The Impact of Frequency on The Relationship between Sales Promotions
offered in Large-Scale Promotional Events and Customer Spending
by
L.V. (Laurens) Kersbergen Student number: 11401257
University of Amsterdam Faculty of Economics and Business
MSc Business Administration – Marketing Track (6314M0252)
23rd of June
1st supervisor: dhr. dr. J.Y. (Jonne) Guyt 2nd assessor: dhr. K.O. (Kristopher) Keller
Statement of originality
This document is written by L.V. (Laurens) Kersbergen who declares full responsibility for the contents of this document. I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Acknowledgements
After an exciting but intensive period of six months, it is with great pleasure that today I write the last section of this study. Without the precious expertise, support and encouragement of several important people, this research would most certainly not have been conducted successfully.
First and foremost, I would like to extend my deep gratitude towards my supervisor Jonne, for his insightful comments and feedback during the entire process of writing this study. He consistently allowed this study to be my own work, but steered me in the right direction whenever he thought I needed it. His keen eye for detail has been a tremendous help, and inspired me to make the most of this study.Above all, he encouraged me to think critically and helped me to understand how I could improve my work. Thank you, Jonne. Second, I would like to thank my girlfriend Marleen for being there for me during the good times and the bad. She created the necessary interludes of distraction to clear my head when I needed these moments the most. Last but not least, I also extend my profound gratitude to my family and friends, for providing me with their unfailing support and encouragement during my entire period as a student, which I have greatly benefited from.
Laurens Vincent Kersbergen Amsterdam, June 2017
Table of contents
1. Introduction ... 6
1.1 Relevance ... 9
2. Literature review and Hypothesis development ... 12
2.1 Conceptual Model ... 12
2.2 Promotion Type and Customers’ Spending ... 14
2.3 Promotion Type and SPI ... 16
2.4 Effect of SPI on Customers’ Spending ... 16
2.5 Frequency of an Event and SPI ... 17
2.6 Frequency of an Event and Customers’ Spending ... 18
3. Methodology ... 20 3.1 Pretests ... 20 3.2 Statistical power ... 21 3.3 Data ... 22 3.4 Design ... 25 3.5 Measurement ... 26 3.6 Analysis ... 27
4. Data & Results ... 31
4.1 Data Preparation ... 31
4.2 Descriptives ... 31
4.3 Model & Hypothesis Testing ... 35
5. Discussion, Limitations and Direction for Future Research ... 39
5.1 Discussion ... 39
5.2 Limitations and Future Research ... 43
Appendices ... 46
Abstract
Large-scale promotional events have become a popular marketing tool used by traditional grocery retailers to improve the price image of the store and increase customer spending. The goal of this study is to empirically address the interplay between different factors of such event settings (promotion type, store price image, and event frequency) and customers’ spending at traditional grocery retail stores. We estimate a moderated mediation model on a dataset comprising 342 U.S. households, who reported their reactions to large-scale promotional events through an online simulated grocery shopping experiment. We find that buy-one-get-one-free discounts (free product offers) during such promotional events especially contribute favorably to the price image of a traditional grocery retail store. Free product offers through large-scale promotional events are also most effective in increasing customer spending at the store. These effects on the price image of the store and customer spending are more pronounced if the large-scale promotional event in which the free product offers are featured has a high event frequency. More specifically, we find that the indirect effect of free product offers during such promotional events on customer spending (through a positive change in store price image), depends on high event frequency. These results have meaningful implications for traditional grocery retailers who seek to improve their store price image and/or to increase customer spending. In short, for grocery retailers who have such intentions, it appears a good practice to (i) employ free product offers by means of large-scale promotional events, (ii) reinitiate these events on a regular basis, and (iii) improve the store price image and customer spending simultaneously. Our findings contribute to academic research by shedding more light on the working mechanism of large-scale promotional events, setting the stage for future study.
Keywords: large-scale promotional events, store price image, promotion type, customer spending, free product offers, price reductions, event frequency
1. Introduction
Nowadays, large-scale promotional events are a popular marketing tool used by grocery retailers to increase their customers’ spending at the store (Garstenveld 2015). Large-scale promotional events can be defined as recurring events with a predefined time span, in which product discounts are emphasized under an overarching theme, to increase current customer spending at the store (Guyt 2015). Such promotional events feature either free product offers or price reductions. Recent examples of large-scale promotional events include two major events held by the Dutch grocery retailer Albert Heijn (subsidiary of Ahold Delhaize): “Hamsterweken” and “Route 99”. During Hamsterweken, customers are encouraged to purchase groceries in bulk through a wide set of buy-one-get-one-free discounts (free product offers). During Route 99, the retailer offers a great variety of products at a discount for 99 cents (price reductions) (te Pas 2012).
Grocery retailers can be categorized based on their business model (or store format), generally into traditional grocery retailers and hard discounters. While traditional retailers (such as C1000 and PLUS) adhere to a high–low price strategy, hard discounters (such as Aldi and Lidl) typically follow a business model with regular low prices rather than promotional deals (Lourenço, Gijsbrechts, and Paap 2015). Large-scale promotional events are mainly used by traditional retailers (Guyt 2015), possibly to compete with the relatively low pricing of hard discounters (te Pas 2012; Zielke 2010). Therefore, this study specifically focuses on large-scale promotional events in traditional stores to ensure the (managerial) relevance of our findings.
With the increased popularity of large-scale promotional events, some academic studies have started examining their impact on current customer spending and have found substantial differences in the effectiveness of various types of such events. Some events show a positive effect on spending, while other events at the same retailer have insignificant effect on spending (Guyt 2015). Still, until now, the underlying reasons explaining the difference in the impact of
various large-scale promotional events, initiated by the same retailer, on current customers’ spending remain unexplored. Therefore, gaining in-depth insights into factors affecting the relationship between large-scale promotional events and customer spending can help us understand the mechanisms of these events.
One such factor may be the promotion type during the large-scale promotional event— whether it is a buy-one-get-one-free sales promotion (free product offer) or a price-per-unit sales promotion (price reduction) (Haans and Gijsbrechts 2011). Thus, in terms of the promotion type, we focus on the format of the promotion and not the size of the benefit, which will be kept constant. Put differently, by using the promotion type, we aim to capture the potential difference in the impact of a buy-one-get-one-free sales promotion and a 50% price reduction, both of which will have the same percentage of discount for customers. Extant research has shown different effects of the promotion type on spending in cases of regular sales promotions (Haans and Gijsbrechts 2011). During such promotions, the effect of free product offers or price reductions on spending varies based on customers’ specific needs (Lee 2016; Mao 2016). While large-scale promotional events feature either free product offers or price reductions, little is known about the difference in the effects of both promotion types on customer spending. Hence, this study compares the effects of these two types of events on customer spending.
