). The interaction effect of high difficulty and roundedness of price was not significant (p=0.11). Agreeing with further elaborations, the control variable usual purchase frequency was excluded from the model due to the interaction effect of moderate size. As a result, the model fit became smaller (adj. R2=0.054), but the interaction effect was significant (p=0.047). The standardized coefficient and thus the effect size were relatively small.

Model Standardized coefficient Sig.

Constant - 0.000

Impulse buying tendency 0.188 0.001

High difficulty 0.007 0.922

Price 0.013 0.825

High difficulty x price 0.159 0.047

Table 9: Regression coefficients

(34)

34

High difficulty Roundedness of price Mean purchase probability (estimated) Difference Sig.

no charm 3.795 -0.080 0.83 round 3.875 0.080 0.83 yes charm 3.851 -1.716 0.02 round 5.568 1.716 0.02

While the graphs given in Figure 9 look quite similar to the ones in the original analysis, the found differences within the high difficulty group are higher in absolute terms. The gap between the purchase probabilities in the charm and the round price condition can clearly be observed in the group which perceived the task as difficult.

Figure 9: Moderating effect of the roundedness of price

Moderator analysis: Product type. Another regression was performed for the moderator

(35)

35

Comparing the means through an ANOVA (equal variances not assumed; Levene’s test: p=0.00) delivered somewhat different results compared to the original analysis, where the utilitarian product resulted in significantly higher means in purchase probability than the hedonic product type across both conditions. The direction of the effect is still the same, but the difference in means is not significant within the group of high perceived difficulty. The results are shown in Table 11.

Parallel to the original analysis, means can also be compared within the combined shelf group. The mean purchase probabilities do not differ significantly between the chocolate and fruit shelf, while the means for the fruit shelf are slightly higher, thereby being in line with previous findings.

High difficulty

Product type Mean purchase probability

(estimated)

no hedonic 2.622 utilitarian -2.353 0.00 combined -1.237 0.01 utilitarian 4.975 hedonic 2.353 0.00 combined 1.116 0.30 combined 3.859 hedonic 1.337 0.01 utilitarian -1.116 0.30 yes hedonic 3.686 utilitarian -1.461 0.31 combined -1.337 0.39 utilitarian 5.147 hedonic 1.461 0.31 combined 0.124 1.00 combined 5.023 hedonic 1.337 0.39 utilitarian -0.124 1.00

(36)

36

Figure 10: Moderating effect of the product type

Combined moderator analysis: Roundedness of price and product type. Finally, another full

model regression was conducted. The control variables impulse buying tendency and usual purchase frequency were first kept in the model, whereas mood was excluded since it did not have a significant effect. The adjusted R2 value was 0.241, thereby rather low. The interaction effects were not found significant in this model. Parallel to the original analysis, the control variable usual purchase frequency was left out in the next step to produce a clearer output, however at a lower adjusted R² value (0.136).

(37)

37 High

difficulty

Roundedness of price

Product type Mean purchase

probability (estimated)

no

hedonic charm 2.814 round 0.359 0.55

round 2.455 charm -0.359 0.55

round 5.315 charm 0.622 0.29

round 4.036 charm 0.352 0.57

yes

hedonic charm 2.654 round -2.320 0.08

round 4.974 charm 2.320 0.08

round 5.627 charm 0.923 0.45

round 5.899 charm 1.835 0.12

Figure 11 illustrates the larger difference between purchase probabilities for charm vs. round prices, especially for hedonic products and can clearly be distinguished from the original plot.

(38)

38

4.6 Discussion

The main construct applied in this piece of research was about the limitedness of resources and their implication on impulse purchasing. Literature findings proof that the resources consumers draw their decision on are limited and that previous acts of using these resources result in lacking resources for following decision situations (Vohs & Faber, 2007; Baumeister et al., 1998). This implies that consumers might not have enough resources to take thoughtful, elaborated decisions, but rather rely on their feeling. In a shopping situation, this might result in impulse purchases. The focus of this research was on the checkout area of supermarkets, a typical impulse purchasing environment. The findings from literature were intended to be reproduced in this context. In this research, the theory on limited resources and impulse purchasing could not be confirmed. As seen in the results section, a higher degree did not lead to a significantly higher mean in purchase probability. Still, a certain trend of such an effect could be seen. This will be further elaborated on in the limitations. The alternative analysis was in line with the before observed trend. It showed that an initial task drawing on processing resources results in a higher purchase probability. Thus, availability of resources might have been diminished by taking a perceived difficult task. This might have resulted in a rather affective than cognitive decision and in an impulse purchase.

