The effect of different types of promotional strategies on firm growth

(1)

The effect of different types of promotional strategies on firm growth

Master Thesis

University of Groningen - Faculty of Economics and Business

Msc BA – Specialization Marketing Management and Marketing Research

August 2013

Frédérique Pigeaud (s1622587)

Roelof Hartstraat 17-3

1071 VG Amsterdam

0619820064

F.Pigeaud@student.rug.nl

(2)

2

Preface

Almost two years ago, I started my master Business Administration in both Marketing Management and Marketing Research at the University of Groningen. Last September, I started my internship at H.J. Heinz, where I had the opportunity to use their databases for the dataset of this master thesis, to complete the two years of my master study and also my study period in Groningen. My supervisor at the Trade Marketing department of H.J. Heinz, Dave Sanders gave me the opportunity to find an interesting subject for my thesis within the organization, and stimulated me to think beyond the borders I thought I had for this thesis. As I was working a lot on category projects and with data on promotions for my internship, this triggered me to go beyond the day to day reasoning at Heinz. During one of the brainstorm sessions with a few colleagues, the idea to look closer at the type of price promotion and the stage of the product lifecycle was born. The managerial relevance for H.J. Heinz to gain more insights in the factors influencing the effectiveness of their promotions was high, as tools to evaluate this were hard to create.

Special thanks go out to Dave Sanders, my supervisor during my internship, for providing me with the (confidential) data and helping me starting up the process .Furthermore I would like to thank the Trade Marketing department at H.J. Heinz for assisting me and providing me with a tool for promotional effectiveness to complement the dataset. Another special thanks to my supervisor from the Rijksuniversiteit Groningen, Maarten Gijsenberg as he provided timely feedback, helped me to look back from where I stood on the process and provided me with great ideas for the modeling part, encouraging me to improve the quality of this thesis. Last but not least I would like to thank my family and friends for their interest and support during the last few months. Special thanks to my parents, who have supported me with my studies and everything around that the past years. Thanks again to everyone and I hope you enjoy reading this thesis.

(3)

3

Managerial Summary

The effectiveness of promotions is a very well researched topic in the marketing literature. Firms in the FMCG business see promotional strategies as a mean to create brand awareness and additional sales, thereby influencing brand and category dynamics, wi th the ultimate goal of increasing firm growth. However, developing an evaluation tool for promotional strategies with a link to the long term performance or firm growth is still a grey area for firms in the FMCG sector. The objective of this study is therefore to increase understanding of how promotions increase firm growth goals like volume targets and market share. According to Dekimpe et al. (2005), price promotions are the most often used form of promotional support. Therefore, the drivers under investigation are two types of price promotions that are used most in the Netherlands, which are Temporary Price Reductions and Multibuy promotions. A multilevel model is used to evaluate the effectiveness of both types of promotions. At level one of the model the variables price, and non-price promotions, namely Display only, Feature only and Display and Feature promotions will be included as well. As organizations like the company of H.J. Heinz often cope with the question which promotions work best for new products versus more mature products, the stage of the product lifecycle will used as a contingency variable at brand level in this research, by translating the theory from category to brand level and including this variable at level two of the model. Furthermore, several other moderating variables, namely category growth, category concentration, brand (reference) price and weighted distribution will be included in the model at level 2. The data is measured on a weekly basis, and provided by ACNielsen and PI_web. 108 weeks are covered, by looking at January 2011 until February 2013. Different brands within six product categories in the Netherlands were analyzed. Also, data on promotions from different supermarkets was used. This extensive dataset therefore provides a good basis for empirical generalizations across brands and categories.

(4)

4

(5)

5

1. Introduction

The effectiveness of promotions is a very well researched topic in the marketing literature. When looking at firms in the FMCG business, promotional strategies are seen as a mean to create brand awareness and additional sales, thereby influencing brand and category dynamics, with the ultimate goal of increasing firm growth. However, developing an evaluation tool for promotional strategies with a link to the long term performance or firm growth is still a grey area for firms in the FMCG sector. There is continuing concern with what happens after the promotion (Gedenk and Neslin, 1999). Also, the optimal level of promotions is seen differently from both the manufacturer and the retailer. Nijs et al. (2001) argue that many leading manufacturers would like to reduce their excessive reliance on price promotions but are reluctant to do so, lest they lose the support of retailers who still appreciate the market expansion power of price promotions, as it is a way of getting more customers into their store and increasing basket size. However, Srinivasan et al. (2004), argue that a price promotion typically does not have permanent monetary effects for either party. Moreover, retailer category margins are typically reduced by price promotions, even when accounting fo r cross-category and store-trafﬁc effects. This would indicate that it might not even be in the interest of retailers to sell in promotions.

In order to explain consumer purchase dynamics, where promotional strategies are an important area, manufacturers in the FMCG industry use observations from household panel and scanner data. How to optimally exploit this data in order to be able to evaluate promotional strategies is still an area in which many things are not yet understood. The availability of supermarket scanner data has demonstrated the powerful short-term sales impact of retailer promotions. It is suggested that, if deals become more effective in the current period and consumers are more price sensitive, promotions should be used more frequently, and if the negative dynamic effect of discounts on sales increases, the optimal level of discounting should go down (Bucklin and Gupta 1999).

(7)

7

price and non-price promotions can be used by both manufacturers and retailers. However, in their research they do not explain the effects of the specific underlyi ng types of promotions. Insights in the area of promotional strategies in relation to firm growth could help FMCG business managers to shape brand strategy, create an evaluation tool for promotional strategies, and shed some more light on the effectiveness of these promotions for both manufacturers and retailers. Therefore, this research attempts to add to the previous research by indeed looking deeper into the different, underlying types of price promotions. By specifying the relationship between the underlying types of both price promotions and non-price promotions, this research attempts to point out which type of promotion is most effective. Furthermore, by including other brand and category variables that have a direct and/or indirect effect on firm growth, this research looks into other factors that might influence the relationship between promotions and firm growth. Finally, this research adds to existing work by incorporating the stage of the product lifecycle in order to find out whether the effect of a type of promotion might be dependent upon the stage of the product lifecycle, as it is relevant whether a specific type of promotion is more suitable for a new versus mature product. Translating research on the PLC to brand level, this research will investigate whether different types of promotional strategies are more effective than others for brands with relatively new product introductions where distribution is still growing, or more mature products with stabilized distributions.

