• No results found

What is the impact of the store flyer composition on store performance and how does the near presence and promotional activities of competitors affect this

N/A
N/A
Protected

Academic year: 2021

Share "What is the impact of the store flyer composition on store performance and how does the near presence and promotional activities of competitors affect this "

Copied!
55
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

Master Thesis

MSc. Marketing Intelligence

HOW TO WIN THE PROMOTIONAL WAR?

What is the impact of the store flyer composition on store performance and how does the near presence and promotional activities of competitors affect this

retailer’s success?

Tom Wielheesen S3638847 Molukkenstraat 90 9715 NW Groningen t.j.p.wielheesen@student.rug.nl

+31 650805043 University of Groningen Faculty of Economics and Business

MSc. Marketing Intelligence

Date: 15-06-2020

Supervisor: Prof. Dr. L.M. Sloot

(2)

ABSTRACT

This study investigates the impact of a supermarket’s weekly door-to-door flyer on the store performance. It is examined how flyer size, promotion depth, and the distribution of flyer space allocated to brand, product and framing types, affects a supermarket’s sales and traffic.

In addition, it is examined how the supermarket’s performance is affected by the competitive intensity, measured by distance effects and promotional intensity. The results indicate that the effectiveness of the flyer composition is dependent on the competitive environment. However, evidence is found that promotion depth and featuring a national brand on the cover increases sales for all the observed competitive situations. Additionally, for some competitive situations, it is found that perishable promotions are generally more effective in increasing sales whereas non-perishable promotions are commonly more effective in increasing traffic. This paper also found some results indicating that the distance effects diminishes when the promotional intensity of competitors increases.

Keywords: Grocery retailing, Retail Promotions, Store flyer, National brands,

Competitive intensity, Distance effects.

(3)

TABLE OF CONTENTS

1. INTRODUCTION………..………. 4 2. LITERATURE REVIEW………...

2.1. Main effects of feature promotions………

2.2. Promotion depth and frequency………..………..……….

2.3. Type of featured brand ……….……….………

2.4. Type of featured product category………..………...

2.5. Framing of promotion………..………….……….

2.6. Composition of cover page………..……….……….

2.7. The influence of competing stores……….………

7 9 9 10 11 12 13 14 3. METHODOLOGY……….……....……….

3.1. Data description………....…..………..……….

3.1.1. Dependent variables………...………..……….

3.1.2. Independent variables……….………..……….

3.1.3. Control variables………..……….

3.2. Data cleaning…..………...………...……….

16 16 16 16 17 19 4. RESULTS………..……..……….

4.1. Descriptive statistics..………...……….

4.1.1. Flyer composition………..……...……….

4.1.2. Competitive intensity………...……..……….

4.1.3. Store performance………...…………..………….

4.2. Empirical results………..………..…………

4.2.1. Multicollinearity issues………...………..……….

4.2.2. Pooled regression model…………..……….…….

4.2.3. Partially pooled regression model.…..………...………...…

4.2.4. Unit-by-unit regression model….………..………...……….

4.3. Model validation……….…...………

22 22 22 24 26 29 29 31 31 32 42 5. CONCLUSION AND DISCUSSION…..………..……….

5.1. Flyer composition………..………...……….

5.2. Competitive intensity………..………..……….

43 44 45 6. LIMITATIONS AND FUTURE RESEARCH………. 48 REFERENCES………...………

APPENDIX A - POOLED REGRESSION MODELS………..……….

APPENDIX B - PARTIALLY POOLED REGRESSION MODELS………..……….

APPENDIX C - DISTRIBUTION PLOTS OF RESIDUALS………..………..

50

53

54

55

(4)

1. INTRODUCTION

Sales promotions are used by many companies to introduce new products, increase sales or to influence the consumers’ decision process (Montaner & Pina, 2008). These sales promotions are offered in many different ways such as price promotions, feature advertising, in-store displays, etc. (Green, 1995). Sales promotions have been considered an effective way to produce a short-term volume gain for the brand being promoted (Blattberg & Neslin, 1990).

However, existing literature on sales promotions has primarily examined the effect of promotions at the brand level. While this information is useful for a brand manager to increase one brand’s sales, sales promotions are mostly at the expense of the other brands. Hence, it is possible that a promotional program is very desirable from an individual brand manager’s perspective, but may in fact lead to a reduction in category sales if much of the products are sold at the promoted price (Raju, 1992). Thus, it is not guaranteed that a sales promotion of a particular brand will result in a sales increase over the entire category. Therefore, retailers are concerned about promotions from a very different perspective than a brand manager. In particular, retailers use promotions in order to increase their store traffic and to increase the spendings once a shopper entered the store. In addition, retailers aim to attract consumers to their store to stimulate consumers to do their entire grocery shopping (including regular priced products) in that particular store (Abraham & Lodish, 1993; Mulhern et al., 1995). Previous studies have shown that by the means of sales promotions, consumers consider a certain store to be temporarily more attractive while another store may be more attractive in a subsequent period. This indicates that consumers are not really loyal to a store regarding grocery purchases and that consumers are likely to switch between grocery stores depending on the best offer (Narasimhan et al., 1996).

In order to attract consumers, Dutch grocery retailers commonly offer a weekly nationwide price promotion which is mainly featured through a weekly leaflet and are supported by in-store displays. For this reason, supermarket chains design and carry out price strategies oriented generally to obtain a positioning of discount prices (Martínez-Ruiz et al., 2010). For instance, retailers with the high-low (Hi-Lo) pricing strategy try to attract the more deal-prone and store-switching consumers by the means of price promotions featured in store flyers.

Besides attracting those deal-prone consumers, feature promotions may also affect the choices

made by the more general customers. Burton et al. (1999) described that exposure to a feature

promotion in a sales flyer results in more than a 100 percent increase in the number of sold

advertised products. Furthermore, featuring promotions in the store’s leaflet may also create an

(5)

important additional source of revenue, since often fees get charged to manufacturers when their brands are advertised in the store’s flyer (Gijsbrechts et al., 2003).

While much literature is pointing towards the positive effects of featured promotions, additional literature is stating that the success of these feature promotions depends on the composition of the promotion flyers. For instance, a distinction can be made between the type of featured brands, the product category, the depth of the promotion and the way promotions are presented. Research on price-tier competition has described that national brands and private labels differ from each other in sense of effectiveness on driving sales. It states that national brands benefit most from infrequent large price cuts and private labels benefit more from frequent small price cuts (Sivakumar, 1996). Furthermore, research has described that a distinction has to be made between product categories since some product classes might respond differently to sales promotions than others (Bemmaor & Mouchoux, 1991). In addition, previous studies have demonstrated that perceived attractiveness of a promotion is influenced by the way the price promotions are presented (Chen et al., 1998). However, these distinctions have barely been studied in the perspective of feature promotions effectiveness. Gijsbrechts et al. (2003) are a few of the pioneers in researching the effectiveness of flyer composition and found that the beforementioned distinctions do have an effect on store performance. However, previous studies hardly took any competitive actions into account (e.g. Gijbrechts et al., 2003;

Luceri et al., 2014).