(39)

39

reduced the need for resources and might also have led to the favorable convenience in impulse-buying situations. All items on the shelf had the same price, so it assumingly was not a big effort to assess the prices. Therefore, round prices might not have been superior in creating higher purchase probabilities in combination with being resource depleted.

The alternative analysis revealed that, after previous efforts, round prices served to produce higher mean purchase probabilities. In this case, the ease of processing round prices might have created a perceived fit in case of being highly resource depleted. Thereby positive feelings towards the product might have been reinforced. This would thus be in line with what was considered by Wadhwa & Zhang (2015). Since the difference in purchase probabilities was not equal among the resource depletion conditions, it can be concluded that round prices are not generally superior to charm prices, but especially in a situation on low availability of resources.

(40)

40

superiority of hedonic goods in a resource depleted situation in the study by Shiv & Fedorikhin (1999) have been only or mainly due to the forced decision between the two options, this might serve as an explanation why their finding could not be reproduced. In the literature review, another study which questioned the findings from Shiv & Fedorikhin (1999) was discussed. Vohs & Faber (2007) distanced their study from the earlier one since they did not force participants to choose one product. They found impulse purchases not to be higher for either product type and rather argued for consumer characteristics to be more important. However, they also exposed participants to both product types at the same time, which limits the overall comparability. Further, purchase probabilities concerning fruit containing shelves might be unnaturally high due the phenomenon of socially desirable responding. This phenomenon says that participants might also be influenced by factors other than those coming from the manipulation, namely by the urge to present themselves in a favorable matter (King & Bruner, 2000). Respondents shown the fruit and thereby healthy shelf might have reported biasedly high purchase probabilities due to meet what seems socially acceptable.

Finally, all constructs were joined in one model to test for their combined effect. In addition to the expectation of hedonic goods to be preferred in a resource depleted situation, round prices were assumed to increase the purchase probability as compared to charm prices. This is due to the affective dimension of both round prices and hedonic products and the induced fit (Wadhwa & Zhang, 2015). This effect was not found significant in the original analysis, yet a certain tendency could be seen. In the alternative analyses, the effect was marginally significant, thereby confirming the previously found trend.

4.7 Limitations

(41)

41

difficult or might have used facilitating means as a calculator. This limitation has been taken care of by doing the alternative analyses. This analysis can however not replace the original approach. Groups put in contrast strongly differed in size (high difficulty “yes”: n=62; high difficulty “no”: n=261). Also, participants were not equally distributed over the other conditions.

Additionally, an effective depletion of resources was part of the further development of hypotheses and built an important precondition. Being resource depleted was intended to be the basis for affective rather than cognitive processing of information. It should thus serve to make round prices and hedonic product more favorable through an induced underlying fit. Since there is a clear difference between the original analyses and the alternative analyses taking perceived difficulty into account, the resource depletion in general probably was not effectively reached. This precondition would thus not have been completely fulfilled. Thereby, there could have been less “need” for convenient round prices. Also product nature and cues, whether affective or cognitive, might not have made a strong difference.

The choice of products might not have enabled a clear distinction between hedonic and utilitarian products. The selected items were assumed to be hedonic or utilitarian, but perceptions might have differed from this assumption. The perceived hedonic and utilitarian level scores differed between chocolate and fruit in the desired direction, but scores could have been more extreme. Fruit has previously been used as a utilitarian product to study impulse buying behavior (Shiv & Fedorikhin, 1999) but it might not generally be classifiable to be utilitarian as revealed in this study. This of course limits the findings concerning the product type or at least the argumentation based on the hedonic and utilitarian level: being exposed to fruit produced higher purchase probabilities than being exposed to chocolate – however, it cannot be fully relied on that it was due to the hedonic or utilitarian level.