In summary, the existing literature stresses the importance of promotions for companies and the effect it has on different dynamics in relation to firm growth. However, little is documented on the effect of different types of price promotions in relation to firm growth. The objective of this study is therefore to further investigate the influence of different types of promotions on firm growth, looking deeper into the category of price promotions and acknowledging that there are different types of promotions within this category that can have implications on brand and firm growth. The contribution of this study within the research on promotional strategies is manifold. First, by linking the effects of price promotions not only to sales, but to firm growth, the existing resear ch on promotions is expanded. Second, by specifying specific effects of the underlying types of price promotions and non-price promotions, this study tries to get into a more in-depth analysis of promotions. Third, this study is to my knowledge one of the first studies that looks at the different effects that promotions can have in different stages of product development using the literature on the product lifecycle and translating this to brand level, acknowledging that the effect of promotions can be different for brands with many new versus mature products. Furthermore, different product categories will be considered with various brands and manufacturers, which should allow continuing solidifying empirical generalizations regarding the influence of different types of promotions. The ability to generalize cross-category provides an indication as to what type of promotions exert different effects across different categories and if this can be standardized cross -category. Finally, this study attempts to combine datasets from two different sources namely PI-web and ACNielsen for one specific category. By combining this, companies can gain more insights in how to optimally exploit this data and enrich their data analyses, thereby gaining more insights into the F ast Moving Consumer Goods market.

(8)

8

design will be elaborated, followed by a preliminary and a multilevel or Linear Mixed Model analysis. In the results section, implications per category are discussed as well as cross -category generalizations. Finally, conclusions, managerial implications and directions for future research are provided.

2. Theoretical framework

Promotional strategies are seen as a means to create brand awareness and additional sales, thereby influencing brand and category dynamics, with the ultimate goal of increasing firm growth. The growing importance of sales promotions in many categories led researchers to start evaluating their effectiveness about twenty years ago. Blattberg and Neslin (1990) outline four ways in which sales promotions can affect sales. First, promotion induces brand switching because it provides a strong reason (lower price) for the consumer to buy the promoted brand instead of another brand (Gupta 1988). Promotion can also induce store switching, that is, consumers make their purchase in one store rather than another, because promotion schedules differ from store to store (K umar and Leone 1988). Third, category expansion refers simply to the ability of a promotion to increase primary demand for a product category. Finally, purchase acceleration refers to consumer stockpiling behavior, which can take the form of consumers purchasing products in the category earlier than they ordinarily do and/or purchasing larger quantities than usual (Gupta 1988).

However, although much research has been conducted with regard to the short - and long-term effect of price promotions on brand choice or brand sales, it has yielded contradicting results. Blattberg and Neslin (1990) use Koyck modeling to determine the effect of promotions on consumer brand choice. The outcome of their research has indicated an existence of a negative promotion usage effect on consumer behavior, indicating that promotions do not induce consumers to buy or use more of the product. On the other hand, Ailwadi et al. (2001) found that the mere purchase effect of promotions is positive, as promotions induce consumers to ( temporarily) buy more and consume it faster. Dekimpe et al. (1999) used persistence modeling to capture a more permanent effect of promotions on sales. They found that permanent effects of promotions are largely absent. Finally, Pauwels et al (2002) add to previous research by looking at the effects of promotions on brand sales. In terms of quantification of promotion effects, several studies break down the immediate impact on incidence, choice, and quantity. Therefore, Pauwels et al. (2002), quantify brand sales as having three components, namely, category incidence, and brand choice and purchase quantity. Their findings are consistent with those of Blattberg and Neslin (1990), indicating that permanent promotion effects are virtually absent. Overall, brand choice elasticities tend to be negative, whereas category incidence effects are typically positive (Pauwels et al., 2002).

(9)

9

Figure 1: Conceptual model, displaying the factors affecting the relationship between the type of promotion and firm growth

Regarding this relationship, firm growth is the dependent variable. Price, the type of non-price promotion and type of price promotion can be seen as independent variables. Weighted distribution, the stage of the brand-level product lifecycle, the brand reference price, category concentration and category growth are moderating variables in this research.

2.1. Defining firm growth

Managers in consumer and industrial sectors alike seek long-term profitable growth for their products and services. According to Nijs et al. (2001), such growth can be found from three sources: growth in category or product demand, in market share, or in profit margins. Competitive conditions largely indicate which sources of growth can be pursued realistically and for how long. Alth ough much of the market response literature has focused on the effects of various marketing resource allocations on brand sales, the implications for identifying the best sources of pro fitable growth are not yet well understood, even less so in the long run (Nijs et al., 2001). This research will combine different streams of research on firm growth and focus at the first two sources of growth as defined by Nijs, namely growth from category or product demand and growth in market share. Growth in profit margins will not be addressed, as this is usually seen from the perspective of the firm as a whole, while this research intends to look at different levels of the firm in itself, thereby assessing firm growth by looking at both brand growth, and growth within the category in which the brand operates

2.2.1. Growth in brand or category demand

Although promotions have increased in both commercial use and quantity of academic research over the last decade, most of the attention has been focused on their effects on brand choice and brand sales. By contrast, little is known about the conditions under which price promotions expand short

-+ - +/- + + + Weighted distribution Category growth Brand Reference Price Firm growth Category concentration Type of non-price promotion: F/D/FD - + Price Type of price promotion:

(10)

10

run and long-run category and product demand, even though the benefits of category or product expansion can be substantial to manufacturers and retailers alike (Nijs et al., 2001). By looking deeper into which models are used in research on firm growth, variables that can be used as a proxy to firm growth can be identified and used in this research. Abratt et al. (1994) state that the marketing department identifies market share growth followed by profit achievement and volume growth as the most important benchmarks for success. FMCG products typically attract high volume sales at a low monetary value. However, due to inflation, prices of FMCG products increase rather fast. Hence, growth in volume is a better target than growth in value. Moreover, the goal of volume growth is more in line with the goal of increasing penetration and/or purchase frequency, as with volume growth, the focus is on increasing the customer base by adding new customers and triggering customers to buy more. Therefore, volume growth will be used in this research as an indication of growth in category or product demand.

2.2.2. Growth in market share

Yet another important measure for firm growth is the growth in market share. According to Venkatraman and Ramanujam (1986), market share can be seen as an operational indicator of firm growth. The market-share position is widely believed to be a determinant of profitability (Buzzell et al. (1975), and therefore it would be a meaningful indicator of performance in this perspective. In terms of the strategic role of market share, one school of thought is that, given the oligopolistic advantages of large firms, market share may be better regarded as a size variable than a performance variable (Peng & Luo 2000). On the other hand, research by Davies & Geroski (1997) has demonstrated that there is “considerable turbulence in market shares even among leading firms". And also, although it is not clear whether a stronger market share can cause higher profits, thus increase firm growth in this sense as well, these two constructs seem to be highly correlated (Oster, 1994). Therefore, share growth will also be an indicator of firm growth in this research.