The price image of the grocery store could be another factor that explains the different effects of large-scale promotional events on customer spending. Store price image (SPI) is defined here as the holistic construct that customers use to recapitulate how cheap or expensive a grocery retailer is, which comprises price level, price pleasure and price value (Bloch, Brunel, and Arnold 2003; Hamilton and Chernev 2013; Mazumdar and Monroe 1990). Several studies state that customers’ spending at a grocery retail store is the result of a positive perceived price image of the store (e.g., Chang and Wang, 2014; Lourenço et al., 2015). This suggests that
large-scale promotional events may positively influence customers’ perceived SPI, and through this perception, their spending at the store. As such, this study examines the effect of large-scale promotional events on spending by including the perceived SPI as a factor.
A third factor is the frequency of large-scale promotional events, which could explain the effects of such events on customer spending. Frequency is considered here as the number of times a large-scale promotional event occurs within a given period (i.e., the rate of recurrence) (Srinivasan et al. 2004). Earlier studies have found that a high frequency of regular sales promotions strengthens their effect on spending, mainly due to repetition or “reminder” effects—constantly reminding customers of the benefits they can secure, thereby reinforcing future spending (e.g., Leenheer, van Heerde, Bijmolt, and Smidts, 2007). While these studies have found that frequent regular sales promotions are most effective in enhancing customer spending, literature to date offers little insight into the effectiveness of frequent sales promotions by means of large-scale promotional events.
Examining the interplay between the different factors (promotion type, SPI, and frequency) and customer spending is important in order to understand the working mechanism of large-scale promotional events as several caveats arise. First, sales promotions during such events create savings potential that may result in a change in the perceived SPI and in customer spending. However, the impact of these events on the perceived SPI and how the potential change in this image relates towards customer spending remains somewhat unclear. For instance, will the items promoted during large-scale events influence customer spending directly, or will they do so through a positive change in the perceived price image of a store? Or, will the items on promotion during large-scale events influence the perceived price image of the store and customers’ spending at the store separately? Second, we expect to find a higher effect on spending in case of a more frequent event: customers will be repeatedly reminded of the benefits of the event, and as they become familiar with the concept, they will be inclined to
spend more at the store. Yet, while common sense seems to dictate that the increase in spending from sales promotions during large-scale promotional events increases with event frequency, little is known about the magnitude of the effect of this frequency. For instance, will the increase in customer spending due to the sales promotions during a large-scale promotional event be higher when the event is held twice as often? Also, will the potential change in perceived SPI be more pronounced in this scenario?
Unfortunately, to the best of our knowledge, whether or not sales promotions (i.e., free product offers or price reductions) through large-scale promotional events influence customer spending through a positive change in the SPI, and how the frequency of the event affects this mechanism are questions that have not been addressed in earlier studies, thereby hampering deeper understanding of the mechanism of large-scale promotional events. Therefore, this study attempts to fill in the gap in existing academic literature by investigating the unexplored extent of these effects. Accordingly, our research question is as follows: To what extent does the frequency of large-scale promotional events influence the relationship between type of promotion during the event and customer spending during event weeks? In addition, does the SPI mediate the relationship between type of promotion during large-scale promotional events and customer spending?
1.1 Relevance
Large-scale promotional events are thought to stimulate customer spending at traditional grocery retail stores, resulting in a rapid increase in the popularity of such events over the last few years (Garstenveld 2015). Yet, academic research on the working mechanism and effectiveness of large-scale promotional events is still fragmented and limits deeper understanding of this phenomenon in the grocery retail industry. Our paper will indicate how customers’ spending at the store is influenced by large-scale promotional events and therefore
sheds light on the usefulness of such events as a marketing tool. Thus, our paper is of both theoretical and managerial relevance which we will discuss in the following sections.
1.1.1 Theoretical Relevance
We contribute to the academic literature in several ways. First, we will provide valuable insights on which promotion type (i.e., free product offers or price reductions) offered in large-scale promotional events has most impact on customers’ SPI formation and which type on their spending. By comparing the effects of these two types of events on customers’ SPI formation and on their spending, we contribute to deeper understanding of how customers react to such promotional design elements offered in large-scale promotional events. Second, we offer evidence to academics on how customers update their perceived SPI due to sales promotions in large-scale promotional events, and whether or not this change in image explains their spending increase at the store. Looking at the underlying motivation of customers (i.e., the perceived SPI) and its link to their spending at the store, will provide valuable insights into the mechanism of large-scale promotional events. Third, we explore the magnitude of the effect of event frequency on the working of large-scale promotional events. Fourth, literature to date lacks profound causal claims on the difference in the impact of various large-scale promotional events on customers’ SPI formation and spending, due to limited experimental research. Hence, we contribute with the results of a deductive approach testing the effects of large-scale promotional events on customers within an experimental study of high internal validity. By focusing on these particular aspects, we contribute to broader knowledge of the effectiveness of large-scale promotional events.
1.1.2 Managerial Relevance
Besides the academic contribution, this study attempts to provide relevant insights for the grocery retail business in general and for traditional grocery retail stores using large-scale promotional events specifically. From a managerial perspective, these insights will prove valuable for managers of such grocery retail stores. That is, greater knowledge about the working mechanism of large-scale promotional events (and the role of promotion type; SPI; and event frequency in this mechanism) will lead traditional grocery retailers to create more effective large-scale promotional events that will actually stimulate customers’ spending at the store. Moreover, these insights will help traditional grocery retailers to validate whether they should refrain themselves from using certain sales promotions to improve SPI beliefs and customer spending−thereby assessing current conventional practice.
2. Literature review and Hypothesis development
This chapter begins with a broad overview of our conceptual model to get a first feeling of the interplay between the different factors (promotion type, SPI, and frequency) during large-scale promotional events and customer spending. These different factors, their mutual relationships, and their influence on customer spending will be discussed in greater depth in the subsequent sections.
2.1 Conceptual Model of the Relationship between the Promotion Type during Large-scale Promotional Events and Customers’ Spending
According to previous academic research, large-scale promotional events are used by traditional grocery retailers to stimulate customer spending (Guyt 2015). Spending is considered here as the change in the average weekly household expenditure of current customers at the grocery retail store, during the promotional weeks of the event (Hunneman, Verhoef, and Sloot 2015).
Large-scale promotional events offer a multitude of monetary discounts on selected products under a common theme, and take the form of either free product offers (e.g., buy-one-get-one-free) and price reductions (e.g., 50% cut on the original price). In this study, we propose that free product offers have a stronger effect on customers’ spending at the store than price reductions do (Chandon, Wansink, and Laurent 2000). In contrast, price reductions are expected to have the most impact on the perceived SPI (Cox and Cox 1990). As we anticipate that free product offers and price reductions will have different impacts, this study examines both types of promotions and their effects on the perceived SPI and customer spending.