(42)

42

The applicability of findings could further be limited by the method of data collection. Several issues can be addressed here. First of all, it is not clear whether the participants imagined a real shopping situation and if they answered in a way that is conform to their actual shopping behavior. Second, it was the goal to reproduce an impulse purchasing situation from reality to the survey. Whether this was achieved through the study setup can be somewhat doubted since the resource depletion was probably not effective and participants might have not felt an actual urge to buy something as they might in reality. Third, difficulties of imagination and reproduction of a real shopping situation could even be enhanced by choosing a survey to collect data since it does not involve participants as much as for example an experiment in a laboratory. Also, as described before, the sample did not serve to represent the Dutch population very well due to the uneven distribution of age groups and occupation types.

4.8 Suggestions for Further Research

The main suggestion for further research is about ensuring better control on the resource depletion when reproducing this study. The analyses could become more reliable if the initial degree of resource depletion would be assessed. Asking participants questions like how exhausted or tired they feel at the moment could be on option, yet underlies subjective perceptions. Further, mental arithmetic generally seems to be an acceptable mean to achieve resource depletion, but it is strongly recommended to assess the perceived difficulty to have some control on the effectiveness of the resource depletion.

The research has shown that fruit might not generally be classifiable as extremely utilitarian. Although the findings on fruit are still interesting, researchers seeking to make a very clear distinction between hedonic and utilitarian products and connect these to the roundedness of price might have to choose other products. Still, it is recommended to respect a realistic checkout assortment to ensure managerial applicability.

An interesting approach would be to assess if the perception of the hedonic or utilitarian level of fruit differs between buying situations. Fruit might be perceived more hedonic at the checkout than in the regular aisles. Due to the single item presentation it could be rather perceived as a snack and less than something that should be bought due to its functionality.

(43)

43

whether round prices could convince resource depleted participants to buy the product if they like it but usually do not buy it.

The study could be conducted in a laboratory experiment with a reproduction of the checkout setting of a supermarket. This could facilitate imagining an actual shopping situation. The impulse buying situation could be induced by leading participants to the respective shelf and giving them the opportunity to purchase something, but not pointing out that this is the actual purpose of the study. In this case, it would probably be more suitable to offer food that could not be bought somewhere else very easily, e.g. cake and fruit salad.

This study did not implement different ways to present prices at the checkout which consequently leaves a gap for further research. It could be investigated on the difference of the effect of round and charm prices when they are displayed with regular shelf price tags as used in this study or for example with (colorful) stickers attached to the products.

5. STUDY 2: FIELD STUDY

This chapter elaborates on the second method of data collection used in this piece of research: the field study. In the following paragraphs, the results will be assessed and discussed. Finally, limitations to this project and suggestions for further research will be given.

The primary purpose of the second study was to apply the theoretical construct of rounded vs. charm prices to practice. However, the circumstances did not allow an equally precise manipulation as given in the first study. The focus is therefore on the manipulation of the roundedness of price which was effectively realized.

5.1 Design and Procedure

(44)

44

prices were changed to round prices and then charm prices again. Prices of all products at the checkouts were changed to provide a homogeneous layout, however only the sales figures of chocolate bars and fruit were tracked. This was due to two reasons: first, this is in line with the products used in Study 1 and thus enables comparisons to a certain degree and second, chewing gum or comparable products cannot be assumed as hedonic or utilitarian as distinctly as it might be possible with chocolate bars and fruit (Crowley et al., 1992).

In preparation of the actual study, chocolate bar sales figures were analyzed. The data contained sales over a period of two weeks and included the time of the purchase. On this basis, the development of sales during the day could be illustrated. There was no data available for fruit since the checkout items could not be distinguished from those in the “regular” area of the supermarket. For the study, new product numbers had to be introduced. Figure 12 shows the layout of the shelves at the supermarket. An example of an adapted price tag is given in Figure 13.

Figure 12: Shelf layout

ROUNDEDNESS OF PRICE AND RESOURCE DEPLETION: