3
Table of contents
ABSTRACT ... 51. INTRODUCTION ... 6
2. LITERATURE REVIEW ... 9
2.1 BENEFITS OF DYNAMIC PRICING ... 9 2.2 CONSUMERS’ PERCEPTIONS OF DYNAMIC PRICING ... 10 2.2.1 Price fairness ... 11
2.2.2 Uncertainty ... 13
2.3 THE ROLE OF CONSUMPTION OCCASION ... 13 2.4 UNCERTAINTY AVOIDANCE ... 14 3. CONCEPTUAL MODEL & METHODOLOGY ... 16
3.1 CONCEPTUAL MODEL ... 16 3.2 DATA COLLECTION ... 16 3.3 EXPERIMENTAL DESIGN ... 16 3.4 MEASUREMENT ... 18 3.4.1 Measurement perceived fairness ... 18
3.4.2 Measurement of uncertainty ... 18
3.4.3 Measurement of price knowledge & consciousness ... 19
3.4.4 Measurement of uncertainty avoidance ... 19
3.4.5 Measurement of hedonic/functional perception ... 19
3.4.6 Perceived knowledge of dynamic pricing ... 19
3.5 CONTROL VARIABLES ... 19 3.6 DATA ANALYSIS PLAN ... 20 4. RESULTS ... 22
4.1 SAMPLE DESCRIPTION ... 22
4.2 RELIABILITY AND VALIDITY MEASUREMENT ... 23
4
APPENDIX 6A: ANOVA MODEL OF CORRELATION BETWEEN GENDER AND DVS ... 44
APPENDIX 6B: ANOVA MODEL OF CORRELATION BETWEEN AGE AND DVS ... 44
APPENDIX 6C: ANOVA MODEL OF CORRELATION BETWEEN EDUCATION AND DVS ... 44
APPENDIX 6D: ANOVA MODEL OF CORRELATION BETWEEN INCOME AND DVS ... 44
APPENDIX 6E: ANOVA MODEL OF CORRELATION BETWEEN EMPLOYMENT AND DVS ... 45
APPENDIX 7. ONE-WAY ANOVA OUTPUT FOR PRICE FAIRNESS ... 46
APPENDIX 8. ONE-WAY ANOVA OUTPUT FOR UNCERTAINTY ... 47
APPENDIX 9. SUMMARY OF MODERATING RESULTS FROM ADDITIONAL ANALYSIS ... 48
5
Abstract
Discriminatory or dynamic pricing is expected to make broad inbounds in the retail sector as an attempt to maximize consumer surplus. This study investigated consumer perceptions towards dynamic pricing practices for hedonic and functional consumption occasions in grocery shopping. Moreover, the study attempts to explore the influence of dynamic pricing practices on consumers’ perceptions of price fairness and uncertainty, with behalf of their consumption occasion and individual level of uncertainty avoidance. An online survey with 6 experimental conditions was developed, distributed and completed by 213 Dutch habitants. Two one-way ANOVAs and regression were performed to test the hypothesized relationships between the constructs. The findings of the study suggest that consumers perceive a price as less fair and obtain higher feelings of uncertainty when dynamic pricing is applied regardless of the consumption occasion and individual level of uncertainty avoidance. Additionally, consumers’ price knowledge and consciousness did not have moderating effects.
Key words: Dynamic pricing, Price Fairness, Uncertainty, Uncertainty Avoidance, Price Knowledge,
6
1. Introduction
Retailers in a variety of sectors have commonly taken a one-note approach to set their prices, using costs as the primary, and sometimes the only criterion (Andersen 2000). More recently, several retailers are engaging in refined dynamic pricing software that use data from their Internet purchases or ERP systems to determine pricing strategies and set prices (Grewal et al. 2011). Dynamic pricing or discriminatory and customized pricing, often related to inter-temporal price discrimination, has become much more accepted due to developments in and growing importance of online marketing practices (Haws and Bearden 2006; Levy et al. 2004). However, up next, dynamic pricing is expected to make its appearance in brick-and-mortar stores.
7
lives, they tend to worry about fairness. Social psychologists have the propensity to think of uncertainty when one does not understand a certain occurrence well enough to predict it. (i.e. fluctuating prices). However, they often talk about uncertain events, events in which one feels that one does not comprehend meaningful aspects of the event or that one does not have enough knowledge about norms regarding the situation (i.e. are dynamic pricing practices accepted according to social norms?). Consumer researchers (e.g. Campbell 199; Xia et al. 2004), show that both a given price and the reason why a certain price is given may result in perceptions of price unfairness. However, when consumers’ perceive a certain price as unfair, this will result in negative effects for the retailer, including consumers leaving the retailer, spreading negative worth-of-mouth and employing behavior to damage the retailer, or even worse.
However, questions remain concerning the situational influences that can affect whether consumers identify prices as fair (Haws et al. 2006), and whether consumers feel uncertain about prices. In earlier research, Gardner (1971) and Monroe (1973) suggest that consumers not only respond differently to price changes due to individual differences, but also with regard to product category. Researchers investigating consumers’ price sensitivity have tried to characterize individual consumers as belonging to groups of shoppers who are more or less price-responsive, but without regard to purchase situation of purpose (Wakefield and Inman 2003). Hirschman and Holbrook (1982) argue that consumers purchase most grocery and household products mainly on the basis of functional or economic utility. However, consumers of leisure activities and more recreational products base their purchase mainly on hedonic or emotional gratification.
Wakefield and Inman (2003) examined the role of consumption occasion (hedonic and functional) in influencing price sensitivity. They argue that price sensitivity differs as a function of consumption occasion. They found that price sensitivity depends upon the consumption occasion, and their results suggest that consumers are less price sensitive when making a hedonic purchase. Hence we argue that consumers that pursue pleasure and enjoyment from product use are less concerned about dynamic pricing, compared to consumers that shop for products providing them with functional benefits.
One might argue, according to research in social psychology that this effect is similar for consumers’ feelings of uncertainty. Van den Bos and Lind (2002) agree and argue that uncertainty can be formed due to circumstances or situational factors that confront one’s feelings of uncertainty
8
known about whether contextual factors might influence consumers’ feelings of uncertainty and whether individuals’ level of uncertainty avoidance might influence consumers’ perceptions of price fairness, with regard to discriminatory pricing.