2.2. Effect of promotions

(11)

11

According to many different scholars, promotions can be divided into price and non -price promotions. Figure 3 shows the distinction between promotion instruments that retailers use.

Figure 2: Based on “Instruments for Retailer promotions” (Gedenk et al., 2010).

2.2.1. Price promotions

According to Dekimpe et al. (2005), price promotions are the most often used form of promotional support. As such, it should come as no surprise that the eﬀectiveness of price promotions has been studied extensively in the marketing literature. Price promotions are promotions that are related to a discount in price. According to Gedenk et al. (2010), the price promotion instrument used most often is a temporary price reduction (TPR). A TRP is a promotion where the shopper will receive a price reduction on every single product bought. Gedenk et al. (2000), state that in addition to a TPR, a coupon is the most used form of price promotion. With coupons, consumers have to bring the coupon to the store in order to get a discount (Gedenk et al., 2010). However, other forms of price promotions are possible. First, retailers can also use multi -item promotions (e.g., “buy three for x” or “buy two get one free”), in order to receive a product for free, consumers will now have to buy at least 2 or more items, which will lead to a higher margin on the product for the retailer. With a cross category Multibuy, these items can be in different categories, as to stimulate cross-buying. A specific form of Multibuy promotions is the so-called BOGOF (Buy One Get One Free). A BOGOF is a way of making consumers buying one product and receiving an extra (same) product for free.

Furthermore, retailers can use promotion packs or “Free Volume” promotions, i.e., packages with extra content. This refers to the fact that consumers will get a larger package for the same price as their regular package of the product (e.g., “25 % extra”). Loyalty discounts also require the purchase of several units, but the consumer can do this over several purchase occasions. In general, several studies indicate that consumers’ perception and attitudes to sales promotion tools can be very arbitrary (Lammers, 1991; Ong, 1999; Ndubisi and Moi, 2006). Ailawadi and Neslin (1998) find that promotions induce consumers to buy more and consume faster. This would indicate that promotions

(12)

12

have a positive short-term effect on penetration and volume growth, as well as share growth. However, it is debatable if this effect holds in the long term. Shi et al. (2005) did a study on behavioral response to sales promotion tools and found that consumers respond more positively to TPR’s, BOGOF, and coupons than to other tools of sales promotion. Their suggestion is that this may be because these tools are easy to understand and can provide consumers with transaction utility. As this research focuses on the Dutch retail landscape, promotional strategies that are used most in the Netherlands will be taken into account, which are TPR, BOGOF and a Multibuy. It is expected in this research that these types of price promotions will have a positive direct short-term effect on firm growth. However, research on the question of discounts versus free units is inconclusive, which makes it interesting to go deeper into this area of price promotions. Some researchers suggest that a discount is questionable because it is sometimes “translated” as a devaluati on of the product. Key to an effective price discount is the extent to which the consumer is convinced that the price discount is substantial and provides value (Blair and Landon 1981; Blattberg and Neslin 1990). This is usually accomplished by providing a reference price at the point-of-purchase. Diamond and Sanyal (1990) found that participants prefer extra units of the product rather than a price discount. Darke and Chun (2005) looked at the quality perceptions of types of promotions and found that cons umers often consider the discount price rather than the initial price to be the true price of the item. However, when a BOGOF promotion is offered, consumers establish their quality judgment as to the full price of the item without accounting for the value of the gift. Therefore, with a BOGOF, quality perceptions are higher than with a TPR. As a result, the BOGOF increases the value of the deal relative to a TRP.

Conversely, Thaler (1985), Monroe and Chapman (1987), Diamond and Campbell (1989) state that a price discount reduces loss, as it decreases spending, while an additional unit increases gain, namely the benefit from the free unit. Since loss aversion means that losses loom larger than gains of the same amount, discounts are preferable to a free additional unit. These results are supported by the research of Sinha and Smith (2002). In their research, they asked students to rank the transaction value of three different possibilities on a scale of 1 - 5. The possibilities were 50% discount, a BOGOF promotion, and a Multibuy, namely buy two and get a 50% discount on both. Participants preferred the 50% discount over the other two promotions. Further, they found a higher transaction value for the BOGOF promotion over the Multibuy or buy two and get a 50% discount on both offers. Due to data limitations, BOGOF promotions and Multibuy promotions will in this research be seen as one type of promotion, under the header of Multibuy promotions. As stated before, it is expected in this research that both types of price promotions will have a positive direct short-term effect on firm growth, however, a TPR promotion is expected to be more effective than a Multibuy promotion.

12.2.2. Non-price promotions

(13)

13

focus will be on supportive non-price promotions. “Supportive” non-price promotions are communication instruments used to alert the consumer to the product or to other promotion instruments. Very often, they are used to draw attention to price promotions. Thus, the focus is not so much on the brand as on price (Gedenk et al., 2010). The type of “supportive” non-price promotions that will be taken into account in this research is the fact that a product is on Display, and/or Featured in a folder.

The Pearson correlation statistic is used to test the relationship between price promotion and supportive, non-price promotion. As can be seen from table 1, significant negative correlations exist between price and non-price promotions, while positive correlations exist between all price promotions and non-price promotions.

Table 1: Pearson Correlation

PRICECUT MULTIBUY Price

FeatureandDisplay Pearson Correlation ,193** ,327** -,139**

Sig. (2-tailed) 0 0 0

N 4536 4536 4536

Display Pearson Correlation ,079** ,104** -,226**

N 4536 4536 4536

Feature Pearson Correlation ,142** ,333** -,127**

N 4536 4536 4536

** Correlation is significant at the 0.01 level (2-tailed).

Table 1: Pearson correlation of price, price promotions, and non-price promotions

(14)

14

expected to be most effective, followed by a Feature only. A Display only promotion is expected to be the least effective form of a non-price promotion.

In summary, this research will look into price promotions and non-price promotions for the company of H.J. Heinz and several important competitors in each product category. Price promotions will be categorized in TPR promotions versus Multibuy promotions. It is expected in this research that both types of price promotions will have a positive direct short-term effect on firm growth, however, a TPR promotion is expected to be more effective than a Multibuy promotion. Non-price promotions will be categorized in Feature only, Display only or Display and Feature promotion. The expectations of this research are that all types of supportive non-price promotions have a direct positive effect on firm growth, however, a Feature and Display promotion is expected to be most effective, followed by a Feature only. A Display only promotion is expected to be the least effective form of a non-price promotion.

2.3. Brand growth factors

Another brand growth factor that is expected to have both a direct effect on firm growth, and thereby also an effect on the relationship between promotional strategies and firm growth is price.