Based on findings of extant research (e.g., Chang and Wang, 2014; Lourenço et al., 2015), we expect to find that the perceived SPI will have a positive influence on customers’ spending at the store. In fact, we anticipate that the increase in customer spending from sales
promotions offered in large-scale promotional events is the result of a positive change of the perceived SPI (Lichtenstein, Ridgway, and Netemeyer 1993). As such, the mediating effect of SPI on the positive relationship between the promotion type during large-scale promotional events and customer spending will be studied in an explanatory manner.
Due to the discussed repetition effects, we anticipate a positive influence of the frequency of the event on the relationship between free product offers through large-scale promotional events and customer spending (Leenheer, van Heerde, Bijmolt, and Smidts, 2007). Furthermore, we propose that frequent price reductions by means of large-scale promotional events will have a stronger positive impact on the perceived SPI than infrequent versions (Alba et al. 1994). Therefore, we will examine the moderating effect of frequency on both relationships separately.
Fig. 1 presents our framework for studying the relationship between the promotion type during large-scale promotional events and customers’ spending at the store. We will zoom in on the different factors (i.e., promotion type, SPI, frequency), their mutual relationships, and their influence on customer spending in the following sections.
FIGURE 1
Conceptual Model Demonstrating the Relationship between the Promotion Type during Large-scale Promotional Events and Customers’ Spending at the Store
Store Price Image Customers’ Spending Promotion type H1: + H2: + H3: + H4a: + Frequency of event H4b: +
2.2 Promotion Type and Customers’ Spending
During a period of sales promotion, companies can use different types of promotion to advertise products. In general, a distinction can be made between monetary and non-monetary promotions. Monetary promotions directly influence the cost-benefit relation of a product by reducing the price (e.g., 50% discount on original price) or by offering a higher number of products for the same price (e.g., buy one get one free) (Chandon, Wansink, and Laurent 2000; Haans and Gijsbrechts 2011). Non-monetary promotions do not offer a direct monetary benefit but provide customers with other benefits such as entertainment (e.g., free sweepstakes) (Chandon, Wansink, and Laurent 2000). Monetary promotions are particularly effective in producing short-term effects on customers’ spending behavior (Büttner, Florack, and Göritz 2015). Moreover, such promotions have a higher influence on the attractiveness of the offer than non-monetary promotions do (Chandon, Wansink, and Laurent 2000). In large-scale promotional events, traditional grocery retailers typically use monetary promotions on the promoted products (Guyt 2015), either by providing a straight out price cut on selected products across different product categories or by offering more products for the same price (Guyt 2015). Therefore, this study specifically focuses on the effects of monetary promotions on customers’ spending at the grocery store during the promotion period of large-scale promotional events.
Customer’s response with regard to the type of promotion tends to be formed on the basis of the benefits the promotion provides (Bijmolt, van Heerde, and Pieters 2005; Chandon, Wansink, and Laurent 2000). Thus, to predict the effectiveness of a particular type of monetary promotion, such benefits should be considered (Chandon, Wansink, and Laurent 2000). A distinction can be made between utilitarian and hedonic benefits. A sales promotion can be called utilitarian when it helps customers to maximize the utility of their shopping, and hedonic when it offers fun and increases self-esteem (Chandon, Wansink, and Laurent 2000). A grocery retail store is primarily stocked with products that are orientated towards satisfying the
utilitarian needs of customers (Hunneman, Verhoef, and Sloot 2015). By satisfying the utilitarian needs of current customers, grocery retailers can encourage their spending at the store (van Heerde and Neslin 2008).
As mentioned, monetary promotions can be divided into two different promotion techniques: free product offers and price reductions. There are several reasons to expect that free product offers have a higher positive influence on spending than price reductions. First, previous research shows that free product offers have stronger perceived utilitarian benefits than price reductions do (Chandon, Wansink, and Laurent 2000). Providing free product offers to customers in the store will give these customers the possibility to maximize the utility of their shopping. As customers are typically motivated to maximize the utility, they are likely to spend more at the store in the presence of free product offers. Second, price reduction offers are vulnerable to negative quality inferences, while framing a discount in free gifts will maintain quality perceptions and increase deal value (Darke and Chung 2005). Thus, free product offer frames increase the perceived value of a deal, suggesting a positive effect of free product offers on those who already buy at the store (Haans and Gijsbrechts 2011). It is important to note, however, that given the way we construct the “promotion type” variable, 50% price reductions will by default have a lower impact on customer spending than free product offers. After all, in case of a 50% price reduction, the items on promotion will be sold for only half of their regular price, and thereby limiting the impact of price reductions on customer spending. In sum, free product offers by means of large-scale promotional events are expected to have a stronger positive influence on customers’ spending at the store, even if they offer similar economic value. Accordingly, our hypothesis is as follows:
Hypothesis 1. Free product offers have a stronger positive influence on spending than price reductions.
2.3 Promotion Type and SPI
Several studies show a stronger positive effect of price reductions on SPI than free product offers. First, customers perceive a store to have lower overall prices if sales promotion prices are advertised as reductions from higher regular prices (Cox and Cox 1990), instead of free product offers. The advertised references prices in the ad helps to create a low SPI, because customers generalize from the group of advertised prices to make inferences about the store’s overall price level (Cox and Cox 1990). Second, the free product offer format may be considered by current customers as more of a hurdle, because it imposes a quantity restriction (Foubert and Gijsbrechts 2007) and therefore signals a less appealing deal than a straight out price reduction. In this sense, a smaller fraction of current customers will be convinced to engage in a purchase, leading them to spend less at the store (Wansink, Kent, and Hoch 1998). Based on this discussion, we expect to find a stronger positive effect of price reductions offered in large-scale promotional events on SPI than free product offers have. Thus, our second hypothesis is as follows:
Hypothesis 2. Price reductions have a stronger positive influence on SPI than free product offers.