Drawing on theories of price fairness and uncertainty, this present study tries to discuss circumstances that affect consumers’ judgments regarding price fairness and feelings of uncertainty due to dynamic pricing practices in grocery stores and based on consumption occasion and uncertainty avoidance. The objective of this research is to investigate whether implementing dynamic pricing will affect perceptions of price fairness and feelings of uncertainty and, moreover, if these perceptions vary across product categories as a function of consumption occasion, based on work from Haws and Bearden (2006); Wakefield and Inman (2003) and Weisstein et al. (2013), and vary across consumers due to their individual level of uncertainty avoidance, based on work from Hofstede (2001) and Sharma (2010). Therefore, the main research question of this study will be: ‘What is the influence of dynamic pricing on consumers’ price fairness perceptions and feelings of
uncertainty, and what is the effect when consumption occasion and consumers’ level of uncertainty avoidance are considered as moderator?’
Consequently, the aim of this study is to fill the gap by combining the constructs as described in the research question. However, this present study differs from Haws and Bearden (2006), since they discuss the circumstances that influence consumers’ judgments regarding price fairness in different dynamic pricing settings, while in this study we seek to address consumption occasions that influence consumers’ judgments about price fairness in the same dynamic pricing environment. Furthermore, the relations between the other constructs are less clear, since less is known about uncertainty and uncertainty avoidance and its link with dynamic pricing practices. Therefore, this presents study seeks to gain new insights and deliver unique contributions to dynamic pricing, price fairness and uncertainty literature.
This study is continued with a review of prior literature on the constructs as described in the research question. Several hypotheses will follow from the literature, where after the conceptual model and methodology section will be described. Furthermore, the results of the performed analyses will be discussed, and the hypotheses are tested. Lastly, this study will end with a discussion, conclusion, limitations and directions for further research.
9
2. Literature review
We begin with discussing the various forms of dynamic pricing. Then, we discuss consumers’ perceptions of dynamic pricing and highlight the concept of price fairness and uncertainty. Furthermore, we discuss the role of hedonic and functional consumption occasion. Lastly, the concept of uncertainty avoidance will be explained.
2.1 Benefits of dynamic pricing
The concept of dynamic pricing has existed for some time, but its recent expansion could be explained due to advancing technology and the increased availability of demand data (Grewal et al. 2011; Haws and Bearden 2006; Elmaghraby and Keskinocak 2003). Haws and Bearden (2010) define dynamic pricing as “a strategy in which prices vary over time, consumers and/or circumstances, often referred to individual price discrimination”. Dynamic pricing is making inroads in many different sectors, such as apparel, automobiles and even personal services. It has been noted to be especially useful for price optimization, transaction efficiency, and revenue management (Friesen, 2003; Heun, 2001; Kambil, Wilson, & Agarwal, 2002). As explained in the introduction section can the properly implementation of dynamic pricing enhance retailers’ revenues and profits with respectively 8 and 25 percent (Elmaghraby and Keskinocak 2003; Sahay 2007). Retailers can improve their profitability by cutting down the costs of changing prices due to the implementation of ESLs, real-time price testing or to customize prices depending upon customers’ need, purchase frequency and behavior (Kannan and Kopalle 2001).According to Elmaghraby and Keskinocak (2003) there are two broad categories of dynamic pricing: posted prices (PP) that customers can see and accept without change or reject outright; and
price discovery (PD) mechanisms, where customers have input into setting the final price through
their own actions during the transaction. Our focus in this study is on dynamic posted-price mechanisms and specifically demand-based pricing, since prices in grocery stores are not determined by bidding processes such as auctions, and pricing strategies are determined with the goal of balancing demand and inventory to maximize profits (Elmaghraby and Keskinocak 2003).
10
2007). The practice of dynamic pricing has been applied in many industries, especially those industries where it is hard to alter the short-term supply (i.e. airlines and hotels). In retail the short-term capacity is more flexible, however changing prices is more costly. Due to new advances in technology and innovations like electronic shelf labels (ESLs) retailers are able to reexamine their pricing policies and implement dynamic pricing through software technologies for a better demand management (Garaus et al. 2016).
Figure 1. Single pricing vs. Dynamic pricing (Sahay 2007)
2.2 Consumers’ perceptions of dynamic pricing
11
dynamic pricing function, consumers’ price fairness perceptions have to be taken into account. He argued that perceived price fairness is the most important prerequisite that must be secured and monitored to make the practice work.
The mere fact that prices can differ across consumers due to dynamic pricing practices, increases the uncertainty that consumers face and the risk of making a mistake and for instance purchase when others pay a lower price (Richards, Liaukonyte, and Streletskaya 2016). However, being treated equally reduces people’s perceptions of uncertainty (Desai, Sondak, and Diekmann 2010). According to uncertainty management theory, studied by Lind and Van den Bos (2002), is uncertainty closely linked to fairness. Hence, they argue that when one is feeling uncertain or when one starts to pay attention to the uncertain elements in one’s life (e.g. discriminatory pricing) one tends to worry about fairness. Fairness and uncertainty are so closely linked to each other that it is impossible to understand the role of one construct without reference to the other (Van den Bos and Lind 2002). Because of the strong link between uncertainty and fairness we focus on both constructs in explaining consumer perceptions of dynamic pricing practices. 2.2.1 Price fairness Xia et al. (2004) define price fairness as “a consumer’s assessment and associated emotions of whether the difference (or lack of difference) between a seller’s price and the price of a comparative other party is reasonable, acceptable or justifiable”.
The past two decades, researchers have adapted several theories to gain insights of price fairness perceptions from consumers (e.g. Bolton et al. 2003; Campbell 1999; Xia, Kukar-Kinney, and Monroe 2010). A large stream of researchers have adapted equity theory and distributive justice (Gelbrich et al. 2011; Xia et al. 2010), procedural justice (Richards et al. 2016; Xia et al. 2010) the dual entitlement principle (Weisstein et al. 2013; Rotemberg 2011) and social norm theory (Garbarinio and Maxwell 2010; Maxwell and Garbarino 2010) to explain the principle of fairness with each theory addressing a distinct reason for price fairness.