2.3.1. Price

The price of a product is of importance when looking at the effect of promotions on firm growth, as, according to Lichtenstein et al. (1993) it represents the amount on money that must be given up. Therefore, purchase probabilities are negatively affected by higher prices. Therefore, it is expected that price has a direct negative effect on firm growth, with lower price leading to more growth. As with a Multibuy promotion, consumers are likely to perceive the price as being lower in absolute terms than with a TPR promotion, due to the fact that this is often in a buy one get one format, price is expected to have a larger negative effect on firm growth with a Multibuy promotion than with a TPR promotion. + + Weighted distribution Category growth Brand Reference Price Firm growth Category concentration Type of non-price promotion: F/D/FD Price Type of price promotion:

(15)

15

2.4. Moderating brand and category growth factors

First, two brand-growth factors that play a role in the marketing mix, and are likely to have a moderating effect on the relationship between type of promotion and firm growth are included into the conceptual model and described below in more detail. Next, promotional elasticities vary significantly from category to category (Narasimhan et al., 1996). Therefore, in order to investigate the relationship between category characteristics, promotional effectiveness and firm growth, it is important to incorporate category characteristics into the model in this re search.

2.4.1. Stage of the brand-level lifecycle

In order to state promotional effects at brand level, it is likely that the effects of promotions for a brand containing many new products, where distribution is not at an optimum level and still growing, are differing from those of a brand with many mature products, with a stable distribution. Research in marketing as well as strategic management indicates that the Product Lifecycle (PLC) is likely a fundamental variable affecting business strategy.

Research by Anderson & Zeithaml (1984) suggests that differences in strategic variables between stages of the product life cycle (PLC) exist, as well as differences among the determinants of high performance across stages of the PLC. The results support the use of the PLC as a contingency variable during strategy formulation on category level. Hofer (1975) developed the most extensive theoretical profile of the PLC as it affects business strategy. He states that the most fundamental variable in determining an appropriate business strategy is the stage of the product life cycle, and furthermore, major changes in business strategy are usually required during the stages of the lifecycle. In the introduction stage, strategies emphasize a buyer focus, building on advertising, and increasing purchase frequency. Product development is seen as important (Hofer, 1975). In the growth stage, there is a movement toward strategic segmentation and building efficiencies in production and marketing. The capital investment and expenses associated with these strategies may be detrimental to short term profits. Intense distribution is also emphasized. High performance

+ + Weighted distribution Category growth Brand Reference Price Firm growth Category concentration Type of non-price promotion: F/D/FD - Price Type of price promotion:

(16)

16

strategies for the maturity stage are more complex than for the previous two stages because of the larger number of research studies and variables. Hall (1980) provides some empirical evidence for high performers in the maturity stage and derives two distinct, comprehensive strategies. These are: (1) achieve the lowest delivered cost position relative to competition, and (2) achieve the highest product/service/quality differentiated position relative to competition. Hall points out that the best performers use both of these strategies. Finally, relatively little work has been done regarding strategies leading to high performance in the decline stage. Strategies ranged from an immediate exit to increasing investment in the declining business (Anderson and Zeithaml 1984).

As said before, the stage of the PLC can be used as a contingency variable for comparing categories. However, as this research tries to define the impact of promotions at brand level within mature categories, the theory behind the product lifecycle can only be used as a framework. In this research, the knowledge of strategy formulation within each lifecycle stage on category level will be translated to brand-level. As the differences in strategies in each stage of the lifecycle already indicate a possibility for different promotional needs for categories that are in one of these stages , it is very likely that this will hold stages of the lifecycle at brand level as well, and different needs for promotional strategies may exist for brands with relatively a lot of new products versus brands containing many mature products, or declining products. A weighted brand-level product lifecycle variable is therefore created.

Research by Bronnenberg et al. (2000), states that mature markets are less likely than emerging markets to exhibit permanent effects of marketing actions, including promotions. Next to this, Pauwels et al. (2002), find that price promotions are only likely to have permanent effects on category incidence, brand choice and purchase quantity if it concerns a new product category. Permanent effects of promotions seem unlikely in mature product categories. Based on this research, the overall effect of promotions will be more positive in the introduction stages of a brand than in the maturity stage. If this also holds when comparing new and mature brands in mature markets, all types of price promotions will be expected to be most effective in the in the start-up and/or growth stage, followed by the maturity stage. All types of price promotions will be expected to be least effective in the decline stage of a brand’s lifecycle.

(17)

17

lifecycle, while a TPR promotion will have the largest positive effect in the start-up en growth stage of a brand’s lifecycle.

In summary, the stage of the product lifecycle will in this research be used as a contingency variable at brand level, by translating the theory from category to brand level. All types of price pro motions are expected to be effective in the in the start-up and/or growth stage and the maturity stage. All types of price promotions will be expected to be least effective in the decline stage of a brand’s lifecycle. Furthermore, Multibuy promotions are expected to have the largest positive effect in the maturity stage of a brand’s lifecycle, while a TPR promotion will have the largest positive effect in the start-up and growth stage of a brand’s lifecycle.

2.4.2. Weighted distribution

Weighted distribution can be defined as the percentage of stores where a particular product is available divided by the total number of stores. Ataman et al. (2009) showed the importance weighted distribution on firm growth, stating that the more available the product the more opportunity the customers have for buying the product. Greater distribution intensi ty can lead to greater sales volume for the brand, at least in the short run and within limits (Corstjens and Doyle, 1979). In this research, both types of promotional strategies are expected to be more effective when combined with high weighted distribution.

2.4.3. Reference price

Some researchers have argued that price is more complex in its nature. One stream of research investigating this link is based on the notion that the consumer establishes a reference price for a brand or product (Monroe 1979; Winer 1986). The reference price reflects the expectations of the consumer, which are shaped by the past pricing activity of the brand. The consumer then evaluates the future price of the brand in relation to this reference point and his or her response is related to the disparity between the two. Hence, consumer response to an unexpected price decrease, or a

+/- + + Weighted distribution Category growth Brand Reference Price Firm growth Category concentration Type of non-price promotion: F/D/FD - Price Type of price promotion:

(18)

18

"pleasant surprise", is greater than the response to an expected price decrease. The reference price framework is consistent with several psychological theories of consumer behavior and price perception. Empirical work by Winer (1986) and Raman and Bass (1986) supports the presence of general reference price effects in consumer brand choice behavior. Therefore, the average brand price within the 108 weeks, is also accounted for as the so-called brand price in this research. It is expected that the reference price of a brand has a negative effect on firm growth. Moreover, price promotions are likely to be more effective when the reference price is high. Therefore, the brand price is expected to have a negative indirect effect on the relationship between both types of price promotions and firm growth.