2.4 Effect of SPI on Customers’ Spending
According to Lichtenstein and Bearden (1989), price promotion practices of a grocery retailer are an important factor whereupon customers’ form perceived price standards and perceptions towards the grocery retail store. Moreover, the price image of a product has a dominant role in customers’ purchase behavior process (Lichtenstein, Ridgway, and Netemeyer 1993). During the purchase behavior process, customers tend to use a “standard” to assess the attractiveness of an offer by comparing the current price with the reference price (i.e., the standard) (Kalyanaram and Winer 1995). A favorable comparison of prices positively
influences the SPI, because the consumer will evaluate the offer of the store as being a good deal (Neslin and Heerde 2009). Customers’ perception on the benefits rising from a positive price-quality assessment increases their intention to spend at a store (Diallo 2012), because it influences customers’ satisfaction and subsequent purchase behavior (Erdil 2015). Furthermore, Chang and Wang (2014) show that a positive relationship between components of SPI (price level, price pleasure and price value) and customers’ spending at the store exists (Chang and Wang 2014). In addition, Zielke et al. (2010) show that value for money is an important determinant of SPI, and provide empirical support for the notion that value for money positively influences spending at retail stores (Zielke 2010). So, based on the discussion above, we can expect to find that customers of the grocery retail store will be influenced by SPI perceptions in their decision process to start spending more at the store. Therefore, our third hypothesis is as follows:
Hypothesis 3. The SPI relates positively to customers’ spending
2.5 Frequency of an Event and SPI
Another event characteristic that is likely to have an impact on the effectiveness of a large-scale promotional event is its frequency. There are several reasons to expect that the frequency of the large-scale promotional event strengthens the relationship between the type of promotion offered during the event and the perceived SPI. First, frequent price reductions generate more attention from traditional grocery retailers’ current customers, and trigger a more thorough processing of the information presented because of rehearsal (Desai and Talukdar 2003; Kardes 1994). Second, pricing literature indicates that customers take little notice or may not remember (the magnitude of) a price reduction. Thus, the simple awareness and recognition created through frequent price reductions by means of a large-scale promotional event may
have more influence on SPI formation than infrequent price reductions (Chang and Wang 2014). Third, the frequency of a price advantage has a dominant influence on customers’ price perceptions (Alba et al. 1994). That is, the power of repetition could lead to a reputation among customers of being less expensive than competitors, even if this is not really the case (Hamilton and Chernev 2013; van Heerde, Gijsbrechts, and Pauwels 2008). This leads to the assumption that the frequency of an event strengthens the relationship between the promotion type offered in large-scale promotional events and the price image of the store. Therefore, our fourth hypothesis is as follows:
Hypothesis 4a. The frequency of an event strengthens the relationship between promotion type and SPI
2.6 Frequency of an Event and Customers’ Spending
Several studies indicate a positive effect of frequency on the relationship between sales promotions in large-scale promotional events and customers’ spending at the store. First, due to their recurring nature, large-scale promotional events may educate current customers to capitalize on the savings offered during the promotion weeks (Guyt 2015). That is, exposing current customers frequently to in-store ads of the large-scale promotional events may stimulate customers to learn about the concept and timing of the event. As a result, current customers may be inclined to spend more during the event, because they are familiar with the concept. Second, customers are likely to postpone their spending until the event-weeks, if they feel certain that future promotional events will occur (Sun 2005). Third, sales promotions of products at retailers that have not been on a deal for a long period of time have proven to be less effective, because current customers will not be alert to sales promotions at these retailers (Guyt and Gijsbrechts 2014). Thus, by frequently organizing large-scale promotional events
current customers will be stimulated to be more alert on sales promotions at the store, making it more likely that they will spend during the event-weeks. This leads to the assumption that a higher frequency of the event strengthens the relationship between the type of promotion offered by means of a large-scale promotional event and customers’ spending at the store during the event-weeks. Accordingly, our fifth hypothesis is as follows:
Hypothesis 4b. The frequency of an event strengthens the relationship between promotion type and customers’ spending
In addition, the mediating effect of the perceived SPI on the positive relationship between free product offers or price reductions offered in large-scale promotional events and customer spending will be studied in an exploratory manner.
3. Methodology
This chapter elaborates on the methods of the analyses, beginning with a discussion on the pretests conducted prior to the experimental study. The next section outlines the statistical power issues, along with the data collection and sample selection methods. This is followed by a description of the experimental research design and the method of measurement. The final section describes the research method used to analyze the outcomes of our research.
3.1 Pretests
Prior to the main experiment two pretests were conducted to ensure that the results of this study are feasible and realistic, thereby increasing the external validity of this research. The first pretest was conducted to determine which grocery products were most appropriate and suitable to be used in the main experiment. Given the great variety of grocery products, those that were to be considered for the first pretest had to fulfill the following criteria: (1) offered by traditional grocery retail stores in the U.S., (2) featured at least once in a previous large-scale promotional event, and (3) listed in the top ten of most frequently purchased products– at least one traditional grocery retailer. Moreover, we selected at least two products from each product category to avoid potential biases on the outcomes of our experiment. Based on these criteria, a pre-selection of a total of 25 different grocery products was done. Next, a total of 39 participants were acquired by distributing the survey through Amazon’s Mechanical Turk (MTurk) platform. Participants were exposed to the 25 preselected grocery products and were asked to rate the extent to which these products were relevant to them on a personal level. All questions were anchored on a 7-point Likert scale (1 = very irrelevant, 7 = very relevant). Based on the outcomes of the pretest, the eight most relevant grocery products (see Appendix A) were considered suitable and hence included in the main experiment.
Shortly after the first pretest, the second pretest was conducted to test the proposed manipulations and measurement scales, and to rule out potential errors and ambiguities of the experiment. Based on the outcomes of the second pretest, several minor adjustments were implemented to improve the manipulations of event frequency and the two promotion types offered during the large-scale promotional event (free product offers and price reductions).
3.2 Statistical power
While the topic of statistical power is not covered in all academic research studies, its importance is extensively emphasized by several methodologists (e.g., Cohen, 1992 and Fritz and MacKinnon, 2007). Hence, we determined the minimum final sample size needed to detect the proposed effects and to reach sufficient statistical power. According to Cohen (1992), a statistical power of .80 is most commonly used, meaning that a probability of 80% is reached to detect certain relations between variables, if this relation really exists.
To determine the minimum sample size that leads to .80 statistical power at a .95 confidence interval when estimating least squares regression models, the first step is to set the population effect size (Cohen 1992). Following Cohen (1992), the population effect size for least squares regression equals 𝑓# = %#
&'%# . As this part is particularly difficult to specify, we
use Cohen’s conventional criteria (Cohen 1992). Cohen (1992) states that a medium effect size is “an effect likely to be visible to the naked eye of a careful observer. It has been noted that a medium effect size approximates the average size of observed effects in various fields.” Hence, in this study, we assume a medium effect size (.15), which leads to R = .1304 when solving the equation as mentioned above. Thus, the minimum sample size for estimating least squares regression to determine the direct effects of free product offers or price reductions on SPI and customer spending is 84.
Fritz and Mackinnon (2007) demonstrated the minimum sample size for .80 statistical power in cases of moderated mediation models with the percentile bootstrapping method. The minimum sample size depended upon the effect size for the paths ai and bi (see Fig. 2),
corresponding to the criteria of Cohen (1992) (Fritz and MacKinnon 2007). In this study, we assumed that the effect of free product offers or price reductions on customer spending through SPIs will be of medium size. Using the empirical power tables of Fritz and Mackinnon (2007), a minimum sample size of 78 or larger, if measurement error is present, was required for .80 power. It is important to note, however, that if the effect size is small, the minimum sample size rises to at least 408.
3.3 Data
To examine the effects of free product offers or price reductions offered during large-scale promotional events on customer spending, we conducted an experimental study of the grocery retail market of the U.S. In this market, where multiple supermarkets can be visited relatively easily, consumers have high visit and purchase frequencies (Food Marketing Institute 2016). Hence, the grocery retail market of the U.S. presents a legitimate field, based on which the effectiveness of large-scale promotional events can be measured.