From an academic perspective, this study will contribute to the literature by shedding more light on the effect of store flyer composition on the overall store performance. While feature promotions are considered to be an important marketing instrument within grocery retailing, papers examining its effectiveness on overall store level performance are lacking, as existing studies barely take any competitive activities into account (e.g. Gijbrechts et al., 2003;

Luceri et al., 2014). Regarding this, the purpose of this study is to better understand how the composition of a story flyer affects a store’s traffic and sales and how this is related to competitive activities. Therefore, this study will analyze the moderating effect of the competitor’s promotional activities on the relationship between store performance and distance to the nearest competitor.

This study is organized as follows: first, an overview of relevant literature regarding

flyer composition and competitive intensity will be presented in section two. Besides reviewing

(6)

third section contains the methodology which addresses the question how the proposed

hypotheses are going to be tested. Therefore, the different datasets will be discussed along with

some descriptive statistics. Furthermore, it is discussed which adjustments are executed to make

the data suitable for analyses. The paper continues with the result section, in which extensive

analyses are performed in order to gain more detailed knowledge about the data and

subsequently to test the hypotheses by the means of regression analyses. Lastly, this paper ends

with the discussion, which entails the conclusion, limitations and directions for future research.

(7)

2. LITERATURE REVIEW

Within the marketing literature, leaflet promotions have been studied in various ways.

Many of these papers examine the impact of featured promotions on the sales of the particular promoted product or the related product category (e.g. Abraham & Lodish, 1993; Lee & Brown, 2009). In general, these studies have found results indicating a strong positive effect between feature advertising and brand or category sales. While these findings are extremely useful from a brand manager’s perspective, the interest of supermarkets for doing feature advertisement might somewhat deviate. For instance, supermarkets adopting a Hi-Lo pricing strategy generally offer deep featured price promotions to attract the price sensitive consumers to do their entire grocery shopping (including regular priced products) in that particular store (Abraham & Lodish, 1993; Mulhern et al., 1995; Gijsbrechts, 2003). This is in line with the findings of Narashimhan et al. (1996) who demonstrated that by the means of sales promotions, consumers consider a certain store to be temporarily more attractive and another store may be more attractive in a subsequent period. This indicates that consumers are not loyal to a store regarding grocery purchases and are likely to switch between stores depending on the best offer.

Store flyers are considered to be one of the most important media for featuring retail and manufacturer promotions (Arnold et al., 2001). A store flyer can be described as “a frequently distributed free printed matter, part of the mass communication marketing from the sender(s), with a minimum of four pages, immediately readable, targeted at private households”

(Christiansen & Bjerre, 2001). Comparing these feature promotions in store flyers to promotions in newspapers and magazines, store flyers usually compromise the promotions for a larger number of products. Furthermore, the composition of flyer promotions entails a difficult trade-off between promoting high margined private labels and earning fees by featuring national brands. Therefore, it is important for retailers to know which composition of the leaflet is most effective in boosting sales. A study conducted by Gijsbrechts et al. (2003) has examined the effect of changing the composition of a leaflet on store performance. They found that the store’s traffic and sales are affected by leaflet composition characteristics, such as the average discount, the type of category featured on the cover page (specialties vs. produce vs. fish/meat, the share of space assigned to food and private label items, and leaflet size (number of pages)).

They have succeeded in demonstrating that altering the composition of a leaflet based on these

components can increase the traffic and sales of a store with 5.7 percent and 7.3 percent,

respectively. These numbers clearly indicate that optimizing the composition of a flyer may

(8)

unquestioned, there are some limitations associated with this study. First, since this study has been conducted around 17 years ago and in the meantime many developments have taken place within grocery retailing context, this study could be considered as outdated. Furthermore, the analyses of this study are confined to one chain and store format, Hi-Lo supermarkets, over a period of one year, which means that the external validity of this study can be discussed. As such, Gijsbrechts et al. (2003) conclude that additional research should be conducted to explore the store flyer effects for other store types and to what extent altering the composition of the store flyer improves a long-term performance. This is where the present study aims to step in.

Building upon the work of Gijsbrechts et al. (2003), the present study will focus on a number of leaflet characteristics that may affect the leaflet effectiveness over a period of two years.

Furthermore, this study will take a number of additional competitive characteristics into account, such as the competitor’s promotional depth and frequency and the distance between the competitor and the focal supermarket. Figure 2.1 provides a graphical overview of the variables included in this study.

Figure 2.1: Proposed conceptual model

H1 (+

)

H7 H6 H5 H4 (+) H3 (+) H2 (+)

H8 (+ )

H9 (-) STORE LEAFLET

COMPOSITION

Leaflet size

Number of promo’s

Promotion depth

Share of NB promo’s

Composition of product categories

Composition of cover page

Store performance Promotions competitor

Framing of promo

Store sales

Store traffic Distance nearest

competitor

(9)

2.1 Main effects of feature promotions

The prevailing hypothesis is that feature promotions have a positive impact on store traffic and sales. However, a distinction has to be made between the additional customers derived from feature promotions and the customers who would visit the store anyway. Since feature promotions tend to attract consumers in the store, these promotions in itself will affect traffic and sales. Once in the supermarket, consumers rarely buy only the promoted articles, but also purchase considerable amounts of non-promoted items which will result in an increase in sales. However, among customers who would visit the store anyway, the use of feature promotions can be considered a bit more ambiguous. A negative effect of price promotions could be that customers might choose to purchase the discounted products instead of the regular priced products and thus might negatively affect sales. On the other hand, featured promotions can also result in higher sales due to increased consumption. In addition, since feature promotions are often used in combination with shelf-tags as a reminder of feature promotions, regular consumers who have not seen the store flyer might be influenced as well (Gijsbrechts, 2003). Based on this reasoning, it is expected that the impact of featured promotions on store sales moves in the same positive direction as that of store traffic, but the intensity of the impact might potentially deviate. Therefore, the sales and traffic impact of feature advertisements will be assessed separately.