2.4.4. Category concentration

Category concentration defines the market share of the main brands within a particular category. The closer a market is to being a monopoly, the higher the concentration. In order to measure category concentration, this research uses the Herfindahl - Hirschman index. The Herfindahl - Hirschman index (HHI), is defined as the sum of squared market shares of firms in a market and thereby provides an easily interpretable measure of concentration. (Leijsen, 2004). It is calculated by squaring the market share of each firm competing in a market, and then summing the resulting numbers. The HHI is expressed as:

Highly concentrated categories do not contain many different brands, which could mean that there is less competition within the market. Economic theory suggests that in concentrated markets, proﬁt margins are higher (Steenkamp et al. 2005). Companies may be less motivated to engage in a price war in such markets because it dissipates attractive high margins (Ramaswamy et al. 1994). This encourages ﬁrms to substitute nonprice forms of competition such as advertising for price competition (Lipczynski and Wilson, 2001). Moreover, Putsis and Dhar (1998) foun d that noncooperative response to price promotions is more likely in less concentrated markets.

Category concentration is thus expected to have positive indirect effect on the relationship between promotional strategies and firm growth. It can be expected that the effect of both types of promotions is higher in low concentrated categories versus high concentrated categories.

2.4.5. Category growth

(19)

19

Although previous research is inconclusive, in this research, it is expected that catego ry growth positively influences the relationship between both types of price promotions and firm growth, as high-growth categories are likely to exhibit more promotional activity than low-growth categories.

2.5. Research question

Having introduced the academic and managerial relevance , as well as the theoretical framework, the research question can be formulated:

“What is the effect of the type of promotion on firm growth and how does this differ for brands with many new product introductions relatively to mature brands or declining brands?”

Methodology

3.1. Sample description

In order to address the research question “What is the effect of the type of promotion on firm growth and how does this differ for brands with many new product introductions relatively to mature brands or declining brands?”, a sample of different product ranges in categories in which the company of H.J. Heinz is active will be used for analysis. Data is gathered in the Netherlands and the hypotheses are tested using a large set of 1210 price promotions in six product categories, namely Fruitbased Drinks, Breadtoppings, Italian Pasta, Pastasauces, Soups and Snacksauces, within a total of 108 weeks, ranging from the first week of 2011 until the fourth week of 2013. The product categories that are used for analysis were selected upon two different criteria. First, these categories are known for a fair amount of promotions for all brands, in order to account for a large numbe r of observations. Secondly, there have been new product introductions over the past two years in these categories, in order to account for the moderating variable, namely, the stage of the product-level lifecycle. + - +/- + + + Weighted distribution Category growth Brand Reference Price Firm growth Category concentration Type of non-price promotion: F/D/FD - + Price Type of price promotion:

(20)

20

3.2. Dataset development

For every brand, data on the different types of price promotions was obtained from the ACNielsen retail scanner database in combination with the PI-web promotional database. Data on non-price promotions and other brand level constructs as well as category-level constructs was also obtained from the ACNielsen retail scanner database. Data on price promotions was collected from the six main supermarkets in the Netherlands, namely Albert Heijn, Jumbo, C1000, Plus, Coop and Spar. The effect of promotions on firm growth will be assessed at an aggregated level of retailers, excluding hard discount formulas like Aldi and Lidl. Hard discount formulas will be excluded first of all because the focus of this research is on A-level brands, and Aldi and Lidl only sell B-level or private label brands. Furthermore, the data in ACNielsen of these formulas is not typical scanner data. Instead of scanner data, sales of products at hard discount are assessed by a sample of receipts from these formulas. Therefore, data of hard discount formulas is less reliable and these formulas will not be included in the analysis.

3.3. Data limitations

As can be seen from table 2, the majority (55%) of the promotions are TPR’s. Table 3 describes the different types of price promotions that were used in the sample. The majority (39%) of the non-price promotions are Display promotions. However, it does not seem to vary very much over categories, what type, and how many promotions are used. The sample could pose problems when performing a regression analysis, because each variable approximately needs 15 (10-20) observations (Hair et al., 2006). In other words, the sample needs to be representative of the population for the inference prediction. As firm growth is dependent upon 12 variables, the category-promotion type combination should obtain at least 180 observations for each type of promotion. As can be seen from table 2, this requirement might not hold for every category-promotion type combination. Table 2: Category – type of promotions

Table 2: Promotion types per category

Category Price Promotions Multibuy TPR

(21)

21

3.4. Model description

3.4.1. Model type

Regression and Generalized Linear Models assume error terms are independent and have equal error variances. However, as the data used in this research are nested or cross-classified by groups, namely different product categories, individual-level observations from the same upper-level group will not be independent but rather will be more similar due to such factors as shared group history and group selection processes. While random effects associated with upper-level random factors do not affect lower-level population means, they do affect the covariance structure of the data (Garson, 2012). Therefore, a multilevel model will be used for analysis, which looks at both brand-level, and category level dynamics. Multilevel models are used for the analysis of data with complex patterns of variability, with a focus on nested sources of variability (Mason et al, 1983, Goldstein 1987). These types of models have also been called Hierarchical Linear Models (Raudenbush, 1993) random coefficients models (Rosenberg, 1973), covariance components models (Dempster et al., 1981) and unbalanced models with nested random effects (Longford, 1987). This class of models combines the advantages of the mixed-model ANOVA, with its flexible modeling of fixed and random effects, and regression, with its advantages in dealing with unbalanced data and predictors that are discrete or continuous (Raudenbush, 1993). Multilevel models recognize the existence of data hierarchies by allowing for residual components at each level in the hierarchy. Thus the residual variance is partitioned into a between-variable component (the variance of the brand-level residuals) and a within-variable component (the variance of the week-level residuals). The brand residuals, represent unobserved brand characteristics that affect weekly outcomes. It is these unobserved variables which lead to correlation between outcomes within the same brand. They are useful for explaining excess variability in the dependent variable, and included in the second level of the model. The model used in this research, exists of two different levels and will be applied to several brands in different product categories of the company of H.J. Heinz, and will look into the dynamics that influence firm growth for every brand in specific categories in which the company of H.J. Heinz is active.

3.4.2. Level one variables

At level one of the model, time-varying brand level characteristics that explain the relationship between the type of promotion and firm growth will be included into the model. As the literature on firm growth has indicated, firm growth can be defined by both growth in category or product demand and growth in market share. Volume will be used as a proxy for growth in category or product demand. Data on the volume and market share of the category over the years 2011 and 2012 will be obtained from the ACNielsen retail scanner database, which makes it possible to collect data on a weekly basis and assess the effect of promotions on firm growth over 108 weeks. The promotional strategies described in chapter 2 will be analyzed for these specific brands .