Data was collected by means of a realistic online store experiment, which has several advantages over scanner data or traditional surveys (such as greater control and flexibility at lower costs; Burke 1996). In terms of external validity, a growing number of academic researchers show that online simulated shopping experiments can be used as good predictors of actual buying behavior (e.g., Burke et al. 1992; Campo, Gijsbrechts, and Guerra 1999).
The online store experiment consisted of three modules: (1) a short introduction to provide pre-purchase instructions (2) a purchase simulation module and (3) a short post-purchase questionnaire to collect data on SPI and demographics (age, gender, income). In the
first module, participants were welcomed and provided with a definition of large-scale promotional events. General information on the length and data privacy issues of the experiment was also outlined. To stimulate response without endangering the representativeness of our sample, participants were told they were eligible for winning the selected grocery products on a lottery basis. Next, participants were randomly assigned to just one of the five different treatments and told to put themselves in a hypothetical scenario in which they had to purchase groceries during four weeks at a fictitious retail store called “Alpha Market”. Respondents were informed about their shopping budget of $80 ($20 per store visit) and told that they could take the remaining amount of the money to one of the subsequent shopping trips (to increase realism, as consumers can now “overspend”, but are also encouraged to shop wisely, more accurately reflecting an average grocery trip). Depending on the scenario, the corresponding promotion type (free product offers vs. price reductions) Alpha Market normally offers during large-scale promotional events was emphasized repeatedly in the instructions, as well as the average frequency of these events (three weeks vs. six weeks). The instructions also explicitly stated the likelihood of event reoccurrence (extremely unlikely vs. extremely likely) to reduce the cognitive effort of participants to determine the event frequency. During the online purchase simulation module, participants were asked to purchase grocery products by dragging and dropping the items into an online shopping basket. Although these purchases were fictitious and thus not restricted by any financial or time constraints, several cues were provided to increase realism. First, participants were reminded not to be obliged to buy every week. Second, the eight preselected products were shown to participants in a random order, and only 50% of these products were at a discount. In case of a large-scale promotional event, either a free product offer (buy-one-get-one-free) or a price reduction (50% discount) sign was embedded in the product image using the overarching theme of the event (see appendices B1 and B2). Third, the likelihood of another event during one of the subsequent
shopping trips was again stated in the instructions. Fourth, the preselected most popular items were equally divided over the four shopping lists that were exposed to participants.
In the third module, participants were asked how the likelihood of enjoying discounts through a large-scale promotional event at Alpha Market arose during one of their next store visits. Next, participants were told to provide information on their perception of the SPI of Alpha Market as well as some demographics. In addition, the relevance of the products bought during the shopping process was noted.
The experiment was randomly distributed through an online survey among U.S. consumers who together form a representative sample of the U.S. population. To get a representative sample, respondents were contacted through Amazon’s online crowdsourcing platform, MTurk. Researchers (termed ‘requesters’ by Amazon) upload tasks to MTurk, that are to be completed by respondents (so-called ‘workers’) in return for a small monetary compensation. MTurk is currently the most popular crowdsourcing platform for research purposes and has been proven extensively to be a reliable, efficient and cost-effective source of high-quality consumer data (Peer et al. 2017). Several studies show that, compared to traditionally collected samples (such as typical college samples), ‘workers’ at MTurk are more demographically diverse and a better representation of the general population (e.g., Crump, McDonnell, and Gureckis 2013; Goodman, Cryder, and Cheema 2013).For these reasons, we considered MTurk as the most appropriate platform to recruit participants for our research.
The total sample consisted of 411 respondents. In some cases, the data recorded seemed unreliable, as some respondents completed the experiment in less than two minutes. Moreover, most of these respondents answered (almost) all of the Likert statements with the same answer. These respondents were considered as outliers, because they probably answered the questions without even reading the statements. Therefore, respondents who had a duration of two minutes (or less) and those who gave the same answers to all the Likert statements 90% of the time were
excluded from the dataset. Following this rule, 59 respondents were considered as outliers and thus removed from the dataset. Although this decreased the sample size by 14.4%, removing those outliers expectedly increased the data quality, which in turn led to more reliable results.
In addition, several outliers were identified and removed from the sample by using the “outlier labeling” rule of Hoaglin and Iglewicz (1987). Hoaglin and Iglewicz (1987) provide a resistant rule of thumb in order to identify potential outliers in datasets by utilizing the lower and upper fourths 𝐹) and 𝐹* (i.e., approximate quartiles), where any observations below 𝐹)− 1.5(𝐹*− 𝐹)) or above 𝐹*+ 1.5(𝐹* − 𝐹)) are labeled as “outlier”. Following this rule, a total of 10 outliers had been identified and removed from the sample, leading to a final sample of 342 respondents (Hoaglin et al. 1986).
Out of the 342 respondents in the final dataset, 65.2% were female and 34.8% were male. Most of the respondents were between 26 and 35 years of age (34.2%). These characteristics are comparable to the sample profiles in earlier studies on customer spending at traditional grocery retail stores (e.g., Degeratu, Rangaswamy, and Wu 2000; Rohm and Swaminathan 2004). The distribution of respondents on household income is presented in Appendix D.
3.4 Design
The research design for examining the impact of frequency on the relationships between free product offers or price reductions offered in large-scale promotional events and customer spending was a 2 (free product offers vs. price reductions) X 2 (high vs. low frequency) + 1 control group between-subject factorial design. Thus, the experiment consisted of four manipulated groups and one control group. The control group was not exposed to a large-scale promotional event. Table 1 presents an overview of the research design used to examine the
impact of event frequency on the relationships between free product offers or price reductions by means of large-scale promotional events and SPI or customer spending.
TABLE 1
Research Design Demonstrating the 2 (Free Product Offer vs. Price Reduction) X 2 (High vs. Low Event Frequency) + 1 Control Group between-subject experiment
Promotion frequency
Promotion type High Low
Free product offer high event frequency Free product offer, low event frequency Free product offer, Price reduction high event frequency Price reduction, low event frequency Price reduction,
Control group: no event condition
3.5 Measurement
This study entails the independent variables “promotion type” and “frequency”, the mediating variable of “SPI”, and “customers’ spending” as the dependent variable. The variable “SPI” was directly measured, whereas “promotion type”, “frequency” and “customers’ spending” were manipulated in the experiment. The operationalization and measurement of the variables will be discussed in greater depth below.
Promotion type. We manipulated the promotion type by exposing participants to a list
of grocery products images, some of which were embedded with either free product offers (buy-one-get-one-free discounts) or a price reduction (50% off discount), reflecting the overarching theme of the large-scale promotional event. The control group was not manipulated with any large-scale promotional event theme. Instead, several regular free product offers or price reductions were shown to participants in the control group by embedding the discounts in the grocery product images.
SPI. As discussed, the perceived SPI comprised customers’ perceptions of price level
on the academic scales ofZielke (2006), as presented in Appendix C. All items were anchored on 7-point Likert scales (1= strongly disagree, 7= strongly agree). All items were tested in terms of internal consistency (α = .843), which showed high internal consistency. Removing an item did not increase the internal consistency. Therefore, a new variable was created by taking the average of the items.