2.2 Promotion depth and frequency

Various studies have shown that increasing the variety of featured items has a positive effect on both store sales and traffic. For instance, Pan and Zinkhan (2006) succeeded in demonstrating that increasing the variety of the assortment increases consumers’ intentions to visit a store and to buy. Furthermore, Luceri et al. (2014) have found that the number of pages and the number of featured products play an important role in attracting customers and enhancing their purchases. This might be a result of consumers using the variety in the store flyers as a salient indication of the variety of a retailer’s assortment. Therefore, it is likely that a more varied store flyer increases consumers’ intentions to visit a store. Furthermore, when a store flyer is more extensive, it becomes more likely that consumers might find a set of promotions which is sufficient enough to visit a particular store. Whereas Gijsbrechts et al.

(2003) have studied the effect of store flyer size on store traffic and sales, they only found a

small significant effect of flyer size on store sales. However, it could be that their results point

(10)

consists of more that 20 pages. However, using the size of the store flyer as an indicator of the amount of featured promotions might be ambiguous since one page could feature one to many promotions. Therefore, the size of the store flyer and the amount of promotions featured will be assessed separately by the following hypotheses.

Hypothesis 1: The number of pages in the flyer has a positive effect on store traffic/sales.

Hypothesis 2: The number of promotions in the flyer has a positive effect on store traffic/sales.

Empirical research has shown that the total amount of discount offered is an important determinant for a flyer’s effectiveness. Mulhern & Leone (1990) have shown that a grocery store featuring many items at small discounts is less effective in increasing a store’s sale than a store that offers a few items at deep discounts. In addition, offering frequent and deep discounts might also lead to a reduction in perceived quality of a brand. This indicates that grocery stores might face a trade-off between offering many small discounts and offering a few products at deep discounts in their flyer (Grewal et al., 1998). Based on this, the following hypothesis is proposed:

Hypothesis 3: The deepness of discounts in the flyer has a positive effect on store traffic/sales.

2.3 Type of brand featured

As mentioned before, the composition of flyer promotions entails a difficult trade-off between promoting high margined private labels and earning fees by featuring national brands.

Therefore, it is important for a retailer to know whether featuring mainly national brands or mainly private labels is the most effective in increasing store traffic and sales. Many papers are pointing towards the idea that featuring national brands is the most effective, since private labels should primarily appeal to customers who are already loyal to a specific store. In addition, Ailawadi et al. (2006) have found that consumers who are loyal towards a national brand are more likely to switch between stores by using out-of-store promotions (e.g. flyer promotions).

Even though much literature is supporting the idea that retailers should mainly feature national

brands to increase sales, empirical evidence shows that many retailers still heavily feature

(11)

private labels (Ailawadi et al. 2006). It is suggested that even though the sales of the private label promotions are smaller than for national brands, the profit impact is probably higher for the private label promotions due to the higher retail margins. Furthermore, literature states that promoting national brands might benefit the so-called ‘asymmetric switching effect’, that is, when these brands are in promotion, they appeal more to customers from the users of private labels than vice versa (Blattberg & Wisnieswki, 1989). Building upon this reasoning, it is hypothesized that the store traffic and sales will increase when a store flyer contains a larger portion of national brand promotions. More formally, the following hypothesis is stated:

Hypothesis 4: The share of national brands in the flyer has a positive effect on the store’s traffic/sales.

2.4 Type of category featured

Another question of interest is how much emphasis should be put on the difference in effectiveness of product categories. Since weekly store flyers typically promote a mix of category types, it is expected that changes in the bundle of category types on feature would affect flyer effectiveness. A logical premise is that primarily featuring category types with high promotion elasticities increase overall flyer impact. Generally, a distinction has been made between perishable and non-perishable goods, where the former seems to be less sensitive to price deals due to the limited stocking up (Blattberg et al., 1981). This is confirmed by the study by Narasimhan et al. (1996), in which they found that promotion elasticities are generally higher for stock items, which have a higher category penetration, shorter interpurchase times and relatively fewer brands. Thus, they suggest that featured promotions are more profitable in those categories and are therefore an important marketing instrument. Contributing to their findings, Eales (2016) found that Dutch retailers often use drinkable products, such as soft drinks and alcoholic drinks, as traffic builders because of its high promotion elasticity. Which therefore suggests that featuring drinks increases flyer effectiveness. On the other hand, a study conducted by Gijsbrechts et al. (2003) found that flyers with a larger portion of space allocated to food items positively affects store performance. This suggests that feature effectiveness is not explicitly dependent on whether the featured item is a perishable on non-perishable product.

By distinguishing the types of food, Gijsbrechts et al. (2003) states that especially fresh food

categories (such as fruit, vegetables, meat and fish) show similarities with regards to deal-prone

characteristics. Their explanation is that these categories contain frequently bought and highly

penetrated items which, indicated by Fader and Lodish (1990), make them major contenders

(12)

for feature promotions. Besides that it might result in brand switching, it is likely that featuring these categories might result in an increase of consumption.

There seems to be a contradiction in explanations regarding the feature effectiveness between food and non-food products, which is most likely due to the fact that non-food products are on average less frequently and impulsively bought than food products. On the other hand, non-food products are non-perishable and therefore better to stockpile. Reviewing existing literature indicates that there is no unambiguous answer for the issue of which product category is most effective for featuring. However, it is expected that since non-perishable promotions are often related with deeper discounts, consumers are more likely to consider it worthwhile to visit a store when non-perishable products are featured. On the other hand, since perishable products are frequently bought and highly penetrated, it is expected that these promotions would be effective in upselling. Building upon the beforementioned reasoning, the following hypotheses can be proposed:

Hypothesis 5a: the share of non-perishable product promotions in the flyer has a positive effect on store traffic

Hypothesis 5b: the share of perishable product promotions in the flyer has a positive effect on store sales

2.6 Framing of promotion

Previous studies have demonstrated that the perceived attractiveness of a promotion is influenced by the way the price promotions are presented (Chen et al., 1998). For instance, Dutch grocery retailers commonly use price promotions such as a single price offer (SPO; 25 percent off), gift promotion (Buy A, get B for free) and a bundle promotion (Buy A and B for

€) (Peters, 2012). Because the framing of a promotion influences the way consumers perceive the value of a promotion, many retailers consider this activity as one of the most important things in store flyer management. Cox & Cox (1990) found that when a supermarket offers a single price offer, it creates a lower price image than when they frame it as a bundle promotion.

In addition, Lee & Yi (2019) demonstrated that gift promotions are more effective than bundle

promotions. However, another point of view is that bundle promotions would increase

consumers’ expenditures since consumers can only receive the offered discount by buying

multiple items. Nevertheless, it could be reasoned that a bundle promotion decreases the

probability of a consumer buying that product at all, since the consumer might not want to

(13)

purchase in bigger quantities. Even though empirical research on this framing type is lacking, it is expected that this type of framing negatively affects the purchase intention. While various studies succeeded in demonstrating that SPO and gift promotions increase the purchase intention of the promoted article, no further research is conducted regarding the effect of the combination of different framing types on total store performance. Since a lot of research stresses the importance of the way a promotion is presented, it is expected that the way a set of promotions are framed has an effect on store traffic and sales. Which leads to the following hypotheses:

Hypothesis 6a: The share of promotions framed as a SPO has a positive effect on store traffic/sales.