12
al. 2004). Consumers will regard prices as more fair if they are aware of the rules used to set prices (Maxwell 2002). The principle of dual entitlement proposes that retailers are designated to a quotation profit, while consumers are designated to a quotation price (Rotemberg 2011). Unfairness perceptions arise if either a retailers’ or consumers’ entitlement is infringed (Bolton et al. 2003; Kahneman, Knetsch, and Thaler 1986). If consumers believe that a price increase is driven by higher demand – sunny weather raising the demand of ice cream – then the price is more likely to be viewed as unfair than if it were driven by higher costs of selling ice cream.
Social norm theory explains why airline passengers or hotel guests do not appear to mind paying different prices from others, while having nearly identical seats or rooms, while Amazon was forced to abandon their attempt to price DVDs and MP3 players the same way in 2000 (Garbarino and Maxwell 2010; Richards et al. 2016). Social norms are the comprehended rules for both consumers and retailers, and they serve as guidance for the behavior of both parties (Heide and John 1992; Maxwell 1999). A violation of the social norms has been empirically shown to lead to judgments of unfairness (Maxwell 2002). And thus violating a norm should lead to lower perceived fairness.
13
price fairness. Ordóñez, Connolly, and Coughlan (2000) predicted that fairness responds differently to advantaged and disadvantaged inequality, based on prospect theory’s prediction that “looses loom larger than gains” Kahneman and Tversky 1979). The results of their study indeed show that advantaged inequality increased fairness perceptions and disadvantaged inequality reduced it. More recently, Campbell (2007) also found evidence that a price increase is more likely to be perceived as unfair than a price decrease.
2.2.2 Uncertainty
14
the store visitors’ price sensitivity. However, Wakefield and Inman (2003) argue that consumers’ price sensitivity is dependent upon the situation or consumption occasion. Therefore, it might be expected that consumers’ reactions to dynamic pricing practices depend upon consumption occasion. Consumption has usually been represented in terms of its functional and hedonic nature (e.g. Wakefield and Inman 2003). According to Hirschman and Holbrook (1982) are hedonic goods primarily distinctive from functional/utilitarian goods on four dimensions: mental constructs, product classes, product usage, and individual differences. Batra and Ahtola (1990) explain why consumers purchase goods and services, namely for “consummatory affective hedonic gratification from sensory attributes and instrumental, utilitarian reasons.” Accordingly, in this present study we use the term functional to describe products in which consumption is mainly valued on the basis of utilitarian or functional facets. Secondly, to describe products in which consumption is mainly valued on the basis of hedonic or affective facets, we use the term hedonic (e.g. Batra and Ahtola 1990; Voss, Spangenberg, and Grohmann 2003).
To draw back to the study of Wakefield and Inman (2003) where they examined the expected effect of hedonic and functional consumption occasion on consumers’ price sensitivity. Their results revealed that consumers are less price sensitive when making a hedonic purchase compared to when making a functional purchase. However, for this study it might be expected that consumers who make a hedonic purchase are less concerned (i.e. higher perceived price fairness and lower uncertainty) about dynamic pricing practices compared to consumers who make a functional purchase. Thus, according to the theory the following is hypothesized:
H3a: Hedonic (vs. functional) consumption occasions will weaken the negative effects of dynamic pricing on price fairness
H3b: Hedonic (vs. functional) consumption occasions will weaken the negative effects of dynamic pricing on uncertainty
2.4 Uncertainty avoidance
15
16
3. Conceptual Model & Methodology
3.1 Conceptual model
Following the literature we outline the proposed conceptual model based on the variables that are considered to be related to dynamic pricing. The independent variable is considered dynamic pricing. Consumption occasion and uncertainty avoidance variables are controlled for moderating effects, whereas the measured dependent variables are perceived fairness and uncertainty.
Figure 2: Conceptual Model
3.2 Data Collection
The data of this study is gathered through developing a web-based Qualtrics survey with questions based on the variables discussed in the conceptual model. The survey is distributed among a sample of Dutch respondents through the Internet and made compatible for mobile phones (see Appendix 1-2). To ensure that all respondents have read all questions properly an “attention check” has been added. However, this attention check will be explained in more detail in the following sections.
3.3 Experimental design
17
the following anchors: “practical purpose/just for fun, purely functional/pure enjoyment” and “for a routine need/for pleasure”. As suggested by Wakefield and Inman (2003), a 7-point scale was obtained for each question. A high score indicates a higher perceived hedonic nature of the product, while a lower score indicates a higher perceived functional nature of the product. For this test we used 22 different product categories obtained from the IRI Marketing Data Set (30 categories, 47 U.S. markets and 5 years of weekly data (2001-2005)), chosen to maximize variety on a number of dimension (e.g. dollars spent per 1000 households, percent of households buying, etc.), based on the sixteen 7-point evaluative semantic differential scale identified by Osgood, Suci, and Tannenbauw (1957) (Bronnenberg, Kruger, and Mela 2008). The survey was digitally distributed among Dutch respondents. Not all product categories were included for this present pre-test, because several product categories were perceived as vague (e.g. “Frankfurters”, “Photography supplies”) or not compatible for the Dutch grocery market (e.g. “sugar substitutes”, “pest control products”). The final sample had an average of 48 respondents per product category (similar as reported by Narasimhan, Neslin, and Sen 1996). As a result “a box of chocolates” was perceived as most hedonic in nature (e.g. scored highest on the three items), while “toilet paper” was perceived as most functional (e.g. scored lowest on the three items)."
After conducting the pre-test the results were used as input for the study to test the hypotheses conducted from the literature. For the study, a 3 ✕ 2 factorial design with the three-leveled factor price level (disadvantage/no dynamic pricing/advantage) and the two-three-leveled factor consumption occasion (hedonic/functional) manipulated between-subjects was employed (see table 1). While uncertainty avoidance is tested for a moderating role. As mentioned earlier, to allocate respondents to one of the six conditions, a web-based Qualtrics survey with a randomized block design was developed and distributed among a sample of Dutch respondents. In each condition, participants were shown one of the six scenarios which involved the purchase of a box of chocolates (hedonic good) or toilet paper (functional good), each at a price of €2.00, based on averaging prices at different grocery stores in the Netherlands. Following the scenario respondents were told that when visiting the same grocery store the next day, the price of the box of chocolates or toilet paper was lower (€1.60), the same (€2.00) or higher €2.40. The three different price levels include a comparison price 20% lower (price disadvantage), 20% higher (price advantage), and identical to the price paid by the respondent. Since Blattberg, Briesch, and Fox (1995) argue that if all elasticities were reported at 20% the ability to compare and generalize results would be greatly improved.