(22)

22

TPR promotions. For each type of promotion, it will be assessed which percentage of the total volume is sold in this type of promotion. In order to do so, the availability of the promotion will be assessed. The retail format will be accounted for by evaluating the units sold in a promotion on a basis that is specific for each the retail format. This is done by looking at the so-called “lift-factor” of a promotion, by using a measure which is referred to as the Promotion Efficiency Index (PEI). The PEI assesses which percentage of items sold in one week, was sold in promotion. It calculates the percentage of what the units sold in promotion have added to by the so-called promoted base units that would have been sold anyway even if there was no promotion. As this factor is calculated per retailer, it automatically accounts for the size of the retailer. Promotional data will be collected from the PI web database, the ACNielsen retail scanner database , and a so-called Account tracker tool. Data on Display, Feature, or Display and Feature promotion will also be collected from the ACNielsen retail scanner database. Within each product category, the promotions of three or four major competitive brands will be taken into account.

Also, it can be expected that the effect of an advertising or promotional campaign does not end when the promotion is over. Leeflang et al. (2000) have proved that the effect of advertising last longer then only in the period of advertisement or the advertising campaign. Promotions create lagged effects, consistent with the idea that consumers engage in stockpiling (van Heerde et al. 2000). Stockpiling effects are decreases in post-promotion category sales in an extended time period excluding the week of the promotion. Therefore, it can be stated that the effect of a promotion, or part of it, will remain perceptible for some future periods, affecting brand sales to be lower in the week after a promotion to be below average. A lagged effect of price promotions will be taken into account, by looking at the PEI of the previous week. Also, as many different retailers were taken into account, and many writers have seen that the lagged effect wears out over the second week after a promotion, the lagged effect in this research will only last one we ek.

Furthermore, the level one model will include price and seasonality. Price will be measured by looking at the regular or base price in euros per liter or kilogram at which the product is sold. Seasonality can be described as the fluctuation in output and sales, not necessarily standard, through the year, caused by changes in the weather, the calendar, and/or timing of decisions (Brendstrup et al., 2002). It can be assumed that product sales fluctuate throughout the year, affecting firm growth this way. Therefore, seasonality will have an effect on the relationship between both types of promotional strategies on firm growth. To investigate this effect, all weeks are divided into the four quarters of a year, and included as dummy variables for each season into the model, and used for analysis.

3.4.3. Level two variables

Finally, after assessing the effects that promotions have for each specific brand within a specific category, level two of the model will include non-time-varying variables that explain part of the constant (ß 0 ) of level one. First, brand-level variables will be included at this stage. One brand-level

(23)

23

The effect of the stage of the product-level lifecycle will be assessed, by including this as a moderator variable. A weighted PLC variable will be included for each brand, calculated by multiplying the growth in the weighted distribution of every SKU within a brand, with the market share within that brand of each separate SKU. Afterwards, the average of these numbers per SKU will be used to determine the brand-level stage of the PLC. A continuous PLC variable will therefore be created Brands can also be categorized into start-up/growth, mature, or declining products, dependent on whether the multiplication of the average weighted distribution and brand share is positive (growing) for a brand in the introduction or growth stage (brands growing with more than 1% over the period of 108 weeks), around zero for a mature brand, or declining for a brand in the decline stage. According to these criteria, 33% of the brands in the start-up/growth stage of the PLC, 65% were in de mature stage and 2% of the brands were in the decline phase.

Furthermore, at level two of the model, cross-category results will be looked at by taking into account two category-level variables as well. First, category concentration will be looked at by means of the HHI index, or the sum of squared market shares of firms in a category over 108 weeks. Secondly, the growth of the category will be assessed by looking at the overall category average growth percentage over the 108 weeks.

Table 3: Model constructs

Construct Variable Definition Measurement Database

Promotion variables

TPR Promotion TPR promotion

The percentage of the total volume that was sold in this type of promotion

PEI PI-web,

ACNielsen Multibuy Promotion MB

promotion

The percentage of the total volume that was sold in this type of promotion

PEI PI-web,

ACNielsen

Feature promotion F The percentage of distribution in this type of promotion

% of Weighted distribution

ACNielsen Display promotion D The percentage of distribution in

this type of promotion

ACNielsen Feature and Display

promotion

DF The percentage of distribution in this type of promotion

ACNielsen

Lagged effect TPRt-1/MBt-1 The “after-promotion” effect, measured by the PEI in the previous period

PEI t-1 PI web, ACNielsen

Brand-level variables

Price P Base Price of the brand, calculated when it is not in promotion, by dividing total euro sales by litres or kilograms sold

Base Price per L/KG (€) ACNielsen

Reference Price BP The average Base price of the brand, calculated average over 108 weeks

The average Price per L/KG

ACNielsen

Distribution WD The percentage of stores where the brand is available divided by the total number of stores, calculated average over 108 weeks

Average weighted distribution (%)

(24)

24

Seasonality S 13-week periods, basically the four quarters of a year

Dummy 4 Quarters of the year Weighted brand level

product lifecycle

PLC Weighted PLC, calculated by multiplying the growth in the weighted distribution of every SKU within a brand, with the market share of each SKU, and then taking the average of all SKU’s within a brand.

Average(ΔWD SKU * Market share SKU)

ACNielsen Category-level variables Category concentration

CC The sum of squared market shares of firms in a category over 108 weeks

Herfindahl index ACNielsen

Category growth CG The overall category average growth percentage over 108 weeks

Volume growth ACNielsen

Dependent variables

Firm Growth (Volume) of the category

FGV Percentage of growth in Volume (Litres or Kilograms) sold of the brand

Volume growth (%) ACNielsen

Firm Growth (Market share) of

the category

ix

= α

0

+ ε

i

(25)

25

Table 4: Level one and two variables

Level of inclusion Variable Abbreviation

Level one TPR promotion TPR it

Multibuy promotion MB it

Display promotion D it

Feature promotion F it

Display and Feature promotion DF it

Lagged effect L it-1

Price P it

Seasonality dummy Residual

S it εit

Level two Weighted distribution WD i

Brand price BP i

Weighted Brand-level product lifecycle PLC i

Category growth CG i

Category concentration Residual

CC i εi

Dependent Firm Growth FG it

Table 4: Level of inclusion of model variables

4. Analysis

In this chapter, the dataset will analyzed by using several techniques. First, a preliminary analysis will inspect the data and explore the nature of the variables, and shed light on some interesting findings. Second, a multilevel analysis is performed in MLwIN, in order to state the relationships between the independent and dependent variables. As stated in chapter 3, the Linear Mixed Models procedure expands the general linear model so that the error terms and random effects are permitted to exhibit correlated and non-constant variability. The linear mixed model, therefore, provides the flexibility to model not only the mean of a response variable, but its covariance structure as well. In Linear Mixed Models, the independent variables are entered into the equation in the order specified by the researcher based on theoretical grounds. Variables or sets of variables are entered in steps, with each independent variable being assessed in terms of what it adds to the prediction of the dependent variable after the previous variables have been controlled for. Once all sets of variables are entered, the overall model is assessed in terms of its ability to predict the dependent measure. The relative contribution of each step of variables is also assessed (Pallant, 2010). In this research, a two-level model is used for analysis.