Hypothesis 6b: The share of promotions framed as a gift has a positive effect on store traffic/sales.

Hypothesis 6c: The share of promotions framed as a bundle has a negative effect on store traffic/sales.

2.7 Composition of cover page

While various studies focus on the importance of the physical stimuli, promotion value, brand type and product category, there are only a few studies addressing the impact of the composition of the cover page of the flyer (e.g. Gijsbrechts et al., 2003; Ailawadi et al., 2006).

These few studies have demonstrated that the content exposed on the flyer’s cover page is an

important factor for attracting the customer’s attention and whether they would respond to a

store flyer or not. The importance of the composition of the cover page is most likely mainly

due to the prominence of the location and has been recognized by many retailers who typically

use the cover to announce their ‘deal of the week’. Ailawadi et al. (2006) even concluded that

front-page features have an 89 percent higher net unit impact than when products are not

featured at all. Even though featuring a promotion on the cover page seems to be extremely

successful, a distinction should be made between the product categories, brands and the

promotion depth of the product featured on the cover page (Gijsbrechts et al., 2003). As

discussed before, it is expected that since non-perishable promotions are often related with

deeper discounts, consumers are more likely to consider it worthwhile to visit a store when non-

perishable products are featured. Moreover, as discussed earlier, it is expected that the

(14)

effectiveness of a promotion is highly related to the type of brand featured and the promotional depth. Especially when these promotions occupy the most prominent position of the flyer such as the cover page. Therefore, the following hypotheses are proposed:

Hypothesis 7a: Featuring non-perishable products on the cover page has a positive effect on store traffic/sales.

Hypothesis 7b: Featuring a national brand on the cover page has a positive effect on the store’s traffic/sales.

Hypothesis 7c: Featuring deeper discounts on the cover page has a positive effect on store traffic/sales.

2.8 The influence of competing stores

Bell and Lattin (1998) have found that the choice for a store highly depends on the distance to a store. The likelihood of switching between stores decreases when the perceptual distance between two stores increases. On the other hand, when the perceptual distance between two stores decreases, the likelihood of consumers switching between stores will increase (Leszczyc et al., 2000). This refers to the findings of Narasimhan et al. (1996), who found that consumers seem not to be loyal to a store regarding grocery purchases and that consumers are likely to switch between grocery stores depending on the best offer. However, when the distance between competitors increases, it is likely that consumers experience higher costs of switching between stores. Therefore, the distance to a nearest competitor is considered as a highly important determinant for store sales and traffic. Hence, it is hypothesized that the distance to the nearest competitor has a positive effect on store performance.

Hypothesis 8: The distance to the nearest competitive outlet has a positive effect on the traffic/sales of the focal store.

In the Dutch grocery retailing environment, major grocery chains are competing with

each other to attract the consumer by means of price promotions. Hansen & Loy (2007) found

that when a competitor offers deeper and more frequent promotions, this will consequently

decrease the sales of the other stores. Furthermore, they found that the discount offered by

competitors has a negative impact on the category sales of the others. Which indicates that

consumers switch between stores depending on which store offers the best bundle of deals. This

(15)

is in accordance with Pesendorfer (2002) who found that, especially for products with low price levels, the competitors’ set of prices positively affects the overall demand in other stores. Once competitors (temporarily) decrease their prices, it will cause the other stores to decrease their prices as well and vice versa (Richards, 2006). However, as discussed earlier, the effectiveness of competitors’ promotions is highly related with the proximity of a competitive outlet.

Therefore, it is hypothesized that when competitors offer deeper and more frequent promotions, consumers are considering it worthwhile to travel further. In this way, it is expected that the positive effect of distance weakens when a competitor’s promotional activities increase. More formally, this results in the following hypothesis:

Hypothesis 9: Competitor’s promotional activities negatively affects the relationship

between the distance to the nearest competitor and the traffic/sales of the focal store.

(16)

3. METHODOLOGY 3.1 Data description

While, due to confidential matters, the name of the supermarket chain cannot be identified, some descriptions regarding the supermarket’s formula will be given to increase the understandability of the context. This study is conducted in the Dutch grocery retailing environment where data from various outlets of the same supermarket chain were made available. With around 270 outlets around the country, this chain belongs to the top 5 supermarket chains in the Netherlands. Together with four other supermarket chains, this chain is responsible for around 80 percent of the total market share. Where these top retailers apply different pricing strategies, the focal chain applies a Hi-Lo pricing strategy which is centrally managed and promoted by means of, among other things, a weekly store flyer. The period in which this study has taken place is from January 2018 until December 2019 (covering a period of 24 months).

3.1.1 Dependent variables

The response variable within this study is the store performance, which is measured by the store traffic and the store sales. To measure these dependent variables, the retailer provided a dataset consisting of sales data of 272 of their outlets including 1) number of transactions per outlet and 2) the sales per outlet. The aggregation level of these variables is on a weekly basis.

3.1.2 Independent variables

In order to measure the store leaflet composition, the nationwide distributed weekly

flyer is manually coded over a timespan of 24 months (covering 104 flyers). Table 3.1 provides

a descriptive overview regarding the characteristics on which these store flyers have been

coded. For pragmatic reasons, it is decided to only distinguish the product groups featured on

the covers by perishability. In order to measure the explanatory variables regarding the

(17)

promotional activities of competitors, data from Superscanner.nl is used. The dataset of Superscanner.nl consists of daily collected data about prices and promotions. In order to measure the competitive activities, the average percentage discount and the share of the assortment in promotion (proportional discount) of the major competitors are used. This data is aggregated on a weekly level to match it with the aggregation level of the flyer composition and the corresponding sales data. For the analysis, two major competitors are selected since these belong to the largest supermarkets in the Netherlands based on market share. Furthermore, these supermarket formulas are most corresponding with the focal supermarket, since they can all be considered as traditional supermarkets. In order to measure the moderating variable

‘distance to the nearest competitor’, two different data sources were used. Firstly, the ZIP codes per outlet of the focal supermarket chain are made available by the supermarket chain itself.

Secondly, to obtain the ZIP codes of the different outlets of the two major competitors, the EFMI Supermarket database is used. Based on these ZIP codes, the distance between the nearest outlet of both competitors and the outlet of the focal supermarket is computed. Furthermore, the statistical analyses in this study are executed in R-studio (version 1.2.1578).