18
making fairness judgments with regard to dynamic pricing practices and to examining respondents’ knowledge about and consciousness regarding prices in general. This is in line with Bolton et al. (2003), they suggest that consumers’ knowledge of prices account for judgments of price fairness.
Furthermore as mentioned in the former section an “attention check” was added to the survey. With this check the respondents were asked if the price of the box of chocolates or toilet paper was lower, higher or the same when visiting the same grocery store the next day. Respondents who did not select the right answer that matches their condition were deleted from the data set (i.e. respondents in the “advantage” condition had to select the “higher” option). Lastly additional information on (socio)demographic variables was collected. Table 1. Factorial design Consumption occasion Price level
Higher/disadvantage The same Lower/advantage
Functional 1 2 3
Hedonic 4 5 6
3.4 Measurement
To increase the validity of the results of the survey, several measurement scales from previous research have been used for measuring the variables related to the discussed constructs. Except for the manipulated scenario’s, where the respondent is exposed to one of the 6 conditions, all the other questions respondents were exposed to are the same across the conditions. An overview of all measurements constructs, items and sources can be found in Appendix 2.
3.4.1 Measurement perceived fairness
In order to measure the perceived fairness of respondents when being exposed to the purchase scenario, a scale developed by Weisstein et al. (2013) has been used, based on past research and adapted from dynamic pricing literature (e.g. Bolton et al. 2003; Xia et al. 2004; Grewal et al. 2004; Grewal and Baker 1994; Dark and Dahl 2003). The scale consists of three items measured on a 9-point Likert scale (from (1) strongly disagree to (9) strongly agree) (α = .92).
3.4.2 Measurement of uncertainty
19
3.4.3 Measurement of price knowledge & consciousness
To measure respondents price knowledge and consciousness six statements employed by Magi and Julander (2005) and adapted from Flynn and Goldsmith (1999) have been used. The scale consists of three items measuring price knowledge (α = .80) and three items measuring price consciousness (α = .83). All items are measured on 7-point Likert scales (from (1) strongly disagree to (9) strongly agree).
3.4.4 Measurement of uncertainty avoidance
For the uncertainty avoidance construct, the validated individual level items of Sharma (2010) for uncertainty avoidance and risk aversion have been adopted. The eight items are measured on a 7-point Likert scale (from (1) strongly disagree to (9) strongly agree) (α = .84).
3.4.5 Measurement of hedonic/functional perception
Given the subjective perception each individual has about the extent to which a product is perceived as hedonic or not and functional or not, a hedonic/utilitarian perception scale is included for the moderation analysis, in case the experimental measure does not pass the manipulation check. The hedonic/utilitarian (HED/UT) scale of Voss et al (2003) was included in the survey with a five-item hedonic scale (α = .86) for the hedonic good scenario and a five-item utilitarian scale for the functional good scenario (α = .85), both measured on a 7-point multi-item scale.
3.4.6 Perceived knowledge of dynamic pricing
In the researchers interest a five-item scale, to measure respondents perceptions of dynamic pricing practices, was developed, based on an April 2018 survey from Forrester Consulting and Revionics (Garcia 2018) about dynamic pricing practices in grocery stores. They asked consumers about which reasons they would accept when being confronted with dynamic pricing practices. Their outcomes were adjusted and used as items for this present study. The items were measured on 7-point Likert scales (from (1) strongly disagree to (9) strongly agree), to give additional insights about respondent’s perceptions of dynamic pricing.
3.5 Control variables
In addition to dynamic pricing practices, demographical characteristics of consumers could also affect price fairness and uncertainty perceptions. To ensure that there is consumer heterogeneity, the variables of gender, age, income, education and employment are therefore added as control variables.
20
3.6 Data analysis plan
The aim of this study is to obtain evidence of a causal relationship between the discussed constructs. In detail, the study tries to find evidence for the assumption that dynamic pricing practices (effect) will lead to increased negative perceptions of price fairness and uncertainty (effect), however this causal relationship might depend on the consumption occasion and/or level of uncertainty avoidance. After gathering all the data with Qualtrics, the raw data has been exported to SPSS Statistics 24. The first step in the analysis was to remove outliers, missing values and respondents who did not finish the whole survey from the dataset. Thereafter, the manipulation check output was analyzed and a descriptive analysis was performed to give an overview of the sub-sample per condition, as resulted in table 2. Table 2. Summary of the resulting data with means Conditions 1 2 3 4 5 6 Frequency (n) 28 40 38 37 25 33 13.9% 19.9% 18.9% 18.4% 12.4% 16.4% Manipulation
Hedonic No No No Yes Yes Yes
Functional Yes Yes Yes No No No
21
22
4. Results
This chapter will provide the results of the data analysis and describes all the steps prior to hypotheses testing. As explained in the methodology section a two-way MANOVA was initially meant to be used for testing the hypotheses. However the typical assumptions of a MANOVA need to be checked prior to hypothesis testing. First it is key to test if there is a linear relationship between the pair of dependent variables for each of the six experimental conditions. A scatterplot for each of the six experimental conditions revealed that there is no linear relationship between price fairness and uncertainty (see Appendix 3).Second, absence of muliticollinearity should be checked for each of the six experimental conditions by conducting correlations among the dependent variables. Pearson’s correlation values were below the threshold value │r│< .7, hence the two dependent variables are only moderately correlated (See Appendix 4). In addition to this, the variance inflation factors (VIF) and tolerance statistics for the dependent variables have been taken into account for each of the six experimental conditions. The VIF values for the six conditions were all well below 10 and the tolerance statistics were all well above .2 in line the threshold (Bowerman and O’Connell 1990; Myers 1990). Hence, based on these measures it can be concluded that there is no collinearity in the model. Therefore, since both assumptions of a MANOVA were not met, two separate one-way ANOVAs were considered a better fit for testing the model (Malhotra 2010).