Customers’ Spending. The total number of items bought and the unit price were
recorded. Based on the average hypothetical basket size, customer spending was calculated (i.e. unit price multiplied by the number of units purchased).
Frequency. Frequency was manipulated by explicitly stating in the pre-purchase
instructions the average frequency (every three weeks vs. six weeks) of large-scale promotional events at Alpha Market. Whereas large-scale promotional events that were normally reinitiated every six weeks by Alpha Market were considered to have a low event frequency, those that were held every three weeks were operationalized as events with high event frequency. To reduce the cognitive effort of participants to determine whether the large-scale promotional event had a low or high event frequency, the pre-purchase instructions also explicitly indicated the likelihood (extremely unlikely vs. extremely likely) of a large-scale promotional event during the next grocery shopping trip at Alpha Market. Our manipulation of frequency was based on Guyt (2015), who documented the average event frequency across multiple traditional grocery retailers, by using GfK scanner panel data and verification of industry experts.
3.6 Analysis
We considered a moderated mediation model (see Fig. 2) to estimate the effect of free product offers or price reductions offered in large-scale promotional events on customer spending directly, as well as indirectly through SPI, with both direct and indirect effects moderated by event frequency. The effect of free product offers or price reductions by means
of such promotional events on SPI was modeled as moderated by event frequency. Doing so yielded the direct effect and first stage moderation model as documented by Edwards and Lambert (2007), or “Model 8” as discussed in Hayes (2013). The model was given by:
𝑆𝑡𝑜𝑟𝑒𝐼𝑚𝑔 = 𝑎9+ 𝑎&:𝑃𝑟𝑖𝑐𝑒𝑅𝑒𝑑 + 𝑎&@𝐹𝑟𝑒𝑒𝑃𝑟𝑜𝑑 + 𝑎#𝐹𝑟𝑒𝑞 + 𝑎B:𝐹𝑟𝑒𝑒𝑃𝑟𝑜𝑑 𝐹𝑟𝑒𝑞 + 𝑎B@𝑃𝑟𝑖𝑐𝑒𝑅𝑒𝑑 𝐹𝑟𝑒𝑞 + 𝑒 (1) 𝐶𝑆𝑝𝑒𝑛𝑑 = 𝑏9+ 𝑐&:G 𝐹𝑟𝑒𝑒𝑃𝑟𝑜𝑑 + 𝑐 &@G 𝑃𝑟𝑖𝑐𝑒𝑅𝑒𝑑 + 𝑐#G𝐹𝑟𝑒𝑞 + 𝑐B:G 𝐹𝑟𝑒𝑒𝑃𝑟𝑜𝑑 𝐹𝑟𝑒𝑞 + 𝑏 &𝑆𝑡𝑜𝑟𝑒𝐼𝑚𝑔 + 𝑒 (2)
where both 𝑎9 and 𝑏9 are intercept terms and 𝑒 are normally distributed errors; CSpend is the
total spending per customer at the store during event-weeks in US dollars; PriceRed is a binary dummy variable indicating whether price reductions were offered through a large-scale promotional event (1) or not (0); FreeProd is a binary dummy variable indicating whether free product offers were presented by means of a large-scale promotional event (1) or not (0);
StoreImg indicates the perceived SPI comprising price level and price value; Freq is a binary
dummy variable indicating whether a large-scale promotional event was offered in a high (1) or low (2) frequency.
FIGURE 2
Moderated Mediation Model (no. 8 from Hayes, 2013) Applied to This Study
X = Promotion type (FreeProd or PriceRed) M = Store Price Image (StoreImg) Y = Customer Spending (CSpend) W = Event- frequency (Freq)
In this moderated mediation model, the conditional direct effect of promotion type on customer spending was captured by 𝑐&G + 𝑐
BG𝐹𝑟𝑒𝑞, while the indirect effect of promotion type
on customer spending through SPI was the product of the conditional effect of promotion type on SPI from Eq (1) and the effect of SPI on customer spending controlling for promotion type in Eq (2): (𝑎&I + 𝑎BI𝐹𝑟𝑒𝑞)𝑏&I. Thus, these direct and indirect effect were conditional, as both effects depend on the value of the moderator (Hayes 2013).
As our independent variable was multicategorical after including the control group as covariate into our analysis (free product offers, price reductions, no-event) and dummy-coded as such, we adopted the terms relative direct effect and relative indirect effect when referring to 𝑐KG and 𝑎
K𝑏. That is, the results of our analysis always quantified the effect of being in one
treatment group relative to the reference group. For instance, in our dummy coding system, (𝑎&I + 𝑎BI𝐹𝑟𝑒𝑞)𝑏&I was the indirect effect of free product offers or price reductions offered in large-scale promotional events on customers spending via SPI, while being in group 𝑖 relative to the reference group. As such, the one parameter estimate could not be interpreted as the total effect of X. Instead, the relative total effect was estimated by analyzing the difference between 𝑘 − 1 dummy variable groups on Y, relative to the reference group (Hayes and Preacher 2014). Although the most widely-used method to test for hypotheses about intervening variable effects is the causal steps approach of Baron and Kenny (1986), this method has also been criticized recently (see e.g., Fritz and MacKinnon 2007). Above all, the fact that “the existence of an indirect effect is inferred logically by the outcome of a set of hypothesis tests” hindered researchers in actually detecting indirect effects through the causal steps approach (Hayes 2009). To tackle this issue, Hayes (2009) recommended to use the bootstrapping method, which bases the inferences on an estimate of the indirect effect itself (Hayes 2009). Among many other reasons, the bootstrapped method is preferred mainly because: (i) it makes no assumptions on the shape of the distribution of the indirect effect, (ii) no standard error is needed to make
the inference, (iii) it can be used for making inferences about indirect effects in any intervening model no matter how complex the relations between X and Y (iv) it is implemented in simple-to-understand software (Hayes 2015). Moreover, numerous academic studies have demonstrated that the bootstrapping method is, indeed, the most valid and powerful method for testing inferable effects (see e.g., MacKinnon, Lockwood, and Williams 2004; Williams and MacKinnon 2008). Based on the discussion above, we decided to conduct the bootstrapping method to test for the moderated mediation effects in this study, with bias-corrected 95% confidence intervals of 5000 bootstrapped samples. In doing so, we used the PROCESS macro for SPSS developed by Hayes (2013). The PROCESS macro was designed to estimate conditional process models in a single command, and produces a bootstrap confidence interval for the index of moderated mediation. According to Hayes (2015), when confidence intervals do not include zero, one can statistically infer that the relation between the indirect effect and the moderator is not zero−indicating moderated mediation (Hayes 2015). Thus, if the indirect effect of sales promotions offered in large-scale promotion events on customer spending (through SPI) depends on event frequency, the confidence interval will not include zero.