3.1.3 Control variables

In addition to the beforementioned variables, some control variables will be included in the analyses. By controlling for 1.) store size, 2.) national holidays, 3.) share of voice, 4.) opening hours and 5.) urbanization rates, the irrelevant variables are kept constant which therefore facilitates stronger testing of the hypotheses. The first control variable is store size, which is included as outlets may dramatically vary in store size. According to Haans &

Gijsbrechts (2011), large stores have a bigger set of potential customers, which indicates that

more consumers are prone to be attracted by the featured promotion. Furthermore, stores with

a bigger selling surface generally attract consumers on major, stockpiling shopping trips. Since

this type of consumer typically reacts more heavily on sales promotions, it is required to control

for these differences in stores. The second control variable is national holidays, which is

included because the weeks before Christmas and Easter increase supermarkets’ revenues

tremendously (Warner & Barsky, 1995). Next, flyer advertisements are in general part of a

bigger campaign ran by a supermarket. In order to isolate the effect of flyer effectiveness, it is

important to control for the share of impressions a campaign has fetched compared to the other

competitors. Since this is measured by the share of voice (SOV), it is necessary to control for

the SOV. The fourth control variable is opening hours, which is included as the opening hours

(18)

Therefore, the opening hours may have an impact on the variance in store traffic and sales and is therefore an important variable to control for. Lastly, in order to account for the nationwide differences in degree of urbanization and therefore the difference in customer base per supermarket, the degree of urbanization is included as a control variable. Data to measure the control variables ‘store size’, ‘share of voice’, ‘opening hours’ and ‘loyalty programs’ is made available by the focal supermarket chain. Furthermore, in order to control for the nationwide differences in degree of urbanization and therefore the difference in customer base per supermarket outlet, data from the Dutch Central Bureau of Statistics (CBS) was used. To give a brief overview of all the used data sources in the current study, the sources are presented in table 3.2 combined with the corresponding obtained information.

Data source Obtained information

Superscanner.nl Promotional activities competitors

EFMI Supermarket Database ZIP codes of the closest outlet of each major competitor

CBS Degree of urbanization in the Netherlands aggregated on ZIP code Focal supermarket chain’s flyer Flyer composition

Focal supermarket chain Store sales, store traffic, store size and ZIP code per outlet Table 3.2: Overview of the used data sources

Furthermore, to give a brief overview of the beforementioned dependent and explanatory variables, table 3.3 presents these variables complemented with the descriptive statistics. However, it should be noted that these descriptive statistics have already been corrected based on section 3.1.4.

Variable Measurement Median Mean SD

Store performance

Store traffic Weekly amount of receipts in an average store 10236 10769 10169.8 Store sales Weekly total sales of an average store in euros 173594 186390 488.4 Flyer composition

Flyer size Number of pages in the flyer 28 27.12 4.6

Number of promotions Number of promotions in the store flyer 78 79.74 10.8

Relative promotion depth Average percentage discount in the flyer 32.5% 32.7% 5.1

Absolute promotion depth Absolute discount in the flyer in euros 131.92 147.70 72.9

(19)

Variable Measurement Median Mean SD

National brand share Share of space allocated to national brand 56.1% 55.8% 5.6 Fresh food share Share of space allocated to produce promotions 33.8% 34.8% 6.0 Dairy share Share of space allocated to dairy promotions 12.5% 12.6% 2.4 Frozen share Share of space allocated to frozen promotions 3.9% 3.8% 1.1 Drinks share Share of space allocated to drinks promotions 12.9% 13.7% 4.6 Other/Nonfood share Share of space allocated to other/nonfood promo 28.0% 28.4% 4.0 Share of SPO promo’s Share of promotions framed as SPO 57.7% 58.1% 10.8 Share of gift promo’s Share of promotions framed as gifts 8.1% 7.7% 4.6 Share of bundle promo’s Share of promotions framed as bundle 23% 22.9% 6.9 Share of percentage promo’s Share of promotions framed as percentage 6.4% 9.4% 10.2 Non-perishable share on cover Share of non-perishable promotions on cover 50% 54.3% 36.1 Absolute discount cover Absolute discount on cover in euros 4.93 5.52 3.2 National brand cover Share of national brands featured on cover 100% 71.2% 36.1 Number of non-monetary gifts Amount of non-monetary gifts featured 9 9.3 4.3

Competitive variables

Proportion competitor A Share of comp. A’s assortment in promotion 9.8% 10% 2.3 Proportion competitor B Share of comp. B’s assortment in promotion 7.7% 7.1% 1.7 Depth competitor A Percentage discount offered by comp. A 31.0% 31.5% 2.9 Depth competitor B Percentage discount offered by comp. B 24.4% 24.8% 1.5 Distance competitor A Distance to nearest outlet of comp. A in meters 1558 2770 2650 Distance competitor B Distance to nearest outlet of comp. B in meters 2169 3088 2560 Distance to nearest own outlet Distance to nearest outlet own outlet 4688 5868 5054

Control variables

Period of the year Dummy for Christmas and Easter period Urbanization level Dummy 1 (very urban) – 5 (very rural)

Share of voice Share of voice of focal supermarket 16.3% 16% 6

Store size Store size in square meters 1050 1105 262.1

Opening hours The total amount of opening hours per week 81 81.99 8.1 Table 3.3: Overview of the variables with the corresponding measurements

3.2 Data cleaning

Since missing values and inconsistencies can create an unrealistic view of the database,

the data should be checked for missing values and outliers before proceeding with the analysis.

(20)

The original merged dataset consists of 27.437 observations gathered from 272 outlets. By analyzing the dataset, it became apparent that the data of multiple stores contained missing values. A plausible explanation for these missing values could be that these outlets have been renovated during the observation period. Stores which did not include the full 104 weeks of observations were therefore excluded from the dataset since these missing values are hard to impute properly. Hence, this reduces the size of the dataset to 24.960 observations which is gathered from 240 outlets. In addition, one missing value was found regarding the composition of the cover page in week 51 of 2018, which is likely due to a typographical error during the manual coding process. This problem is resolved by manually imputing the right values of the store flyer.

Based on the descriptive overview in table 3.3, it is striking that the mean of both the sales and receipts is a lot higher than the median. A large difference between these values might indicate that there are some outliers in the dataset. To easily identify these outliers, the sales and traffic are plotted over the time period in figure 3.1 and 3.2. Based on these plots, it is noticeable that the outliers for sales and traffic appear around the same period of both years.

These periods correspond with the weeks around Easter and Christmas for which dummy variables were included.