4.1. Preliminary analysis

4.1.1. Descriptive statistics

(26)

26

Table 5: Descriptive statistics

N Mean Std. Deviation TPR 4536 82,510 258,836 MULTIBUY 4536 56,510 193,501 PEI 4536 139,020 338,511 Lagged 4536 144,420 381,665 Price 4536 2,688 1,597 FeatureandDisplay 4536 6,043 12,457 Display 4536 3,974 6,101 Feature 4536 6,848 13,816 PLCoutcome 4536 0,792 1,122 Weighteddistribution 4536 88,209 24,090 BrandPrice 4536 2,688 1,555 Categorygrowth 4536 0,010 0,003 Categoryconcentration 4536 964,715 630,996 Valid N (listwise) 4536

Table 5: Descriptive statistics of model variables

Additionally, histograms have shown that both dependent and independent variables do not follow a normal distribution curve, which could indicate non-normality. However, in current studies, attention is not focused as much on normality of variables, but more on the normality of the error terms.

4.1.2. Outliers and missing values

First, Boxplots were inspected to detect outliers. Many outliers can be detected for all types of price and non-price promotions, as well as for the weighted distribution. Next, the Descriptives table is checked to get an indication of how problematic these outliers are likely to be. The 5% Trimmed Mean and mean values are compared. As they are very different, the outliers have to be investigated further, however, all scores appear to be genuine. Therefore, after close consideration of the outliers for each variable, no data had to be removed from the dataset. Furthermore, as can be seen from the descriptive statistics table, no missing values were found in the dataset.

4.2. Model Assumptions

(27)

27

4.2.1. Linearity

The linear mixed model is an extension of the general linear model, in which factors and covariates are assumed to have a linear relationship to the dependent variable. If the relationship is not linear, the results of the regression analysis will underestimate the true relationship (Osborne & Waters, 2002). This is also the case with linear mixed models. Partial regression plots are used to check for linearity. Based on these plots, the relationships between the independent variables and the dependent variable are not considered linear. Therefore, for several continuous variables, namely price promotion, non-price promotion, price, brand price, and category growth, the natural logarithm of this variable was included into the analysis (LN(X+1)), which also implies more linear relationships between the independent, continuous variables and the dependent variables.

4.2.2. Normality

Another assumption of regression is that the disturbances are normally distributed. To test if the disturbances are normally distributed, a normal probability plot or Normal Q-Q Plot of the unstandardized residuals can be used. In this plot, the observed value for each score is plotted against the expected value from the normal distribution.As the line is not reasonably straight, this would suggest that the disturbances are not normally distributed. The Kolmogorov-Smirnov test can also be used, with a non-significant result (sign. >0,05) indicating normality. However, in this case, the significance value is .000, suggesting violation of the assumption of normality. Fortunately, the problem of non-normality is only relevant with small sample sizes. In this research a large sampl e size is used, therefore signs of non-normality can be neglected. Nevertheless, non-normality does have an effect on the reliability of the p-values. Hence, if non-normality is detected, p-values close to the signification level of 0.05 will have to be assessed with extra care.

4.2.3. Multicollinearity

(28)

28

Nevertheless, researchers have shown that how variables are centered when testing MLM has implications for interpreting the results. In MLM analyses, which are regression based, the results of the Level 1 analysis (i.e., the intercept and slope) become the outcome variables in the Level 2 analysis—the Level 1 results must have a clear meaning (Raudenbush & Bryk, 2002). A meaningful zero-point for Xij is necessary for the MLM to have a meaningful interpretation. Consequently, meancentering will be used in this research to increase interpretation of the variables.

4.3. Model estimation

Model estimation requires an iterative computational procedure to estimate the parameters optimally. Maximum likelihood (IGLS in MLwIN) and Restricted Maximum likelihood (RIGLS in MLwIN) estimation methods are most often used for multilevel models. ML provides a description of the fit of the full model, whereas REML only takes into account the random parameters. ML determines the optimal population values for parameters in a model that maximize the probability or likelihood function, that is, the function that gives the probability of finding the observed sample data, given the current parameter estimates (Hox, 2002).

4.3.1. Empty model

First, the so-called empty model is specified, which has no predictors, other than the intercept. This model, Displayed in the appendix, is very useful as it shows the percentage variance to be explained at the different levels. In other words, at how much of the variance in firm growth is due to differences between weeks, and how much is due to the fact that these weekly observations belong to different brands. For the dependent variable Firm growth volume, first, the Beta coefficient for the intercept is 0.087, with a standard error of 0.020, which is significant. Next, the partition of the variance explained across the two levels, weeks and brands can be inferred. Brand level variance explained here is 0.004, while week level variance explained is 1.274. Total variance explained is therefore 1.278. The percentage of the total variance explained at week level is therefore 99,69%, variance to be explained at brand level, or intraclass correlation is 0.31%. For the dependent variable Firm growth market share, the percentage of the total variance explained at week level is therefore 99,84%, variance to be explained at brand level, or intraclass correlation is 0.16%. In summary, for both the model for firm growth market share, and for firm growth volume, the variation between occasions within brands is large and overwhelms the variation between brands.

4.2.3. Full model

(29)

29

from the output Displayed in the appendix. For Firm Growth Market Share, brand level variance explained here is 0.000, while week level variance explained is 2,389. Total variance explained is therefore 2,389. The percentage of the total variance explained at week level is therefore 100%, variance to be explained at brand level, or intraclass correlation is 0.0%. For the dependent variable Firm growth market share, brand level variance explained is 0.000, while week level variance explained is 1,221. Total variance explained is therefore 1,221. The percentage of the total variance explained at week level is therefore 100%, variance to be explained at brand level, or intraclass correlation is 0.00%. Therefore, the fact that level-1 explanatory variables were made random at level-2 has indeed explained part of the error term at brand level.