4.1 Sample description
This study has reached a total of 332 respondents, however 119 respondents did not completed the entire survey and were therefore removed from the dataset. Furthermore, following the manipulation check section outcome, the respondents that did not identify the correct price according to their condition (12) were removed from the data, hence the sample size was reduced to
n = 201. Of the total sample, 91 (45.3%) are male and 110 (54.7%) are female, with an average age of
29 (Mage = 29.19 , SD = 10.55). In table 2, the demographics summary per condition is shown.
23
Table 3. Demographics overview Conditions 1 2 3 4 5 6 Age 27.9 28.5 31.1 28.5 29.9 29.3 Gender Male 57.1% 42.5% 47.4% 32.4% 56.0% 42.4% Female 42.9% 57.5% 52.6% 67.6% 44.0% 57.6% Education (%) High school 7.1 10,0 15.8 13.5 4.0 6.1 Vocational education 7.1 10,0 7.9 5.4 4.0 6.1 University of applied Sciences (HBO) 32.1 25.0 31.6 32.4 60.0 27.3 University (bachelor) 32.1 27.5 28.9 35.1 20.0 27.3 University (master) 21.4 27.5 15.8 13.5 12.0 27.3 University (PhD) 6.1 Income (%) <€20,000 60.7 62.5 52.6 59.5 48.0 51.5 €20,000 - €34,999 17.9 17.5 26.3 16.2 28.0 18.2 €35,000 - €49,999 14.3 17.5 10.5 16.2 16.0 21.2 €50,000 - €64,999 7.1 2.5 2.6 5.4 4.0 3.0 €65,000 - €79,999 2.6 6.1 >€79,999 5.3 2.7 4.0 Employment (%) Student 35.7 50.0 47.4 44.4 32.0 45.5 Unemployed 10.7 5.3 2.8 12.0 Part-time employed 17.9 27.5 18.4 25.0 8.0 12.1 Full-time employed 32.1 22.5 23.7 27.8 48.0 36.4 Retired 3.6 5.3 6.1
4.2 Reliability and Validity Measurement
Not all constructs used in this research for data collection have been used in prior studies, therefore a factor analyses with the Principal-Component-Analysis (PCA) technique was performed to test the validity and reliability of the constructs. All coefficients of the reliability and validity measurements used in this paragraph can be seen in table 4. The PCA technique determines the number of factors on their eigenvalue. The factor analyses revealed the simultaneous presence of 7 factors from all the 27 items. However, several assumptions have to be checked to see if the PCA is appropriate to use.24
analyses, it can be determined that there are seven different factors with a bigger eigenvalue than one. The total variance explained by these factors was 69.15%, which is above the threshold of 60%. Another technique used to test how well the measured variables represent the constructs, and similar to the PCA is the confirmatory factor analysis (CFA). The CFA is used to verify the factor structure of the set off observed variables. Since it is not possible to run a CFA through SPSS Statistics, the related reliability and validity are computed manually with behalf of the data through Excel. The CFA revealed that almost all factor loadings are higher than the threshold value of. 5 (3 out of 27 items loaded lower than .5). In addition, the composite reliability (CR) was computed to define the total value of the true score variance in regard to the total score variance. The composite reliabilities are .7 or higher and meet general guidelines, except for uncertainty (.606), however according to Malhtora (2010) are estimates between .6 and .7 considered acceptable (also: Hair Jr., Black, Babin and Anderson 2014). Furthermore, the average variance extracted (AVE), used to assess convergent validity, was calculated for each construct. An AVE of .5 or more indicates satisfactory convergent validity, which is the case for five of seven constructs. However, on the basis of CR alone, one may conclude that the convergent validity of a construct is acceptable, nonetheless more than 50% of the variance is due to error (AVE) (Malhotra 2010).
Lastly, by using Cronbach’s Alpha the reliability of the multi-item measures can be assessed According to the output are the seven constructs reliable and all meet the threshold of .600 (Malhotra 2010). Table 4. Reliability and Validity coefficients Construct KMO Bartlett’s test Cronbach’s alpha Average variance extracted Composite reliability No. of items Price fairness .759 .000 .92 .696 .872 3 Uncertainty .690 .000 .72 .300 .606 5 Price knowledge .721 .000 .80 .597 .816 3 Price consciousness .701 .000 .83 .507 .755 3 Uncertainty Avoidance .854 .000 .84 .418 .844 8
Hedonic perception .774 .000 .86 N.A.* N.A.* 5
Functional
perception .842 .000 .85 N.A.* N.A.* 5
*Not available since both constructs are not measured for whole population (AVE = .546 and CR = .857)
4.3 Assumptions for testing the model
25
However, conducting a parametric test like a one-way ANOVA for hypotheses testing also concerns several assumptions that have to be met (Field, Miles, and Field 2012; Malhotra 2010).
First, the presence of one or more categorical independent variables and the presence of one or more continuous dependent variables. In this case, the study contains one dichotomous experimental variable as independent variable and two continuous dependent variables.
Second, normally distributed data is required for the performance of the one-way ANOVAs. Therefore the Skewness and Kurtosis method is applied to test for normality (i.e. the further the value of skew and kurtosis is from zero, the more likely it is that the data are not normally distributed). For each measured dependent variable (see Appendix 5), all six sub-samples meet the -2/+2 value for Skewness and Kurtosis for normal distribution (George and Mallery 2010).
Last, the third assumption of parametric testing is in respect of homogeneity of variance. This assumption suggests equality of variances in price fairness and uncertainty scores across the six conditions. Using Levine’s test for equality of variances, it was established that there is a homogeneity of variances for price fairness F(5, 195) = .931, p = .529, and uncertainty F(5, 195) = .649, p = .663 (i.e. p-value indicating equal variances between groups). Concluding from the above discussion is that the data collected for this study meet all the assumptions of parametric testing and is fit for performing one-way ANOVAs.
4.4 Control variables
As previously mentioned in the methodology section is the independent variable dynamic pricing expected to correlate with the dependent variables, perceived price fairness and uncertainty. However, in this study, the variables age, gender, income, employment and education are used as control variables to account for the heterogeneity of the sample. Therefore ANOVAs were performed to test if the suspected correlation holds and the control variables need to be included in the model (see Appendices 6a-e). For all the control variables, ANOVAs revealed there is no significant difference between groups for both price fairness and uncertainty (p > .05). For this reason, the variables were not controlled for further analysis.