Figure 3.1: Sales plotted over time

Figure 3.2: Traffic plotted over time

(21)

Furthermore, a striking observation is that the mean of the total absolute promotion depth (€147,70) is a lot higher than the median (€131,92). Having a large difference between these two statistics might also indicate that there are some outliers within this variable. By plotting the absolute promotion depth over time in figure 3.3, it becomes clear that there were six weeks in which the absolute promotional depth was over 250 euros. Five of these outliers correspond with promotional weeks in which the supermarket chain promoted wine with many and deep discounts. Additionally, correlation analyses show a significant correlation (p <.01) of 0.87 between the total absolute discount and being a wine promotion week. Since these promotions do tremendously affect the total absolute discount and therefore might have an influence on the effect of the regular absolute discount on store performance, it is decided to split the promotions during the wine promotional weeks. Hence, a new variable is created for the total absolute discount (excl. wine promotions during wine promotional weeks) and the promotional discount during these weeks. This issue will be further discussed in section 4.2.1.

Furthermore, the sixth peak is a regular promotion week with high promotions which is considered as a meaningful outlier and does therefore not require any dummy substitution. After accounting for the discussed missing values and inconsistencies, the dataset is considered clean for further analyses.

Figure 3.3: Total absolute discount plotted over time

(22)

4. RESULTS

In this chapter, the descriptive statistics and empirical results will be discussed. To create a good understanding of the data, descriptive statistics are presented by graphs and boxplots. Furthermore, the empirical results of this study will be presented in the last subsection.

4.1 Descriptive statistics

In order to get some feeling with the data, this chapter will present the descriptive statistics of the data along with the presentation of some explorative charts. Where chapter 3 already presented the mean, median and standard deviation per variable, some elaboration on these statistics will be given to create a better understanding of the context.

4.1.1 Flyer composition

The total dataset for the current study consists of 24.960 observations gathered over a period of 104 weeks. In every week of those two years a promotional flyer was active with an average price discount of 32.7 percent (relative promotion depth). Those flyers featured on average 79.7 promotions. Furthermore, the total absolute discount has a mean of €147.70.

However, as discussed in section 3.1.4, this mean is highly affected by the promotions during the wine promotional weeks. This becomes clearly visible in the average absolute discount after subtracting the wine promotions, since this adjustment changes the average value to €134.01, which is much closer to the original median of €131.92.

Furthermore, as displayed in figure 4.1, there seems to be a rather constant distribution between the featured brand types in the promotional flyer. With a mean of 55.8 percent, the majority of the featured promotions consists of national brand promotions. Subsequently, 38.2 percent of the featured promotions consists of private label promotions. With a high

Figure 4.1: Distribution of brand types in promotion

(23)

underrepresentation, the unbranded category takes up an average of 6 percent of the total featured promotions. This type of brand is often found in the fruit and vegetables category where the product is generally not related to a brand.

The share of promotions per product category is displayed in figure 4.2 and shows a clear difference between the amount of featured promotions per category. For instance, the average share of promoted Fresh food products is 34.8 percent. This indicates that over a third of the total featured promotions consists of Fresh food promotions. Furthermore, dairy and drinks are almost equally promoted with an average share of 12.6 and 13.7 percent, respectively. With the second highest share, the other and nonfood category is promoted rather often with a mean share of 28.4 percent. This finding is in line with Blattberg et al. (1981) and Narashiman et al. (1996), since they state that non-perishable products are generally more often used as traffic builders because of their ability to be stockpiled. On the other hand, the frozen product category is promoted way less with a mean share of 3.8 percent. This low value is considered somewhat contradicting to the theory of Blattberg and Narashiman, since these products are generally not considered as perishable goods when it remains frozen. However, this difference could also be designated to the fact that the product category itself contains many fewer stock keeping units (SKUs) than for instance the other and nonfood category.

Furthermore, figure 4.3 shows how much each type of framing is used for the featured promotions. Based on this graph, it is noticeable that the SPO framing is used the most with an average of 58.1 percent. By using the perspective of Cox & Cox (1990), the supermarket probably offers this framing type in order to create a lower price image than when they offer bundle promotions. However, the figure shows that bundle promotions are also commonly used in the store flyer with an average share of 22.9 percent. This points towards the reasoning that supermarkets use these types of framing to stimulate consumers to increase their consumption quantity and therefore their total spending. Lastly, the graph shows that the percentage discount and the gift promotions are used significantly less with an average share of 7.7 and 9.4 percent, respectively.

Figure 4.2: Share of promotions per product category Figure 4.3: Distribution of how promotions are framed

(24)

Since the content exposed on a flyer’s cover page is considered as an important factor in attracting the consumers attention, it is decided to analyze this content separately from the in-flyer content. Based on table 3.3, it noticeable that the cover page is dominantly featuring non-perishable promotions with an average of 54.3 percent. This statistic reflects the idea that a consumer might be more attracted to non-perishable promotions due to their ability to be stockpiled and that subsequently supermarkets use these as traffic builders. Furthermore, the total absolute discount on the cover is €5.52. However, it should be noted that this value is the sum of the discount of the front and backside of the cover and it should therefore be divided by two to get the average absolute discount per promotion. This results in an average absolute discount of €2.75 per promotion, which is higher than the average in-flyer discount of €1.85 per promotion. This difference seems logical since generally attractive promotions are featured on the cover. Furthermore, the majority of the time national brands are featured on the cover page with an average share of 72.1 percent. A high correlation of 0.75 is found between the national brand share and the share of non-perishable products and a correlation of .11 with the promotional depth. These correlations denote that often a set national brands are featured in combination with non-perishable products and deep promotions, which is in line with the reviewed theory that the cover should be composed in such a way that it attracts the consumers’

attention.

4.1.2 Competitive intensity

As described before, the intensity of the promotions of competitors can be measured by the actual offered discount and by calculating the proportional amount of promotions of one’s entire assortment. As displayed in figure 4.4 and 4.5, the intensity of these variables varies across the supermarkets, which seems to be in line with the general pricing strategies across the

Figure 4.4: Proportional amount of promotions per chain Figure 4.5: Average discount per chain

(25)

supermarkets. For instance, the pricing strategies of the focal supermarket and competitor A can be described as a Hi-Lo strategy. This strategy is characterized by many fluctuations and emphasizes on offering deep and frequent discounts (Ellickson & Misra, 2008). The difference in strategies becomes clear by looking at the difference in means of the average discount, since the focal supermarket (32.2%) and competitor A (31.4%) offer much deeper discounts than competitor B (24.8%). The lower average discount offered by competitor B is also corresponding with the pricing strategy of this supermarket, since its pricing strategy can be described as an EDLP (everyday low pricing) strategy. This is characterized by offering relatively stable prices across many different products (Ellickson & Misra, 2008). On the other hand, it should be noted that despite of the Hi-Lo strategy of the focal supermarket, its proportional amount of promotions of the focal supermarket (4.9%) is significantly lower compared to competitor A (10%) and competitor B (7.1%). However, this difference might also be a result of competitor A and B having a broader assortment compared to the focal supermarket. To get a better understanding of the promotional intensity over time, the development of the average discount per supermarket chain is displayed in figure 4.6. When comparing the lines in the graph, it again shows some clear results. This clarifies that competitor A and the focal store are quite comparable regarding their offered discount over time. On the other hand, competitor B offers a quite constant but much lower discount over time.