Next, fixed effects were added for each level-2 variable, by using cross-level interactions for each level-1 variable for price, price promotions and non-price promotions and each level-2 variable and incorporating this into the model. Hereby, the indirect effects of each level-2 variable on firm growth are added to the model. Level two effects are expected to explain part of the residual and improve model fit. The covariance parameter for the residual reflects the within-subjects variance, and as such is the unexplained variance in the model. As the covariance parameter has decreased for both the model for firm growth volume and the model for firm growth market share, this indicates that the level two variables indeed have an effect on the level 1 intercept and the level 1 slope, and explain part of the excess variability in the dependent variable. The final model is displayed in the appendix.

4.2.3. Checking assumptions

Several assumptions had to be checked furthermore, before stating the final model. First, the continuous PLC variable was checked for non-linearity, as, according to the theory on the product lifecycle stage, it is very likely that a "U"-shaped link exists between the residuals and an explanatory variable. When a residual plot shows a "U"-shaped link between the residuals and an explanatory variable, the fit of the model can be improved by introducing the square of that explanatory variable as an additional variable in the model. A squared term of the continuous PLC variable was therefore also included into the model. As the deviance of the model with PLC squared did not decrease the deviance, the continuous PLC variable can be considered linear. However, in order to make the interpretation of this variable easier, dummy variables for each separate stage of the brand-level product lifecycle were included into the model. Adding dummy variables to the model instead of the continuous PLC variable has even slightly decreased the deviance. Therefore, the final model will be the model with dummy variables.

(30)

30

4.3.4. Model evaluation

A -2log-likelihood value is the probability of obtaining the observed data if the model were true and can be used in the comparison of two different models. (Rasbash, et al., 2005). The deviance (-2log-likelihood value) of the simpler model (Ds) minus the deviance of the more complex model (Dc) provides the change in deviance (ΔD = Ds− Dc). In large samples, the difference between the deviances of two hierarchically nested models is distributed as an approximate chi -square distribution with degrees of freedom equal to the difference in the number of parameters being estimated between the two models (de Leeuw, 2004). In particular, the full model contains the parameters of the empty model, and k additional parameters. Then, under the null hypothesis that the full model is the true model, the difference between the deviances for the two models follows an approximate chi-squared distribution with k-degrees of freedom. At a 0,10 acceptance level, this would mean that the change in deviance should be between 90% and 110% of K. As K is 4036 in this case, the difference between the empty model and the full model deviance should be between 3632,4 and 4439,6. Therefore, the chi-squared test does not hold. However, commonly, the deviance of the more complex model must be lower than (or as low as) that of the simpler model. The Deviance values of both the empty model and the full model are displayed in table 6. The deviance of the models has decreased to 13777,353 (Firm Growth Volume) and 16823,371 (Firm Growth Market Share),and therefore, the random coefficients have indeed led to increased model fit. The MLwIN output of both the full model, and the empty model can be found in the appendix . Table 6: Deviance Full model

Model Dependent Deviance

Empty model FG Volume 13982,518

FG Market Share 16898,199

Full model FG Volume 13777,353

FG Market Share 16823,371

Table 6: Model fit (Deviance)

4.3.5. Hypotheses evaluation

The hypotheses were tested for different weeks within different brands. Statistical significance can be deduced from the relationship between the coefficient for a variable and the standard error by means of a Wald test, or dividing the coefficient by the standard error. The significance was tested at both a .05, and a 0.10 significance level. The following section will discuss the results as presented in the MLwIN output in the appendix.

Overall model

(31)

31

Price promotion

In this research, price promotions are categorized in TPR promotions versus Multibuy promotions. Expectations were that that both types of price promotions have a positive direct short-term effect on firm growth, however, a TPR promotion was expected to be more effective than a Multibuy promotion. Furthermore, expectations of this research were that all types of price promotions would be most effective in the in the start-up and/or growth stage, followed by the maturity stage. All types of price promotions were expected to be least effective in the decline stage of a brand’s lifecycle. Also, Multibuy promotions were expected to have the largest positive effect in the maturity stage of a brand’s lifecycle, while a TPR promotion will have the largest positive effect in the start -up and growth stage of a brand’s lifecycle. The stage of the brand level PLC was measured by incorporating three dummy variables into the model. The direct effects of the variables TPR and Multibuy on Firm Growth cannot be assessed separately, as they will be altered by incorporating level -2 variables into the model. Therefore, these effects are displayed in the tables below.

The p-values that are displayed in this tables are assessed at a (one -sided) 0,10 level of significance. Before discussing these variables more in-depth, it should be stated that almost no variables were significant. This is perhaps not surprising, given the fact that, as stated on page 28, for both the model for firm growth market share, and for firm growth volume, the variation between occasions within brands is large and overwhelms the variation betwee n brands. However this has as a consequence that the analysis of these variables can only be done by explaining the directions of the parameter estimates only, without looking at significance.

For Firm Growth Market Share the first two tables above display how the effects of both types of

promotions depend on the level-2 effects. First, TPR promotions will be analyzed.

FGMS Parameter estimate Standard deviation Wald test P value

TPR 0,004 0,148 0,027 0.489 *PLC1 0,003 0,153 0,020 0.492 *PLC2 -0,001 0,152 -0,007 0.502 *PLC3 0 0 0,000 - *BP -0,022 0,04 -0,550 0.708 *WD -0,009 0,065 -0,138 0.554 *CC -0,006 0,027 -0,222 0.587 *CG 1,791 5,684 0,315 0.624

(32)

32

promotions seem to be less effective when combined with high weighted distribution, although these effects are not significant. Therefore, the hypothesis that both types of promotional strategies are more effective when combined with high weighted distribution, cannot be accepted. Fourth, Category concentration was expected to have positive indirect effect on the relationship between promotional strategies and firm growth. However, the parameter estimates of category indicate a negative effect on the relationship between both types of promotions and firm growth. However, as this effect is not significant at a 0,10 level, the hypothesis cannot be confirmed. Finally, in line with expectations of this research, Category growth positively influences the relationship between both types of price promotions and firm growth. Parameter estimates indicate a quite large influence, however, as this effect is not significant at a 0,10 level, the hypothesis cannot be accepted.

FGMS Parameter estimate Standard deviation Wald test P value

MB 0,02 0,175 0,114 0,455 *PLC1 -0,012 0,179 -0,067 0,527 *PLC2 -0,014 0,077 -0,182 0,572 *PLC3 0 0 0,000 - *BP -0,006 0,047 -0,128 0,551 *WD 0,017 0,05 0,340 0,367 *CC -0,005 0,028 -0,179 0,571 *CG 1,006 6,586 0,153 0,439

The effect of different types of promotional strategies on firm growth