4. Data & Results
This chapter elaborates upon the results of the analysis, and begins with describing the preparation of data. We then discussed some descriptives of the data, which sets the stage for testing our model to empirically address the hypotheses of this study.
4.1 Data Preparation
Prior to the analysis, the data was prepared through several steps. As a first step, a frequency test was conducted in order to check for missing or invalid data. All 342 participants of the final data set had correctly completed the entire experiment. Thus, no additional participants were excluded from the analysis. In the second step, the variables included in our analysis were encoded as discussed in the methodology section of this study. Moreover, to organize the final dataset, each experimental condition was allocated to a treatment group number (e.g., condition 1 = treatment group 1). Appendix D presents an overview of the sample distribution, resulting from the random allocation of participants to the groups. The third step involved testing the mediation and dependent variables on heteroscedasticity and normality. Levene’s test for equality of variables was conducted to test for heteroscedasticity and no problems were detected (see Appendix G). Normality plots of both variables are shown in Appendix G, approx. indicating a normal distribution. Thus, the data seemed appropriate for further analysis and no transformation of variables was required.
4.2 Descriptives
Table 1 reported some basic descriptives for the data. It presented the mean and standard deviations of the purchase volume and spending for all five treatment groups. On average, customers that were not exposed to any large-scale promotional event bought roughly two items, while those that were exposed to such events typically purchased more than four items.
Moreover, customers not only purchased more items during event weeks, but also spent more on average. The average amount spent during event-weeks ranged from a minimum of $14.08 to a maximum of $26.84, and differed markedly between the two promotion types offered. For instance, customers exposed to free product offers spent between $19.30 and $26.84 on average, whereas the average spending of those that were shown price reductions ranged from $14.08 to $16.17. This underlines the need to take the different promotion types during large-scale promotional events into account while examining the impact of such events on customer spending.
TABLE 1:
Descriptives for Purchase Volume and Customer Spending (N=342)
Treatment groupa
No. of Items Purchased Customer Spending
Mean St. Dev. Mean St. Dev.
1 5.3250 1.14781 26.8378 3.49060 2 4.9103 .74181 19.2963 1.82925 3 4.1000 1.09545 16.1743 1.72617 4 3.9948 1.08869 14.0751 1.77332 5 2.1194 .94606 9.8419 3.01307 Total: 4.0899 1.01847 17.2612 6.04382
a1(free products, high frequency), 2(free products, low frequency), 3(price reductions, high frequency), 4(price reductions, low frequency), (5 no event: control)
A closer look into the descriptive statistics of customer spending also revealed differences across treatment groups depending on event frequency. On the whole, it seemed that customers spent relatively more when large-scale promotional events were likely to reoccur frequently, compared to situations in which such promotional events hardly (if ever) occurred. For instance, the average customer spending in cases of frequent free product offers through large-scale promotional events was $26.85, whereas customer spending during infrequent
versions of the same event was $19.30. Interestingly, customers that were exposed to such promotional events appeared to spent even more ($6.85) than their initial mental shopping budget of $20. This may point to unplanned or impulse purchases, possibly because these customers were unable to plan all the items they needed to buy (Stilley, Inman, and Wakefield 2010), leading them to rely more on external cues such as those generated by free product offers during large-scale promotional events. On the contrary, customers that were not exposed to sales promotions through large-scale promotional events spend roughly half of their shopping budget.
TABLE 2:
SPI Descriptives (N=342)
Treatment groupa
SPI
Meanb Median St. Dev.
1 5.0990 5 .80929 2 4.6650 5 .89428 3 4.4634 5 1.06350 4 4.3541 4 .96216 5 3.9462 4 1.03305 Total: 4.5855 5 .97384
a1(free products, high frequency), 2(free products, low frequency), 3(price reductions, high frequency), 4(price reductions, low frequency), (5 no event: control) b 7-point Likert scale where 1 = extremely negative SPI, 7 = extremely positive SPI
Table 2 displayed the mean rating of the perceived SPI of the five treatment groups. Overall, customers exposed to sales promotions through large-scale promotional events seemed to have a more positive SPI than those who were not exposed to such events: SPI increased by at least .4079 units, from 3.9462 to 4.3541 on a 7-point scale. Clearly, the average SPI varied across treatment groups, ranging from a minimum of 3.9462 to a maximum of 5.0099. Whereas the observed average SPI tends to be most positive if customers are exposed to free product
offers by means of large-scale promotional events, offering price reductions through such events appears to hamper the positive SPI formation. This underscores the need to take both types of promotions into account when analyzing the effects of large-scale promotional events. To give a feel of the interplay between promotion type (FreeProd or PriceRed), frequency (Freq), SPI (StoreImg), and customer spending (CSpend), we calculated the correlation coefficients for these variables (see Appendix E2). FreeProd were strongly positively correlated with CSpend, r = .759, p <.05, and PriceRed were moderately negatively correlated with CSpend, r = -.308, p <.01− indicating a stronger correlation between FreeProd and CSpend compared to PriceRed and CSpend. FreeProd were moderately positively correlated with SPI (StoreImg), r = .310, p <.01, while PriceRed were weakly negatively correlated with StoreImg, r = -.163, p <.05, implying a stronger correlation between FreeProd and StoreImg than PriceRed and StoreImg. StoreImg was strongly positively correlated with
CSpend, r = .533, p <.01. Freq was weakly positively correlated with FreeProd, r = .149, and PriceRed, r = .131, moderately positively correlated with StoreImg, r = .359, and strongly
positively correlated with CSpend, r = .550 (all ps <.05). NoEvent was stronger negatively correlated with StoreImg (r = -.188) and CSpend (r = -.579) than PriceRed, both ps <.01.
In sum, the descriptives showed that (i) did customers spend more during event-weeks (ii) customer spending differed across the promotion types that were presented (iii) price perceptions appeared most positive after sales promotions through large-scale promotional events (iv) frequent sales promotions seemed most influential (v) promotion type, event frequency, SPI and customer spending appeared to be linked. Yet, these rough descriptives do not reveal to what extent sales promotions through large-scale promotional events influence customers spending directly, or through a positive change in the perceived SPI, and whether event frequency strengthens these relations−issues we will shed light on by estimating our model.
4.3 Model & Hypothesis Testing
To empirically address our hypotheses, we estimated the previously described moderated mediation model (no. 8 from Hayes) with bias-corrected 95% confidence intervals of 5000 bootstrapped samples, using the PROCESS macro for SPSS developed by Hayes (2013). Our predictor variable was multicategorical with three levels: (1) free product offers in large-scale promotional events, (2) price reductions offered through such promotional events, and (3) regular sales promotions presented without a large-scale promotional event (neutral condition)1. Table 3 reported the results of the moderated mediation model.