According to Narasimhan et al. (1996), the distance to the closest competitor is important since consumers seem not to be loyal to a store and generally switch between stores depending on the best offer. To get an overview of how close the competitors’ outlets are located from the focal store, a boxplot is presented in figure 10. Based on this graph, it is noticeable that the

Figure 4.7: Distance to nearest outlet per

Figure 4.6: Development of Average discount per chain over time

(26)

widely. On average, an outlet of competitor A is the closest to the focal supermarket with an average of 2770 meters. The distance to an outlet of competitor B seems to be a bit farther away with an average of 3088 meters. However, the variance is widely spread since the closest outlet of competitor A is 80 meters away and the furthest outlet is almost 16 kilometers away.

Furthermore, the closest outlet of competitor B is 90 meters away and the furthest outlet is approximately 13 kilometers away. Since these differences are likely a result of the difference in level of urbanization, figure 4.8 shows the different distances plotted between the five urbanization levels. It appears that in the very urban, urban and semi-urban areas the distance to the nearest competitor is around 1 and 2 kilometers, while the distance to the nearest competitor for the rural and very rural areas is around 3 and 5 kilometers. Since the distance to a supermarket is considered as an important determinant for the selection of a supermarket, it is expected that a high level of heterogeneity is present between the observed stores in the current study.

4.1.3 Store performance

By having a closer look at the distribution of sales and traffic over the observation period, some remarkable findings are discovered. To give an overview of the distribution over time, these dependent variables are plotted in figure 4.9 and 4.10 As presented in these graphs, there is a wide variance in sales and receipts which indicates that the performance of each outlet differs substantially. For instance, the minimum observed weekly sales of an outlet is €2.430 and the maximum observed value is €588.089. In addition, this wide variance is also found in the receipts where the minimum amount of weekly receipts is 1108 and the maximum is 33996.

While these extreme values could also be due to variance over time, the wide variances in the graphs suggest that there is also a high level of heterogeneity between the different outlets.

Figure 4.8: Distance to competitor’s outlet per urbanization level

(27)

Figures 4.9 and 4.10 show the aggregated store performance for all the different urbanization levels, but the store performance may also vary between the different urbanization levels. To clarify how the performance per store differs, figure 4.11 shows the differences in sales aggregated on urbanization level. With a mean exceeding €200.000, it appears that outlets in the semi-urban areas have in general the highest sales numbers. Followed by the urban and rural areas, whereas the very rural and very urban areas seem to provide the lowest sales in general. This could be related to the fact that in competition in the very urban areas is really high and on the other hand, the customer base in the very rural areas is really low. Furthermore, figure 4.12 shows the differences in receipts aggregated on urbanization level. With a mean of around 15.000 receipts, the outlets in the very urban areas seem to have the highest traffic compared to the other urbanization levels, followed by the outlets in the urban and semi-urban areas, whereas the outlets in the rural and very rural areas have the lowest amount of receipts in general. By comparing both figures, it is striking that outlets in the very urban areas have more customers than all the other areas but the sales are substantially lower than the others.

This indicates that the average spending per customer is much lower in the very urban area compared to the other areas. A possible explanation is that the distance between various stores

Figure 4.11: Sales per urbanization level Figure 4.12: Receipts per urbanization level Figure 4.10: Variance of traffic plotted over time

Figure 4.9: Variance of sales plotted over time

(28)

is generally much smaller which stimulates consumers to visit different stores to get the best offer. This would be in line with the findings of Narasimhan et al. (1996) that consumers are not loyal to a store and that consumers are likely to switch depending on the best offer.

However, since the distances between stores in the more rural areas are generally greater, the costs for getting the best deal also increase. This results in consumers purchasing all of their groceries in the store which offers the best deal or is the closest. These findings indicate that the effectiveness of flyer promotion is heterogeneous across different stores.

To get a first impression of how the store performance is affected, a very basic model is proposed including flyer size, total absolute discount, national brand share and distance effects.

The results of this model are presented in table 4.1. Both models are highly statistically significant, implying that the included variables improved the fit of the model. Furthermore, the model reveals that the direction of the estimates of the included variables are conform the hypothesized. However, this model is just to give an impression of the correlation of variables but no official conclusions can be drawn from this model yet, since not all explanatory variables are included and it has to be assured that no model assumptions are violated. This will be discussed further in the next section.

Sales (SE) Receipts (SE)

Intercept -66562.98** (0.00) 380.68 (0.00)

Flyer size 243.92** (0.02) 20.73** (0.03)

Absolute promotion depth 89.86** (0.04) 1.58* (0.01)

National brand share -48.27 (-0.00) 5.23* (0.01)

Distance competitor A 0.27* (0.01) 0.01 (0.00)

Distance competitor B 2.64** (0.10) 0.13** (0.08)

Control variables

Christmas Dummy 34160.39** (0.07) 388.40** (0.01)

Easter Dummy 19168.00* (0.04) 267.70* (0.01)

Share of voice -211.45** (-0.02) -7.53** (-0.01)

Store size 148.99** (0.68) 5.54** (0.43)

Opening hours 948.94** (0.11) 93.89** (0.19)

Urbanization level -3414.70** (-0.07) -1421.06** (-0.49)

Observations 24960 24960

R2 0.542 0.622

Adjusted R2 0.542 0.621

Residual std. error (df = 24948) 45600 2421

Table 4.1: Simple linear regression model Note: * p<0.05, ** p<0.01

(29)

4.2 Empirical results

In order to test the proposed hypotheses, different models will be discussed in this section. Before building these models, some model assumptions will be discussed.

Subsequently, different regression models will be discussed based on the level of pooling. After selecting the most appropriate pooling method, the hypotheses will be tested. Lastly, it will be discussed whether model assumptions are violated with the proposed model.

4.2.1 Multicollinearity Issues

When an Ordinary Least Squares regression is applied to obtain estimates for the parameters in the model, some assumptions about the model should be satisfied. One of them is that the variables should be free from multicollinearity. Multicollinearity arises when a high correlation is found between the predictor variable and one or more other predictor variables included in the model (Leeflang, Wieringa, Bijmolt, & Pauwels, 2015). Hence, when having multicollinearity in the model, the regression might give unreliable parameter estimates which might even be in the opposite direction than expected. A way to measure whether the data suffers from multicollinearity is by computing the Variance Inflation Factor (VIF). In addition, a correlation matrix of the predictor variables will help to detect the correlation between variables.