26
for each of the six conditions (see table 2), it may be assumed that respondents perceive a box of chocolates as hedonic and toilet paper as utilitarian/functional.
4.5 Testing the model
In the following subchapters, all the hypothesized relationships are tested upon significance and the fit of the models used for this purpose is assessed. The direct effect of the categorical independent variable was measured through between-subjects ANOVAs for each dependent variable, whereas the moderating effect was measured through a linear regression.4.5.1 One-way ANOVA price fairness
The one-way ANOVA between-subjects effects for price fairness revealed that there was a significant main effect of dynamic pricing, F(2, 200) = 11.666, p = .000 (see Appendix 7 for output). However, as hypothesized this effect is supposed to be stronger for disadvantaged consumers compared to advantaged consumers. To test which groups significantly differ from each other, the Post-Hoc test of Tukey was applied, since it was assumed that there is homogeneity of variances for price fairness. The results of the Tukey HSD revealed that price fairness scores significantly differ between disadvantaged consumers and no dynamic pricing, and advantaged consumers and no dynamic pricing (see Appendix 7 for output). Consumers’ price fairness scores are on average .745 higher for consumers in the no dynamic pricing condition compared to consumers in the disadvantaged condition, and .416 higher than consumers in the advantaged condition. The price fairness perceptions of disadvantaged consumers and advantaged consumers do not significantly differ (p = .079). Therefore, H1 can only partially be accepted (a). The partial acceptance of the hypothesis is caused by the inability to find significant differences between consumers in the advantaged and disadvantaged condition (b).
4.5.2 One-way ANOVA uncertainty
27
condition, and .641 lower than consumers in the advantaged condition. The uncertainty scores of disadvantaged consumers and advantaged consumers do not significantly differ (p = .294). Therefore H2 can only partially be accepted (a). The partial acceptance of the hypothesis is caused by the inability to find significant differences between consumers in the advantaged and disadvantaged condition (b).
4.6 Moderation analyses
The ANOVAs revealed a significant main effect from dynamic pricing, on both, price fairness and uncertainty. Hence, to find evidence for the other hypotheses (H3 and H4) moderating analyses
were performed. As reported, there were no multicollinearity issues with all the VIF scores smaller than 10 and tolerance statistics larger than .2 (Field et al. 2012; Malhotra 2010). Furthermore, all the obtained regression models were significant (p < .05). A summary of the unstandardized coefficients for the interaction effects between the constructs and their p-value can be seen in table 5. Table 5. Summary of moderating results
Hypothesis Dependent variable Interaction name Expected sign Unstandardized coefficient p-value H3a Price fairness Hedonic x Dynamic
pricing
+ -.010 .972
H3b Uncertainty Hedonic x Dynamic
pricing + .129 .483
H4a Price fairness Uncertainty avoidance x
Dynamic pricing – .002 .987
H4b Uncertainty Uncertainty avoidance x
Dynamic pricing – -.001 .839
The high p-values for all the interaction constructs communicate that the interaction effect is not significant. Hence, consumption occasion and uncertainty avoidance did not moderate the impact of dynamic pricing in this study and the hypotheses H3a, H3b, H4a and H4b are rejected. To summarize, only hypotheses H1a and H2b were accepted, all the other hypotheses were rejected.
4.7 Additional analyses
28
can be found in Appendix 9. The high p-values for all four interaction constructs communicate that the interaction effect is not significant. Hence, price knowledge and price consciousness did not moderate the impact of dynamic pricing on price fairness and uncertainty. Therefore the assumptions of Haws and Bearden (2010) and Bolton et al. (2003), that consumers’ price knowledge and consciousness may influence price fairness, cannot be proven within this study
4.8 Insights of dynamic pricing
Additional analyses regarding respondents’ perceptions of dynamic pricing practices revealed no significant interactions between the control variables and dynamic pricing practices. However, when comparing the individual means of the five items (see Appendix 10), one may assume that two out of the five items score relatively high (i.e. respondents strongly agree), while the other three items score relatively low (i.e. respondents strongly disagree). It seems that consumers agree with retailers lowering prices due to a surplus of stock (M = 5.06) or when products reach its sell-by-date (M = 6.12). However, the high scores on these items can, again, be explained by Prospect Theory’s prediction, that people place more weight on perceived gains versus perceived losses (Kahneman and Tversky 1979).
29
5. Discussion and conclusion
Building on the dynamic pricing trend and the gaps in the literature, this study aimed to identify the impact of dynamic pricing in different contexts. However out of the eight tested hypotheses of this study, six were rejected due to lack of interaction effect. But the lack of statistically significant interaction effects in this study does not mean the interaction effect does not exist in the populations (Aiken, West and Reno 1991). Nonetheless, delivering the main insights and findings of this study will over several implications for consumer research and marketing theory.
Firstly, dynamic pricing was proven to have a significant effect on price fairness and uncertainty, regardless of the consumption occasion and regardless of the level of uncertainty avoidance. Accordingly, this effect significantly differs between consumers exposed to a single price (no dynamic pricing) and consumers exposed to dynamic changing prices (advantaged and disadvantaged). Consumers perceive a one-price strategy as more fair compared to a dynamic pricing strategy. Similarly, consumers feel less uncertain when exposed to a one-price strategy compared to a dynamic pricing strategy. According to Prospect Theory, individuals do care more when they are disadvantaged compared to when they are advantaged in equity (Kahneman and Tversky 1979). Ordóñez et al. (2000) showed that advantageous inequality increased fairness, while disadvantageous inequality reduced it, echoing Prospects Theory’s prediction that ”losses loom larger than gains”. A similar pattern was expected to be found in this study, however, a significant difference between disadvantaged consumers (i.e. significantly lower scores) and advantaged consumers has not been supported by our data, while the concept of asymmetric inequity is ubiquitous in the price fairness literature (Richards et al. 2016).