TABLE 3
Results of Moderated Mediation Analysis (N=342)
Consequent
StoreImg CSpend
Antecedent Coeff. SE p Coeff. SE p
FreeProd 𝑎& .1596 .1480 .0498 𝑐′& 5.0727 .3671 .0000
NoEvent -.1079 .1572 .0473 -4.1327 .3896 .0000 StoreImg -- -- -- 𝑏& .9311 .1349 .0000 Freq 𝑎# .2355 .1579 .1368 𝑐′# 1.8799 .3923 .0000 FreeProd × Freq 𝑎O .9567 .2216 .0210 𝑐′B 4.5516 .5636 .0000 constant 𝑖& 4.5541 .1050 .0000 𝑖# 9.8349 .6670 .0000 R2 = .2389 F(3,337) = 26,441 p<.001 R 2 = .8596 F(5,336) = 44,111, p<.001
1 To estimate the direct and indirect effects of all k X variables, we conducted PROCESS k times while we put one X
i in the model as X and
the remaining 𝑘 − 1 X variables as covariates so that we obtained direct and indirect effects on Y relative to our reference group (PriceRed). In addition, we tested a model with NoEvent as reference group, which showed that FreeProd and PriceRed differed accordingly (Hayes and Preacher 2014).
Frequency Unstandardized Boot Effects Boot SE Boot LLCI Boot ULCI
Conditional indirect effect at Spending for levels of frequency
Low Freq (0) .1486 .1439 -.1168 .4551
In our model of CSpend (Eq. (2)), estimated coefficients of FreeProd and NoEvent corresponded to the differences in customer spending between the free product offers and no-event conditions, relative to the reference group of price reductions. The overall model explained 86% of the variance of CSpend, which was highly statistically significant (p <.001). Results showed that the relative direct effect of free product offers on customer spending was positive and statistically significant, b = 5.0727, t(336) = 13,818 p <.001 (patch c1’, Table 3).
Thus, free product offers increased customer spending by 5.0727 units more than price reductions during event-weeks, while regular sales promotions increased customer spending by 4.1327 units less than price reductions in large-scale promotional events, b = -4.1327, t(337) = -10,6086 p <.05. This means that, during event weeks, free product offers have a stronger
positive effect on customer spending than price reductions. Moreover, sales promotions through large-scale promotional events are more effective in increasing customer spending than regular sales promotions. Hence, hypothesis 1 is supported.
In our model of StoreImg (Eq. (1)), the estimated coefficients of FreeProd and NoEvent represented the differences in SPI between the free product offers and no-event conditions, relative to the reference group of price reductions. The overall model explained 24% of the variance of StoreImg, which was highly statistically significant (p <.001). The relative direct effect of free product offers on SPI was positive and marginally statistically significant, b = .1596, t(337) = 1.0780, p <.05 (path a1, Table 3). So, those assigned to the free product offers
condition had SPIs that were .1596 units more favorable than those assigned to the price reductions condition. Although only marginally significant, this result indicates that free product offers during large-scale promotional events have a stronger positive effect on SPI than price reductions through such events. Thus, hypothesis 2 is not supported. Furthermore, SPIs of those exposed to regular sales promotions were .1079 units less favorable than the price reductions group, b = -.1079, t(337) = -.6862, p <.05−indicating that sales promotions through
large-scale promotional events are more effective in positively influencing SPI than regular sales promotions.
As expected, a highly significant and positive effect of SPI on customer spending was confirmed, b = .9311, t(336) = 6,9026, p <.001 (path b1, Table 3). So, a customer’s perception
on SPI that is 1 unit more favorable, is estimated to increase customer spending by .9311 units. This means that the perceived SPI positively influences customer spending, which is in line with hypothesis 3. Hence, hypothesis 3 is supported.
Results showed that event frequency does positively influence the relationship between free product offers and SPI, as revealed by a statistically significant interaction between free product offers and event frequency in the model of StoreImg (path 𝑎O= .9567, p < .05). More specifically, the indirect effect of free product offers on customer spending via SPI was only significant in cases of events with a high frequency (effect = 1.0393, SE = .2259, CI: .6497 to 1.5573), rather than of events with a low frequency (effect = .1486, SE = .1439, CI: -.1168 to .4551). In other words, the indirect effect of free product offers on customer spending through SPI, is stronger if large-scale promotional events reoccur frequently. Thus, hypothesis 4a is supported.
Finally, the results indicated the relative direct effect of free product offers on customer spending to be contingent on event frequency (path 𝑐′B, Table 2), as evidenced by a highly significant interaction between free product offers and event frequency (p <.001) in the model of CSpend (Eq. (2)). Again, a closer look at the conditional effects revealed that the direct relationship between free product offers and customer spending appeared as significant only if large-scale promotional events were initiated frequently (effect = 1.0393, SE = .2259, CI: .6497 to 1.5573), compared to infrequent large-scale promotional events (effect = .1486, SE = .1439, CI: -.1168 to .4551). Put differently, a high event frequency positively influenced the relationship between free product offers and customer spending. So, if large-scale promotional
events reoccur on a regular basis, this enhances the effectiveness of free product offers through such promotional events. Thus, hypothesis 4b is supported. Table 4 provided an overview of the findings of this study.
TABLE 4
Findings of this Study Summarized
Hypothesis Outcome
Free Product Offers → Customer Spending H1 (+) Supported
Price Reductions → SPI H2 (+) Not supported
SPI → Customers Spending H3 (+) Supported
Free Product Offers × Event frequency in relation to SPI H4a (+) Supported Free Product Offers × Event frequency in relation to Customers Spending H4b (+) Supported
5. Discussion, Limitations and Direction for Future Research
This chapter begins with a discussion on the implications of the findings that emerged from our research, in the context of existing literature. We then describe the implications of our findings while taking a managerial perspective. Lastly, we describe the limitations of our study whereupon further research directions will be given.
5.1 Discussion
In this study, we address the effect of sales promotions through large-scale promotional events—a recent promotion phenomenon that is catching on popularity among traditional grocery retailers in Europe and the U.S.—on customer spending and, more specifically, the role of SPI and event frequency. We argue that customers may be inclined to spend more at the store during event weeks through a positive change in the perceived SPI, and that the effects may be more pronounced in case of a high event frequency. We discuss the differential effects of two sales promotion types offered in large-scale promotional events, and reflect on how these types may improve SPI and increase customer spending. We propose a methodology to test this framework, which we applied to 342 households in the U.S. As such, we shed light on the working mechanism of large-scale promotional events as well as the underlying drivers of SPI formation and customer spending during event weeks.
Our main findings are as follows. The results highlight the differential role of free product offers versus price reductions offered through large-scale promotional events in stimulating customer spending and SPI. We find that free product offers by means of such promotional events are a stronger driver of customer spending than price reductions. This substantiates the premise that quantity-based discounts especially offer utilitarian benefits (Chandon, Wansink, and Laurent 2000) that stimulate stockpiling behavior (Guyt 2015), and thus increase customer spending. Interestingly, unlike previous studies on regular sales