When estimating the model, some multicollinearity issues (VIF > 5) were found in

several variables. First, multicollinearity was found in the absolute promotion depth (VIF =

9.3). This variable was highly correlated with the number of promotions. Hence, when the

number of promotions was higher, the total absolute promotion depth increases as well. This

issue is solved by reformulating the total number of promotions into separate parameters which

contained the number of promotions per product category. A negative consequence of this

solution is that it is no longer possible to test H2 formally. However, by assessing the number

of promotions per product category, it is expected that more valuable insights will be gathered

from the analysis. Furthermore, by inclusion of counts per product category, new issues arise

with the share of promotions per product category. Therefore, it is decided to also remove the

share variables from the model. Whilst not formally hypothesized, to gain richer information

about the promotional effects per product category, it is decided to also include the average

percentage discount per product category. Furthermore, as discussed in section 3.2, new

variables are created to account for the high correlation between the total absolute discount and

(30)

the wine promotional weeks. These variables explained the regular amount of discount and the additional discount during the wine promotional weeks. However, a high level of multicollinearity was found in this new variable (VIF = 47.2). By inspecting the correlation matrix, it appears that it was highly correlated with the number of drinks in promotion. This issue is solved by subtracting the number of drinks in promotion by the amount of wine promotions during these promotional weeks, which solved this multicollinearity issue.

Multicollinearity was found between the shares of the way promotions were framed (VIF > 16).

These variables create issues since their sum is 100 percent which results in a high level of correlation. This issue could only be solved by deletion of some of the framing types. Therefore, only the share of promotions framed as a SPO or gift are left over in the model.

Lastly, a correlation is found between the number (VIF = 4.76) and average percentage discount (VIF = 5.71) in the Fresh food category. These variables show a significant (p < .01) correlation of - 0.48, which might refer to the promotional weeks in which the supermarket promoted many fresh food products but with small percentage discount. Since solving this issue by deletion might result in the omitted variable bias, it is decided to follow Curto & Pinto (2011) by also considering VIF values smaller than 10 as acceptable. The reformulation of the variables resulted in a model consisting of 34 variables, which might create overspecification issues. In order to reduce the number of variables, it is decided to reformulate the competitive intensity by multiplying the promotional depth by the proportional amount of promotions into one variable called ‘promotion intensity’. This decision is made since the proportional discount did not contribute much in the explanation of the variance in the dependent variables, but remains important in describing the differences between stores. By reconstructing the discussed variables, the multicollinearity issues in the explanatory variables are solved (table 4.2).

Variables VIF Variables VIF

Flyer size 2.44 Number of non-monetary gifts 1.46

Abs. promotion depth (excl. wine promo) 2.73 Share non-perishable cover 1.74

Abs. wine promotion depth 1.90 Share of national brands on cover 1.45

National brand share 3.28 Abs. discount on cover 1.31

Number of Fresh food promotions 4.76 Promo intensity competitor A 1.55

Avg. perc. discount Fresh food 5.71 Promo intensity competitor B 1.54

Number of Dairy promotions 1.98 Distance competitor A 3.13

Avg. perc. discount Dairy 3.25 Distance competitor B 4.63

Number of Frozen promotions 1.69 Distance nearest outlet focal supermarket 1.25

Avg. perc. discount Frozen 2.15 Christmas Dummy 2.65

Number of Drinks promotions 2.22 Easter Dummy 1.53

Avg. perc. discount Drinks 2.32 Share of Voice 1.44

(31)

Variables VIF Variables VIF

Number of Other/ Nonfood promotions 1.81 Store size 1.26

Avg. perc. discount Other / Nonfood 2.62 Urbanization level 4.80

Share SPO 2.70 Opening hours 1.29

Share Gift 2.10

Table 4.2: VIF values after transformation

4.2.2 Pooled regression model

Firstly, two pooled models are built to analyze the effects of the flyer composition and the competitive variables on the sales and traffic per store. Since the pooled regression model assumes homogeneity across all entities, this approach specifies exactly the same model for each entity. The advantage of this approach is that is allows for the use of all available data points as efficiently as possible (Leeflang, et al., 2015). The estimations of the pooled regression models for predicting sales and receipts can be found in Appendix A. Both the models are statistically significant (p<.01) which means that the included variables improved the fit of both models. The adjusted R

2

indicates that the sales model explains about 88.6 percent of the variance and the receipts model about 91 percent of the variance.

In order to test whether this level of pooling is statistically appropriate for testing the hypotheses, the Chow test is performed. This test is used to examine whether there is homogeneity in the slopes and the intercepts for both models. The Chow test for the pooled model for the sales prediction is statistically significant (F = 26; p <.01), indicating that pooling in this model is not statistically appropriate. Furthermore, the Chow test for the receipts prediction model also extremely exceeds the critical F-value (F = 239; p <.01), indicating that pooling for this model is not allowed either.

4.2.3 Partially pooled regression model

In order to account for the differences in the intercepts, a partially pooled model is

created with different intercepts for 1.) the near presence of competitor A, 2.) the near presence

of competitor B and 3.) the near presence of competitor A and B. By using a partially pooled

OLS regression model, also known as Ordinary Least Squares with Dummy Variables, it is

possible to strike a balance between the benefits of pooling (increased statistical efficiency) and

to accommodate the heterogeneity between the variables by including separate intercepts

(Leeflang, et al., 2015). The estimations of the estimation models for predicting sales and

Referenties

GERELATEERDE DOCUMENTEN

For the case of this study, the perspective of Colombian journalists regarding the hard news paradigm versus a more interpretative style of journalism is relevant as it influences

How does Chinese Outward Foreign Direct Investment (OFDI) compare to OFDI from the three other main FDI source countries to the SADC region, and Zambia more specifically, and how

The main objective of this research was to understand the relationship between CSR and sales and the moderating effects of brand equity and GDP on the relation between CSR and

Therefore, it might be expected that these more task-oriented customers buy less in terms of quantity (e.g. only the lamp). At the same time, we can expect that

[r]

The influence of assortment variety (independent variable) on clothing store choice likelihood (dependent variable) has been tested and this relationship may be influenced by the

How are store characteristics related with customer loyalty behavior, including the moderating effects of different shopping motivations and fashion involvement, focused

In addition, we therefore analyzed the effects a more hedonic brand attitude has on the individual components of Customer Performance, which showed that a brand store with a