However, our results can be explained by equity theory, which posits that any kind of inequality (i.e. no significant difference between advantaged and disadvantaged consumers) is perceived as unfair (Adams 1965; Gelbrich 2011). Despite, Prospect Theory’s prediction, a significant difference between disadvantaged consumers (i.e. significantly higher scores) and advantaged consumers was neither found.
30
retailers with regard to dynamic pricing strategies for hedonic products. Nevertheless, the limitations of our study will be discussed in the next chapter.
Thirdly, respondents in the dynamic pricing conditions were asked to think about possible reasons for the price change (i.e. lower/higher next day). A large stream of respondents named “discount” or “promotion” as possible reason, but reasons as “inflation” and “change in supply and demand” were also named frequently. However, one respondent noted: “some consumers do not bother monitoring the price if they have full intend to buy the product anyway”. For the grocery sector and the retail as a whole it is key to find and define those consumers, and maximize consumer surplus through dynamic pricing practices (Sahay 2007).
Lastly, the interaction of uncertainty avoidance and dynamic pricing neither had a significant effect for price fairness and uncertainty. The expectation that consumers high in uncertainty avoidance would feel more uncertain and would obtain lower perceived price fairness scores, regarding dynamic pricing, was not supported by our data. According to the literate on uncertainty management theory is uncertainty closely linked to fairness and are fairness effects stronger under conditions of high uncertainty (Lind and Van den Bos 2002). According to the experiment from Van den Bos et al. (1998) had fairness a greater impact when people felt uncertain. Therefore, it may be assumed that fairness effects result from consumer’ uncertainty regarding a certain cause (i.e. dynamic pricing), which might explain the insignificant effect found in our study. However, in the next section the limitations will be discussed in more detail
31
6. Limitations and directions for further research
Within this study certain limitations have been identified, which will be discussed in this chapter, as well as directions for further research. The first limitation of this study lies in the amount of variance explained by it. A bigger and a more demographically diversified sample might solve this problem. The number of respondents of the survey with 201 might be rather low to draw conclusions regarding the reliability of the data. In addition, due to the convenience of sampling in a digital environment, most of the respondents are young adults, students (43.8%) with a low income (56.2%) and may be less experienced in grocery shopping than middle-aged adults or seniors. Secondly, future research needs to expand beyond the hedonic and functional goods used in this study. Other consumer goods, beyond chocolates and toilet paper, may be found to moderate the discussed relationship between dynamic pricing and price fairness or dynamic pricing and uncertainty. However, according to Hunneman, Verhoef, and Sloot (2016) are fast-moving consumers goods mostly utilitarian, associating that hedonic consumption barely exist in a grocery store context. Instead, they distinguish between three types of shopping trips based on the known role of situational factors in consumer behavior. This assumption might explain the insignificant interaction effect between hedonic consumption occasion and dynamic pricing in our study. Hence, a recommendation for future research papers would be to study shopping trip type as moderating variable, instead of consumption occasion. However, another limitation of this study is the fact that it tackles solely two product categories out of a dozen product categories available in grocery stores.
Third, as suggested by uncertainty management theory are uncertainty and fairness closely linked to each other. However, in our study both constructs are studied separately, which might explain the insignificant results. Hence, a recommendation for future research would be to study the (possibly) dual relationship between uncertainty and price fairness.
According to Lal and Rao (1997) is discriminatory pricing the solution to cope with the inefficiencies inherent in traditional systems of promotional pricing. Finding solutions to some of the problems, as reviewed in this study, and implementing discriminatory pricing across a wider range of categories may be welfare improving for the (grocery) retail sector as a whole (Richards et al. 2016).
32
References
Adams, J. Stacey (1965), “Inequity in Social Exchange,” Advances in Experimental Social Psychology, 2, Berkowitz Leonard ed. New York: Academic Press, 267–299 Adamy, Janet (2000), “E-Tailer Price Tailoring May Be Wave of Future,” Chicago Tribune, (September 25), 4Aiken, Leona S., Stephen G. West, and Raymond R. Reno (1991), Multiple regression: Testing and
35
Journal of Business Strategy, 23, 40–43
Kannan, P. K., and Praveen K. Kopalle (2001), “Dynamic Pricing on the Internet: Importance and Implications for Consumer Behavior,” International Journal of Electronic Commerce, 5 (3), 63–83 Kimes, Sheryl E. (1994), “Perceived Fairness of Yield Management,” The Cornell H.R.A. Quarterly, 35 (February), 22–29 Kopalle, Praveen K., Dipayan Biswas, Pradeep K. Chintagunta, Jia Fan, Koen Pauwels, Brian T.
Ratchford, and James A. Sills (2009), “Retailer Pricing and Competitive Effects,” Journal of
36
Monroe, Kent B. (2003), Pricing: Making Profitable Decisions, 3rd ed. Burr Ridge, IL: McGraw- Hill/Irwin Myers, R. (1990), Classical and modern regression with applications, 2nd ed. Boston, NY: Duxbury Narasimhan, Chakravarthi, Scott A. Neslin, and Subrata K. Sen (1996), “Promotional Elasticities and Category Characteristics,” Journal of Marketing, 60 (April), 17–30 Ordóñez, Lisa D., Terry Connolly, and Richard Coughlan (2000), “Multiple Reference Points in Satisfaction and Fairness Assessment,” Journal of Behavioral Decision Making, 13, 329–44 Opp, Karl-Dieter (1982), “The Evolutionary Emergence of Norms,” British Journal of Social Psychology, 21 (2), 139–49 Osgood, Charles E., George J. Suci, and Percy H. Tannenbaum (1957), The Measurement of Meaning. Urbana: University of Illinois Press Rao, Raghunath S., and Richard Schaefer (2013), “Conspicuous Consumption and Dynamic Pricing,” Marketing Science, 32 (5), 786–804 Reinartz, Werner J. (2001), “Customizing prices in online markets,” European Business Forum, 6, 35– 41 Richards, Timothy J., Jura Liaukonyte, and Nadia A. Sreletskaya (2016), “Personalized pricing and price fairness,” International Journal of Industrial Organization, 44, 138–153 Rotemberg, Julio J. (2011), “Fair Pricing,” Journal of the European Economic Association, 9 (5), 952– 981 Rogers, E.M. (1995), Diffusion of Innovations, 4th ed. New York, NY: The Free Press Sahay, Arvind (2007), “How to Reap Higher Profits With Dynamic Pricing,” MIT Sloan Management Review, 48 (4), 53–60 Sanburn, J. (2012). “Delta appeared to overcharge frequent flyers for weeks-was that legal?”
37
Sweeney, Jillian C., and Geoffrey N. Soutar (2006), “A Short Form of Sweeney, Hausknecht and Soutar's Cognitive Dissonance Scale,” In Proceedings of the 20th Annual Conference of the Australian and New Zealand Academy of Management, J. Kennedy, and L. Di Milia, eds. Sydney, Australia: Central Queensland University. Thibaut, John W., and Laurens Walker (1975), Procedural Justice: A Psychological Analysis. Hillsdale, NJ: Erlbaum Associates Van den Bos, Kees, Riel Vermunt, and Henk A. M. Wilke (1997), “Procedural and Distributive Justice: What is Fair Depends More on What Comes First Than on What Comes Next,” Journal of
Personality and Social Psychology, 72 (January), 95–104 Van den Bos, Kees, and E. Allan Lind (2002), “Uncertainty management by means of fairness judgments,” Advances in Experimental Social Psychology, 34, 1–60 Voss, Kevin E., Eric R. Spangenberg, and Bianca Grohmann (2003), “Measuring the Hedonic and Utilitarian Dimensions of Consumer Attitude,” Journal of Marketing Research, 40 (August), 310–20
Weisstein, Fei L., Kent B. Monroe, and Monika Kukar-Kinney (2013), “Effects of price framing on consumers’ perceptions of online dynamic pricing practices,” Journal of the Academy of
38
Appendices
Appendix 1. Experimental conditions
Imagine that you are in a grocery store to buy toilet paper. You have decided exactly which toilet paper you will buy and found it for a price of €2.0039
Appendix 2. Overview of constructs, items, measurements and sources
Constructs and items Measure Source
43
Appendix 4. Pearson’s correlation between Price Fairness and Uncertainty
Group Condition Pearson’s Correlation Degree
1 FDAdv -.655 Moderate 2 FNDP -.387 Moderate 3 FAdv -.325 Moderate 4 HDAdv -.424 Moderate 5 HNDP -.675 Moderate 6 HAdv -.494 Moderate
Appendix 5. Skewness and Kurtosis values for normal distribution assumption
DV: Price fairness DV: UncertaintyGroup Condition Skewness Error Kurtosis Error Skewness Error Kurtosis Error
44
Appendix 6a: ANOVA model of correlation between gender and DVs
Sum of Squares df Mean Square F Sig.
Price fairness Between Groups .084 1 .084 .098 .755 Within Groups 171.911 199 .864 Total 171.995 200 Uncertainty Between Groups 1.065 1 1.065 2.288 .132 Within Groups 92.571 199 .465 Total 93.635 200
Appendix 6b: ANOVA model of correlation between age and DVs
Sum of Squares df Mean Square F Sig.
Price fairness Between Groups 29.535 38 .777 .884 .664 Within Groups 142.460 162 .879 Total 171.995 200 Uncertainty Between Groups 16.493 38 .434 .911 .620 Within Groups 77.142 162 .476 Total 93.635 200
Appendix 6c: ANOVA model of correlation between education and DVs
Sum of Squares df Mean Square F Sig.
Price fairness Between Groups 4.172 5 .834 .970 .438 Within Groups 167.823 195 .861 Total 171.995 200 Uncertainty Between Groups .546 5 .109 .229 .950 Within Groups 93.089 195 .477 Total 93.635 200
Appendix 6d: ANOVA model of correlation between income and DVs
Sum of Squares df Mean Square F Sig.
45
Appendix 6e: ANOVA model of correlation between employment and DVs
Sum of Squares df Mean Square F Sig.
46
Appendix 7. One-way ANOVA output for price fairness
Sum of Squares df Mean Square F Sig.
Between Groups 18.132 2 9.006 11.666 .000 Within Groups 153.864 198 .777 Total 171.995 200 Cell means of dynamic pricing N Mean Disadvantage 65 -.3570 Advantage 71 -.0284 No dynamic pricing 65 .3881 Total 201 .0000 Tukey HSD (I) Dynamic Pricing (J) Dynamic pricing Mean
difference (I-J) Std. Error Sig.
95% Confidence Interval Lower bound Upper bound Disadvantage No dynamic pricing -.745
*
.155 .000 -1.110 -.380 Advantage -.329 .151 .079 -.686 .029 No dynamic pricing Disadvantage .745*
.155 .000 .380 1.110 Advantage .417*
.151 .018 .059 .774 Advantage Disadvantage .328 .151 .079 -.029 .686 No dynamic pricing -.417*
.151 .018 -.774 -.059 *. The mean difference is significant at the 0.05 levelOne-way ANOVA output for price fairness with two groups (Adv + DAdv combined)
Sum of Squares df Mean Square F Sig.
47
Appendix 8. One-way ANOVA output for uncertainty
Sum of Squares df Mean Square F Sig.
Between Groups 23.214 2 11.607 32.634 .000 Within Groups 70.422 198 .356 Total 93.635 200 Cell means of dynamic pricing N Mean Disadvantage 65 .3109 Advantage 71 .1576 No dynamic pricing 65 -.4831 Total 201 .0000 Tukey HSD (I) Dynamic Pricing (J) Dynamic pricing Mean
difference (I-J) Std. Error Sig.
95% Confidence Interval Lower bound Upper bound Disadvantage No dynamic pricing .794
*
.105 .000 .547 1.041 Advantage .153 .102 .294 -.088 .395 No dynamic pricing Disadvantage -.794*
.105 .000 -1.041 -.547 Advantage -.641*
.102 .000 -.882 -.399 Advantage Disadvantage -.153 .102 .294 -.395 .088 No dynamic pricing .641*
.102 .000 .399 .882 *. The mean difference is significant at the 0.05 level. One-way ANOVA output for uncertainty with two groups (Adv + DAdv combined)Sum of Squares df Mean Square F Sig.
48
Appendix 9. Summary of moderating results from additional analysis
Dependent
variable Interaction name Expected sign Unstandardized coefficient Significance level
